oral delivery of poorly soluble actives – from drug discovery to marketed products ·...

®

Oral Delivery ofPoorly Soluble Actives– From Drug Discoveryto Marketed Products

Tokyo, Japan • June 6, 2003

Ora

l Del

iver

y of

Poo

rly S

olub

le A

ctiv

es –

Fro

m D

rug

Dis

cove

ry to

mar

kete

d Pr

oduc

ts

Oral Delivery ofPoorly Soluble Actives– From Drug Discoveryto Marketed Products

Tokyo, Japan • June 6, 2003

Contents

Introduction by Roland DAUMESNIL, Capsugel

5

Opening Remarks – An overviewProfessor Shinji Yamashita

7

1. Special Lecture

Professor Yuichi SugiyamaFactors determining the bioavailabilityof drugs: interplay between drugmetabolizing enzymes and transporters

13

Dr. Christopher A. LipinskiPoor aqueous solubility - an industrywide problem in ADME screening

31

2a. Invited lecture

Dr. Hirokazu OkamotoNovel approaches for oral deliveryof poorly soluble drugs

49

Dr. Hiroshi KikuchiImportance of dose numberand absorption test in formulationoptimization: an industrial case

63

Dr. Akira KusaiHow to handle practically insolubleAPIs: our experience at Sankyo

73

2b. Invited lecture

Dr. Soon-Ih KimAccelerating discovery and developmentof poorly water-soluble actives by theAqueous Solubilizing System (ASS)

83

Dr. Hassan BenameurFormulation of poorly soluble actives:how to make it an industrial reality

91

2

2c. Invited lecture

Professor James PolliIn-vitro dissolution: methodconsiderations and relations to in-vivo

101

Dr. Lawrence X. YuWater insoluble drugs: scientific issuesin drug development and drug regulation

115

3. Open discussion

Roland DaumesnilCase studies/ marketed products

127

Closing Remarks by Roland DAUMESNIL, Capsugel

139

3

5

Introduction

FFERING A PLATFORM for the exchange of scientific ideas is a very well-known andwell-established Capsugel commitment. We have been doing this for 12 years and

the best examples has been the series of symposia we organized around theBiopharmaceutic Classification System (BCS) and the optimization of oral drug delivery

systems.

Today, we are going to look at oral delivery of poorly soluble actives, from drug discovery tomarketed products. Poor aqueous solubility is an industry wide problem in ADME screening.Newer drugs are larger, more lipophilic and less permeable. Whatever is the reason for thispoor solubility, excessive lipophilicity or crystal packing, formulating this kind of componentsis a real challenge.

Thus we are here to assess where we are today and to discuss the factors determining thebioavailability of these drugs. What can be the novel approaches to formulate class IIcomponents? As usual some industrial experts will share their experience and expertise bypresenting real cases. The drug regulation of poorly soluble actives will also be reviewed.

In 2002, the market share of poorly soluble actives was 26% or 110 billion USD. Theyrepresented 58% of the NCE's launched. It is well accepted that 40 - 50% of product underdevelopment are class II products. No need to say that we need to try different approachesto circumvent the issues to formulate such components.

Some industrial examples will be presented at the end of the symposium and willdemonstrate that new industrial technologies are now available to achieve an acceptableabsorption.

Our ultimate goal in organizing and publishing these symposia is to provide timely input tobuild the science around the formulation of poorly soluble actives. Your feedback will tell us ifwe achieved our objectives.

Looking forward to hearing from you.

Roland Daumesnil

Director, Global Business Development - Pharmaceutical

O

7

Opening remarks – An overview

Professor Shinji YAMASHITA

9

Noriko Yamanuchi, Capsugel Japan: Ladiesand gentlemen, I am Noriko Yamanouchi of CapsugelJapan. We are very pleased to open this symposiumon the Oral Delivery of Poorly Soluble Actives. I'd liketo welcome you. The Masters of Ceremony for todayare myself, and Roland Daumesnil of Capsugel.

As for today's program, in the morning there will belectures and a discussion and in the afternoon our in-vited lecturers are scheduled for an open discussionforum. We have to go through a very tight scheduletoday and we really would hope for your co-operationin this. On behalf of the organizers, I would like to askProfessor Yamashita of Setsunan University to say afew opening words.

Professor Yamashita, please.

Chair, Professor Shinji Yamashita, SetsunanUniversity: Good morning, I am very pleased to behere and on behalf of the organizers would also like tothank the audience for coming. Our topic today is thediscovery and design of poorly soluble actives andthis is the latest in this series of symposia. I'd like tothank Roland Daumesnil, Phil and other colleagues forCapsugel Inc.'s involvement. I would like to extendour hearty appreciation for your support of this sym-posium.

Before I open the symposium I must inform youthat my co-organizer, Professor Gordon Amidon ofMichigan University, who was very much looking for-ward to coming to this symposium, is unable to do sothrough unavoidable circumstances. He has askedme to convey his best regards and sincere apologies

to you. In his place and speaking on his behalf wehave another distinguished lecturer, Professor JamesPolli. Dr. Polli is a young professional scientist whospecializes in this area and I believe you will really en-joy hearing what he has to say.

With your permission I would now just like to tellyou about the purpose of the symposium and the to-pics we will be discussing. The title of today's sympo-sium, as mentioned in the introduction, is the OralDelivery of Poorly Soluble Actives: from Drug Disco-very to Marketed Products, and we will be exploringthe subject comprehensively. Let me give you anoverview of why we have chosen this topic.

In the past 10 years, drug development strategyhas undergone drastic changes. Now that we areusing combinatorial chemistry or high-throughputscreening methods, we have been able to produce

Opening remarks– An overview

Professor Shinji Yamashita

Setsunan University

Target ID(Genomics)

Target Validation(FunctionalGenomics)

Lead ID(HTS

Primary Screening)

Lead Optimization(ADME-Tox)

Formulation Study

ADME Screening

BCSBiowaiver of

in vivo BE study

Clinical Trials RxDx

Compound Library

(CombinatorialChemistry)

Figure 1.

IMPORTANCE OF SOLUBILITY AND PERMEABILITYASSESSMENT IN DRUG DEVELOPMENT

10

numerous candidates which can be further screened(Figure 1). High-throughput screening (HTS) andcombinatorial chemistry are targeted at pharmacologi-cal efficacy, or the activity, and so the lead optimiza-tion stage, or the ADME screening stage, could turnout to be a barrier to further development. How canthese issues be resolved, particularly with oral formu-lation?

Of course, solubility is important. Once the drug isdissolved it then permeates through the intestinalmembrane, which means both the permeabilityandsolubility have to be evaluated before reaching formu-lation stage. Once solubility and permeability havebeen determined and assessed, we have to narrowdown the NCE candidates during early stage develop-ment, so screening becomes important. These para-meters continue to be important at formulation stage.

Also, when those drugs are in clinical trial or atpost-marketing stage, we have to comply with regula-tions such as the US guidance brought out by theFDA (Food and Drug Administration), which includesthe Biopharmaceutical Classification System (BCS).Where this particular standard is concerned wewould, of course, like to obtain a waiver of the in-vivobioequivalence study (Figure 2). Therefore, solubilityand permeability have to be evaluated at every stageof development and by using those parameters wecan achieve a shorter time to market for new chemi-cal entities. That is the topic for today.

With both early-stage development and regulatoryadherence, the parameters are the same, eventhough the criteria and methodology may be different.

In fact, it is important to look at the total scientific pic-ture. Figure 3 is a well-known chart showing the dis-solution rate and solubility. To arrive at the permeabi-lity, three parameters have to be evaluated (An, Doand Dn). The Fmax, or the absorption, can then be de-termined by adjusting the three parameters.

At every stage of pharmacological development,specialist scientists have to look at those parametersto optimize the formulation and maximize the efficacyand safety of the drug. How to achieve this is the ba-sis of the discussion that we are going to have.

With regard to the associated regulatory issues,Professor Lawrence Yu will be discussing the BCS,his area of expertise at the FDA regulatory agency. Hewill be focusing on the important parameters of solu-bility, permeability and dissolution and discussing theirrelevance and the criteria involved (Figure 4). Usingthose parameters you may be able to re-examineyour candidate drugs with a view to obtaining a bio-

Guidance for Industry

Waiver of In-Vivo Bioavailability and BioequivalenceStudies for Immediate-Release Solid Oral Dosage Forms

Based on a Biopharmaceutics Classification System

Additional copies are available from: Office of Training and Communications Division of Communications

Management Drug Information Branch, HFD-210 5600 Fishers Lane Rockville, MD 20857

(Tel) 301-827-4573 (Internet) http://www.fda.gov/cder/guidance/index.htm

U.S. Department of Health and Human ServicesFood and Drug Administration

Center for Drug Evaluation and Research (CDER)August 2000

BP

Figure 2.

Pharm Res., 12,413-420 (1995)

Fmax = 0.86

F

Dn

Do

An = 1.0

Figure 3.

FRACTION DOSE ABSORBED VS Dn AND Do AT An = 1.0

Class boundary in BCS

Solubility• Highly soluble when the highest dose strength is solu-

ble in < 250mL water over pH range of 1 to 7.5 at 37°C.

Permeability• Highly permeable when extent of intestinal absorption

in humans is > 90% of dose.

Dissolution• Rapidly dissolving when > 85% dissolves within 30

min in 0.1 HCl, pH 4.5, and pH 6.8 buffers.

Figure 4.

11

waiver of the in-vivo studies or you may be able tosimplify the later stage of the development process.

It is important to consider these different applica-tions from early development stage (Figure 5). Indeed,even at drug discovery stage. For instance, it can beuseful in meeting the regulatory requirements for thecompounds, or it might help in coming up with thescientific rationale behind the candidates or in selectingthe most suitable compounds for future development.

At late formulation stage you will again need to usethese parameters to determine the cause of poor ab-sorption, for instance. In fact, you will be relying onthose parameters and their different criteria throu-ghout the later stages of drug development, and oneof today's topics will be about the most efficient andeffective ways of utilizing them at each stage of deve-lopment.

Figure 6, for instance, shows the relationship bet-ween solubility and permeability. At an early stage, atleast, 20 percent absorption seems to be the cut-off

value. At this point, you can choose the candidates li-kely to have those properties. Then at a later stagewhen you perhaps might want 80 percent absorption,you might select the compounds within the right-handquadrant of Figure 7. Of course it depends upon par-ticle size and so forth, and on how well the formula-tion technology works in terms of delivering the maxi-mum range of drug efficacy.

Likewise, if we apply the BCS in discovery stageand find by formulation stage that it is a Class I drug,meaning that it has high solubility and high permeabi-lity, there is no problem (Figure 8). But if the drug isClass II, with low solubility, then a formulation studymay be needed to increase absorption; even aClass II drug can be dealt with at pre-formulationstage. So these are the issues to be incorporated intoyour drug development strategies.

BCS in Drug Discovery and Development

Drug Discovery• Scientific insights to rank order compounds

Solubility, Permeability, Stability, etc……

• Selection of compounds for future evaluation

Drug Development• Determination of causes of poor oral drug absorption• Formulation strategies

Figure 5.

100

0.11 100 1000

Solubility (µg/mL)

Dose :100 mg Suspension: particle radius = 10 µm

Pef

f (hum

an)(c

m/s

ecx1

0-4)

10

1

10

Fa = 80%

Fa = 20%

Fa = 50%

Solution 1.24 cm/sec x 10-4

0.49

0.15

Figure 6.

PERMEABILITY - SOLUBILITY RELATION IN DRUG ABSORPTION

100

0.11 100 1000

Solubility (µg/mL)

Dose :100 mg Suspension: particle radius = 1 µm

Pef

f (hum

an)(c

m/s

ecx1

0-4)

10

1

10

Fa = 80%

Fa = 20%

Fa = 50%

Figure 7.

PERMEABILITY - SOLUBILITY RELATION IN DRUG ABSORPTION

High Solubility

No problem

Class I

➞

Low Solubility

• Drug absorption can be improved by appropriate formulation

Class II

Hig

h P

erm

eabi

lity

• Drug absorption is regulated by permeability

• Difficult to improve by formulation study

Class III

Low

Per

mea

bilit

y

No way orVery difficult for

oral use

Class IV

Figure 8.

BCS FOR DRUG DISCOVERY (OUTLINE)

12

If I may take a little more of your time… Last monthI attended a Pharmaceutical Profiles' symposium onSelection in Drug Discovery, held in the US, andheard the talk given by Professor Borchardt of theUniversity of Kansas (Professor Lipinski was anotherof the speakers). I was rather impressed with Profes-sor Borchardt’s talk, analyzing the educational andcommunication issues associated with integratingand applying drug data during discovery stage. Thetalk illuminated the importance of educating the per-sonnel involved. I was so impressed that I e-mailedProfessor Borchardt after I returned to Japan, askingwhether his talk could be communicated to Japanesecolleagues.

I would just like to use some of the ideas in his talkto illustrate how important the human element is in thedrug development program. For instance, when weare talking about technologies such as solubility as-sessment and how to improve absorption, we have tostart with the question: what do we need for develop-ment? And the correct answer is, what we need isthe right scientists. What do we mean by 'the rightscientists'? Well, Professor Borchardt drew up a list ofthe necessary qualifications and the two most impor-tant are, scientific depth and scientific breadth.

Scientific depth means having a profound unders-tanding of your own scientific discipline, and scientificbreadth means understanding related disciplines aswell. So both are needed – breadth and depth. Untilnow, the educational focus has been solely on scien-tific depth. As a result, even if we have great scien-tists they only understand their own discipline. Thereare many drawbacks in this attitude which must beovercome in trying to meet the US protocol for thedevelopment program.

The current educational program fails either at dis-covery stage or at formulation stage. For instance, inthe course of development, the way to deal with pre-cisely the same compounds will vary. Maybe youneed 80 percent purity at the initial stage but in the la-ter stage you need 99 percent purity. Unless you un-derstand the different criteria involved there will beconflict between scientists in different disciplines.

Professor Borchardt would therefore like scientiststo have both types of understanding because, in theknowledge-based future, the focus will be on a multi-disciplinary approach incorporating scientific depthand breadth. Whether your background is in chemis-try, biology, or pharmaceutics, multidisciplinary trai-ning is necessary so that we can achieve a more effi-cient and more effective drug development programin the future.

Although this is a one-day seminar, if you are achemist, biologist or pharmaceutical scientist, pleaselook at the other disciplines so that your scientificbreadth can be expanded. With this I conclude myopening remarks. Thank you very much for your at-tention.

Well, let us get into the first morning's session,which I am chairing. In this morning's session wehave two special speakers. The first honorable lectu-rer is Professor Yuichi Sugiyama of the University ofTokyo. For his career background, please refer to thefirst page of your handout. Actually, he needs no in-troduction, therefore I will be very brief. I would like togive him plenty of time to discuss the topic, while lea-ving some time for discussion. His lectures are al-ways insightful and stimulating, and his talk incorpo-rates both scientific depth and breadth. So –Professor Sugiyama.

13

Factors determining thebioavailability of drugs: interplay

between drug-metabolizingenzymes and transporters

Professor Yuichi SUGIYAMA

15

Professor Yuichi Sugiyama, University ofTokyo: Thank you very much for the introduction. Firstof all, let me express my gratitude to my colleagueRoland Daumesnil for holding this stimulating sympo-sium. Of course I also appreciate Professor Yamashita'swork as an organizer – he is a great organizer. Also,yesterday our speakers and chair persons got togetherand I would like to thank all the staff from Capsugel.

As Professor Yamashita has already mentioned,intestinal absorption depends on a membrane perme-ability, rate and extent of dissolution and metabolic sta-bility. In recent years, various drug transporters in theliver and the small intestine have been analyzed to deter-

mine the pharmacokinetics of the compounds – togetherwith, of course, metabolizing enzymes. Hepatic enzymes,such as the P450, transporter play an important role.Today, I will be talking about the transporters in the gas-trointestinal (GI) tract and the liver, and the interplay ofthese important transporters.

I'll start with the GI tract transporters but before I doso I'll just mention xenobiotic detoxification. Figure 1shows the liver. Various xenobiotics come into the liverfrom the vessel, the parent compound is metabolizedand then an efflux system comes into operation toremove them. In the GI tract the efflux transporter andthe influx transporter may interplay, working together todetoxify the xenobiotics in those organs.

Now I'll talk about P-glycoprotein (P-gp) and theCYP3A4 enzyme. Figure 2 is a very well-known schemafrom 1996 by Les Benet. It is a comparison of cyclo-sporine A and midazolam. If you run a pharmacokineticstudy in human with these orally administered drugs,bioavailability is about 30 percent. But within the gut theclearance, or the metabolism, is similar to that of theliver or maybe higher. This was the first occasion thatthey were able to demonstrate this. Here in this caseboth drugs are the 3A4 substrates, and Cyclosporine isa good substrate also for P-glycoproteins. What LesBenet frequently says is that CYP3A4 and P-glycopro-tein have substrate features in common; they share thesame substrate and the same inhibitors. However, thereare exceptions, so this is just a rule of thumb and not auniversal rule.

Figure 3 shows that it can work as the substrate.From the lumen the drug is taken up by the epithelial

Factors determining the bioavailability ofdrugs: interplay between drug-metabolizingenzymes and transporters

Professor Yuichi Sugiyama

University of Tokyo

-OH

-OX

ATP

ATP

ADP

ADP

OCT 1NTCP

OATP-8

MDR1

MRP6

MRP2

SPGP/BESP

OATP-B OAT2

ATP

ADP

ATPADP

ADPATP

MRP3OATP-

A, D, E?OATP-2/OATP-CLST-1

Phase I

Phase I

Figure 1.

XENOBIOTICS DETOXIFICATION SYSTEM IN HUMAN LIVER

16

cells of the GI tract and if there is P-glycoprotein theefflux system works, and then after efflux there is move-ment downward the intestinal lumen, and the uptakeand efflux recycle many times. Without P-glycoprotein,of course, the drug would be more efficiently absorbedin the bloodstream. However, the presence of the P-glycoprotein means increased exposure with a highprobability of the metabolism because of the recyclingthrough the joint effect of the P-glygoprotein andCYP3A4.

But first of all, what is P-glycoprotein? Let me showyou a rather good animation (Video not supplied). Sup-pose this is the layer of the epithelial cell, this is thelumen side and this is the cytosol side. The drug or thexenobiotic is transported in the lumen and then it ispermeated through the lipid bilayer. Within that layer P-glycoprotein recognizes the drug and the conforma-tional change of P-glycoprotein takes place, and then

the efflux system works and the xenobiotics are takenout to the lumen. In this model, of course, the substratebinds to the P-glycoprotein within the lipid bilayer mem-brane and then it is taken out so the P-glycoprotein isworking as a flippase here. Now I am going to talk aboutthe work done by Dr. Kiyomi Ito, who is now at KitasatoUniversity. She and I heard Les Benet's talk and wethought that we might be able to come up with a math-ematical model to illustrate the concept which Dr. Benetdescribed, and we started on this work. We used thepartial differential equation with boundary conditionshown in Figure 4. A boundary condition is the condi-tion at the membrane wall of luminal side, which thenincorporates the metabolism/diffusion/influx and the P-glycoprotein mediated efflux. The Figure shows the ini-tial conditions and the boundary conditions which explainthe interplay of these two molecules, that is P-glyco-protein and CYP 3A4. The work was published in Phar-maceutical Research.

To make a long story very short, if you look at theleft-hand side of Figure 5. Only the inhibition of the P-gly-coprotein affect the intestinal absorption to some extent.So did the inhibiton of only CYP 3A4. However, whenboth are inhibited, the change in absorption is quitelarge, larger than with the single inhibition, so a syner-getic effect is clearly shown in our mathematical model.

Although we were able to calculate this, it is not soeasy to understand intuitively so I asked Dr. Ito to comeup with the schema in Figure 6. In simple language, thetop part represents the lumen side, the bottom part theblood vessel side. After absorption within a cell you cansee the diffusion. If the diffusion is slow enough, fol-lowing uptake the drug is effluxed out by the P-glyco-

Calculated values for gut and hepatic extraction in human

Foral ERG ERH

Cyclosporin A(1) 0.22~0.36 0.43~0.66 0.24~0.27Midazolam(2) 0.30±0.10 0.40±0.24 0.44±0.14

(1) C-Y. Wu et al., Clin. Pharmacol. Ther.,58, 492- (1995) (2) K.E. Thummel et al., Clin. Pharmacol. Ther.,59, 491- (1996)

Gut LiverFabs

ERG ERH

Foral

Fabs x FG Fabs x FG x FH

Figure 2.

SCHEMATIC DIAGRAM DEPICTING THE EFFECT OF ABSORPTION,GUT AND HEPATIC FIRST-PASS EXTRACTION ON DRUG

ORAL BIOAVAILABILITY

K.Ito et al. Pharm.Res. 16: 225-231 (1999)

Lumen

Epithelialcell

Initial condition

Boundary condition

Lumen

Epithelial cell

δC(x,t)δt

QAr, L

δc(x,t)δx

PSinfVL

flow influx efflux

VeVL

C(x,t) + PSeff x Ce(x, 0, t)= - x

δCe(x,y,t)δt

δ2Ce(x,y,t)δy2

diffusion

D

C(x,0) = 0

Ce(x,y,0) = 0

CLm x Ce(x,y,t)= -

metabolism

δCe (x,0,t)δy

PSinf x C(x,t) = PSeff x Ve x Ce(x,0,t) - D x Ar,e

CLab x Ve x Ce(x,M,t) = -D x Ar,eδCe (x,M,t)

δy

Luminal

Basolateral

Figure 4.

Calcium-channel blockersDiltiazem, Nicardipine,Verapamil

Chemotherapeutic agentsEtoposide, Paclitaxel,Vinblastine, Vincristine,Vindesine

HormonesDexamethasone, Estradiol,Hydrocortizone

ImmunosuppressantsCyclosporin, FK506,Rapamycin

OtherDigitoxin

V.J. Wacher et al., Molecular Carcinogen., 13:129-134 (1995)

Flow

Epithelialcells

Small Intestine

Absorption

P-gp

CYP 34A

Circulation

Circulation

Flow FlowD D

D

D

D

D

Metabolism

Metabolism

Figure 3.

OVERLAPPING SUBSTRATES FOR BOTH CYP3A AND P-GP

17

protein; so the P-glycoprotein role is obvious. If the dif-fusion process is very fast or there is good permeabilityto the vessel, then once it is taken in the cells, the effluxby P-glycoprotein may not take place so efficiently. Thisis the interplay confirmed by the mathematical modelthat we have generated.

Maybe in the future we will be able to reach the sameconclusions without such a difficult mathematical model.If we consider the epithelial cell as a simple compartment,then we should be able to arrive at a simpler equation.From the lumen, a compound is taken up and if the per-meability is very high then total absorption is limited bythe influx rate, therefore there is no metabolic effect.However, if the permeability is low enough, then notonly the influx but the efflux capacity or the metabolicpathway would have a major direct effect on absorp-tion. You can explain these phenomena with the simplemathematical model and so, perhaps, instead of usingthat first very complicated mathematical model, maybewe can use this simple equation to illustrate the phe-nomena.

The concept of “enzyme and transporter interplay”may be demonstrated by using specific inhibitors whichwe recently identified (Figure 7).

There is some supportive evidence for the idea ofinterplay between CYP3A4 and the P-glycoprotein. Thesame transcription factor regulates the expression ofboth 3A4 and the P-glycoprotein. It is the PXR/RXRalpha heterodimer, which has been found to be the co-

00

PSefflux or CLm (% of control)

50 100

500

250

P-gp inhibit

PSefflux = 0.1CLm = 0.2

J Ato

B (%

of c

ontr

ol)

00 50 100

500

250

CYP3A4 inhibit

PSefflux = 0.1CLm = 0.2

00 50 100

500

250

P-gp,CYP3A4 inhibit

PSefflux = 10CLm = 0.2

Inhibition K.Ito et al. Pharm.Res. 16: 225-231 (1999)

Figure 5.

EFFECT OF INHIBITORS ON THE APICAL TO BASAL ABSORPTION OFSUBSTRATES FOR BOTH CYP3A4 AND P-GP: SYNERGETIC EFFECT

Diffusion istoo rapidto comeback

Diffusion isslow enoughto comeback Drug taken up into the epithelial cell

Absorbed drug

Portal vein

Metabolism

MetabolismMetabolism

Metabolism

LumenRapiddiffusion

Slowdiffusion

Intracellular

Figure 6.

Comparison of IC50 values for CYP 3A4 and P-gp

CYP 3A4 P-gp CYP3A4 / P-gp ratio

Ketoconazole 0.01 - 0.04 > 1 0.01 - 0.04

Verapamil 10 - 20 > 30 0.33 - 0.67

L754, 394 0.006 - 0.04 > 3 0.002 - 0.01 CYP3A4 Specific

PSC833 (Valspodar) 4 - 7 0.03 - 0.1 40 - 233 P-gp Specific

Figure 7.

18

upregulator of the expression of those two molecules(Figure 8). Therefore, both the transporter and theenzyme may be working together to accelerate detox-ification, or to inhibit intestinal absorption of the com-pound.

Next, I'll talk about the functions of the P-glycoprotein.Well, again to make a long story short, we used a P-glycoprotein knock-out mouse, mdr1a(-/-), to look atdistribution to the brain.

Using the knock-out mouse we can see the increasein drug distribution in the brain (Figure 9). For instance,number 12, which is quinidine, shows a 25-fold increasein BBB (blood brain barrier) transport. With number 11,digoxin, the increase in BBB transport is 11-fold but if youlook at number 1, diazepam, or number 2, proges-terone, they do not offer any P-glycoprotein substrateand there is no change in the brain distributon in themdr1a(-/-), knock-out mouse.

In Figure 10, the Y-axis shows the in-vitro result ofP-glycoprotein function, using an expression system.The X-axis shows the P-glycoprotein function in-vivo(assessed by using intestinal perfusion system), andthe in-vitro/in-vivo correlation is very good. So the in-vitro performance can predict the in-vivo result.

Figure 11 is the model that I used for the quantita-tive analysis, and I'd just like to give you the result of thisstudy, where we compared the performance of P-gly-coprotein in the normal mouse vis-à-vis the knock-outmouse. Here in this model, it was assumed that themembrane permeabilities which came from other mech-anisms than P-glycoprotein mediated transport are thesame between normal cells/mouse and the knock-outcells/mouse. We must look at the proportion of the bal-ance between the two functions. Even when the com-pund is a P-glycoprotein substrate, if passive diffusion is

higher than the P-glycoprotein mediated transport, the P-glycoprotein does not function, as you may understandfrom the final equation shown in the bottom of the line.

Since our topic is GI tract absorption, using a knock-out mouse we looked at the intestinal perfusion result aswell as the result of the normal mouse, and in Figure10 the X-axis shows the absorption change and the Y-axis shows the in-vitro results.

In general, there is some correlation; however, whatI would like you to note here is number 11, quinidine.Digoxin, number 7, the maximum increased absorptionin the GI tract in the knock-out mouse was just two-fold,so the change is very small. In the case of quinidine, thechange is eight-fold. However, with all the other com-pounds, even if their transports to the brain were changedvery much in the knock-out mouse, GI absorption doesn'tchange greatly; though there are some changes.

Rifampicin etc.

PXR/RXRαheterodimer

CYP3A4

CYP3A4

P-gp

Rifampicin etc.

PXR/RXRαheterodimer

P-gp

Geick A, et al., J Biol Chem. 2001 May 4;276(18):14581-7.Goodwin B, et al., Mol Pharmacol. 1999 Dec;56(6):1329-39.

Figure 8.

TRANSCRIPTIONAL REGULATION OF CYP3A4 & P-GP BY PXR/RXRα

PSa-to-b ratio(in vitro transcellulartransport)

1. Cimetidine2. Progesterone3. Diazepam

4. Dexamethasone5. Cyclosporin A6. Vinblastine

7. Digoxin8. Verapamil9. Daunomycin

10. Loperamide11. Quinidine

PSinf ratio(in situ perfusion)

Figure 10.

CORRELATION OF P-GP FUNCTION DETERMINEDIN THE TRANSCELLULAR TRANSPORT STUDIES AND

IN SITU INTESTINAL PERFUSION STUDIES

1 + PSP-gp, u / PS2, u(transcellulartransport studies)

1 + PSP-gp, u / PS2, u (in vivo studies)

1.Diazepam2.Progesterone3.Daunomycin4.PSC833 5.Dexamethasone6.Ondansetron7.Loperamide8.Verapamil9.Vinblastine10.Cyclosporin A11.Digoxin12.Quinidine

Figure 9.

CORRELATION OF P-GP FUNCTION DETERMINEDIN THE TRANSCELLULAR TRANSPORT STUDIES AND THAT IN

IN-VIVO STUDIES IN NORMAL AND MDR1A(-/-) MICE

19

Now I would like to show you the clinical data thatmatches the in-vitro data. Figure 12 is Dr. Hoffmeyer'sdata. It shows some polymorphism and in some patientsthe GI tract P-glycoprotein expression level is very low,down to one-tenth. Even if the expression level is one-tenth of the normal, if you look at digoxin's AUC (thearea under the curve), you can see that it has reduced

only two-fold. If you remember, P-glycoprotein is defec-tive in the knock-out mouse, and yet absorption does notchange greatly. So the clinical result matches the dataobtained with knock-out mouse.

Of course, the function of the P-glycoprotein is mod-eled in the intestinal absorption. However, P-glycopro-tein would actually work best in the permeability of drugs

PSP-gp

PSinf

PSeff

PS1, u

PS2, u

PS3, u

blood brain Kp, brain in normal mice =

normal mice

PS4, u

PS4, u • (PS2, u + PSpg-p, u)

PS1, u+ PS3, u

fu, plasma • PS1, u •

=

=

fu, cell • (PS2, u + PS2, u + PSpg-p, u + PS3, u)

fu, cell • PS3, u

fu, plasma • PS4, u • fu, cell • (PS2, u + PSpg-p, u + PS3, u)

fu, cell • (PS2, u+ PSpg-p, u)

PSinf

PSeff

PS1, u

PS2, u

PS3, u

blood brain Kp, brain in mdr1a (-/-) mice =

mdr1a (-/-) mice

PS4, u

PS4, u • PS2, u

PS1, u+ PS3, u

fu, plasma • PS1, u •

=

=

fu, cell • (PS2, u + PS3, u)

fu, cell • PS3, u

Kp, brain in mdr1a (-/-) mice

Kp, brain in normal miceKp, brain ratio 1 + = =

fu, plasma • PS4, u • fu, cell • (PS2, u + PS3, u)

fu, cell • PS2, u

PS2, u

PSpg-p, u

Figure 11.

BRAIN TO PLASMA CONCENTRATION RATIO (KP RATIO) IN MDR1A(-/-) AND NORMAL MICE

P-g

p

MDR1 expression in the duodenum

A mutation in MDR1 results in...AUC ~ x 20

Cmax ~ x 1.4

1 400

MDR1 genotype in exon 26

1 200

1 000

800

600

400

200

0

N = 3CC (wt) CT TT

4 5

Dig

oxin

AU

C (g

*h /

L)

AUC

80


60

40

20

0

N = 3CC (wt) CT TT

4 5

Dig

oxin

C

Cmax

2.8


2.4

2.0

1.6

1.2

0.8

0.4

0

N = 7CC (wt) TT

4

S.Hoffmeyer et al.Proc. Natl. Acad. Sci. USA , 97, 3473-3478 (1999).

Figure 12.

AN MDR1 POLYMORPHISM AFFECTS THE PHARMACOKINETICS OF ORALLY ADMINISTERED DIGOXIN

20

in the BBB. We have to look at P-glycoprotein's differentcapacity in different tissues compared with the passivediffusion mediated permeabilities.

Let me talk about the interplay of these transporters.

P-glycoproteins in humans and in the rat areexpressed in the luminal side of the intestine, thereforethey are thought to be involved in the efflux of drugs.According to a recent study, the MRP2 transporter andthe BCRP transporter are also involved and expressedon the luminal side, and involved in the efflux of thecompounds. And on the basolateral side, i.e. vesselside, there is MRP3. We were the first researchers whoidentified this transporter. It is expressed in the baso-lateral side and involved in the function there.

What is most interesting are the OATP families. Theyare also on the lumen side and perhaps they may beworking as the influx transporter. Although we have noactual evidence of this, there are suggestions of suchan influx function.

Together with Dr. Rost we published a report inAm.J.Physiology in 2002 (282 : G720-726 (2002). Thelocalization of MRP2 is the apical side as well as the P-glycoprotein.. However, MRP3 does not match, clearlyshowing that the localization of the MRP3 is on the baso-lateral side (Figure 13). When we looked for the localiza-tion of MRP2 using Western blots, we found it in theupper tract, in the duodenum or jejunum. However, inter-estingly enough, MRP3 is expressed mainly in the lowerGI tract, in the ileum and colon.

One of the attendees here is Dr. Goto of Kissei Phar-maceuticals, and Figure 14 shows the work he did whilehe was with us in our laboratory. We wanted to seeMRP2 functioning within the intestinal tract. We have a

very good rat model called EHBR;MRP2 is defective inthis strain. We administered CDNB, a lipophilic com-pound which is taken in and conjugated to the glu-tathione conjugate, and excreted to the lumen side. Theefflux of glutathione conjugates into the intestinal lumenis reduced in EHBR, suggesting that this efflux is medi-ated by MRP2. Even if we use the everted sac, a sim-ilar result is obtained. While there is a large variation, inthe jejunum there is a significant difference, and thecontribution of MRP2 is demonstrated.

In Figure 15, we were able to prepare the basolateralmembrane from rat intestine and found that the MRP3substrates such as glucuronide conjugate and bile acidsare transported by this membrane in an ATP dependentmanner. Perhaps MRP3 is working as the efflux systemon the basolateral side. In the past we did not knowanything about transporters at the basolateral side but

Km = 0.82 ± 0.18 (µM) Vmax = 27.9 ± 6.3 (pmol/mg protein/min)Km = 35.4 ± 4.0 (µM)Vmax = 894 ± 47 (pmol/mg protein/min)

00

V0 (pmol/min/mg protein)

E2-17b-D-glucuronide

V0

/S (m

L/µg

pro

tein

/min

)

10

20

30

40

50

60

800600400200

Km = 16.4 ± 1.3 (µM) Vmax = 179.7 ± 10.7 (pmol/mg protein/min)

00

V0 (pmol/min/mg protein)

Taurocholate

V0

/S (m

L/µg

pro

tein

/min

)

2.5

5

7.5

10

12.5

20015010050

Figure 15.

CONCENTRATION DEPENDENCE FOR THE UPTAKEOF E217B G AND TC INTO COLON BLMVS

P-gp Mrp2 Bcrp

MRP 3CYP3A4

Metabolism

Lipophilicity

Absorption Absorption 9-sub

Excretion

Transcellulartransport

Paracellulartransport

Molecular size

Figure 13.

SCHEMATIC DIAGRAM ILLUSTRATINGTHE POSSIBLE FUNCTION OF MRP3 IN THE LIVER

Each point represents the mean ± S.D. of five animals.

Y.Goto et al., J.Pharmacol.Exp.Therap. 292: 433-439 (2000)

00 100

Time (min)

DNP-SGDNP-cysDNP-CGDNP-Nac

SD EHBR

Cum

ulat

ive

inte

stin

alex

cret

ion

(nm

ol)

10

20

30

40

2000

0 100

Time (min)

Cum

ulat

ive

inte

stin

alex

cret

ion

(nm

ol)

CL

perf

usat

e, D

.(n

l/min

/kg)

10

20

30

40

200

20

0DNP-SG

SD EHBR

40

60

80

100

P<0.05

Figure 14.

INTESTINAL EXCRETION OF METABOLITESAFTER IV ADMINISTRATION OF CDNB

21

MRP3 may turn out to be a good candidate for trans-porter involved in the efflux system at the basolateralside. As shown in the interplay schema the metabolicenzymes and the transporters may have a synergeticeffect and in that sense, maybe some common regu-lators are working in there, too.

The MRP families that we are working on, MRP1 orMRP2, accept the conjugates such as glutathione con-jugates and glucuronide conjugates as substrates.Therefore, the conjugating enzymes and also the effluxtransporters, may be co-regulated. For the past twoyears we have been trying to demonstrate this hypo-thesis.

Figure 16 shows the result of work on MRP1 tran-scription factors, carried out by our students. The tran-scription factor, Nrf2 may be associated with theup-regulation of the entire MRP family. It has been alreadyknown that this transcription factor up-regulate glutha-thione S-transferases.

If I may return to intestinal absorption. Figure 17 is ajoint study undertaken by Okudaira-San of Meiji Seika.The company had developed a compound calledME3277. It is an active form. As you can see, it hastwo carboxylic acids, the log-P is very small and per-meability is also low. Therefore, absorption is very low in-vivo. The approach Meiji Seika took with this compoundwas to create the prodrug. Of course, the log-P beco-mes higher and so membrane permeability rises, andthere is the chance that absorption in the GI tract willimprove. The thought was that maybe it would then bereleased as an active form into the blood vessels. Butif you look at the experimental results, actually this didnot help improve absorption.

That is why we started the joint research and wecame to understand that the mechanism involved meansthat the active form produced from the prodrug in the GIepithelium is effluxed out into the lumen by an activeefflux transporter. In Figure 18, we used an Ussingchamber and everted sac to see the results, If you lookat the control, it compares the mucosal to serosal (Mto S), and the serosal to mucosal (S to M) transport,and as you can see, the S to M secretory transportercapacity is higher than the other and if the ATP isdecreased then the difference is narrowed down. There-fore, this transport works by an active transport system.As you may recall, this drug is the anion type; there-fore, I thought, this efflux must be mediated by MRP2 soEHBR (Eisai hyperbilirubinemic rats) where MRP2 ishereditarily deficient was compared with normal rat,because the efflux should be decreased if MRP2 playsa role in the efflux process. However, we live in a worldmore complex than we expect.

O

OOCH2COO-

OCH2COO-

HN

Log P Papp(cm/sec)

< -3 8.50 x 10-8

1.2 1.04 x 10-5

N.Okudaira et al.J. Pharmacol. Exp. Ther., 294: 580-587 (2000)J. Pharmacol. Exp .Ther. 295: 717-723 (2000)

S

O

OOCH2COOnBu

OCH2COOnBu

HN

S

HN

HN

ME3277

ME3229

Figure 17.

CHEMICAL STRUCTURES OF ME3277AND ITS PRODRUG, ME3229

ME3277 concentration in the donor medium: 50µM

a) Ussing chamber b) Everted sac

00 30

Time (min)

Control (M to S)2-DG (M to S)Control (S to M)2-DG (S to M)

Control (M to S)2-DG (M to S)

Am

ount

per

mea

ted

(nm

ol)

0.3

0.5

0.8

1

60 90

Am

ount

per

mea

ted

(nm

ol)

00 20

Time (min)

1

2

3

4

5

40 60

Figure 18.

EFFECT OF GLUCOSE DEPLETIONON THE PERMEATION OF ME3277

Wild type + DEM

yGCS

GSH

ATP ADP+Pi

GS

GST

GS

MRP1

Wild type

yGCS

GSH

ATP ADP+Pi

GS

GST

GS

MRP1

Nrf2 knockout

yGCS

GSH

ATP ADP+Pi

GS

GST

GS

MRP1

Figure 16.

THE POSSIBLE REGULATORY MECHANISM BY NRF2

22

As you can see from Figure 19, there is some direc-tional transport. However, no difference was observedbetween normal and EHBR rats. Therefore, the trans-porter responsible for the efflux of this compound is notMRP2. I didn't know what this candidate transporterwas, so this research was discontinued.

However, six months ago the study was looked atagain because, according to recent research, the BCRPtransporter is known to be located on the lumen side.According to our own study BCRP accepts anion as asubstrate. BCRP is in the ABC transporter family but itis a half-size transporter. MDR1 and MRP2 have twobinding cassettes and after ATP binds to those cas-settes, the hydrolysis of ATP starts and the substrate iseffluxed out into the lumen side. However, BCRP onlyhas the one ATP binding cassette.

In reality the homo-dimer is formed and works as adimer. In any case, the BCRP transporter is focusedand several researchers in my laboratory have beenworking on it. Before I will show our result, let me intro-duce a great work by Jonker et al.

As you can see from Figure 20, there is anothercancer drug called topotecan. It can be the substrate ofthe BCRP as well as the MDR1. GF120918 is a P-glycoprotein-inhibiting and BCRP-inhibiting compound

In order to differentiate the two, they used a P-glyco-protein knock-out mouse. If GF compound inhibits theefflux activity in the P-glycoprotein knock-out mouse,the transporter involved ought to be BCRP. Even in aknock-out mouse with a GF compound, GI absorptionis upgraded and therefore BCRP seems to be workingas a transporter in the efflux system on the lumen side.The reason why topotecan was taken out to the lumen

Jonker JW, et al. J Natl Cancer Inst: 2000

00 120

Time (min)

In intestine and liver, BCRP contributes to the pharmacokinetics of topotecan

Pla

sma

topo

teca

n (n

g/m

l)

200

400

240

mdr 1a/1b (-/-) micep.o.

+GF 120918

vesicle

00 30

Time (min)

cum

ulat

ive

bilia

ry to

pote

can

(% o

f dos

e)

20

40

60

mdr1a/1b(-/-) micei.v.

+GF 120918

vesicle

00 120

Time (min)

plas

ma

topo

teca

n (n

g/m

l)

1 000

500

240

mdr1a/1b(-/-) micei.v.

+GF 120918

vesicle

00 120

Time (min)

plas

ma

topo

teca

n (n

g/m

l)

200

400

240

wt micep.o.

+GF 120918

vesicle

Figure 20.

THE ROLE OF BCRP IN THE BIOAVAILABILITY AND BILIARY EXCRETION

Permeation coefficient: mm/minME3277 concentration: 50µM

N.Okudaira et al.J. Pharmacol. Exp. Ther., 294: 580-587 (2000) J. Pharmacol. Exp .Ther. 295: 717-723 (2000)

Serosal to Mucosal Mucosal to Serosal

ControlEHBR

Mid gut

Jejunum

Ileum

01 1 2234 0

*

Figure 19.

PERMEATION OF ME3277 ACROSS THE SMALLINTESTINAL TISSUE OF EHBR AND SD RATS

23

side was because of the BCRP. BCRP indeed acceptstopotecan as a substrate. There are many anionic com-pounds that can serve as the substrate for BCRP(Figure 21). There is another cancer drug, methotrexate,or DHEAs or the DNP conjugate, and many of them canwork as the anion substrate for the BCRP transporter.

I have maybe 15 more minutes to talk about hepa-tobiliary transport and the functions of OATPs (organicanion transporting polypeptides) and MRP2 in relation toits role in drug disposition in the body.

The drug is taken in and is metabolized. Then ABCtransporters on the lumen side are responsible for theefflux (excretion) of various kinds of anionic compounds.The uptake transporter, OATP2 and the efflux trans-porter,MRP2 are both deeply involved in the hepato-biliary transport of anionic compounds. A part of that isthe example of OATP2 in Figure 22. Incidentally, OATP2is also referred to as OATP-C or LST1. The nomencla-ture is very complex and I am very sorry about that, butwithin our laboratory it is called OATP2, and so I will callit today OATP2. From the Northern blotting system, you

P388/BCRP P388DH

EA

S u

ptak

e (Ķ

l/mg/

5min

)

DHEAS

200

150

100

50

0

P388/BCRP P388TLC

S u

ptak

e (Ķ

l/mg/

5min

)

TLCS

100

75

50

25

0P388/BCRP P388

TCA

upt

ake

(Ķl/m

g/5m

in)

Taurocholate

20

15

10

5

0P388/BCRP P388

TCA

upt

ake

(Ķl/m

g/5m

in)

DNP-SG

100

5

10

15

20

0P388/BCRP P388

TCA

upt

ake

(Ķl/m

g/5m

in)

4-MUG

100

0

5

10

P388/BCRP P388

MTX

upt

ake

(Ķl/m

g/5m

in)

Methotrexate

10

5

0

P388/BCRP P38817ßG

upt

ake(

Ķl/m

g/5m

in)

E217ßG

0

5

10

15

20

25

P388/BCRP P3884MU

S u

ptak

e (Ķ

l/mg/

5min

)

4-MUS

0

150

100

50

200

** ** **

**

**

Figure 21.

THE UPTAKE OF COMPOUNDS INTO MEMBRANE VESICLES PREPARED FROM P388/BCRP

Human hepatocytes Jorg Konig et al, Am J Physiol Gastrointest Liver Physiol ,278,G156-64,2000

OAT2 OCT1OATP 2/OATP-C OATP 8

MDR1 MRP2

OATP-BNTCP

ATP

ATP

ATPbile

bloodpravastatin

BSEP

Figure 22.

LIVER SPECIFIC EXPRESSION OF OATP2 IN NORTHERN BLOTTING

24

can see it is a very liver-specific type of expressionsystem. It is clear that uptake by the liver has a majorimpact on the pharmacokinetics of, for example, pravas-tatin. This transporter accepts a variety of therapeuti-cally important drugs.

Figure 23 is a slide I refer to quite frequently. It showstemocapril from Sankyo, a compound that was launchedin Japan a long time ago, but as the seventh or eighthACE inhibitor. Now, it is one of the top-ranked inhibitorson the market, because of its good pharmacokineticproperties.

It is converted into temocaprilat, an active form whichhas two carboxylic acids. One of its characteristics isthat, whereas with conventional ACE inhibitors theplasma concentration tends to rise steeply in peoplewith renal failures, with temocaprilat the concentrationlevel is not so dependent on renal properties. In thatsense it has a very good in-vivo property. It is usedmainly for hypertension in elderly people, meaning thatinter-subject variability must not be too high as this wouldbe likely to induce a lot of side-effects.

So we became involved in a study where we foundthat with a conventional ACE inhibitor, more than 90 %of dose is excreted into the uriner. If there is a problemwithin the kidney there is no way that it can get out of theblood circulation, and that is the reason why the plasmaconcentration level goes up. I think that makes sense(Figure 24).

But because temocaprilat is also involved in biliaryexcretion as well as in urinary excretion, there is an alter-native path. Even if the kidney is not functioning well,there is another way to exit. That's the reason why theplasma concentration level of temocaprilat is not affectedso much. That's one of the kinetic explanations, whichhas been proved experimentally.

We have also found out why it is only the temocaprilatthat is excreted into the bile. It is taken up by OATP2and is excreted into the bile by MRP2 (Figure 24). Thereare also other ACE inhibitors which use the uptake trans-porter and gets into the liver. However, they are not rec-ognized by the MRP2 so that's why they are not excretedinto the bile. We hope that this will become establishedknowledge.

As for in silico models, we are now working on thoseas well, but as we have limited time I won't go into detailabout that.

Now about the HMG CoA-reductase inhibitor, pravas-tatin. Figure 25 shows that pravastatin is absorbed inthe intesine by a transporter, maybe by OATP-B. Thenit is taken up by the liver by OATP2 from the portal veinfollowed by the biliary excretion by MRP2 (previouslycalled cMOAT). Having three transporters, a very effi-

Discrimination based on the difference in pharmacokinetic profile

temocaprilat

Urine

OATP2 MRP2

Bilecaptoprilenalaprilatetc.

Figure 24.

ACE INHIBITORS

Liver

Enterohepaticcirculation

Small Intestine

Metabolism

Binding toHMG-CoAreductase

Hepatic uptake

Absorption

pravastatin

Biliary excretioncMOAT

ADPATP

oatp

Figure 25.

00 12

Time (hr)

CLcr<4040<CLcr<8282<CLcr

Temocaprilat

Pla

sma

conc

entr

atio

n (n

g/m

l)

20

40

60

80

100

24 36 480

0 12

Time (hr)

Enalaprilat

Pla

sma

conc

entr

atio

n (n

g/m

l)

10

20

30

40

50

24 36 48

Figure 23.

PLASMA CONCENTRATION OF ACE INHIBITORSIN PATIENTS WITH VARIOUS RENAL FUNCTIONS

25

cient enterohepatic circulation takes place.For this drug,this is very efficient because the pharmaceutical targetin this case is the HMG-CoA reductase in the liver. If we have this kind of efficient enterohepatic circula-tion, that means that exposure within the target site in theliver is very high.

OATP is especially active with this type of uptake,and the first pass hepatic uptake after its oral adminis-tration is very high. So only a small portion of drug goesinto the blood circulation and then a severe muscle lysiswhich is known as the side effect caused by statinsmay be minimized. In this case, transporters which playa role in the enterohepatic circulation act as good players

for the drug in terms of the better pharmacological effectand minimum side effects. We need to be very con-scious about this example, because people have beenbelieving that drug transporters have been working asbad players as is the case of P-glycoprotein. That's myview.

Figure 26 shows theMRP2/cMOAT substrates. Thereare a lot of different types of conjugates and even if it'snot a conjugate some compounds may have certainanionic properties. I wanted to point out that more than80 percent of MRP2 substrates are also substrates ofOATP2, though the structures of OATP2 and MRP2 arequite different. OATP2 is an exchanger type of trans-

Substrates for rat cMOAT and human MRP

ratcMOAT/Mrp2 hMRPGlutathione LTC4 ? ?Conjugate

DNP-SG ? ?glutathione bimane ? ?glutathione disulfide (GSSG) ? ?monochloro-monoglutathionyl melphalan N.D. ?

Glucuronide 17ß-estradiol 17-(ß-D-glucuronide) ? ?Conjugate -naphthyl glucuronide ? N.D.

bilirubin glucuronide ? N.D.E3040 glucuronide ? N.D.liquiritigenine (flavonoid) glucuronide ? N.D.glycyllhizin ? flavonoid-glucuronide ? N.D.grepafloxacin (new quinolone) glucuronide ? N.D.SN-38 (camptothecin analogue)s glucuronide ? N.D.glucuronosyl etoposide N.D. ?

Bile acids cholate-3-O-glucuronide ? N.D.lithocholate-3-O-glucuronide ? N.D.tauro/glycolithocholate 3-sulfate ? N.D.taurochenodeoxycholate 3-sulfate ? N.D.6α-Glucuronosylhyodeoxycholate N.D. ?3α-sulfatolithocholyltaurine N.D. ?

Others LTE4 ? ?LTD4 ? ?LTE4NAc ? N.D.dibromosulfophthalein ? N.D.cefodizime (ß-lactam antibiotic) ? N.D.grepafloxacin (new quinolone antibiotic) ? N.D.methotrexate ? N.D.CPT-11 acid form (camptothecin analogue) ? N.D.SN-38 acid form (camptothecin analogue) ? N.D.pravastatin (HMA CoA reductase Inhibitor) ? N.D.temocaprilat (ACE inhibitor) ? N.D.BQ-123 (cyclic peptide; endothelin antagonist) ? N.D.

N. D.: Not Determined

Figure 26.

26

porter, while MRP2 is a primary active transporter uti-lizing ATP hydrolysis as a driving force.

Since the substrate specificity is very similar, these twotransporters may provide vectorial transport of drugsand endogenous compounds such as bilirubin glu-curonide for the liver. This is very useful because if one

compound is recognized by these uptake and effluxtransporters, it will be transported effectively from theblood side to the bile side. We call it vectorial transport.In trying to reconstruct in-vitro the vectorial transport,we recently came up with MDCK2 (Figure 27). It is acell line with polarity, in which we expressed simultane-ously the uptake and efflux transporters on differentsides and the Figure shows the result. I won't go into thedetails but we call this type of cell the double transfec-tant cell. Several pharmaceutical companies are inter-ested in these cells and use them in drug development.

With this double transfectant cell we have establisheda system that's used for Caco-2 cells, and a type of vec-torial transport which has been observed in-vivo has beenreconstructed in-vitro For example, this is the result withpravastatin (Figure 28). This is a single expression type,or vector-control. Without the transporter, no vectorialdirection is seen, but by using this double transfectant, thetransport from basal to apical is much higher comparedwith the opposite direction. We have been able to demon-strate very clear vectorial transport here, with pravastatin.

0

0 30Time (min)

basalto apical

basal to apical

apical to basal

basalto apical

basal to apical

apical to basalapicalto basal apical

to basal

Vector

Tran

spor

ted

prav

asta

tin [µ

L/m

g pr

otei

n]

100

200

300

400

60 90 120

0

0 30Time (min)

OATP2

100

200

300

400

60 90 120

0

0 30Time (min)

MRP2

100

200

300

400

60 90 120

0

0 30Time (min)

OATP2/MRP2

100

200

300

400

60 90 120

Figure 28.

TRANSCELLULAR TRANSPORT OF PRAVASTATIN (1 µM) ACROSS MDCK II CELLS EXPRESSING OATP2/MRP2

0

0 50Time [min]

CsA

Cer

ivas

tatin

[µL/

mg

prot

ein]

500

1 000

1 500

2 000

2 500

3 000

100 150 200

0 µM0.03 µM

0.1 µM

0.3 µM1 µM3 µM

10 µM30 µM

0

0 0.1Concentration of CsA [µM]

Ki = 0.0836 ± 0.0149 µM

PS

BA

A o

f CE

R[µ

L/m

in/m

g pr

otei

n]

2

8

6

4

10

12

14

16

18

1 10 100

Figure 29.

EFFECT OF CSA ON THE TRANSCELLULAR TRANSPORT OF CER IN OATP2/MRP2 EXPRESSING CELLS

MRP2 TRANSWELL

OATP2

MDCK cell

apical

basalIsotope labeled substrate

Figure 27.

SCHEMATIC DIAGRAM ILLUSTRATING THE POLARIZED CELLS CO-TRANSFECTED WITH OATP2 AND MRP2 CDNA

27

Now I want to talk about the drug-drug interactions.Cerivastatin(CER) was withdrawn from the market dueto the side-effects of muscle lysis. One of the majorcause for this serious side effect is drug-drug inter-action.

We used cyclosporine as an interacting compoundand tried to see cyclosporine's inhibiting effect on thecerivastatin transport using the double transfectant cell.As you can see on the right-hand side of Figure 29 theKi value is about 0.1µM, therefore this is something that

is clinically feasible at the cyclosporine concentrationlevel in the plasma. Our further analysis indicated that thisdrug drug interaction is mainly due to the inhibition ofOATP2-mediated uptake process. In this way, cyclo-sporin increased the plasma concentration of cerivas-tatin 5 folds.

Let me tell you about one of the dreams I have. Asshown in Figure 30, in the future I would like to be ableto use the double transfectant cell in drug development.In order to do that, as a first step, a single transfectantcell should be used in finding candidate transporters,such as when you use the CYPisofrm expression sys-tems. Once you have narrowed down the number ofcandidates, at that stage you can refer to this doubletransfectant cell model. This is my proposal. Thethroughput is not very high, just like in Caco-2 system,so you need to have narrowed the candidates downbefore moving on to this kind of double transfectant cellmodel.

We now have about 15 targets – we are trying tocome up with about 15 different types of double trans-fectant cells, and once they are available we can startto evaluate the transport properties of new drug candi-dates in various tissues, such as liver, kidney, intestineand the blood brain barrier. After that, at late-stage ofdrug development you can make the final decision ofwhether it should go into clinical trial or not.

More recently, the OATP2 transporter has been foundto be genetically polymorphic (Figure 31), especially star15 (*15). In this case the allel frequency is about 10 per-cent in Japanese population Patients with this kind ofpolymorphism have very high concentration levels ofpravastatin.

Step 1:Transport assay using human transporter-expressed cells (OATPs, OATs, OCTs, MRPs ...)- Which transporters? Affinity (Km)?

Step 2:Transcellular transport assay using double transfectants- Uptake and efflux clearance?- By comparing the expression level of transporters between human

tissues and double transfectant cells, can the prediction of in vivotranscellular transport activity be possible ?

Examples

apical

basal

model organ basal side apical side

Liver OATP-2/LST-1 MRP2/CMOATNTCP BSEPOCT1 P-gp

Kidney OAT1 or OAT3 OAT-K1 or OAT-K2OCT2 P-gp or OCTN2

Figure 30.

2-STEP SCREENING FOR DETERMINATION OF TRANSPORT PROPERTY OF DRUG CANDIDATES

Allele frequency (Japanese)

OATP2*1a N VOATP2*1b D VOATP2*5 N AOATP2*15

Asn130AspA388G

Extracellular

N terminal

C terminal

Intracellular

D A

188 (35.2%)287 (53.7%)

4 (0.7%)55 (10.3%)

Collaboration with Dr. Ieiri (Tottori Univ.)

130 174

Val174Ala T521C

Figure 31.

OATP2 SNPS IN JAPANESE HEALTHY SUBJECTS

28

Figure 32 is a joint study done with Dr. Ieiri in TottoriUniversity. For those with the heterozygous polymor-phism, the plasma concentration of pravastatin isincreased about two times, therefore the star 15 poly-morphism has some impact on the uptake of pravastatinand that would consequently make a difference to theconcentration level and this has also been verified inindividuals as well.

Dr. Richard Kim and his co-workers used fexofe-nadin. Grapefruit juice is known as an inhibitor of P-gly-coprotein, so they thought that the intestinal absorptionof fexofenadin must increase by adding grapefruit. How-ever, through experiments they showed that grapefruitjuice worked to reduce the concentration level, and it

was the same with orange juice and apple juice. Thismay indicate that the transporter(s) responsible for theintersinal absorption of this drug may be inhibited bythese juices.

Taking this further, they used other types of expres-sion systems, and they also diluted various types ofjuice drinks and tried to find out about their inhibitoryeffect. The findings were that OATP uptake was inhib-ited by this kind of juice.

Now let me tell you about another dream that I havefor the future. Maybe it will take about 5-10 years. I haveabout eight more years until my retirement, but by thenI would like to create a huge database with a simula-

OATP2*1a N VOATP2*1b D VOATP2*5 N AOATP2*15 D A

130 174

00 12

Time after dose (hr)

Plasma concentration of pravastatin

(ng/

mL)

10

20

30

40

50

48

genotype CL non-renal AUC(L/kg/hr) (ng*hr/ml)

*1a/*1a 2.22 60.5*1a/*1b 1.45±0.72 47.2±27.4*1b/*1b 2.01±0.42 44.2±6.38*1b/*15 1.11±0.34 62.1±21.8*15/*15 0.28 111.8

*1b/*1b*1b/*15*15/*15

Figure 32.

EFFECT OF OATP2 SNPS ON THE PHARMACOKINETICS OF PRAVASTATIN IN JAPANESE HEALTHY SUBJECTS

Patients’ geneticcodes

Genotyping

Allelic variants in genes codingenzymes andtransporters

Metabolizing Enzyme #1mut#1…CLm1,#1mut#2…CLm1,#2mut#3…CLm1,#3

Drug Transporter #1mut#1…CL_1,#1mut#2…CL_1,#2mut#3…CL_1,#3

In vivo(CLtot)

Function of wild type (wt)metabolizing enzymes andtransporters (Clmj wt, Cltj, wt)

Tissue(CLH ± CLR)

Cell

f(CLmj,i, CLtj,i)

Enzyme 1(CLm1,i)

Transporter(CLm1,i)

Enzyme 2(CLm2,i)

Predict

Figure 33.

QUANTITATIVE PREDICTION OF IN VIVO DRUG DISPOSITION IN HUMANS WITH GENETIC POLYMORPHISMS– INVOLVEMENT OF POLYMORPHISMS IN METABOLIZING ENZYMES AND TRANSPORTERS –

29

tion capacity. Once we have installed that type of simu-lation software on a database then we could look atwhat happens after the administration of drug candi-dates, as shown in Figure 33.

We could look directly at polymorphism and func-tionalities, and also at drug-drug interaction. You coulduse the system during drug discovery and developmentor get experimental data out of it. All of the in-vitro infor-mation could be useful; for example, you need to takethe biochemical data such as drug metabolism, intestinalabsorption, membrane transport, binding to plasma andtissue proteins into consideration as well. But by usinga mathematical approach to bring together informationon what happens at cell level, at organ level and also atthe whole-body level, it may be possible to create sim-ulation models and to build up these kinds of statistics.

We have worked on this approach for the pastdecade and we have our methodology available and inplace. What remains is the content of the database.We need to get good quality content, plus we need tosupplement some of the -genetic polymorphism relatedinformation. Once we have achieved that we will havea very high-quality database with simulation function-ality. This is one of my objectives before I retire.

If you look at my home page (http://www.f.u-tokyo.ac.jp/~sugiyama/), you will find a logo depicting afat rat - it's a caricature of me (Figure 34). There's amessage in it. First, even when you are taking a napyou have to think of the research. Second is the contentof our research. We are studying the molecular and cellbiology of transporters, but our final goal is to integratethus obtained information into the mathematical modelingto understand and predict the drug disposition in thebody. This is what I have been stressing all the time,and that's the point of the logo.

I have an announcement for you. The Second WorldPharmaceutical Congress is going to be held next yearin Kyoto and I am sure that Capsugel will be heavilyinvolved in it. I will be serving as chairman. Dr. Hashidais responsible for the program and Drs Nagai and Benetare the co-chairs of the scientific advisory committee.I hope you will attend this meeting in Kyoto and thatthere will be a big turnout. We are expecting about3,000 delegates and hope to see you again on thatoccasion.

With this, I would like to conclude my presentation.Please do visit my home page. I now have my secretaryin my lab who is very artistic, so we have renewed ourhome page. Thank you very much for your attention.

Chair, Professor Sjinji Yamashita, SetsunanUniversity: Thank you very much, Professor Sugiyama,for leaving plenty of time for discussion. If you have anyquestions or comments, would you come up to themicrophone. The whole program is being recorded andwill be edited for the proceedings, so would you pleasestate your name and affiliation before you speak.

Kato of Keio University: CYP3A4 and MDR1 areknown to have some common features. I have two ques-tions. What are the common features and which aremore important? Some say size is important, but howabout you? And if it is the size, then if it is below 300 itcannot make a good substrate. A molecular weight ofmore than 400 would be a good substrate for P-glyco-protein.

Professor Yuichi Sugiyama, University ofTokyo: Lipophilicity may also play a part; some level oflipophilicity is needed to service a substrate. While insome exceptional cases the anions can be the sub-strate, either the neutral or the cationic substrate wouldbe a better substrate for the P-glycoprotein.

Having said so, you may think that prediction is quiteeasy. However, for the past few years, people havebeen making in silico predictions of good substrates forP-glycoproteins. There are several publications which I have read and we have had discussions. I understandthat the predictability of in silico models is quite low atthe moment.

Kato of Keio University: I have a second question.For instance, we do CYP3A4 and we also have MDRItransporters as some can serve as the substrate andsome may work as the inhibitors. But there are somediscrepancies. What are the sizes of these discrepan-cies? According to my impression, MDR1 may havethe larger discrepancy.

Professor Yuichi Sugiyama, University ofTokyo: What do you mean by discrepancy – whetherFigure 34.

30

that can serve as the substrate or inhibitor? Some maywork as the inhibitor but not the substrate. That is themeaning of the 'discrepancy', right? When the com-pounds are the substrates, they always work asinhibitors. However, there exist inhibitors which are notthe substrates. There are many compounds like that.P-gp may have that feature more than 3A4. Actually,much of the research on 3A4 has been published. Ofcourse, there are different strengths but with any 3A4-metabolized compounds, inhibition takes place moreor less. In the case of 3A4 there are multiple bindingsites on those enzymes, so it has been shown that sub-strate A enhances the enzymatic activities for substrate B.Thank you.

Question from the audience: The differencebetween inhibition and the substrate, does it comeabout after binding, or is it because the feature is re-ceptor-specific?

Professor Yuichi Sugiyama, University ofTokyo: Even if the binding site is the same for two com-pounds, in order to obtain the transport, conformationalchange has to be present. Therefore, even if it bindsbut that binding does not lead to conformational change,that would never become the substrate; in other words,it is never transported.

Roland Daumesnil, Capsugel Inc., North Ca-rolina, USA: Yuichi, thank you, I always enjoy yourpresentations. I have a question for you. You're tryingto correlate in-vitro permeability with in-situ permeabi-lity in a rat. You have a PS ratio, I assume that is apermeability ratio from apical to basal level. I'm justwondering how you determine this in numbers be-cause in the in- situ I would imagine it's difficult fromthe basal to apical side.

Professor Yuichi Sugiyama, University ofTokyo: I used a knock-out mouse, not the rat, to lookat the in-situ perfusion and the in-vitro, and there is acorrelation curve. Actually, the quantification and deter-mination of the apical to basal flux in a strict sense isvery difficult. Was that your question?

I'm sorry it's not so clear. It is the net apical to basalflux. In other words, this is a steady-state experiment, soat the steady state of intestinal perfusion we measuredthe drug concentration in-rate and also out-rate. Byusing that difference, based on the so-called tube -model we then calculated the PS apical to basal. Inother words, that PS apical to basal flux clearance isthe net clearance. It's not necessarily the real unidirec-tional apical to basal flux. That is my answer.

Chair, Professor Sjinji Yamashita, SetsunanUniversity: Any other questions, Roland San? Do youhave any other comment or questions? Well, I am afraidthe time is up. He has kept to time, and I must thank Pro-fessor Sugiyama for an informative presentation and goon to the next speaker.

I am very pleased to call upon the second speakerin this special lecture session, Dr. Christopher Lipinski.He will talk about poor aqueous solubility – an industry-wide problem in ADME screening. He is a very well-known scientist who actually pioneered this theory andhe is well known in this area I believe that all of us knowhis name and the role that Dr. Lipinski has played. Hewas with Pfizer for many years. Since his retirement lastyear he has participated even more actively in academicsociety meetings world wide, and his topic today willcover the worldwide perspective. Dr. Lipinski, the flooris yours.

31

Poor aqueous solubility – an industry-wide problem

in ADME screening

Dr. Christopher A. LIPINSKI

33

Thank you, Capsugel, for the privilege of addres-sing this excellent conference.

Let me outline the first part of my talk. I'm going tobe talking about property changes over time. I'mgoing to be talking about errors, mistakes, in the earlyyears of combinatorial libraries. I'm going to be talkingabout physico-chemical property changes, changesthat occurred in the medicinal chemistry labs at thePfizer location in Groton, Connecticut. I'll be talkingabout changes in early Pfizer Groton high-throughputscreening leads and then I'm going to switch and I'mgoing to talk about changes, not just at Pfizer, but thatare industry-wide.

Then I'm going to talk about the profiles of clinicalcandidates from the Pfizer Groton Laboratories andthe Merck worldwide organization. I'm going to be tal-king about the changes in molecular weight that havecome about more recently, as well as the more recentchanges in lipophilicity and changes in hydrogen bon-ding. Then I'm going to explain how these changesfor these two organizations have implications for eitherpoor permeability or poor solubility, as problems inpoor oral absorption.

So what were the errors, the mistakes, in the earlyyears of combinatorial libraries? Well, combinatorial-like chemistry is a new technology; it's the automatedsynthesis of many compounds. When you're startinga new technology, you usually try to use technologythat existed earlier, and the earlier technology for theautomated synthesis of chemistry compounds wasMerrifield solid-phase peptide synthesis. So the earlycombinatorial libraries had peptidic backbones and,

just like Merrifield solid-phase peptide synthesis,many of these early combinatorial libraries used solid-phase synthesis.

Now, one of the problems with solid-phase syn-thesis is that oftentimes in the earlier years the quanti-ties were very low, so we could not do much testingwith these compounds. There is an additional pro-blem with hydrolytic instability. We now recognize thatunless you take some extra efforts with the peptidebackbone, these compounds can hydrolyze, can de-grade. We also now recognize that having a peptidicbackbone is very bad in terms of passive intestinalpermeability. You can expect many permeability pro-blems with peptidic backbones.

This is already a series of bad problems, but in ad-dition another problem was superimposed: an error inthe application of the principle of maximal chemicaldiversity. Let me explain. In the period before combi-natorial chemistry, it was very hard to get startingpoints for medicinal chemistry programs. As a result,organizations were very eager to develop methods forgenerating in-vitro actives as starting points, and sothere was a tremendous emphasis on in-vitro activity.So the idea was – using the wrong application of theprinciple of maximal chemical diversity – you mighttake the central part of a compound and attach allkinds of interesting functionality so as to maximize thechance of getting in-vitro activity. And that worked.

Many of these combinatorial libraries generatedmany in-vitro active compounds. But the problemwas that these compounds were very large. It wasvery common, in those early libraries, to get com-

Poor aqueous solubility – an industry-wideproblem in ADME screening

Dr. Christopher A. Lipinski

Pfizer Inc., USA

34

pounds with molecular weight 650, and it was notrealized until later that with these very large com-pounds it was very difficult to obtain oral activity. Somany of these early libraries generated in-vitro activesbut no orally active compounds.

That is true across the industry. Now in Figure 1,illustrates the trends at the Pfizer Groton Labs that re-sult from high-throughput screening. What I haveshown here are the percentage of compounds madeby the Pfizer Groton medicinal chemists that had veryhigh molecular weight, greater than 500. These areshown in cross hatched. You can see that between1985 and 1988 the percentage of very large com-pounds was quite small, between 10 and 15 per-cent.

But all of a sudden, in 1989, something happe-ned: our chemists were making much larger com-pounds. About 20 to 25 percent of them were verylarge. Then, in 1992, the size problem got muchworse. Our chemists were now making 30 to 35 per-cent compounds which had a molecular weight grea-ter than 500. What happened to cause this?

Well, from about 1988 to 1989, the Pfizer GrotonLaboratories began to do high-throughput screeningand from about 1992, the organization was doing agreat deal of high-throughput screening. So the star-ting points for medicinal chemistry were coming fromhigh-throughput screens and we began to get largerstarting points. Therefore everything that the chemistsmade was large and that's why we see this increasein size of the compounds. For comparison, shown ingrey in the Figure are the compounds that Pfizer pur-chased from academic sources. There is no trend,and that's because these academic compoundshave no relationship to high-throughput screening.

Now you can learn a lot about drug-like com-pounds by looking through the literature (Figure 2). Ihad the idea of looking at compounds with an INNname or a USAN name – these are names that aregiven to a compound when it goes into Phase II stu-dies. If you identify compounds with these names orthat actually were marketed somewhere, then youhave identified drug-like compounds because, if acompound has these names, you have eliminated allthe compounds that failed preclinically. You have eli-minated all the very insoluble compounds, all the verypoorly permeable compounds, all the very toxic com-pounds. Now the compound with these names mightnot be efficacious, because that's what you discoverin Phase II studies. But with respect to everything elseexcept efficacy, these are drug-like compounds.

In the next few Figures, I'm going to compare thesedrug-like compounds, these Phase II compounds, 90percent of which were intended for oral use, with what

5

00

Per

cent

MW

T >

500

10

15

20

25

30

35

1985 1986 1987 1988 1989 1990 1991 1992 1993 1994

HTS starts

fully blown HTS%Comm_MWT %Syn_MWT

Figure 1.

LEADS AT PFIZER AND IN THE DRUG INDUSTRY IN GENERAL, NOW TREND TOWARD HIGHER MWT AND LIPOPHILICITY

Computationally comparing libraries.Drug-like vs new drugs

• Use the presence of an INN name or a USAN name ormarketed status as a flag for a compound with “drug-like” properties

• 7 483 Drugs with INN name, USAN name or approvedfor marketing

• Compare to 2 679 New Drugs from the Derwent WorldDrug Index- mechanism field - trial preparations- No CAS registry number, no INN/USAN name,

abstracted in 1997, 1998, 1999

Figure 2.

35

I call new drugs. I've described the characteristics inFigure 2. Essentially, these new drugs are the kinds ofcompounds that medicinal chemists have been ma-king in recent times. They would be reported in themedicinal chemistry journal, reported at the medicinalchemistry meeting… but they are not necessarilydrug-like compounds. There are a lot of badly-beha-ved compounds among these new drugs.

Figure 3 shows a plot distribution of compounds indifferent collections. The Phase II drugs are shown asthe broken line and the dotted line, and the unbrokenline shows the first 190 high-quality, high-throughputscreening hits from the Groton Laboratories. This is adistribution plot, the calculation of lipophilicity, usingthe Moriguchi log-P algorithm. If the curve is shifted tothe right it means that on average these early high-throughput screening hits were more lipophilic. Theaverage shift is about a half a log-P unit over thePhase II orally active drugs.

Now half a log-P unit shift may not seem like verymuch. But if, for example, you have 10 clinical candi-dates, and if you only have a few of these com-pounds with these more difficult properties, it createsa great problem. Our pharmaceutical sciences peopletell us that for these large lipophilic compounds ittakes two or three times longer, it costs two or threetimes more and it takes two or three times morepeople-effort to develop these kinds of compounds.So, just a few compounds among the clinical candi-dates can cause a great many problems in the deve-lopment organization.

Now the second point is that this half a log-P shiftonly occurred in the early years of high-throughputscreening. If you look at the entire high-throughputscreening history of Pfizer you would not see this.

Why not? Because, after the early years, we learnedfrom our mistakes and we put filters in place to cor-rect for this property shift.

So that was Pfizer's high-throughput screening his-tory. But what we saw at Pfizer occurred across theentire pharmaceutical industry (Figure 4). The Figureshows a plot distribution of molecular weight. Thecurves to the left are the Phase II drugs and marketeddrugs, while the broken line shows the newer drugs,the kinds of compounds that chemists are makingnowadays. Note that this curve is shifted to the right.That means that, on average, the kinds of com-pounds that chemists are making across the entirepharmaceutical industry are much larger than the tra-ditional drugs, the type of drugs that got into Phase IIstudies. That's molecular weight.

The same thing holds for lipophilicity (Figure 5).This is the Moriguchi log-P algorithm. The top curve

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

00 1 2 3 4 5 6

Moriguchi Log P

Frac

tion

with

MLo

gP

HTS USAN NCE

On averagemore lipophilic

Data from earlyyears of HTS

Figure 3.

HIGH THROUGHPUT SCREENING HITS ARE MORE LIPOPHILICTHAN PHASE-2 OR MARKETED DRUGS

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0100 200 1000

Molecular Weight Values

Frac

tion

with

Mol

ecul

ar W

eigh

t Val

ue

300 400 500 600 700 800 900

WDI-Drugs INN/USAN-Drugs New Drugs NCE-Drugs

Figure 4.

NEWER DRUGS ARE LARGER

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

00

Moriguchi LogP Values

Frac

tion

with

Mor

iguc

hui L

ogP

Val

ue

1 2 3 4 5 6


Figure 5.

NEWER DRUGS ARE MORE LIPOPHILIC

36

shows Phase II drugs, marketed drugs. The brokenline is the newer drugs that chemists are making, shif-ted in a right shifted downwards direction. The newerdrugs that chemists are making are more lipophilicthan the kinds of compounds that traditionally havereached Phase II studies.

The same thing holds for permeability (Figure 6).On the X-axis is a calculation of hydrogen bond do-nor and acceptor energy values. The further onegoes to the bottom right-hand corner the more diffi-cult it will be for a compound to cross the gastroin-testinal tract for passive permeability. The two topcurves are the Phase II orally active drugs. With thenewer drugs, the kinds of compounds medicinalchemists are making nowadays, the dotted-linecurve is shifted in a right shifted downwards direc-tion. Chemists are making less permeable com-pounds now than they were a number of years ago,less permeable than the kinds of compounds thatused to get into Phase II studies.

I have gone through specific Pfizer history and Ihave gone through the experience of the entirepharmaceutical industry. Now I'm going to gothrough a comparison of clinical candidates fromtwo very productive organizations: the Pfizer Labs inGroton, Connecticut, and the Merck worldwide or-ganization.

Let me explain Figure 7. This is data from the Pfi-zer Groton Laboratories. Each square represents aclinical candidate compound made between about1965 and 1995. On the Y-axis you can read off themolecular weight of the compound. There are lots ofclinical candidates, lots of scatter. The black line isthe best straight line through those points. You see itgoes up, the slope is rising, meaning that on averagethe clinical candidates from the Pfizer Groton Labora-tories have become larger in more recent years. Thenext point to make here is that there are very few cli-nical candidates with molecular weight above about500; most are below 500. So this is the trend withmolecular weight.

Figure 8 gives a similar molecular weight plot forthe Merck Organization, but there are now several dif-ferences from Figure 7. The slope looks larger, andthere are fewer points. The slope is higher becausethe scale is different. The number of points has no-thing to do with productivity. It simply has to do withthe fact that I was only able to capture the data on theMerck compounds from the literature at a later stageof clinical candidacy. But the trend is the same withtime.

The Figure shows the Merck Organization's clinicalcandidates from about 1965 to 1995. In this periodthe clinical candidates are larger. The plot shows lotsof scatter, but do you notice there are not too manyclinical candidates with molecular weight greater than500? Most of them are below 500. So, with respect

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

00

Sum of Raevsky H-Bond Donor and Acceptor Energy Values

Frac

tion

with

Rae

vsky

H-B

ond

Sum

10 20 30 40 50


Figure 6.

NEWER DRUGS ARE LESS PERMEABLE

1200

0

Time Sequence

Mol

ecul

ar W

eigh

t

1000

800

600

400

200

Figure 7.

UPWARDS MWT TREND IN PFIZER, GROTON CANDIDATES

1200

01960 1970

Mol

ecul

ar W

eigh

t

1000

800

600

400

200

1980 1990 2000

Figure 8.

UPWARDS MOLECULAR WEIGHT TREND IN MERCK CANDIDATES

37

to molecular weight as a function of time, Pfizer Gro-ton and Merck have very similar patterns of behavior.

With regard to lipophilicity, using the Moriguchi cal-culation, Figure 9 shows that over time, between1965 and 1995, there is lots of scatter but the trendis upward. So in the Pfizer Groton Laboratories, theclinical candidates in more recent times are becomingmore lipophilic.

Now what about Merck? Figure 10 shows that it isvery different with the Merck clinical candidates. Fromabout 1965 to 1995 there is no upward trend. The li-pophilicity remains about the same and you'll noticethat there are very few candidates for the Merck Or-ganization with lipophilicity in the 5 to 6 range.

So now we have a difference. Merck and Pfizercandidates are similar in molecular weight but theyare different in respect to lipophilicity. Pfizer Grotoncandidates are becoming more lipophilic but Merckcandidates are not.

With respect to hydrogen bonding (Figure 11), Iam simply plotting for each candidate how many oxy-

gens and nitrogens there are in the compound. Soit's a rough index of hydrogen bonding/bond accep-tor ability.

The Pfizer Groton candidates show lots of scatter,but there is no upward rise; there's no trend in recenttimes for compounds to have greater hydrogen-bon-ding functionality. Notice there are very few clinicalcandidates with more than 10 oxygens and nitrogens;most of them have 10 or fewer.

Figure 12, showing the Merck Organization's can-didates, is very different. The trend is upwards withthe Merck Organization, so with recent time the clini-cal candidates have more oxygens and nitrogens.Notice, though, that just like Pfizer, there are very fewMerck clinical candidates with more than 10 oxygensand nitrogens; most have fewer than 10.

We now have two very successful organizationswith very different trends in properties over time. Sowhat is the explanation?

Figure 13 gives my explanation, which I believe is related to how the leads are developed in the two

10

-4

Time Sequence

Mol

ecul

ar W

eigh

t

-2

0

2

4

6

8

Figure 9.

UPWARDS LOG P TREND IN PFIZER, GROTON CANDIDATES

25

0

Time Sequence

Sum

of O

+ N

5

10

15

20

Figure 11.

NO H-BOND ACCEPTOR TREND IN PFIZER,GROTON CANDIDATES

10

-41960 1970

Mor

iguc

hui L

ogP

1980 1990 2000

-2

0

2

4

6

8

Figure 10.

NO INCREASE IN LIPOPHILICITY WITH MERCK CANDIDATES

25

01960 1970

Sum

of O

+ N

1980 1990 2000

5

10

15

20

Figure 12.

INCREASING H-BOND ACCEPTOR TREND IN MERCK CANDIDATES

38

organizations. Let's look at the left-hand column. Thisis the Merck worldwide organization. Now, I have cal-led it 'structure-based' because actually, between1965 and 1995, the Merck Organization generated itsstarting points for programs using every single tech-nique of rational drug design, but not high-throughputscreening, not then. Only in the last eight years hasMerck been doing high-throughput screening.

So, for example, in structure-based drug designthere is a tendency for molecular weight to go up be-cause maybe you are working on peptidomimeticstructures, you're trying to fit three or more bindingsites. Hydrogen bonding goes up because you aretrying to fit hydrogen bonding sites, you're trying to fitsalt bridges. But in structure-based drug design,where you're using an X-ray of the target as part ofyour information, there is no change in lipophilicity inlog-P. With this method of rational drug design, the-re's no selection pressure for log-P to change.

Now, if molecular weight goes up and hydrogen-bonding properties go up, then those two changesvery reliably translate into poor intestinal permeability.So an organization that generates its starting pointsfor its programs by every technique of rational drugdesign except high-throughput screening, will worrymore about permeability problems as being related topoor oral absorption.

The right-hand column shows the Pfizer GrotonLaboratories. Among all the Pfizer laboratories, thosein Groton do more high-throughput screening thanany other Pfizer laboratory in the world. Now if youdevelop your starting points for programs using high-throughput screening, molecular weight goes up be-cause high-throughput screening selects for largersize and lipophilicity, log-P, goes up because high-throughput screening selects for higher log-P.

Let me explain this. What is a high-throughputscreen? It is the empirical search for in-vitro activity.What you get from high-throughput screening reflectsmedicinal chemistry principles. What's the most effi-cient way for a medicinal chemist to improve in-vitroactivity? It's to add lipophilic groups to a compound,that very reliably improves in-vitro activity. Other thingsbeing equal, a high-throughput screen will generateleads that are higher in molecular weight and higher inlipophilicity.

But high-throughput screening does not result inany changes in hydrogen bonding because there'sno selection pressure. Again, this reflects medicinalchemistry principles. In general, if a medicinal chemisthas a starting point, it usually is not very successful tochange the polar functionality. Just leave it alone, andyou add lipophilic functionality. Now, if molecularweight goes up and if lipophilicity goes up, then thosetwo changes very reliably translate into poor aqueoussolubility. So an organization like the Pfizer Groton La-boratories is much more worried about poor aqueoussolubility than about poor intestinal permeability.

This is my explanation for you when you go tomeetings or you read review articles about poor oralabsorption. Sometimes people talk about permeabi-lity, sometimes about solubility. It's not really clear whypeople have such different opinions. I think the em-phasis on either permeability or solubility is really rela-ted to how the organization develops its startingpoints.

Now in this next section I'm going to talk aboutPhase II drug property distributions. I'm going to betalking about the Rule of 5, about improving propertyprofiles, and how Pfizer uses the Rule of 5.

Lead approach, permeability and solubility

• Structure Based

• MWT up- peptidomimetic- fit 3 or more sites

• H-Bonding up- fit _H bond sites- fit salt bridges

• LogP no change- no selection

pressure

• Poor Permeability

• HTS Based

• MWT up- HTS selects for

larger size

• LogP up- HTS selects for

high LogP

• H-bonding no change- no selection pressure

• Poor Solubility

Figure 13.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

00

Rule of Five Parameter Values1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

MLogP MWT/100NO HBD

Figure 14.

DISTRIBUTION PARAMETERS FOR 7483 INN/USAN DRUGSDEFINE THE 90% LIMITS CORRESPONDING TO PROPERTIES

UNFAVORABLE FOR ORAL DRUG ABSORPTION.

39

Figure 14 describes a database-mining exercise,an analysis that I did in early 1995. I was looking atthe property distributions for 7,500 Phase II drugs.Actually, I initially did it on 2,500, but I have extrapola-ted here to 7,500. I looked at four properties – lipo-philicity, molecular weight, the number of hydrogenbond acceptors (just from counting the number of ni-trogens and oxygens in the compound), and thenumber of hydrogen bond donors. Now why did Ichoose these four properties? Because it was veryclear from the literature that these properties werevery important in oral absorption.

How do you understand this graph? Let's startwith the broken-line curve. For these 7,500 com-pounds, a 0.9 fraction, or 90 percent of these orallyactive drugs, will have 10 or fewer hydrogen bond ac-ceptors. For the molecular weight, 90 percent ofthese Phase II drugs have a molecular weight of 500or less. For lipophilicity, if you use MlogP as the cal-culation, 90 percent of these Phase II drugs have avalue of 4.2 or less. If you use the Clog-P calculationit's 5 or less. For hydrogen bond donors, shown onthe line with circles on it, 90 percent of these orallyactive Phase II drugs have five or fewer hydrogenbond donors.

This led me to come up with something I call theRule of 5 mnemonic. Mnemonic is an English wordwhich means saying something so it's easy to re-member. I was trying to come up with something thatwould be easy for our medicinal chemists to remem-ber, because they were making compounds thatwere just very poorly orally absorbed, and I wantedthem to remember something.

So I came up with this rule that if you have morethan five hydrogen bond donors, the molecular weightis over 500, the Clog-P is over five, and the sum ofnitrogens and oxygens, the number of hydrogenbond acceptors, is over 10, you're very likely to getpoor oral absorption or permeation. I called it the Ruleof 5 because the number 5 comes in several timesand I wanted our chemists to remember it.

The way we use this rule is, we made it part of ourcompound registration system. Our chemists had todraw in the structure when they registered the com-pound so that it could be tested biologically. If therewas a problem – if two or more parameters were out-side the desirable range – a computer screen wouldcome up and it would say something like 'You are li-kely to have poor oral absorption because your mole-cular weight is 650 and your log-P is 6.' The chemistscould still register their compound, but we made surethat they could not bypass this warning screen, so

they could not give the excuse that they had neverheard about poor physico-chemical properties.

The next thing we did was to load the results ofthis calculation into our Oracle databases, so that forevery compound we knew we had stored the numberof hydrogen bond donors, molecular weight, lipophili-city and the number of hydrogen bond acceptors.Why this helped is because our managers, especiallyour chemistry managers, knew we had a big pro-blem, especially with aqueous solubility, but theydidn't know where the problem was coming from.They didn't know who these people were who weremaking these very large and very lipophilic com-pounds.

Once we loaded the results in the Oracle data-base, they could check. They could get computerprint-outs and then at meetings they started askingquestions such as, what are the properties of yourcompounds? And the chemists had to answer. Theyhad no excuse, because we made sure that they sawthis. We did this in about the middle of 1995 and wi-thin six months, at the beginning of 1996, we also setup a very high-capacity solubility screen, to test everycompound that the chemists made. Now, the che-mists could no longer say, well, there's not enoughcapacity in the assay to test my compound, becausethere was.

So, did this work? The answer is yes, it worked ex-tremely well. Figure 15 plots for a number of years thefraction of the compounds made by the medicinalchemists that had an alert according to the Rule of 5,meaning that there were two parameters outside ofthe desirable range. The fraction of the alerts is verylow in the years before high-throughput screening.The years 1986 to 1988 represent early high-through-

0.25

0

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1Q97

2Q97

3Q97

4Q97

1Q98

Frac

tion

with

Ale

rt

0.05

0.1

0.15

0.2

Figure 15.

DECREASE IN “RULE OF 5” ALERTS

40

put screening, 1989 to 1991 represent high-through-put screening but before we really had the Rule of 5 orthe solubility assay. We started the Rule of 5 in themiddle of 1995 and by the beginning of 1996 we wererunning the solubility screen and, very quickly, the pro-perty profiles of the compounds the chemists weremaking greatly improved. Now, I've stopped this graphin early 1998 but if you were to continue it, it's flat.

What this says is that if you have the property fil-ters in place and you have the associated experimen-tal screens, you really can change the behavior of thekinds of compounds the chemists are making. Soyou can get the considerable advantages of high-throughput screening without these very bad propertyprofiles. Will we ever get back to the very desirable1988 level of properties? I don't think so. That's be-cause the targets have changed. The targets are no-wadays more complex.

So how does Pfizer use the Rule of 5? Well, I'vetold you that we use it as an on-line alert at compoundregistration and we use it as a fi lter for our high-throughput screening library. Then, about 18 monthsafter I came up with the Rule of 5, we went in andidentified all the compounds in our high-throughputscreening file in the Groton Laboratories that broke theRule of 5, with two parameters out of the range – andwe stopped screening those compounds.

The compounds were still in the file, we just didn'tscreen them any more. The argument was, why workwith those compounds? Even if they come out activein-vitro in screens, they're too difficult to turn intoorally active compounds. We use the Rule of 5 as afilter for purchased compounds; we don't buy anycompounds if they break the Rule of 5.

We use it as a criteria for focused library synthesis.Like many companies in combinatorial chemistry no-wadays, we are making more libraries but with smallernumbers of compounds. We're using the Rule of 5 aspart of the guideline, part of the library design pro-cess. And we use the Rule of 5 as a guideline forquality clinical candidates. We have a whole long listof requirements for our clinical candidates, one ofwhich is the question, does the compound break theRule of 5?

Now it's sometimes allright to break the Rule of 5,but if you do, then the people who are proposing thecandidate have to have detailed experimental proof ofwhy it is OK, why it will not cause problems in deve-lopment. You could, for example, break the Rule of 5if you have a very potent compound or a highly per-meable compound, but you have to explain whythat's OK.

Now I'm going to talk about some specifics onaqueous solubility. I'm going to talk about presentingdata to chemists. What is acceptable solubility?What's the relationship between solubility, potencyand permeability? I'm going to be talking about solubi-lity by types of compound, to give you a realistic ex-pectation of the kind of solubility you can expect inPhase II compounds, in commercial or academiccompounds, and in the kind of compounds thatcome from medicinal chemistry. Then I'm going to in-troduce the concepts of lipophilicity and crystal pac-king as causes of poor aqueous solubility.

Figure 16 is one that we show to our medicinalchemists. The chemists ask us, how much solubilitydo I need in my compound? This is our answer.

Let me explain it by looking at the three middlebars, depicting a one milligram per kilogram dose. Ifthe clinical potency of the compound is thought to beabout one milligram per kilogram (though perhaps adose of 100 mgs may be a little higher), and if thepermeability (Ka) is average, then the chemist needs52 micrograms per milliliter minimum thermodynamicsolubility. If the permeability is in the upper-tenth per-centile, then you maybe only need 10 microgramsper milliliter. And if you are in the bottom-tenth per-centile in permeability – for example, you might havethat in a peptidomimetic – then you need several hun-dred micrograms per milliliter solubility.

Now, chemists who are making insoluble com-pounds see this data, and chemists are always lookingfor explanations, excuses so that their compounds lookbetter. So they look at the bars to the left-hand sideand they say: 'Aha, if my potency was much better, if a

0.10.1 low

Ka0.1 avg

Ka0.1 hgh

Ka1.0 low

Ka1.0 avg

Ka1.0 hgh

Ka10 low

Ka10 avg

Ka10 hgh

Ka

Sol

ubili

ty in

ug/

mL

Projected Dose in mg/kg

1000

10

21

5

207

52

10

2100

520

100

1

Figure 16.

MINIMUM ACCEPTABLE SOLUBILITY IN UG/ML BARS SHOWSTHE MINIMUM SOLUBILITY FOR LOW, MEDIUM AND

HIGH PERMEABILITY (KA) AT A CLINICAL DOSE.THE MIDDLE 3 BARS ARE FOR A 1 MG/KG DOSE.

WITH MEDIUM PERMEABILITY YOU NEED 52 UG/ML SOLUBILITY.

41

dose were very low, if I had a tenth of a milligram per ki-logram in-vivo potency, and if I had very good permea-bility, then maybe one microgram per milliliter solubilitywould be acceptable.' And then we have to explain toour chemists that it's extremely hard to find these highlyin-vivo potent compounds with low dose.

I estimate that among current clinical candidatesthese low-dose compounds are probably not morethan 10 percent, they're probably less than 10 per-cent of clinical candidates. It seems it's mostly a mat-ter of good luck to find these compounds; it does notseem as if there's currently any rational way to disco-ver these compounds. So most of the time, 80 per-cent of the time, chemists have to live with these onemill igram per kilogram candidate compounds, orabout 100 milligram, maybe 200, total-dose com-pounds.

Now there's another point with this Figure, and I'mreally glad that Professor Sugiyama showed some ofhis slides. Professor Sugiyama showed equations,right? Two or three years before I made up this Figureour pharmaceutical sciences people were talking toour chemists and, just like Professor Sugiyama, theyshowed equations. I don't know about chemists inJapan, but American chemists hate mathematicalequations. As soon as the pharmaceutical sciencespeople showed an equation, the chemists would lookaround and not pay any attention at all.

All I did was I took the equations in a pharmaceuti-cal sciences paper and converted them into a graph.The characteristics of chemists are that they are su-perb in pattern recognition, and a graph is a pattern,so they understand this, a chemistry structure is apattern… So the message here is, if you want to talk

to chemists, it's not enough to give the correct mes-sage. You must talk in a language and a way that thechemists really understand.

Now, Figure 17 is a graph that we also show toour chemists and it illustrates the relationships bet-ween the three properties that are important for earlyoral absorption that the chemists can control throughthe chemistry structure. They are: permeability fromworst to best; in-vitro potency from worst to best, andsolubility from worst to best. The square-shaped in-terior surface defines the region of space abovewhich real orally active drugs lie, and the circles arethe starting points, the leads. It's the chemist's job,through changing the structure, to get above this sur-face to obtain an orally active compound.

Now the problem is, if the chemist just movesalong the in-vitro potency arrow lying on the base, it'susually quite possible to get compounds that arehighly active in-vitro, and maybe go from micromolarto nanomolar. But in the process you very often donot improve solubility, nor do you improve permeabi-lity. So you end up in the bottom right-hand corner.You have a very active in-vitro compound. You canget a publication in a medicinal chemistry journal. Butthat compound will never make any money for yourcompany, and it will never help sick people, becauseit's not orally active.

Now I'm going to introduce the concept of the kindof solubilities you can expect for different collectionsof compounds. I was actually able to experimentallytest almost 1,600 Phase II compounds in our experi-mental assays (Figure 18). Here compounds that had20 micrograms per ml or lower solubility, would beconsidered to have poor solubility. About 14 percent

10000

100001000

10010

1

0.06 0.05

Best

Worst

Permeability (min-1)

0.04 0.03 0.02 0.01 0

1000

0.060.05

0.040.03

0.020.01

0

1

10 Worst

Best

potency (mg/kg)

0.1

100

10

1

Wor

st

S

olub

ility

(ug/

mL)

Bes

t

Real Drugs

Leads

Figure 17.

CHEMISTRY SAR OPTIMIZATION, AQUEOUS SOLUBILITY,PERMEABILITY AND ORAL ABSORPTION PROBLEMS

1400

0 5 10 15 20 25 30 35 40 45 50 55 60 65

Solubility ug/mL

200

400

600

800

1000

1200

Figure 18.

14.2 PERCENT OF 1597 PHASE II COMPOUNDS HAVELOW AQUEOUS SOLUBILITY (<20 UG/ML)

42

of these Phase II compounds had poor solubility. Ithink this is absolutely the best one could ever get,sort of a gold standard among compounds: 14 per-cent poor solubility.

What about compounds that came from commer-cial, from academic sources? I was able to test about2,200 of these compounds experimentally (Figure 19),and now the poor solubility rate is much higher, it'sabout 31 percent. These commercial compoundswere the ones that went into the high-throughputscreens, and the actives from those screens were thestarting points for our medicinal chemistry. So whatdid medicinal chemists do to the solubility, with thesekinds of starting points?

Well, the solubility got worse. Now, around 40 per-cent of the compounds had very poor solubility (Fi-gure 20). This level of solubility in general would notresult in any in-vivo activity in a biology assay andwould result in little or no systemic exposure in a drugmetabolism assay. So the chemists have made thecompounds less soluble. That makes a great deal ofsense because, when chemists increase in-vitro acti-vity, as I told you, they typically add to molecularweight, they add to lipophilicity, and that makes thecompounds less soluble.

The other point here is that I think the medicinalchemists in Groton are very good chemists, so this40 percent poor solubility is a very realistic number.It's probably the best you will ever achieve nowadays.If you have good organization, good planning, thenmaybe you can keep the poor solubility at 40 per-cent. Now that sounds like a lot, and it is, but it's stillpossible to get quality clinical candidates when che-mists are making 40 percent poorly soluble com-pounds. But if you are not very, very careful, this per-

centage of poor solubility easily can go up to 50 or60 percent or higher.

Figure 21 is a very busy and complicated Figure,so I will give you the message just from the title. As ofvery recently, the Pfizer Groton Laboratories screen-tested more than 60,000 compounds for aqueoussolubility. What we have found very consistently isthat in about 45 percent of the compounds that areexperimentally poorly soluble, we can detect a pro-blem of excessive lipophilicity. In fact, we found that ifthe calculated log-P of the compound is above theRule of 5 limits, above 5, about 75 percent of thosecompounds will be experimentally insoluble, poorlysoluble.

But in about 55 percent of the compounds thatare experimentally poorly soluble, the computer doesnot detect any kind of a problem. These compoundsdo not break the Rule of 5, and so the poor solubilitymust be due to a crystal packing issue. What thispoints out is that it's extremely important to have anexperimental solubility screen, because currently

1400

0 5 10 15 20 25 30 35 40 45 50 55 60 65

Solubility ug/mL

200

400

600

800

1000

1200

Figure 19.

31.2 PERCENT OF 2246 COMMERCIAL COMPOUNDS HAVELOW AQUEOUS SOLUBILITY (<= 20 UG/ML)

14000

0 5 10 15 20 25 30 35 40 45 50 55 60 65

Solubility ug/mL

2000

4000

6000

8000

10000

12000

Figure 20.

39.8 PERCENT OF 33093 MEDICINAL CHEMISTRY COMPOUNDSHAVE LOW SOLUBILITY (<= 20 UG/ML)

9000

0

1000

2000

3000

4000

5000

6000

7000

8000

PoorSolubil

Assays

Solubil>65

Solubil>65MLogP>4.15

MLogP>4.15

PoorSolubilMLogP>4.15

Figure 21.

POOR AQUEOUS SOLUBILITY, 45% DUE TO EXCESSIVELIPOPHILICITY, 55% DUE TO CRYSTAL PACKING

43

there is no computer program that will reliably predictpoorly soluble compounds due to crystal packing.

Now I'm going to go into a new section and talkabout how compounds differ in aqueous and DMSO(dimethylsulfoxide) solubility based on crystalline form.Let me explain. There was a German gentleman na-med Ostwald who came up with something called theRule of Stages in about 1880. It says that the se-quence of compound batch isolation proceeds to-wards the thermodynamically most stable form.

What that means is that when chemists first isolatea compound and they don't know much about theisolation method, it's very common that the first mate-rial isolated is amorphous; that's the highest energysolid form. As they learn more about the isolationconditions they gradually begin to isolate crystallinematerials, initially the highest energy polymorph, andfinally ending up with the lowest energy polymorph.The significance of that is that, uniformly, the amor-phous materials in the highest energy form are themost soluble, both in water and in DMSO.

Now I'm going to show you that amorphous com-pounds are becoming extremely common. I'm goingto talk about changes in compounds' solid-state pro-perties. I'm going to be talking about changes in puri-fication methods. I'm going to tell you that in many or-ganizations, melting points have disappeared, andthat predominantly the industry is now working onamorphous compounds.

In early discovery changes are occurring that affectcompound purity. There is a great deal of pressure onchemistry to increase output, to make more com-pounds. The managers tell the chemists, make morecompounds. How are the chemists going to do it?Well, one of the ways is to stop carrying out steps inthe chemistry process that take a lot of time.

First, crystallizing a compound to a sharp meltingpoint takes a lot of time, and you don't need a meltingpoint to know that you have made the correct com-pound because an NMR (nuclear magnetic reso-nance) spectrum or a mass spectrum will tell you that.Secondly, in combinatorial chemistry, combinatorialcompounds are now being purified by automatedprocedures, by automated HPLC (high performanceliquid chromatography), so nobody crystallizes combi-natorial compounds. These are the current, across-the-industry purity criteria for combinatorial com-pounds purified by automated procedures.

This level of purity is lower than when chemistscrystallize compounds, because crystallization is avery good way of removing impurities. So the conse-quence of this is that nowadays compounds appearmore soluble than before because they're predomi-nantly isolated in the amorphous state, and that's thestate that's more soluble, and also because theycontain more impurities and those impurities typicallyenhance solubility.

Well, Figure 22 is my proof, at least within the Gro-ton Laboratories, that melting points have disappea-red. It shows all of the compounds synthesized bytraditional medicinal chemistry from 1970 to the year2000. In 1970, 90 percent, or a 0.9 fraction of thecompounds our medicinal chemists made had a mel-ting point. Then from about 1995, melting points di-sappeared. In the year 2000, zero percent, zero frac-tion melting points. Actually, it wasn't zero, we had 18compounds with a melting point.

So what happened in the year 2000? Our pharma-ceutical sciences people started a class to teach ournew chemists how to crystallize compounds. Wetaught the chemists about the importance of solid-state properties, about solid forms. Why was that?

The reason was that it's perfectly acceptable towork with amorphous compounds in early discoverybut when you get close to the late discovery/develop-ment interface, and especially if you're interested inoral absorption, you must compare crystalline mate-rials. We had to get the chemists to change their che-mistry, change how they isolated the compounds, sothat in the late stages they would start to really makecrystalline compounds. Since they didn't know howto do this, we had to train them. This is just the newerchemists, the older chemists knew how to do it.

What are the consequences of amorphous com-pounds? Amorphous aqueous solubility is always hi-gher than when the compound is crystalline, whichagain reiterates the point. The crystalline state is veryimportant in late discovery because you oftentimes

0.90

0.001970 1975 1980 1985 1990 1995 2000

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

Figure 22.

FRACTION OF PFIZER GROTON COMPOUNDSHAVING MELTING POINT FIELD INFORMATION

44

can get poorer oral absorption when a compound iscrystallized.

We actually had one clinical candidate in the Gro-ton Labs that we chose when it was amorphous, andthen it was crystallized. The solubility dropped by overa factor of 100 and it caused all kinds of problems.We were able to rescue that compound with formula-tion technology but it was very, very difficult. That'sthe kind of situation you would really like to avoid.

Now, amorphous compound solubility in DMSO isjust like it is in water; it is always higher than when thecompound is crystalline. What that means is thatmost of these amorphous compounds dissolve relati-vely easily in DMSO. In my laboratory, we dissolvedover 40,000 compounds in DMSO at quite highconcentration, 60 millimolar and in general we hadvery little trouble dissolving the compounds.

So we were able to dissolve the compounds inDMSO. But you are more likely to get errors in thescreening data because, if the compounds precipi-tate from DMSO solution then the screening concen-tration may be much, much less than you think it is.And if the compounds stay in DMSO then you aremuch more likely to test insoluble compounds. So bytesting DMSO solutions you end up with many moreinsoluble compounds as leads, but if you had startedby testing powdered crystalline compounds you'd ne-ver have been able to get them in solution, and there-fore you'd never have detected the activity in an as-say.

Now this is something I believe very strongly. Atthe Groton Laboratories we maintain two kinds of so-lubility assays: one intended for very early discovery,one intended for late discovery and the developmentinterface.

The early discovery assays start from the DMSOstock solution; we actually use a light-scattering end-point. This kind of assay always overestimates the so-lubility versus the thermodynamic and it's relevant toearly discovery in-vivo SAR (structure activity relation-ships). We use this kind of assay very early in disco-very when the chemists are changing their structureand they're just trying to get a little bit of activity in-vivo, a little bit of systemic exposure.

But when you come close to the development in-terface then you have to do a proper thermodynamicsolubility assay. We've actually built a robot which willspeed that process up, but we demand that the che-mists give us crystalline materials because if youscreen amorphous materials in a thermodynamic so-lubility assay you could be completely wrong. The

thermodynamic assay has an equilibrium endpoint.It's essential at the clinical candidate stage and we tellour chemists that the distinction between these twotypes of solubility assays is absolutely critical. Thelight scattering assay is intended for early discoveryand must never be used late in discovery, when thethermodynamic assay should be used instead.

Compound solubility in water and DMSO is reallydetermined by two factors: solvation energy and crys-tal disruption. For any one compound there's a sort ofcontinuum in the importance of each. Maybe for onecompound, solvation energy is more important, foranother it could be crystal disruption. For the low mel-ting point, large lipophilic compound, solvation ismore important, and for the high melting point hydro-philic compound, crystal disruption is more important.

For solubility in water and DMSO, if you have alarge, lipophilic compound and it's maybe not verysoluble in water, then DMSO greatly helps the solubi-lity. But a very crystalline compound may show nocomputational problem, no Rule of 5 violation. It's in-soluble in water. It might have a high melting point –and that would be a clue to this kind of a problem, ifyou had melting points. Probably you have strong in-termolecular crystal lattice interactions. For that kindof a compound, DMSO just does not help improvesolubility. So if a compound is insoluble in water be-cause of this crystallinity issue, then ultimately it isvery likely the compound will come out as a crystalli-zed precipitate from DMSO.

Suppose you are an end user, a person who usescompounds dissolved in DMSO, and the compounddisappears from the DMSO solution. You might say,well, I don't care what the explanation is. But youreally should care because the explanation is impor-tant in understanding what to do to solve the pro-blem.

If the problem for the compound disappearing fromDMSO is a chemistry problem, a chemical integrityproblem, meaning that the compound is changing itsstructure, it's going to something else, then it's veryclear what you need to do. You need to keep the so-lution of compound in DMSO cold and frozen, youneed to avoid oxygen and you need to keep it dry.Why do you need to keep it dry? Because hydrolyticinstability requires water. If there's no water in DMSOyou cannot get hydrolysis.

If the problem though is precipitation crystallization,if that's what's removing the compound from DMSO,then keeping it cold and frozen is the absolute worstchoice possible. If you think about it, how does thechemist crystallize compounds? He / she takes so-

45

mething in solution and cools it down. You want toavoid that, and you want to avoid freeze/thaw cycles.

What we found in our laboratory is that there is a ti-ming factor in compound DMSO solubility. First of all,once a compound crystallizes from DMSO, very oftenit will not easily re-dissolve, and people don't unders-tand that. But what's happening is that when the com-pound is first dissolved in DMSO it is amorphous, so itgoes in at high solution. But when it crystallizes out, it'scoming out as crystalline material which is inherentlyless soluble and therefore you will probably have a lotof problems re-dissolving that compound in DMSO.

We also noticed a narrow working time-window forkeeping most compounds dissolved in DMSO. Whatwe found out is that generally there was no problemin keeping compounds in solution for the first day ortwo. But past that time, compounds started to preci-pitate. This explains why compounds are active whensolutions in DMSO are freshly made from powders,but not when compounds are stored for a long periodof time in DMSO. They have probably precipitatedand, again, freeze/thaw cycles increase the probabi-lity of crystallization.

Here is some very practical advice we give to theend user of compounds in DMSO. Figure 23 is aphase diagram between liquid DMSO and solidDMSO. The X-axis shows how much water there is inthe DMSO; at the extreme right of the diagram is verydry DMSO, no water. The melting point of dry DMSOis 18 degrees Centigrade. So the scientists think theycan stop crystallization by putting the compoundsdissolved in DMSO into the refrigerator, which isusually at about 4 or 5 degrees Centigrade, and theDMSO will be frozen, and you cannot get precipita-tion if the DMSO is frozen.

That argument is incorrect. The reason is that themelting point of DMSO is incredibly dependent on theamount of water in DMSO. If you just have 9 percentwater in DMSO, the melting point goes down from 18degrees to 4 degrees Centigrade. And 4 to 5 de-grees Centigrade is the temperature of the non-free-zer part of a laboratory refrigerator. So scientists arestoring samples in DMSO, they're still liquid, they arevery cold, these are ideal conditions for crystallization.

How easy is it to get 9 percent water? Incrediblyeasy. If you get the compounds in DMSO from thecentral organization that produces them, you probablyhave 5 percent water in them. Just open thesamples, especially in humid weather as you havehere (in Tokyo) at this time of year, and you easilyhave 9 percent or more. So the typical biology proce-dure of storing samples in DMSO in the non-freezerpart of a lab refrigerator is very bad practice.

This advice is for short-term storage, from a fewdays to a few weeks. Do not store samples in DMSOin a refrigerator. Keep them in a chemistry glass des-sicator at room temperature, keep dry nitrogen or ar-gon over the samples, maybe throw a towel over thedessicator to keep light out, just keep them at roomtemperature. This is what you should do for short-term storage.

Now I am in my last section, talking about solubilityand combinatorial chemistry. I'm going to tell you thatamong combinatorial libraries, solubility is a major pro-blem but permeability is seldom a problem.

Figure 24 is my illustration of this point. It shows acollection of about 48,000 compounds from a veryreliable source of combinatorial compounds. On thisCD-Rom there were 30 chemical sub-series and Iworked out the average calculated solubility for each

LiquidDMSO

4° C

-20° C

-73° C

9% H2O18% H2O

Ice

Ice + FrozenDMSO.3H2O

20

10

0

-10

-20

-30

-40

-50

-60

-70

-80

-90

-100

-110

-120

-130

100 20 30Weight per cent DMSO

DMSO containing 9% wateris unfrozen in the typicallab refrigerator

Nature (1969), 220, 1315-1317

Very bad practice. Typical biologyprocedure of storing samplesin DMSO in the non-freezer partof a lab refrigerator

40 50 60 70 80 90 100

FrozenDMSO

Figure 23.

DMSO – WATER PHASE DIAGRAM

70

01

Ave

rage

Cal

c S

olub

ility

60

50

40

30

20

10

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Calculated average aqueous solubility in ug/mL for 47,680 combinatorial compounds from the same commercial source made according to 30 different synthesis protocols. Solubility <= 20 ug/mL (calculated using the Pfizer turbidimetric solubility model) is generally incompatible with oral activity.

Figure 24.

SOLUBILITY PROFILES CAN BE VERY DEPENDENTON THE CHEMISTRY SYNTHESIS PROTOCOL

46

of these sub-series. Now in this calculation less than20 ug/ml (micrograms per ml) represents bad solubi-lity, anything higher is approaching towards accep-table. The point here is that there's a lot of variation.This is the hallmark, the characteristic of combinatoriallibraries: aqueous solubility is very dependent on thechemical series, it varies a lot across chemical series.Now, what about permeability?

Figure 25 is very different. For each of these 30chemical series, I've calculated the average polar sur-face area. If you have a polar surface area of 150square angstroms or more, you likely are going tohave passive intestinal permeability problems. Here,only a single series, series 25 maybe has a problem,everything else is OK. Again, that's the characteristicof combinatorial libraries: you generally do not have apermeability problem.

Why is that? It's entirely a chemistry issue. In com-binatorial chemistry it's extremely easy to make large,lipophilic compounds. It's difficult to make hydrophilicpolar compounds. The automated purification proce-dures like reverse-phase HPLC that you use in com-binatorial libraries work much better for the large, lipo-philic compounds, but they work much more poorlyfor the hydrophilic compounds. So it's entirely a che-mistry issue.

Figure 26 gives another way of illustrating it. Eachof the squares is a collection of compounds and, foreach of those collections, I've calculated what per-cent of the compounds have a high polar surfacearea greater than 150 square angstrom. So if youmove out towards the right, on average those com-pounds will be less permeable.

I've also done a calculation for each library collec-tion of average poor solubility. To the lower left-hand

side is where 7,500 Phase II drugs lie; here are 330drugs with human fraction absorbed. This region ofchemistry space is what I call traditional drug space.This is the region of space occupied by nice, well-be-haved compounds. If you are working with com-pounds in this region of space you probably are notgoing to be very interested in predictive models or hi-gher-throughput experimental screens – for example,for solubility and permeability – because you don'tneed them.

Where do combinatorial libraries lie? They lie in theregion outside the traditional drug space, very consis-tently. I think the realistic goal for combinatorial libra-ries is to have about 40 percent poor solubility, justlike I showed you in the medicinal chemistry in earlierexamples. But if you do not have good library design,if your chemistry is bad, then poor solubility can justgo way up. It's very, very easy to get 50 to 60 percentpoor solubility in badly-designed, badly-made combi-natorial libraries. Actually, in the worst permeability re-gion to the right, among many combinatorial libraries Iwas only able to find three that had a permeabilityproblem. It's almost impossible to find combinatorialcompounds that have problems with permeability.

So here's my summary. I think it's very important touse the appropriate aqueous solubility assay. We usea kinetic assay in early discovery and at that stage it'sperfectly OK, in fact it's probably efficient, to workwith amorphous compounds. In late discovery weuse a thermodynamic assay, and we test crystallinecompounds only. We tell our chemists, and we teachthem, to worry about crystallinity in late discovery.Chemists need to know something about solid-stateproperties, that the crystalline state is important toaqueous solubility.

180

01

Ave

rage

pol

ar s

urfa

ce a

rea

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

20

40

60

80

100

120

140

160

Set of 47,680 combinatorial compounds from the same commercial source made according to 30 different synthesis protocols. There is little variation in average polar surface area across the protocols.

Figure 25.

PERMEABILITY PROFILES ARE NOT VERY DEPENDENTON CHEMISTRY SYNTHESIS PROTOCOL

50.00

0.000.00

Percent Polars Surface Area > 150 A^2

Per

cent

Poo

r S

olub

ility

10.0020.00

30.0040.00

50.0060.00

70.0080.00

90.00100.00

5.00

10.00

15.00

20.00

25.00

30.00

35.00

40.00

45.00

phase 2drugs Suggests 30% of drugs

are transportersubstrates

FA_manBest

Worst

Traditionaldrug space

Figure 26.

PERMEABILITY, SOLUBILITY GRID FOR CHEMICAL LIBRARIES

47

The need for collaboration between chemists andpharmaceutical scientists is also a very importantpoint. I'm glad that Professor Yamashita showed theslide from Professor Ronald Borchardt presented ear-lier this year at an ADMET conference in Whippany,NJ USA because collaboration between chemistsand pharmaceutical scientists and drug metabolismscientists is absolutely essential.

End users of compounds in DMSO need to bealert for compound precipitation from DMSO. A verypractical message: if you are storing compounds inDMSO for short periods of time, a few days to a fewweeks, do not store those DMSO compound stocksin the refrigerator.

And the final point is, I really believe that pooraqueous solubility is here to stay. I speak at many,many meetings. I have not seen anything that wouldtell me that the solubility problem is going to improve.In fact, I think it's very realistic to say that about 40percent poor aqueous solubility is probably the bestyou can expect. You have to be really careful itdoesn't get much worse, because it's incredibly easyto get 50 to 60 percent – even worse than that –poor aqueous solubility.

Lastly, I'd like to thank the Pfizer Organization fortheir support of me in my post-retirement activities,and I'd like to thank you for your attention.

Professor Yuichi Sugiyama, University of To-kyo: Thank you, Chris, for your very exciting talk. It'snearly lunchtime now but I would like to take a few ques-tions from the floor, so please ask any questions. Japa-nese is also acceptable as we also have translators.

Ibuke, Fujisawa Pharmaceuticals, Japan:You made a number of interesting points. One wasthat Merck uses a structure-based approach and Pfi-zer a high-throughput approach, and there are diffe-rent issues associated with each. But which strategywas better at obtaining new product output?

Dr. Christopher A. Lipinski, Pfizer Inc., USA:Which strategy is better for new product output? Is iteverything except high-throughput screening, or is ithigh-throughput screening? I have to be honest here.The Pfizer Laboratories in Sandwich in the UnitedKingdom generate their starting points almost exactlylike the Merck Organization and in the 1990s all thedrugs from Pfizer came from the English laboratories.They did not come from the Pfizer Laboratories inGroton, Connecticut, so you could probably draw aconclusion from that.

Ibuke, Fujisawa Pharmaceuticals, Japan:The reason I asked this question is for pharmaceutical

scientists. Poor solubility is manageable in our view,through various solid dispersion methods or so-callednon-aqueous systems, using solvents and dissolvingthose. On the other hand, poor permeability is intrin-sic to the nature of the compound itself, and so wecannot overcome this issue. So, for me, poor per-meability is a larger challenge. That was my impres-sion. Do you have any comments on this?

Dr. Christopher A. Lipinski, Pfizer Inc., USA:What can you do about poor permeability? Let's fo-cus on chemistry. There actually is something youcan do about it in chemistry but there's a problem inthat you can't computationally predict it. The literatureis very, very clear. If you can form an intramolecularhydrogen bond within a compound, that will improvepermeability by at least a factor of 10, maybe 20 or30. So for those kinds of compounds that are confor-mationally flexible, that can form an intramolecular hy-drogen bond, it's maybe a good idea to have an ex-perimental assay that will pick up that property; aparallel artificial membrane permeability assay, forexample.

Now the problem from the library design perspec-tive is that there is no computational program that canpredict very well, either for a single compound or formany compounds, whether in terms of energy a com-pound forms an intramolecular hydrogen bond. Andin many of the meetings at which I speak, I tell peopleI hope that there are people in the audience who de-velop software. I tell them that this is a big opportunitybecause if an organization had a software method forpredicting an intramolecular hydrogen bond, then thatwould be a great advantage.

Now once you get away from intramolecular hy-drogen bonding, then the questioner is completelycorrect; there is nothing you can do in formulation.The only thing you can do is change the chemicalstructure, make a prodrug. So if you have to deal witha problem in terms of poor oral absorption, it's muchbetter to deal with poorly soluble compounds than todeal with poorly permeable compounds because youreally don't have any formulation rescue technologythere to help you.

Question from the audience: Thank you verymuch. Can I ask a question in English? Figure 15clearly shows that the number of compounds withRule of 5 alerts decreases with time. Is this becauseof the introduction of so-called focused library synthe-sis, or are there other reasons?

Dr. Christopher A. Lipinski, Pfizer Inc., USA:You saw that over the years chemists were makinglarger and more lipophilic, less soluble compounds.

48

We can correlate that with the use of high-throughputscreening. If the starting points for the chemistry areleads from high-throughput screening and if you donot deliberately correct for that, then inevitably every-thing the chemist will make will be larger and more li-pophilic, because their starting points are larger andlipophilic. In the normal chemistry optimization of in-vi-tro activity, it's very common that the molecular weightincreases by 75 units and the lipophilicity increasesby a log-P of 1 or 2. So it's not only if the startingpoint is difficult, but also that when the chemistschange the structure they always make the structuremore difficult in terms of properties for absorption.

Question from the audience: I have another re-lated question. Actually, the percentage of alerts issomething like 20 percent, right? My question isabout the strictness of the Rule of 5. For example, ifyou increase the strictness of each rule then, ofcourse, the percentage of compounds with alerts willincrease, and then you can decrease the high-throughput screening effort. Therefore my point is thatthe strictness and the effectiveness of the Rule of 5always depends on the money allocated to high-throughput screening. When you initially proposed theRule of 5, how did you determine the strictness foreach parameter?

Dr. Christopher A. Lipinski, Pfizer Inc., USA: I did it based on the 90th percentile, and I chosethose values because I wanted to be extremelyconservative. I was trying to convince chemists tochange their behavior and it would not have been agood idea to come up with rules where there werevery many exceptions, so I made the rules very strict.So, if you break two parameters in the Rule of 5 the-re's a very high probability that you will have a greatdeal of difficulty obtaining an orally active compound.

Now for the first part of your question. There is adifference in the viewpoint about the Rule of 5 acrossdifferent Pfizer laboratories. At our English laboratorieswhich, like Merck, do very little high-throughputscreening, they say it's bad to have just one parame-ter wrong.

At the Groton Laboratories, that does a lot of high-throughput screening, we allow two parameters; wesay it gets bad if two parameters are in the wrongrange. I can also tell you that for very well-designedcombinatorial libraries, we are able to hold the num-ber of compounds with two parameters in the wrongrange to below 1 percent, but for one parameter outof range it's around 20 percent.

Professor Yuichi Sugiyama, University of To-kyo: Thank you. It is now time to close the morningsession, so thank you again for your great contribu-tion to this symposium.

49

Novel approachesfor oral delivery of

poorly soluble drugs

Dr. Hirokazu OKAMOTO

51

Noriko Yamanouchi, Capsugel Japan: Gentle-men, we would like to begin the afternoon session. Dr. Ibuki will chair Session II-a, Invited Lectures.

Chair: Dr. Rinta Ibuki, Fujisawa Pharmaceu-tical Co., Ltd, Japan : I am Ibuki from FujisawaPharmaceuticals. Section II-a will include three pre-sentations, talking about various approaches to poorlysoluble drugs. The speakers are all very experiencedexperts in this field. Rather than give a formal lecturethey want to be provocative and stimulate discussion.They will each give a short presentation so that wecan have a longer discussion period.

The first speaker is Dr. Hirokazu Okamoto fromMeijo University. Dr. Okamoto, as you know, hasworked in both industry and academia. He is very ex-perienced and I am sure we will learn a lot from him.

Dr. Hirokazu Okamoto, Associate Professor,Meijo University, Nagoya, Japan: Thank you verymuch, Dr. Ibuki, for the introduction, and thank you forgiving me this opportunity. I would like to thank the or-ganisers and everyone involved. I will be talking aboutnovel approaches for oral delivery of poorly solubledrugs and from the title you can see it is rather ab-stract.

I'll be talking about solubility and absorption im-provement for poorly soluble drugs and will present anumber of methods. Dr. Yamashita said that we needdeep and wide knowledge, but I won't go into detailstoday. I won't go into the deep issues, but I'd like totalk about the wider issues.

Dr. Lipinksi mentioned HTS (high-throughputscreening) – in other words, combinatorial screeningprograms. He said that 40% of the compoundsscreened are poorly soluble in water, and talkedabout how to develop them into oral active products.

The issues associated with poorly soluble drugsinclude poor bioavailability; highly variable bioavailabil-ity between patients; highly variable bioavailability inthe fed or fasted states, and a slow onset of action.

Flux (J) is calculated as shown in Figure 1 and canusually be obtained by multiplying permeability (P) bythe drug concentration at the absorption site (C) andR. But, as Dr. Sugiyama mentioned, with recent com-pounds we now know that an increasing number gothrough first-pass metabolism, or efflux by a trans-porter such as P-glycoprotein; they are fluxed out,thrown out in other words. So flux would be repre-sented by the multiplication of these three factors, P,C and R.

Novel approaches for oral deliveryof poorly soluble drugs

Dr. Hirokazu Okamoto

Associate Professor, Meijo University, Nagoya, Japan.

Theoretical consideration for improvedabsorption of poorly soluble drugs

J = P x C x R

P: intrinsic permeability coefficient (R=1)

C: drug concentration at the absorption site

R: parameter between 0 to 1 representing the extent ofthe absorbed drug avoiding the first-pass metabolismand/or efflux by a transporter such as p-glycoprotein

Figure 1.

52

Figure 2 summarizes the possible methods avail-able that can enhance the oral bioavailability of suchpoorly soluble drugs. One method is the chemicalmodification of drug molecules by using prodrugs orsalt formation. For the physical modification of bulkdrugs, we have polymorphism selection and particlesize reduction. With modification of formulation ordrug products, we have co-solvents, emulsions,complexation – cyclodextrins, for example are used incomplexation – and also solid dispersion. So for eachof these, by increasing the solubility, the dissolutionrate can be improved. In addition, with formulationmodification the permeability can be improved or, insome cases, efflux and metabolism can be restrictedand this will improve the oral absorption of poorly sol-uble drugs.

I would now like to talk about prodrugs and particlesize reduction, co-solvents, emulsion and solid disper-sion. These techniques will each be explained in turn.

First of all, concerning prodrugs. In general, we aretargeting poorly soluble drugs and so they need to bewater-soluble. But if they are hydrophilic then theirpermeability tends to be reduced or sacrificed. Sohow can we increase absorption, or the water solubil-ity of the drug, while maintaining a high level of per-meability? That is our challenge.

Figure 3 shows one of the strategies to cope withthat issue. The prodrug itself has hydrophilicity, but inthe small intestine it passes through the membraneround the brush border and is metabolized by thebrush border enzyme (BBE). Then it changes to theparent drug and solubility is improved. Now, when thedrug passes through the membrane, the permeabilityremains at a high level, and so this prodrug strategycan improve solubility as well as maintain permeability.

The prodrugs shown in Figure 4 were designedaccording to this strategy. Cam 4451 was a parentdrug with low aqueous solubility. Leucine, dimethyl-glycine or phosphate ester was used for developingthe prodrug. The original solubility is less than 2 mi-crograms per milliliter and the solubility of each pro-drug is 0.1, 3 and more than 61 mill igrams permilliliter, and so solubility is significantly improved.

The table in Figure 4 contains some numbers. Thehalf-life in rat intestine perfusate is shown within thethird column. The next column is the half-life in brushborder membrane (BBM) homogenate that includesthe BBM enzymes. The selectivity ratio is in the mostright column. The larger this number is, the more sta-ble the prodrug is within the GI tract; it is also metabo-

Possible methods to improve the oral bioavailability of poorly soluble drugs

Intrinsic P C R

Chemical modification of drug moleculesProdrug ± ↑ ±Salt selection ± ↑ ±

Physical modification of bulk drugsPolymorph selection ± ↑ ±Particle size reduction ± ↑ ±

Formulational modification of drug productsCo-solvent ↓ ↑ ± or ↑Emulsion ± or ↑ ↑ ± or ↑Complexation ± or ↑ ↑ ± or ↑Solid dispersion ± ↑ ± or ↑

Figure 2.

PRODRUG

DRUGDRUG

DRUG

PRODRUG

PROGROUP

distance

co

nc

en

tra

tio

n

Fig. 1. Soluble prodrug provides a greater concentrational driving force for absorption in the intestinal lumen. Cleavage of the progroup by a membrane-bound brush border enzyme releases the lipophilic parent drug in the vicinity of the mucosal membrane.

D. Fleisher et al. Adv. Drug Delivery Rev.,19, 115, 1996.

Figure 3.

PRODRUG STRATEGY TO ENHANCE ORAL ABSORPTIONOF POORLY SOLUBLE DRUG

53

lized close to the membrane. So in these prodrugs thephosphate ester has high solubility, plus it is selec-tively hydrolyzed at the brush border membrane.

Turning now to focus on the right-hand side,graph b in Figure 5. This indicates the prodrug withthe phosphate ester and the parent drug. After ad-ministering it to the rat, the phosphate ester concen-tration was 10 times higher. So the solubility andbioavailability of a poorly soluble drug can be im-proved with a prodrug.

Going on to the next topic, Figure 6 is about parti-cle size reduction strategy. Reducing particle size hasbeen a conventional method and the dissolution ratecan be calculated according to the formula shown inthe slide. By reducing the particle size the surface areawill increase and so the dissolution rate is accelerated.Another advantage of doing this is that several microm-eters at nano level means a higher saturated solutionand better solubility. Solubility is shown by the upward-facing arrows in the Figure, and the dissolution rate in-creases with smaller particle size, as you can see.

Table 1. Reconversion Half-Lives and Selectivity Ratio of Prodrugs

Fig. 1. Chemical structures of Cam-2445, Cam-4451, prodrugs, and internal standard

Prodrug Pro-group t1/2 in perfusate t1/2 in BBM (min) Selectivity ratio

Cam-4580 Leucine 5.34 ± 0.98 148 ± 38.4 0.036Cam-4562 Dimethylglycine 17.3 ± 5.3 70.5 ± 6.5 0.245Cam-5223 Phosphate 5.62 ± 2.5 0.0668 ± 0.0062 84

Note: Selectivity ratio = t1/2 in perfusate / t1/2 in BBM.

R

O

O

ONH NH

N

N

0

0

0

0R

NH NH

R CamH 2356 (internal standard)CH3 2445

O

O

N

O

O

NH2

R CamOH 4451

4562

4580

5223

O

PO OH

OH

O. H. Chan et al. Pharm. Res., 15, 1012, 1998.

Figure 4.

DESIGN OF WATER-SOLUBLE PRODRUGS TARGETED TO INTESTINAL BBM ENZYMES

Fig. 2. Mean (± SD) plasma Cam-4451 concentrations in male Wistar rats after administration of a single dose, (a) Cam-4451 PO ( ), Cam-4451ID (x). Cam-4562 PO ( ), or Cam-5223 PO ( ) in ethanol/PEG 400 cosolvent (Inset: Cam-4451 IV), (b) Cam-4451 PO ( ) or Cam-5223 PO ( )in methylcellulose. Cam-4451 concentrations were normalized to a 20 mg/kg dose-equivalent of parent compound PO or ID.O. H. Chan et al. Pharm. Res., 15, 1012, 1998.

10.000

1.000

0.00 2

Time (h)

a

Dos

e-no

rmal

ized

Cam

-445

1co

ncen

trat

ion

(µg/

mL)

0.100

0.010

4 6 8 10

0.001

Cam

-445

1 co

ncen

trat

ion

(µg/

mL)

2

0.0100.100

1 00010 000

100 000

12 14

4 6 8 10 12 140

10.000

1.000

0.00 2

Time (h)

b

0.100

0.010

4 6

R Cam4451

Cam

5223

OH

R

OHO POH

O

8 10 12 14

Figure 5.

IMPROVED ABSORPTION BY WATER-SOLUBLE PRODRUGS TARGETED TO INTESTINAL BBM ENZYMES

54

It is necessary to have homogeneous particles. Ifyou have a mixture of large and small particles, thesmaller ones have high solubility and the dissolveddrug is used for growth of the larger particles, so thisis not a good strategy. Instead, the drug should bemade up of small, fine and homogeneous particles.

Figure 7 is an example of nanocrystalline drug par-ticles. To disperse them evenly you need to add anumber of stabilizers. Stabilizers are usually additivessuch as the cellulose type of polymers, polyvinylalco-hol and PVP.

Figure 8 shows the apparatus used for the reduc-tion of particle size. The drug and the stabilizer are in-corporated in a suspension and this is put into themedia milling machine. The media and the polymerare mixed together at high speed and a sheer force isused for reducing the particle size. After 30 to 60 min-utes of processing it is brought down to nano-ordersize. The smaller the size of the particle, the higher theplasma concentration. With a reduced particle size,the absorption rate goes up. Not only that, but thereare a number of other absorption advantages; the ef-fect of food is not as large as before, for example.

Nanocrystalline dispersions consist of water, drug and stabilizer.

The stabilizer must be capable of wetting the surface of the drug crystals and providing a steric or ionic barrier.

E. Merisko-Liversidge et al. Eur. J. Pharm. Sci., 18,113, 2003.

Fig. 1. Nanocrystalline drug particles. The transmission electron micro-graph of a NanoCrystal® Colloidal dispersion magnified 35,000 x. The insert provides a visual description of the crystalline nanoparticles generated using wet milling technology. The nanoparticles are typically less than 400 nm and are physically stabilized with a polymeric excipient.

Figure 7.

NANOCRYSTALLINE DRUG PARTICLES

A: effective surface area of solid drugD: diffusion coefficient of the drug in GI fluidL: effective thickness of diffusion boundary layerCs & C: solubility & drug concentration in GI fluid

Coarse andheterogeneousparticles

small A

small Cs

small V

large A

large Cs

large V

Fine andhomoneneousparticles

V = (Cs – C)AD L

Figure 6.

PARTICLE SIZE REDUCTION STRATEGY TO ENHANCEORAL ABSORPTION OF POORLY SOLUBLE DRUG

Fig. 2. The Media Milling Process is shown in a schematic representation. The milling chambercharged with polymeric media is the active component of the mill. The mill can be operated in a batch orre-circulation mode. A crude slurry consisting of drug, water and stabilizer is fed into the milling chamber and processed into a nanocrystalline dispersion. The typical residence time required to generate a nanometer-sized dispersion with a mean diameter < 200 nm is 30-60 min.

E. Merisko-Liversidge et al. Eur. J. Pharm. Sci.,18, 113, 2003.

Coolant

Motor

Screenretaining

milling mediain chamber

Nanocrystals

Media

MillingChamber

LargeDrug

Crystals

Chargedwith drug.

waterstabilizer

Re-circulationChamber

MillingShaft

Figure 8.

MEDIA MILLING PROCESS

55

Figure 9 shows danazol, a poorly soluble drugs.With the marketed product, post-prandial absorption– the bioavailability – is six times higher than pre-pran-dial. This is because bile salts improve the absorp-tion. But if you make a nano crystalline, the solubilityis higher from the start, so the role of bile salts in im-proving solubility is reduced. As a result, the bioavail-ability before and after food is not so variable; it isclose to a ratio of 1.

I have talked about the drug substance being re-duced in size, but now let me talk about the actualformulation being reduced. In Figure 10 they used ananoparticles with an enteral-coated additive. Theparticle size is 300 nanometers and the X-ray diffrac-tion data b means that it has become amorphous.Spray-dried microparticles 10 micrometers in size

were used as a control, and you can see that X-raydiffraction c is also amorphous.

The blood concentration data is shown in Figure 11.The pattern of plasma concentration is as shown. Thepharmacokinetic parameters are summarized in thetop chart. If you compare the dose-normalized AUC(the area under the curve) the control, which is theconventional formulation, is 1.9, and in comparisonthe nanoparticles AUC have risen to 6.3. With thespray-dried microparticles it is a little lower than that.So by reducing the particle size to 300 nanometersyou can increase the solubility and bioavailability aswell. Compared to the control and microparticles thevariability of AUC is smaller and so precision in thedosing can be achieved by the nanoparticles.

RR01: aqueous solubility = 0.09 mg/LEudragit L100-55: soluble intestinal fluid above pH 5.5Nanoparticles (300 nm): emusification-diffusion method with PVA followed by freeze dryMicroparticles (10mm): spray dry

2 Theta (deg)

Rel

ativ

e in

tens

ity

Figure 4. X-ray diffraction pattern of (a) compound RR01, (b) Eudragit L100-55 nanoparticles (drug loading 50%), (c) Eudragit L100-55 microparticles (drug loading 50%), (d) physical mixture of compound RR01 and Eudragit L100-55 (50% of compound RR01), and (e) Eudragit L100-55.

F. De Jaeghere et al. AAPS Pharmsci, 3, article 8, 2001.

Figure 10.

PREPARATION OF NANOPARTICLES OF A POORLY SOLUBLE DRUG

0

Rat

io o

f Bio

avai

labi

lity

Fed

to F

aste

d S

tate

MarketedProduct

NanoCrystallineCapsule

NanoCrystallineDispersion

6

5

4

3

2

1

Fig. 5. Nanocrystalline particles can reduce absorption variability resulting from the presence or absence of food. The data compare the performance of various dosage forms of danazol administered to volunteers in the fed and fasted state at a 200 mg dose. The variability observed in the commercial product was significantly lowered when danazol was formulated using nanoparticles and administered as a liquid dispersion or a dry-filled capsule.

E. Merisko-Liversidge et al. Eur. J. Pharm. Sci., 18, 113, 2003.

Figure 9.

REDUCED ABSORPTION VARIABILITY OF A POORLYSOLUBLE DRUG BY PARTICLE SIZE REDUCTION

Table 2. Pharmacokinetic Parameters of the Test Drug Administered Orallyto Fasted Beagle Dogs as Control Suspension and Incorporatedinto Eudragit L100-55 Particles; Mean (CV, n=6)

Formulation Dose Tmax Cmax AUC0-48h Normalized AUC0-48h

(mg/kg) (h) (µg/mL) (µg.h/mL) (µg.h/mL)(mg/kg)

Control 3.5 4 - 24 0.3 (113) 6.5 (116) 1.9 (116)

Nanoparticles 4.3 1.5 - 6 1.5 (18) 27.1 (19) 6.3 (19)

Microparticles 4.1 1.5 - 6 1.1 (48) 17.7 (40) 4.3 (40)

Figure 5. Plasma concentration profiles of compound RR01 after oral administration of the reference formulation ( ), Eudragit L100-55 nanoparticles ( ), and microparticles ( )(drug loading 5%) to fasted dogs; mean ± SD (n=6).

F. De Jaeghere et al. AAPS Pharmsci, 3, article 8, 2001.0

0 8

0.4

0.8

1.2

1.6

16 24 32 40 48

Figure 11.

IMPROVED ABSORPTION OF A POORLY SOLUBLE DRUG BY NANOPARTICLES

56

Those are the advantages of micronization andnow I would like to talk about the use of co-solvents.In our view, five of the merits of using co-solvents are:increased drug solubility in co-solvent/GI fluid mixture;increased permeability; decreased metabolism and/orefflux transport; fine particles if precipitation occurs,and micelle formation with surfactants. There may besome other advantages, too.

Figure 12 is the only example of work that comesout of our laboratory. We administered indomethacinto rat intestine and noted the absorption. We usedtwo capsule forms, one including PEG 400 solution,

and the other containing indomethacin powder withsome additives. The results show that with the solu-tion concentration goes up significantly, and the ab-sorption obtained is also higher, as calculated by thedeconvolution method. So by using PEG the result isthat we can get better absorption at a faster pace.

There is a further advantage to using co-solvent –and not just as a solution that enhances absorption.Another reason is that, depending on the compound,the co-solvent itself may have an inhibitory effect onthe metabolism.

6

00 60

Time (min)

conc

entr

atio

n (µ

/mL)

Abs

orpt

ion

(%)

Time (min)

120 180 240 300 360

PEG400

Powder

PEG400

Powder

8

4

2

Formulation AUC Extent of Absorption(µgram•min/mL) (%/360min)

PEG400 solution (IM 2.0mg+PEG400 31.5mg)/cap 1829±507 27.4

Powder (IM 2.0mg+lactose 27.5mg +Ac-Di-Sol 1.5mg)/cap 357±63 5.5

25

00 60 120 180 240 300 360

30

20

15

10

5

Figure 12.

IMPROVED ABSORPTION OF INDOMETHACINFROM PEG400 SOLUTION

Fig. 1. Amprenavir solubility as a function of vitamin E-TPGS concentration in pH 7 phosphate buffer, ionic strength 0.15M. The cmc of vitamin E-TPGS was determined to be 0.2 mg/mL from the inset figure. The solubility of amprenavir was unchanged below the cmc and increased linearly with increasing vitamin E-TPGS concentratin above the cmc. The data point and error bars represent the mean ± SD of three replicates.

L. Yu et al. Pharm. Res., 16, 1812, 1999.

Vitamin E-TPGS: d-a-tocopheryl polyethylene glycol 1000 succinate

0.8

0.7

0.00 5

Vitamin e-TPGS concentration (mg/mL)

R2= 0.999

Am

pren

avir

solu

bilit

y (m

g/m

L)

0.6

0.5

0.4

0.3

0.2

0.1

10 15 20 25

Figure 14.

VITAMIN E-TPGS SOLUBILIZE A POORLY SOLUBLE DRUG

00.00 0.05

% excipient in incubation

Propylene glycol

Cyclodextrin

PEG400

% a

ctiv

ity

25

50

75

100

125

0.10 0.15 0.20 0.25 0.30 0.50 0.75 1.00

Dimethylacetamide

N-methylprrolidone

Oleic acid

Tween 80 Tween 20

00.00 0.05

% excipient in incubation

% a

ctiv

ity

120110100908075

50

25

0.10 0.15 0.20 0.25 0.30 0.50 0.75 1.00

Soybean oil

Vit. E:PEG400

Cremophor:PEG400

Glycocholic/lecithin

Cremophor: waterTaurocholate/lecithin Vit. E:water

Miglyol

Fig. 1. Effect of increasing concentrations of various excipients on the activity of cDNA expressed human CYP3A4. Inhibition of CYP3A4 activity by the various excipientswas calculated based on the inhibition of the conversion of the probe CYP3A4 substrate 7-benzyloxy-4-(trifluoromethyl)-coumarin (BFC) to the fluorescent metabolite7-hydroxy-4-(trifluoromethyl)-coumarin (7-HFC), as described by Crespi et al. (1997). R. J. Mountfield et al. Int. J. Pharm., 211, 89, 2000.

Figure 13.

INHIBITORY EFFECT OF FORMULATION INGREDIENTS ON CYP3A4

57

As was mentioned by Dr. Sugiyama this morning,with CYP3A4 there are various inhibitory effects andthey have studied those. In Figure 13 the horizontalaxis shows the excipient concentration in incubationand the inhibitory activity is shown on the vertical axis.From this you can see that maybe Tween 80, oleicacid, or by using various additives, CYP3A4 activity isinhibited. That's what this diagram shows.

Figure 14 is another example, using vitamin E-TPGS. It was used for the formulation of amprenavir(Agenerase). Dissolution of the drug is dependent onthe concentration of vitamin E-TPGS, and this wouldimprove the solubility. On the Y-axis of the insetgraph, first there is a horizontal line and then a slight

upwards bend; this is where the vitamin E-TPGS con-centration reached its CMC. In this case it's at 0.2mg per milliliter.

Figure 15 is a CACO-2 cell culture that was usedto demonstrate flux and efflux from the apical to thebasal and from the basal to the apical. What kind ofabsorption was seen? The horizontal line shows thevitamin E-TPGS concentration; as the concentrationlevel goes up, so the flux direction rises. But once itgoes above the CMC then micellar uptake takesplace, and so the apparent permeability is reduced.On the other hand, vitamin E-TPGS inhibits flux fromthe basal to the apical, and as the drug is a P-glyco-protein substrate, vitamin E-TPGS is also inhibiting itsactivity. So using this kind of co-solvent in a formula-tion not only solubilizes the drug, it also improvesmembrane absorption permeability by reducing efflux.

And now turning to another subject, to emulsionstrategy. As for the emulsions themselves, severalother speakers will be touching on this, so due totime constraints I would like to skip some of the ex-planations. One type that is currently hotly debated isSEDDS, or self-emulsifying drug delivery systems.SEDDS is an emulsion system that uses oils, surfac-

35

0

Vitamin E-TPGS concentration (mg/mL)

App

aren

t per

mea

bilit

y x1

06 (cm

/sec

)

5

2

10

15

20

25

30

L. Yu et al. Pharm. Res.,16, 1812, 1999.

Fig. 3. Effect of vitamins E-TPGS on transport of amprenavir.The data point and error bars represent the mean ± SD.

0.020.0002 0.002 0.20

AP > BLBL > AP

Figure 15.

INCREASED FLUX AND DECREASED EFFLUX BY VITAMIN E-TPGS

Fig. 5. Examples of phase diagrams. The mixture of the equal amount of MCG and DCPG was employed as an oil phase. Ringer's solution was used as an aqueous phase. The surfactant and temperature were as follows: (a) Tw80, 25°C; (b) Tw80, 37°C; (c) C12E9, 25°C; (d) C12E9, 37°C. ME phases are shown as the shaded area. The dark regions represent the highly-viscous phase (see text).

K. Kawakami et al. J. Cotrol. Release, 81, 65, 2002.

C12E9: polyoxyethylene(9) monolauric ether (BL-9EX)

MCG: glycerol monocaprylic ester (Homotex PT)

DCPG: propyleneglycol dicaprylic ester Sefsol-228)

MCG/DCPG

Tw80

Ringer's solution

(a) 1

0.80.2

0

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

0.6

0.4

0.2

0

W/0

0/W

MCG/DCPG

Tw80

Ringer's solution

(b) 1

0.80.2

0

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

0.6

0.4

0.2

0

W/0

0/W

MCG/DCPG

C12E9

Ringer's solution

(c) 1

0.80.2

0

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

0.6

0.4

0.2

0

MCG/DCPG

C12E9

Ringer's solution

(d) 1

0.80.2

0

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

0.6

0.4

0.2

0

W/0

0/W

W/0

0/W

Figure 17.

PHASE DIAGRAMS FOR SEDDS DILUTED WITH WATER

System Surfactant Cosurfactant Oil Km Ratio(% wt/wt) (% wt/wt) (% wt/wt)

System 1 60 20 20 3:1

System 2 40 30 30 1.3:1

Surfactant: dioctyl sodiumsulfosuccinate (aerosol OT)

Cosurfactant: glycerolmonooleate (Monomuls 90-018)

Oil: a blend of Miglyol 812and oleic acid in 1:1 ratio

Table 1. Composition of the 2 Investigated Microemulsion Systems

Figure 5. Effect of dilution on (D)system 1 and (E) system 2

H. M. El-Laithy. AAPS PharmSciTech,4, Article 11, 2003.

Figure 16.

CLEAR AND TRANSPARENT FORMULATION OBTAINEDBY DILUTION OF SEDDS WITH WATER

58

tants and drugs, and by administering it in the gas-trointestinal tract, O/W (oil in water) emulsion will bespontaneously generated.

Figure 16 is a comparison of two formulations. Theone at the top, system 1, is the SEDDS; system 2, iswithout SEDDS. If we add water then (D) shows the

result for system 1. In (D) we get a very clear andtransparent formulation, whereas in (E), which is with-out SEDDS, it is white and you get a homogeneouslayer. What the diagrams in Figure 17 are showing isthe relationship between oils, surfactants and the GIfluid, and the kind of water emulsion that can beformed.

Figure 18 shows the relationship between dropletsize and the release rate. As the concentration levelgoes up, the droplet size goes down and the releaserate is also improved. So, with SEDDS, it's better touse smaller particles in order to obtain higher absorp-tion.

Figure 19 is an in-vitro dissolution test and an in-vivo absorption test using an SEDDS solution, withdrug in a PEG 400 solution as control. In the dissolu-tion test, SEDDS had a worse score than the PEGsolution, but after it was administered to a dog,SEDDS performed better than the PEG. Dissolutionwas not good, but absorption was higher withSEDDS. Maybe some recrystallization takes place inthe PEG solution, maybe that's the reason why ab-sorption might be more difficult, whereas the SEDDScreates a clearer solution.

In Figure 20 the three bottom formulations in theTable are SEDDS-related formulations, along with anMC (methylcellulose) suspension. Chart (a), normal,indicates the fed state and chart (b) the fasted state.

Fig. 8. Effect of concentration of emulsifier, Labrafac CM 10 BM 287, on partition coefficient ( ) droplet size ( ) and the release rate ( ) of Ro 15-0778.

N. H. Shah et al. Int. J. Pharm., 106, 15, 1994.

Drug: RO15-0778 (Cswater<0.01mg/mL, Cspeanut oil = 95 mg/mL)Emulsifier: Labrafac CM10 BM287 (HLB=10)Oil: peanut oil

40

0 00 20

Emulsifier Concentration (%)

10 10

20

30

40

20

30

40 60 80

Release rate k x 10

3 (Min

-1)

Par

titio

n co

effic

ient

(PC

oil/w

ater

)D

ropl

et s

ize

(u)

Figure 18.

SMALLER DROPLET SIZE AND FASTERRELEASE RATE BY A SURFACTANT

Formulation Cmax (µ.g/ml) fmax (h) AUC (µ.g h ml-1) % relativebioavailability

Self-emulsified solution (SEDDS) 5.57 2.50 29.77 389.0

Drug solution in PEG 400 (control) 1.44 2.00 7.64 100.0

Capsule form. of wet-milled spray dried powder 0.78 3.00 2.69 35.3

Tablet form. of micronized drug 0.58 2.00 1.32 17.2

Table 3. Pharmacokinetic parameters of Ro 15-0778 from different formulations in non-fasting dogs

% r

elea

sed

PEG 400SEDDS SEDDS

N. H. Shah et al. Int. J. Pharm., 106, 15, 1994.

Fig. 12. In vitro dissolution drug profile from different formulations.( ) SEDDS ; ( ) 1.2% PEG 400 ; ( ) wet milled spray dried powder ;( ) micronized drug.

00 20

Time (Min)

20

40

60

80

100

120

40 60 80 100 120 140

SEDDS

Pla

sma

conc

entr

atio

n (µ

g/m

l)

Fig. 13. Mean plasma concentration of Ro 15-0778 (in non-fasted dogs, after oral administration of four different formulations. ( ) SEDDS ;( ) 1.2% PEG 400 ; ( ) wet milled spray dried powder ;( ) micronized drug.

0.010 5

Time (Min)

0.1

1

10

10 15 20 25 30 35

PEG 400

PEG 400SEDDSPEG 400

Figure 19.

REDUCED FED/FASTED VARIABILITY OF BIOAVAILABILITY BY SEDDS

59

In the fasted state, absorption from the MC suspen-sion is very poor; the difference in the fed state isabout 20-fold. But with the microemulsion system,absorption is very good in both cases, whether it's afed or fasted situation.

Figure 21 shows two well-known cyclosporine for-mulations, Sandimmune and Neoral. By using a rangeof cyclosporine dosages it was possible to calculate

the AUC ratio against the dose. With Neoral, the AUCagainst dose is constant and there is linearity,whereas with Sandimmune there is no constancy, noproportion can be seen, there is no linearity. As a re-sult, Neoral with a smaller droplet size allows moreprecise administration.

The final example I would like to discuss is thesolid dispersion system (Figure 22). This is a three-component solid dispersion system using YM022,TC-5E and HCO-60, and it's been formulated byspray drying technique. The drug becomes a soliddispersion system and, if it is dispersed in water, thenthe particle size is about 160 nanometers and we geta colloid type of particle. I won't go into details due tolack of time, but the drug and HCO-60 interact witheach other, whereas TC-5E would act as the frame-work to form that colloid.

Figure 23 shows YMO22 formulations that wereadministered to dogs. The concentration level is veryhigh when drug is administered as solid dispersionsystem (SD5), and also when it's in suspension(SD5W). Compared to the physical mixtures (APM) orthe drug alone (AP), absorption is highly improvedwhen solid dispersion system (SD5) are used. Theformulation is dispersed in the intestine where it formsa colloid, and absorption is improved.

Formulation Tmax (h) Cmax (µ.g/ml) AUC (µ.g h/ml) AUC ratio

Normal Fasted Normal Fasted Normal Fasted (normal/fasted)

MC suspension 8.0±0.0 1.5±0.5 0.23±0.03 0.04±0.00 1.03 0.05 21.4

Oil solution 4.0±0.0 3.5±1.7 0.53±0.09 0.30±0.09 2.55 1.71 1.50

Tw80 ME 1.3±0.7 1.3±0.7 0.36±0.00 0.59±0.12 2.09 2.50 0.84

C12E9 ME 4.3±3.8 3.0±1.0 0.22±0.19 0.08±0.00 0.58 0.47 1.25

HCO60 ME ≥8.0 >7.0 ≥1.45 ≥1.44 7.70 6.43 1.20

Pharmacokinetic parameters of oral administration study of nitrendipine

Tmax : Time to reach maximum drug concentration (average ±S.E.), Cmax : maximum drug concentration (average ± S.E.),AUC: area under on line curve from 0 to 8 h calculated by the trapezoidal method.

Pla

sma

Con

c. (µ

g/m

l)

(a) Normal

K. Kawakami et al. J. Cotrol. Release, 81, 75, 2002.

Fig. 2. Plasma concentration profiles after oral administration of nitrendipine formulations to rats under (a) normal and (b) fasted condition.Each data point is an averaged value, and the error bars represent the standard errors. Formulation type: ( ) MC suspension, ( ) oil solution,( ) Tw80 ME, ( ) C12E9 ME, ( ) HCO60 ME. The formulation prescriptions are shown in Table 1.

x

xx xxxx0.0

0 2Time (Hrs)

0.5

1.0

1.5

4 6 8

Pla

sma

Con

c. (µ

g/m

l)

(b) Fasted

0.00 2

Time (Hrs)

0.5

1.0

1.5

4 6 8x x x x

Figure 20.

REDUCED FED/FASTED VARIABILITY OF BIOAVAILABILITY BY SEDDS

Sandimmune®: internal dropletsize upon dilution = 864 nm

Neoral®: internal dropletsize upon dilution = 39 nm

AU

C/D

ose

[ng

•h/m

L p

er m

g]

Dose [mg]

E. A. Mueller et al. Pharm. Res., 11, 301, 1994.

Fig. 1. Boxplot comparing dose-normalized cyclosporine AUCb at four dose levels following single oral administrations of the reference formulation (top) and the test formulation (bottom) to healthy volunteers. The box, which contains 50% of the observations ( ), is divided at the median and has a height equal to the interquartile range. The tails indicate the range of the upper and lower 25% of observations. Not depicted is an outlier value (33.4 ng•hr/mL per mg) for the test formulation at the 200-mg dose level.

20 30

20

10

0200 400 600 800

Dose [mg]200 400 600 800

15

10

5

0

Figure 21.

IMPROVED DOSE-LINEARITY IN CYCLOSPORINEPHARMACOKINETICS BY SEDDS

60

If you look back at Figure 2, it will remind you ofthe possible ways to improve oral bioavailability. Oneof the directions we are moving towards is nanotech-nology. For example, particle size reduction, emul-sion, solid dispersion, those are the new technologiesthat are now being exploited. Very small, controlledand uniform particles represent the kind of means bywhich solubility and absorption can be improved. It'slikely that in the future a lot of our researchers will be

working in this area. Thank you very much, this con-cludes my presentation.

Chair: Dr. Rinta Ibuki, Fujisawa Pharmaceu-tical Co., Ltd, Japan: Thank you very much, Dr.Okamoto. You covered a number of approaches. Iwould like to invite questions or comments from thefloor. Please state your name and affiliation.

Dr. Akira Yamamoto, Professor, Kyoto Phar-maceutical University, Kyoto, Japan: I am fromKyoto Pharmaceutical University. Thank you verymuch for your talk. We are looking at the effects ofvarious additives, and what you said about P-glyco-protein was very interesting. You also mentioned theco-solvent effect on amprenavir HIV protease inhibitor,which I found very interesting.

With regard to the vitamin E derivative, TPGS, if youincrease its concentration the efflux of amprenavir de-creases and so P-glycoprotein must be involved. I un-derstood that. But in Figure 13,you said that CYP3A4activity is also inhibited by excipients. In other words,amprenavir HIV protease inhibitor probably is a sub-strate of CYP3A4, so the effect is not present in thiscase. But this is only according to the literature, andaccording to the literature on P-glycoprotein, the effectof the additive on P-glycoprotein is the only focus;they didn't study the effect of CYP3A4. However, if itis serving as a substrate, CYP3A4 would be inhibitedand permeability would be improved. In the in-vivo sit-uation, that is a possibility.

Also, with your indomethacin example, you used aPEG in the formulation. In the case of indomethacin it

Table 3. Composition of Fractionned Colloidal Particles Formed from SD5

Fig. 4. Particle Size Distribution Curves of Various Solid Dispersions in Water after Stirring for 30 min.

YM022

Centrifugal Total weight Composition ratio(a)

condition (rpm) Weight of YM022 YM022:TC-5E:HCO-60

5000 1.9 1.0: 0.5 :0.3

10000 1.5 1.0: 0.2 :0.3

50000 1.6 1.0: 0.1 :0.3

(a) determined from IR spectra (I1077/I1097).

SD5

YM022: aqueoussolubility <1mg/mL 10

TC-5E: hydroxypropyl-methylcellulose 2910 35

HCO-60: polyoxyethylenehydrogeneated caster oil 60 5

K. Yano et al. Chem. Pharm. Bull.,45, 1339, 1997.

100

50

00.02 0.1 1 10

Particle Diameter (µm)

Cum

ulat

ive

%

100 1000

CO

NHCONH

CH3

CH3

ON

N

Figure 22.

DESIGN OF A 3-COMPONENT SOLID DISPERSION SYSTEM

Time (h)

Pla

sma

Con

cent

ratio

n of

YM

022

(ng/

ml)

YM022 TC-5E HCO-60

SD5 10 35 5 CH2Cl2/MeOH, spray dry

SD5W 10 35 5 Aqueous suspension of SD5

SD5 10 35 0 CH2Cl2/MeOH, spray dry

AP 10 0 0 CH2Cl2, spray dry

APM 10 35 5 Mixture of AP, TC-5E & HCO-60

00 2

5

10

15

20

25

4 6 8 10

Fig. 1. Plasma concentration of YM022 in Beagle Dogs after Oral Administration of Various Formulations at a Dose of YM022 10 mg/Body. : SD5, : SD 5W, : SD 4.5A, : AP, : APM (n=4, mean ± S.E.).

K. Yano et al. Chem.Pharm. Bull., 45, 1339, 1997.

Figure 23.

IMPROVED ABSORPTION OF A POORLY SOLUBLE DRUGBY SOLID DISPERSION SYSTEMS

61

does not become a substrate for p-glycoprotein andit will not become a substrate for CYP3A4 either. Thatis the only improvement of this ability.

Dr. Hirokazu Okamoto, Associate Professor,Meijo University, Nagoya, Japan: Using PEGmeans it is a dissolved situation. It is diluted in waterand, of course, some precipitation will occur. But withthe addition of water the precipitate probably is a finerparticle compared to the actual drug, the originaldrug. So even if there is precipitation the particle sizeis reduced, and the absorbability is improved.

Dr. Shinzi Yamashita, Professor, SetsunanUniversity, Hirakata, Japan: You presented vari-ous types of methodologies. I listened to this and Iam now wondering, which is the best? Surely it hasto do with the drug itself? This morning Dr. Lipinskispoke about poor solubility. But when people speakabout poor solubility in this general way they could bereferring to the molecule itself, or to its lipophilicity, orto the fact that it's poorly soluble. It could be thatcrystallization leads to poor solubility. I think therecould be a number of quite different reasons.

When we talk about poor solubility itself, onemethodology may be appropriate for highly lipophilicproducts, but it may not be applicable to the crystaltype, where another methodology may be better. Andin the crystalline system, yet another methodologymay be better. We speak of a poorly soluble sub-stance as a drug, but it may be indicating two differ-ent things and if you could give me some kind of sug-gestion in that regard…

Dr. Hirokazu Okamoto, Associate Professor,Meijo University, Nagoya, Japan: This is my per-sonal view. Micronization is the first option for allcompounds in drug development. You don't addanything, and if it increases the absorption level,that's desirable. Apart from that, at what phase in de-velopment would we look into the matter of poor sol-ubility? I think it has to do with the stages of develop-ment. At the initial stage we would be looking at thesalt selection and a prodrug, but in the latter stages

of development, some kind of additive may be re-quired. In that case you would have to look into itfrom the formulation point of view, the dosage pointof view.

Chair: Dr. Rinta Ibuki, Fujisawa Pharmaceu-tical Co., Ltd, Japan: Now, the final question.

Dr. Yuichi Sugiyama, Professor, University ofTokyo, Tokyo, Japan: In the very first part of yourpresentation you talked about prodrugs. Together withthe * company we are developing a prodrug, and thechallenge always is whether cleavage should occur inthe GI lumen or at the membrane. There is no signifi-cance or value in cleavage in the lumen. But accord-ing to your presentation today you can have cleavageat the brush border membrane and I think it is good tocompare cleavage there with cleavage at the lumen.One point I want to ask you is, when you use an esterwith the brush border membrane, where is the activesite? I think the surface is on the lumenal side. Soeven if you use BBM, it's not inside the membraneand if it's water-soluble to start with, then the ab-sorbability that enables it to pass through will not beimproved. It's not limited to esters, but the brush bor-der membrane topology of that cleavage enzyme isimportant. Do you have any data on that?

Dr. Hirokazu Okamoto, Associate Professor,Meijo University, Nagoya, Japan: Well, which di-rection is it facing? I don't have much data but what Ihave presented today is that the enzyme is bound tothe BBM, the prodrug dissolves and comes close tothe BBM, and then it is changed to the parent drugand goes into the membrane. But when it goes intothe membrane, depending on where the active site is,it has a totally different meaning, and that data is veryimportant in this type of meeting. It is a focused strat-egy, but at the micro level the mechanism has notbeen announced or published yet.

Chair: Dr. Rinta Ibuki, Fujisawa Pharmaceu-tical Co., Ltd, Japan: Thank you, I would now liketo end this session.

63

Importance of dose numberand absorption test

in formulation optimization:an industrial case

Dr. Hiroshi KIKUCHI

65

Chair: Dr. Rinta Ibuki, Fujisawa Pharmaceu-tical Co., Ltd, Japan: The next speaker is Dr. HiroshiKikuchi of Daiichi Pharmaceutical Co. Ltd. As you mayknow, Dr. Kikuchi is the company's principal investi-gator, and just as professionally active outside the com-pany as within it. I believe that his talk will be bothstimulating and interesting.

Dr. Hiroshi Kikuchi, Principal Investigator, Dai-ichi Pharmaceutical Co. Ltd, Japan: Thank youvery much for your introduction. I would like to thankProfessor Yamashita, the organizer, and also Dr. Amidonfor inviting me as a speaker, and I would like to thankCapsugel for this opportunity. I feel quite nervousbecause liposomes or gene delivery are the topics I amused to speaking about, and on this auspicious occa-sion I am speaking about the importance of dose numberand the absorption test for the first time.

Ten years ago I stayed at Dr. Amidon's laboratory inthe University of Michigan for one year to study oralabsorption, and since coming back to Japan I havebeen involved in the oral formulation project in additionto the DDS project (liposomes, gene delivery, etc.).Though the research results of our oral formulation pro-ject have not yet been in open domain, today I can dis-close one of the results of our project to a limited degree.I have also been involved in Dr. Yamashita's researchfor the past decade, but although it has produced agreat deal of information, this cannot be published orput into open domain. After my today’s presentation Dr.Yamashita will be able to refer to the results of clinical trialwhich will be disclosed here. Dr. Yamashita’s researchas to clinical prediction is linked to these results of clin-

ical trial. I will only be talking about some preliminarydata, for the benefit of our colleagues in the industry. Iam sorry that I cannot disclose the names of the newchemical entities or give their molecular weight and soforth.

Let me start. Our company operates in the world-wide market, so development is usually started over-seas. Five years ago, however, we started to handlethe drug discovery process in Japan. We determined thedose range for a product I shall call D-compound at100 mg, given as two 50 mg tablets. The CV (coeffi-cient of variance) value of the area under the curve (AUC)was very good, so the dose was raised to 200 mg andwe still had no problem with bioavailability. However, wefound out that the dosing regimen was different overseas.

Overseas, the dosing regimen was 500 mg (as two250 mg tablets), a very large dose compared to theJapanese therapeutic dose, although the formulationwas pretty much the same. But when we started a pilotstudy in the UK population, we began to have prob-lems; we began to see lower absorption and a higher CV(Figure 1). Given that the site was in a different country,we thought the analysis methodology might have beendifferent. We measured the Japanese data, while theoverseas clinical research organization measured theUK data, so there might have been a difference in themethodology. But we found that we had a major problemat 400 mg: the mean bioavailability was 71 percent,and the CV was 49 percent. In a parallel developmentwe had an overseas dose of 250 mg, and when thiswas given, as with the 500 mg dose, absorption wasreduced to 67 percent and the CV also went 38 percent.

Importance of dose number andabsorption test in formulation optimization:an industrial case

Dr. Hiroshi Kikuchi

Principal Investigator, Daiichi Pharmaceutical Co. Ltd, Japan

66

As you may know, the US Food and Drug Adminis-tration (FDA) would like to maintain the CV below 30 per-cent, although it is not a written requirement – I evenchecked yesterday that these are the requirements in theUK or at the FDA, but I could not find out the writtenrequirement. Anyway, we were told by the overseasdepartment that we had to reformulate the product foroverseas markets. It might only be an internal rule at theFDA that the CV has to be below 30 percent, but wemade it our target, along with a target of 80 percentabsolute bioavailability (BA). Of course, 90 percent would

have been better but since 80 percent was the absorp-tion achieved in Japan we targeted the same BA for theoverseas market. To do that, a new formulation wasneeded and as the clinical trials were already in progress,the UK site asked to have the study drug within oneyear.

Why did we have these variations in the UK? We didan analysis, but ultimately we do not know. We found anumber of phenomena, but none of them was defini-tive.

There was a gender difference, for one thing. Figure 2shows the absorption (Cmax and AUC) after adminis-tration of the oral solution or the tablet to the sameperson (male and female). In males, the oral solutionshowed almost no variation in absorption and the tabletalso showed a small variation. But in the females,although the solution did not show any variation, therewas quite wide variation with tablet administration. Wehypothesized that there may be some female-specificreasons for this variation; there are various stomachshapes, for instance, such as a J or an L shape. Thatwas our hypothesis, but we were unable to reach aconclusion.

We determined pH solubility profile as shown inFigure 3. This 100 mg administration was the dose thatwas used in the clinical trial in Japan. According to thecalculation of the dose number, or the dose divided bythe solubility, divided by 250 ml, the dose number was0.4 at pH5 and 4.4 at pH7 (below 5 in the all pH region).So it was confirmed that there was no problem in sol-ubility of this drug from the view point of dose number.

However, when the dose number was calculated forthe 500 mg dose, the value was 22 at neutral pH andso we were faced with a major problem in terms of the

Absorption of D-Compound in Clinical Trial

Product used Dose Site BAmean (%) CV (%) N

Tablets 50 mg* 100 mg Japan 95 20 6

200 mg Japan 89 22 6

100 mg UK 88 29 12

200 mg UK 78 33 12

400 mg UK 71 49 12

Tablets 250 mg* 500 mg UK 67 38 24

Solution, po 400 mg UK 86 21 12

IV 400 mg UK 100 24 24

* Similar components were used in both formulations.

Figure 1.

00

Oral Tablet

Cmax (µg/mL)

Male (n=6)

50mg Tablet x 8T vs 400mg Oral Solution (PRT 018)

Ora

l Sol

utio

n

105

10

5

00

Oral Tablet

AUC 0-∞ (µg•hr/mL)

Ora

l Sol

utio

n

5025

50

25

00

Oral Tablet

Cmax (µg/mL)

Female (n=6)

Ora

l Sol

utio

n

105

10

5

00

Oral Tablet

AUC 0-∞ (µg•hr/mL)

Ora

l Sol

utio

n

5025

50

25

Figure 2.

RETROSPECTIVE COMPARISON OF ABSORPTION (MALE & FEMALE)

67

solubility. Yet all the tablets we were testing were basedon a similar formulation. Though we knew ‘dose number’at that time, but honestly speaking we did not regard itas important. But looking at the result of the clinical trialwe recognized the real importance of dose number –because we failed.

Figure 3 also shows that the permeability of this com-pound is good from the loop method and Caco-2 per-meability test. As a result, we classified this compoundas a Class II drug in BCS. As a Class II drug has lowsolubility, so when we formulate such compounds wehave to look at the dissolution rate. However, the 500 mg

10.000

104 6

pH

pH Solubility Profile Loop Method

Dose: 0.94 mg/kg (rat) to each loop

Caco-2 Permeability

very high Class 2 in BCS

Sol

ubili

ty (m

cg/m

L)

Rem

ainn

ing

Rad

iore

activ

ity (%

of D

ose)

100

1 000

8 10 12

100

0.0

20

40

60

80

Stomac

h

Duoden

um

Jejun

umColo

nIle

um

Dose No.: (100mg administration) (500mg administration) pH 5: ca. 0.4, pH 7: ca. 4.4 pH 5: ca. 2, pH 7: ca. 22

Figure 3.

New Oral Solid Dosage Formulation

Ingredient Component Form. A Form. B Form. Cmg/tablet mg/tablet mg/capsule

Active ingredient D-Compound 266.5 266.5 266.5(as Anhydrate) (250) (250) (250)

Diluent AAAAAA aa --- ---Diluent BBBBBB bb --- ---Disintegrant CCCCCC c1 c2 ---Binder DDDDDD d1 d2 ---pH adjustor Organic acid x1 x2 x3pH adjustor Sodium bicarbonate --- y2 ---Wetting agent Polysorbate 80 --- --- z3Glidant EEEEEE --- --- eeLubricant FFFFFF f1 f2 f3Lubricant GGGGGG g1 g2 ---Coating agent HHHHHH h1 h2 ---Solvent Water, purified * q.s. q.s. q.s.Solvent Dehydrated ethanol * q.s. q.s. ---

Total 536 478 320pale yellow film-coated whiteto yellow tablet opaquesize: 11.1 mm size: 10.1 mm size: no. 0

* removed during the manufacturing process

- Amount of organic acid for improvement of solubility.- Effervescent ingredient (organic acid and sodium bicarbonate); totally 25% in Form. B core tablet.- Polysorbate 80 is added as a wetting agent in Form. C.

Figure 4.

68

dose was prepared using the original formulation, andthat caused the problem in the clinical setting.

To find out the reason, we began with the hypothesisthat the high variance of AUC/Cmax might be causedby lack of solubility in the digestive tract, and/or by rapidmovement through the digestive tract, so preventingdisintegration or dissolution from taking place. In ourreformulation strategy we aimed to increase the disso-lution rate and, if possible, the absorption level. Theconcept was to produce a drug where rapid disinte-gration and dissolution was independent of pH and oftransit time in the digestive tract.

We looked at nine points in our approach to refor-mulation. But we had to place this study drug in clin-ical trial within a year, so time was limited. We thoughtthat changing the disintegrants and perhaps addingorganic acids might be a good way of improving disin-tegration/dissolution. The second point involved usingeffervescents to produce quick disintegration within thestomach, and the third was to use a hard capsule for-mulation; compared to a tablet, it would have been pos-sible to achieve faster dissolution and disintegration.

The fourth way was to add surfactants to improvewetting. In the worst-case scenario, maybe we wouldturn to the fifth way and make an oral aqueous formu-lation. Well, it is not possible in Japan. Western author-ities may accept this dosage form, but for us the optionswere soft capsule formulations containing liquid/dis-persion, or to make amorphous forms with polymers,or – as the eighth choice – to grind the particle size tomicrometers.

The ninth approach was to make a change in thechemical form of the drug. However, if the chemicalform was changed to another one, we had to go backto the pre-clinical stage. So the ninth approach wasjudged impractical, and we relied on the first eight.

Figure 4 is a sort of conclusion to all this. Organicacids were added to the formulation of two tablet forms(Form A and Form B), and sodium bicarbonate was alsoadded to Form B for effervescent effect. Organic acidand wetting agent (surfactant) were added to the cap-sules (Form C). We screened them with several tens offormulations and these were the promising candidates.Let me skip the details.

Dissolution was more or less adequate with all the var-ious types of solutions. One example is shown inFigure 5, which gives the result of a dissolution studyrun at pH 1, using the old 250 mg tablet, an organicacid tablet (Form A), an effervescent tablet (Form B), ora capsule (Form C). All these formulations had noproblem at pH 1.

But at pH 5 you can see some difference in the oldformulation, as shown at the bottom of Figure 6. Only30 percent was dissolved after an hour. Both the mod-ified formulations, the organic acid (Form A) and theeffervescent tablet (Form B), showed a high dissolutionrate, while the capsule was just medium. With a neutralpH (pH6.8) the old tablet had the lowest dissolutioncurve. As opposed to this, the improved formulationtablets had faster dissolution, with the capsule comingsomewhat below the tablets.

As already mentioned, using an oral aqueous for-mulation, or a soft capsule formulation containingliquid/dispersion, or making amorphous forms with poly-mers, or micro-grinding were also considered as can-didate approaches. But the conclusion was that theywere not desirable.

The D-compound is quite acid in an oral aqueoussolution, so dissolution itself is not a problem, but ittastes very bitter. I tried it myself, but as soon as thesolution hit the tongue it felt as if my tongue was burning,so it was not possible to use it as a product.

100

00 40

Time (min)

% D

isso

lved

80

60

40

20

302010 50 60

0.1N-HCl (pH 1), paddle method,50rpm, 900mL, 2 units/vessel

Effervescent tab. x2 (Form. B)Capsule x2 (Form. C)

Organic acid tab. x2 (Form. A)

Old 250mg tab. x2

Figure 5.

DISSOLUTION CURVES (PH 1)

100

00 40

Time (min)

% D

isso

lved

80

60

40

20

302010 50 60

1 x 10-5 N-HCl (pH 5), paddle method, 50rpm, 900mL, 2units/vessel

Effervescent tab. x2 (Form. B)

Capsule x2 (Form. C)Organic acid tab. x2 (Form. A)Old 250mg tab. x2

Figure 6.


69

We also have to ensure a certain dose, and we foundwith the soft capsule that the concentration was toohigh to be in a stable state as a solution/dispersion form.We would have had to use a large number of capsules,or a large-sized capsule. We would also have had tooutsource capsule manufacturing, and the cost wouldhave been higher. Today, we can fill a solution/disper-sion form in the hard capsules on our filling machines sowe would be able to use a hard capsule. But five yearsago, Capsugel's products (Capsule Liquid Filling &

Sealing Machine; CFS 1000) were not available, andbecause of higher cost and other issues we gave upon the soft capsule idea.

The amorphous form with organic polymers did notimprove the dissolution. As for the amorphous form, wehad already pulverized the particles down to 10-20 micrometers. We also tried to grind particles intoeven more minute sizes, but there was no improvement.

To summarize our reformulation results. Addingorganic acids, adding effervescence, and using organicacids and surfactant in hard capsules all improved dis-solution and solubility. These three approaches wereselected as candidates. With the others – an oral solu-tion, soft capsule forms, making amorphous forms withorganic polymers, and grinding the particles to min-crometer order – we felt that there was no chance ofbringing them into the clinical trial environment, so wegave up on them.

We conducted a comparison study on dogs, usingthe old 250 mg tablets with and without pentagastrin,and formulation A, the organic acid reformulation, withoutpentagastrin (Figure 7). As you know, the pH of thedog’s stomach is higher than that of the human's. So ifwe could tilt the gastric environment to an acid envi-ronment with pentagastrin in dogs, we could make anacidic pH environment similar to human’s stomach, andthat would lead to higher oral absorption of this D-com-pound. Even with the higher pH in dogs, the reformulated

0.00 1

Time (hr)

Con

cent

ratio

n (m

cg/m

L)

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

2 3 4 5 6 7

Fig. Concentration of D-Compound in plasma after single oraladministration to male dogs (dose: 250mg/head).

* Data are expressed as the mean ±SD of six animals.

Old 250mg x1 (with pentagastrin)Form. A x1 (without pentagastrin)

Old 250mg x1 (without pentagastrin)

Figure 7.


Pharmacokinetic parameters of D-Compound in plasma aftersingle oral administration to male dogs (dose: 250mg/head)

Sample AUC0-6h Cmax Tmax(µg x hr/mL) (µg/mL) (hr)

Old Tab. 250mg x1 (without pentagastrin) 15.62 ± 11.31 (CV 72.4%) 3.72 ± 2.58 (CV 69.3%) 1.63 ± 1.07

Old Tab. 250mg x1 (with pentagastrin) 30.76 ± 3.70 (CV 12.0%) 7.96 ± 1.43 (CV 18.0%) 1.04 ± 0.51

Form. A 250mg x1 (without pentagastrin) 29.87 ± 3.90 (CV 13.1%) 7.31 ± 1.86 (CV 25.5%) 1.58 ± 0.20


Pharmacokinetic parameters of D-Compound in plasma aftersingle oral administration to male dogs (dose: 250mg/head)

Sample AUC0-6h Cmax Tmax(µg x hr/mL) (µg/mL) (hr)

Form. A 250mg x1 (without pentagastrin) 29.87 ± 3.90 (CV 13.1%) 7.31 ± 1.86 (CV 25.5%) 1.58 ± 0.20

Form. B 250mg x1 (without pentagastrin) 28.85 ± 1.26 (CV 4.4%) 7.25 ± 0.61 (CV 8.4%) 3.17 ± 0.75

Form. C 250mg x1 (without pentagastrin) 20.77 ± 6.34 (CV 30.5%) 5.27 ± 1.61 (CV 30.6%) 1.42 ± 0.58


Figure 9.

Figure 8.

70

drug without pentagastrin showed very good oral absorp-tion.

Figure 8 shows the pharmacokinetic (PK) parame-ters. This is the same data as in the previous Figure,the old tablets with and without pentagastrin, and for-mulation A without pentagastrin. The comparison showsthat the AUC and the Cmax for formulation A (withoutpentagastrin) was almost twice as compared with thoseof the old tablets (without pentagastrin), and the CVvalues were decreased very much.

We then tested the effervescent tablets and the cap-sules without pentagastrin (Figure 9). For formulation Aand the effervescent tablets (formulation B) the AUCswere almost 30 µg·hr/mL in dogs, and the capsuleswere 21µg·hr/mL. The CV for the capsule (formula-tion C) was very high in the dog – 30 percent – whereasthe CVs for the tablets were lower. Normally, I think thecapsule candidate would have been dropped at thispoint, before moving on to clinical studies. But becauseof manufacturing difficulties with formulation B, the effer-vescent tablet, it was dropped instead and we decidedto continue carrying out clinical studies with formula-tions A and C.

Please keep this in mind. Under normal circum-stances, when you have these outcomes from a dogstudy you would choose only formulation A, the organicacid reformulation, with high absorption and a low CV,because the capsule (formulation C) had a higher CV andlower absorption. Pharmaceutical scientists, you maythink that we should be making predictions based onanimal studies. But if you select only formulation A, andconduct clinical studies only on formulation A, you willnever know whether formulation A was the best formu-lation.

That is why we consider the next part of our projectas very important. Usually it is not possible, but we werefortunately able to study both formulation A and formu-lation C in human trials, and we obtained very inter-esting results.

Figure 10 shows the results from actual clinicalstudies, using tablets (formulation A) and capsules (for-mulation C). With the old tablets, absorption was 67 per-cent and the CV was 38. But with the formulation Atablets there was almost 100 percent absorption andthe CV was 19 percent, much lower than the target of30 percent. As for the capsule, with dogs we had apoor figure but in humans we had even better resultsthan with the tablets. So in the end, in the clinical studywe decided to use capsules, as in formulation C. I havenot been able to spend enough time on this Figure in theinterests of time, but I would like to return to this issuein the panel discussion.

So what are my conclusions? In the case of the100 mg tablet dosage, we had a very good outcome inJapan. But when we took it to the UK and gave it as a500 mg dose, absorption was poor and there was largervariance. With a little self-reflection, we now realize thatwe were not wise in using a similar formulation for the100 mg, despite the dose number being high. How-ever, we paid a lot of attention to dissolution and solu-bility, and we used organic acids and surfactants and thisled to better clinical results.

Because of this experience we feel that we have topay attention to the dose number and the Biopharma-ceutical Classification System (BCS) in formulationstudies, especially for Class II drugs. When I studied atthe University of Michigan 10 years ago, ProfessorAmidon was speaking about BCS in his lectures. At thattime, I was wondering why it was important to classifydrugs into four categories and I did not pay very muchattention. But as I gained more experience I came torealize the importance of the BCS classification, espe-cially with Class II drugs.

I was with a development team for a few years andthat is when I did the work I have been talking abouttoday. We had to complete the project within a year andwe actually completed it in nine months. There were26 people working with me and half of them were moreor less dedicated to this project. We worked into the

Absorption of D-Compound in Human PK Study

Product used Dose Site BAmean (%) CV (%) n

Form. A (Tablet) 250 mg 500 mg UK 104 19 9

Form. C (Capsule) 250 mg 500 mg UK 108 9 9

Solution, po 500 mg UK 93 8 9

IV 500 mg UK 100 10 9

cf: Old Tablets 250 mg 500 mg UK 67 38 24

Figure 10.

71

early hours, to complete the development and start theclinical trials. I would also like to thank the manufac-turers.

Normally, we conduct screening with animal studiesand we can only use the best candidate in human trials.But we do not know whether that best candidate outof an animal study is the best candidate for humans.We will never know the truth, because we can only useone formulation and we have to make a prediction ofhow effective the drug is in humans. We therefore haveto accumulate data to make better predictions. We werelucky enough to be able to use several candidates inhumans and, interestingly, we were able to show that acandidate that did not do so well in animals did betterin humans.

Chair: Dr. Rinta Ibuki, Fujisawa Pharmaceu-tical Co., Ltd, Japan: Thank you very much on behalfof Capsugel. It has really been a wonderful talk. I wouldlike to open the floor for discussion.

Question from the audience: Thank you verymuch for the interesting talk. I'm really envious thatyou could run two formulations in a clinical study. Onbehalf of all my industry colleagues I would ask youabout the discrepancy between dogs and humans.How do you explain this difference?

Dr. Hiroshi Kikuchi, Principal Investigator, Dai-ichi Pharmaceutical Co. Ltd, Japan: Well, after thisproject I went back to the discovery team and what Itold them was, we have this clinical data and we havesome data in dogs, but we need to test it in other animalmodels like monkeys and other species. The reasonwhy we used dogs is that we cannot administer thestudy drug formulation to monkeys because of the sizeof tablets and capsules. But with the clinical outcome weattained I thought it was necessary for us to go backagain to animal tests. By doing that I believed our pre-dictability accuracy would be improved.

Dr. Shinji Yamashita also helped us to look into dis-solution in the GI tract, and we used simulation models.We really wanted to bring about better predictability ofthe best candidates by using those data for the benefitof the whole industry and the therapeutic setting, sothat this might be used as the starting point to link it tohuman data.

Chair: Dr. Rinta Ibuki, Fujisawa Pharmaceu-tical Co., Ltd, Japan: I have a question for you. Youare saying that the FDA has an internal rule that the CVshould remain within 30 percent. Could you commenton that, please?

Dr. Hiroshi Kikuchi, Principal Investigator, Dai-ichi Pharmaceutical Co. Ltd, Japan: I am not awareof any internal insight or guidance regarding the coeffi-cient variance on bioavailablity or bioequivalence. Obvi-ously, higher bioavailability leads to a low coefficientvariance, and usually this is more likely to be successfulin the clinic. But it does not necessarily mean that ahigh CV cannot be approved. In our case the overseasdepartment demanded that the coefficient variance hadto be less than 30 percent because it was not easy forpatients to use this drug.

Dr. Lawrence X. Yu, Food and Drug Adminis-tration, USA: I think the best example is Phosmax fromMerck. The bioavailability is 1 percent, the CV I don'tknow, it's a couple of hundred percent or somethinglike that. But the drug's still approved and still you tryto protect human health. So I think we should look at theoverall picture based on safety and efficacy, not somuch at a specific rigid rule based on specific numberssuch as the coefficient valence of bioavailability.

Chair: Dr. Rinta Ibuki, Fujisawa Pharmaceu-tical Co., Ltd, Japan: Thank you very much. Let memove on to the third speaker. Dr. Akira Kusai of SankyoCompany Ltd will talk on How to handle insoluble APIs:our experience at Sankyo.

73

How to handle practicallyinsoluble APIs: our experience

at Sankyo

Dr. Akira KUSAI

75

Chair: Dr. Rinta Ibuki, Fujisawa Pharmaceu-tical Co., Ltd, Japan: After Dr. Kusai's presentationthere will be a discussion with all three speakers. Dr.Kusai, please, are you ready?

Dr. Akira Kusai: Thank you very much for invitingme today. As you can see, the word 'practically' isadded to the title announced before. In principle Iwould like to deal with the issue of practically inso-luble pharmaceutical active ingredients: how we atSankyo have handled them in the past and today.

I would also like to mention several topics to acti-vate our discussion later.

Dr. Okamoto has already discussed a number ofways to improve dissolution behavior of APIs; salt for-mation, solvation and surface active agents, whichhave been used for several decades. As he mentio-ned, cyclodextrin derivatives of enhanced aqueoussolubility have been available recently. Then this ap-proach is currently increased. The dispersion in oilybase is another method, and I believe Dr. Benameurwill discuss this point at his presentation lator. Moreand more often these days, the dispersion in oilybase is filled into capsules. However depending onthe regulations in each country, some types of oilybase can be permitted to be used while others can-not, or they are restricted in their use. So we have tobe very careful in selecting the oily base.

Micronizing APIs to nano-size is good approach,which has been getting much attention, and I will dis-cuss this in detail later. The increase in surface areamay increase the dissolution rate, but the dispersed

state of the nanoparticles has to be maintained afterformulation process and during storage. It may coa-gulate during storage. To keep the particles from coa-gulation, various additives are incorporated. However,this may result in high viscos dispersion. The recentmeeting of the Society for powder technology, Japanfocused on this area. They say the 100-nanometer le-vel was achieved, but this is actually not nano levelbut submicron level. Co-grinding or mix grinding isanother approach, where the carrier and the drug aremixed and ground together. As long as it is conduc-ted at labo-scale, everything goes well. But taking thepowder out, powder adhesion to the equipment, pro-ductivity and how to handle the generated heat be-come issues, when it is conducted at the largerscale, pilot sale and production scale. Solid disper-sion is another existing technology, and today I amgoing to focus on it as an example.

With respect to nanoparticles, Dr. Ibuki, who is ser-ving as Chairman today, presented why nanonizationis so effective at the 19th Symposium on ParticulatePreparations and Designs at fall in 2002. The reasonshe stated are: improvement of the dissolution rate, re-duction in crystallinity, supersatulation and the mobilityinhibition in the GI tract, i.e. longer retention time. Fur-ther the direct contact with the GI tract is also inclu-ded. The factors are related to each other in a com-plex manner and interact with each other to cause theeffect. Figure 1 is kindly offered from Dr. Naito, Profes-sor at the Joining and Welding Research Institute,Osaka University. It is very interesting, and illustrateshow many molecules are on the side, how many onthe surface, the total number of molecules and the ra-

How to handle practically insoluble APIs:our experience at Sankyo

Dr. Akira KUSAI, Director, Pharmaceutical Development Laboratories

Sankyo Company Ltd

76

tio of the number of molecules on the surface againstthem. If particle size is reduced by one-tenth, then na-turally the total number of molecules per particle willbe reduced by three orders of magnitude, and theones that are on the surface will be reduced by twoorders of magnitude. Then the ratio of surface againsttotal is increased by 10 times. With particles of200 nanometers or less, this ratio becomes important,for molecules on the surface are obviously magnifica-ted. If you micronize or create particles of 200 nanome-ters or less, they are expected to exhibit very interes-ting properties. As Dr Okamoto mentioned, thenanocrystals still stay at several hundred nanometers.New types of technology and equipment are then re-quired to reduce them by one digit with well controlledmanner. A lot of subjects may be overcome. But itcould lead us to the new world of quite interesting phe-nomena, as we can easily imagine from these data.

Now I would like to focus on the solid dispersionand present on troglitazone as a case study. Ourcompany developed it, and marketed for a couple ofyears. But due to the unexpected adverse event , it iswithdrawn from the market. Some of the data will bepresented to facilitate discussion later.

Figures 2 and 3 shows the physico-chemical pro-perties of troglitazone. It is basically within the rangethat Dr. Lipinski described, but aqueous solubility isvery low, at microgram/mL level with difficulty of nu-merical expression. Under the physiological condi-tions it does not practically dissolve, although its solu-bility increases a little under the alkaline condition. Wecannot achieve the required level of dissolution under

Physico-chemical Properties (2)

Appearance White to yellowishwhite crystalline powder

Melting point around 175°C

Dissociation around 6.1 (pKa1),constant around 12.0 (pKa2)

Partition coefficient 2.7 (n-octanol /(log PC) pH7 phosphate buffer)

Solubility pH 1.2 (JP1) < 0.01µg/mLpH 6.8 (JP2) < 0.01µg/mLpH 7.0 0.4 µg/mL pH 8.0 0.9 µg/mL pH 10.1 29.0 µg/mL

ethanol 5.3 mg/mLethyl ether 8.3 mg/mLacetonitrile 9.5 mg/mLacetone 105 mg/mL

Figure 3.

Percent of Molecules on Surface vs Particle Size

No. of Molecules No. of Molecules Total No. Surface/ Particleon a Side on Surface of Molecules Total (%) Size

2 8 8 100

4 56 64 87.5

10 488 1,000 48.8 2 nm

100 58,800 1 x 106 5.9 20 nm

1,000 6 x 106 1 x 109 0.6 200 nm

10,000 6 x 108 1 x 1012 0.06 2 µm

100,000 6 x 1010 1 x 1015 0.006 20 µm

1,000,000 6 x 1012 1 x 1018 0.0006 200 µm

Prof. M. Naito, Joining & Welding Research Institute, Osaka University

Figure 1.

Generic Name: Troglitazone ➞ Mixture of 4 isomers L: RR, SS H: RS, SR

Chemical Name: (±)-5-[4-(6-hydroxy-2,5,7,8-tetramethylchroman-2-ylmethoxy) benzyl]-2,4-thiazolidinedione

Molecular Formula: C24H27NO5SMolecular Weight: 441.55

* asymmetric carbon

MeMe O

HOMe

MeO O

NH

S

O

**

Figure 2.

PHYSICO-CHEMICAL PROPERTIES (1)

77

the physiological pH within the GI. We are able toachieve order of milligram/mL as solubility in organicsolvents.

So we compared a number of approaches howmuch extent they can improve the bioavailability. Fi-gure 4 gives those data obtained with beagle dogs;the oral suspension, the tablet made of jet-milled tro-glitazone at the order of several microns as particlesize, the tablet prepared by solid-adsorption method,and tablets made of solid-dispersion prepared by aco-grinding method, a melting method and a solventmethod. Melting was performed by heating. The sol-vent method was conducted by dissolving troglyta-zone with excipients in organic solvent and followedby evaporating the solvent to provide the solid mass.The solid dispersions prepared by the melting methodand the solvent method exhibit bioavailability equal tothat of the solution. If the co-grinding had provided

solid dispersion at 0 % crystallinity, it could haveshown the comparable level of availability. It is notedthat it is very difficult to 100% reduction of crystallinitywith the co-grinding method.

Figure 5 summarizes the comparison of twoaqueous soluble polymers, PVP and HPMC, on pre-paring solid-dispersions (SD). We finally selected PVP.Here I would like to explain why we did so. We com-pared the dissolution profiles of SDs prepared withPVP and HPMC. One factor was what kind of organicsolvents can be used; ethanol and acetone can beused for PVP, but chlorine-type organic solventsshould be applied for HPMC. So PVP is preferable interms of safety for environment and workers at opera-tions. It is also taken into consideration how muchsolvent is required to obtain the solid dispersion, i.e.maximal dissolved amount of solid in solvent. Further,it must have practical physical stability during storage,the amorphous state and the dissolution behavior. Asyou can see, PVP's dissolution rate is sperior. AsHMPC is less wettabe, and then SD with HPMC is in-ferior in wetting property and dissolution rate to thoseof PVP. Taking all these factors into consideration, weselected PVP.

Figure 6 displays the effect of troglitazone ratioagainst PVP on the dissolution behavior and the crys-tallinity pattern measured with a powder X-ray diffrac-tion (PXRD). The drug substance itself dissolvesabout 30 percent in 20 minutes, but the dissolutionrate is enhanced with PVP. The right-hand side of theFigure shows the PXRD crystallinity patterns. Thereexisits still some crystallinity at the ratio of 4 to 1. Atthe ratio of 2 to 1, no crystallinity is observed at all.

Figure 7 demonsrates the effect of initial crystalli-nity on physical stability. We stored samples of diffe-rent level of initial crystallinity at accelerated condition.

Suspension

Tablet (Jet-milled)

Tablet (Solid adsorption)

Tablet (Solid Dispersion (Co-grinding))

Tablet (Solid Dispersion (Melting))

Tablet (Solid Dispersion (Solvent))

Solution

Absolute Bioavailability (%)(Beagle dog: 5mg/kg)

0 10 20 30 40 50 60

Figure 4.

BIOAVAILABILITY (1)

100

80

60

40

20

00 20 40 60 80 100 120

Time (min) ➞ PVP is selected Dissolution profiles of troglitazone S.D.

paddle speed: 200 rpm100mg / 900ml of pH9

Dis

solu

tion

(%)

Troglitazone:PVP(1:1)

Troglitazone:HPMC(1:1)

PVP HPMCOrganic Solvent Safety Amount

Solid Dispersion Physical Stability Dissolution Rate

x~x

Figure 5.

TROGLITAZONE SDAQUEOUS SOLUBLE POLYMERS: PVP VS HPMC

100

80

60

40

20

00 5 10 15 20

Time (min)

20 (degree)

Dissolution profiles of troglitazone S.D.

Figure. PXRD of troglitazone S.D

paddle speed :100 rpm10mg / 900ml of pH9

Dis

solu

tion

(%)

10 20 30

1:2

1:0

1:1

2:1

4:1

9:1

Troglitazone : PVP

Figure 6.

EFFECT OF AMOUNT OF PVP ON DISSOLUTIONBEHAVIOR & CRYSTALLINITY

78

Samples at 0% and 17% of initial crystallinity did notchange their initial crystallinity. That at 48% ended upat increased crystallinity after storage. Thus it is evi-dent that the initial crystallinity is very important.

Figure 8 shows the electron probe micro-analysisof troglitazone physical mixture (PM) and SD, whichindicates how the sulfur distributes in the mass. Youcan see the heterogeneous distribution of sulfur in theleft-hand photo (SD). However, the homogeneousdistribution is observed in the right-hand photo (PM).

When solid dispersion is manufactured with thesolvent method, there are a number of options onequipments to be used. The spray dryer is one candi-date, and the others are the vacuum drying system,the fluidized bed granulator and the vacuum granula-tor. To select the most favorable equipment, certainmeasures were estimated. First is the safety issue;what kind of safeguard against explosion do theyhave? Second is whether they require a carrier. Fur-ther measures are the residual solvent level, the den-sity and the flowability of the SD. Those are the mea-sures to determine the equipment. When the solventis sprayed onto the carrier, we found it is rather diffi-cult to control and reduce the residual solvent level.That's the reason why we chose the spray-dryer asan equipment.

Figure 9 displays the appearance of SD preparedwith the spray-drying method. They are fairly largespherical particles. Figure 10 compares the dissolu-tion amount of troglitazone itself and its SD at 20 minin the various dissolution media. It is clear that disso-lution property is markedly improved with SD.

Effect of Amount of Initial Crystallinity on Physical Stability

Initial 60°C 40°C/75%RH Crystallinity2W 4W 6W 1M 2M 4M 6M Change

~0% ~0%* ~0% ~0% ~0% ~0% ~0% ~0%

17% 19% 14% 21% 15% 19% 16% 15%

48% 67% 85% 89% 53% 75% 83% N.T.

* Apparent crystallinity (measured by PXRD) Test tablets were stored in glass bottle with desiccant.

Figure 7.

➞➞

➞

Mapping of S (troglitazone crystalline)

10 µm

Troglitazone - PVP P.M.100 : 60

Troglitazone - PVP S.D.100 : 60

Figure 8.

ELECTRON PROBE MICRO-ANALYSISTROGLITAZONE PHYSICAL MIXTURE & SD

100 µm

Troglitazone : PVP =100 : 60 + surfactant

Figure 9.

TROGLITAZONE SD MANUFACTURED WITH SPRAY DRYING METHOD

Troglitazone SDDissolution Behavior

Comparison between dissolved amount of Troglitazonefrom Troglitazone and Troglitazone S.D.

pH Dissolved amount*Solid Dispersion** Troglitazone

1.2(JP1) 3 µg/ml < 1 µg/ml

6.8(JP2) 28 µg/ml < 1 µg/ml

8.0 112 µg/ml 6 µg/ml

8.5 392 µg/ml 13 µg/ml

9.0 881 µg/ml 41 µg/ml

9.7 1678 µg/ml 151 µg/ml

* after shaking for 20min at 37°C** Troglitazone : PVP =100 : 60 + surfactant

Figure 10.

79

Figure 11 illustrates the effect of PVP ratio againsttroglitazone. I'd first like to compare the three sets offigures at the bottom in the left-hand box with 100 mgtroglitazone, where the total tablet weight is fixed at250 mg. On the right-hand side the dissolution per-centages at 20 minutes against various relative humi-dity as storage condition have been plotted, wherethe dotted line suggests the temporal target of disso-lution level after storage. You can see that physicalstability against moisture increased with the amount ofPVP. If you increase the amount of troglitazone to 110mg for the hypothetical worst case of its purity at 90%with keeping the amount of PVP at 50 mg, it be-comes rather sensitive to moisture. This demons-trates that the amount of PVP is very important; thephysical stability increases with PVP ratio.

Next, we have to optimize thr amount of the diluentand disintegrant (Figure 12). The experimental condi-

tions are the same as for the previous Figure. As youcan see, if the diluent weight is kept constant at 23mg, the physical stability increased with the amountof disintegrant. When the total tablet weight is fixed at250 mg, the stability increased with the amount of di-sintegrant. The higher the total weight, the better isthe physical stability. It is a matter of course, but thetablet size should be appropriate for human oral ad-ministration.

Figure 13 summarizes the formulation and manu-facturing process. Noscal 100, the tablet of 100 mgstrength of troglitazone is formulated with 60 mg ofPVP and other excipients such as disintegrant, diluentand film former to make 260 mg as total weight. Withrespect to the process, the specific volume of SDprepared with the spray-drying method is too large todirectly compress into tablets. Then the dry granula-tion process was introduced to provide denser mass,followed by tableting and film coating.

Figure 14 demonstrates the dissolution profiles ofthe tablet prepared with SD and control tablets. Webasically used the same excipients for the control ta-blets. However, in order to obtain the same disinte-gration time, the amount of disintegrant is increasedwith reduction of the amount of diluent. It is evidentthat the dissolution % is improved by five-fold with theSD method.

Finally, Figure 15 summarizes the bioavailabilitydata after oral administration to beagle dogs, which isimproved by about three-fold with the SD method.

Now I’d like to close my presentation. Thank youvery much for your attention.

100

95

90

80

80

7020 40 100

Relative humidity (%)

Dis

solu

tion

% a

t 20

min

75

60 80

Tablets are manufactured with Dry GranulationMethod. Dissolution % at 20min after Storageunder various relative humidity, 40C for 95hrs.

Troglitazone PVP Total(mg) (mg) (mg)

110 50 250100 50 250100 55 250100 60 250

PVP ➞ Physical Stability

Figure 11.

FORMULATION DESIGNTROGLITAZONE /PVP RATIO

Table. Formulation of Noscal 100 and 200

Noscal 100(mg/Tab) Noscal 200(mg/Tab)Troglitazone 100 200PVP 60 120Others 100 200(disintegrant, diluent, film former etc.) Total 260 520

Manufacturing process of Noscal 100 and 200 S.D. (solvent method)

Dry granulation

Tableting

Film-Coating

Figure 13.

FORMULATION & MANUFACTURING PROCESS WITH TROGLITAZONE SD

100

95

90

80

80

7020 40 100

Relative humidity (%)

Dis

solu

tion

% a

t 20

min

75

60 80

Tablets are manufactured with Dry GranulationMethod. Dissolution % at 20min after Storageunder various relative humidity, 40C for 95hrs.

Diluent Disintegrant Total(mg) (mg) (mg)

23 20 20023 40 22023 70 25041 22 22068 25 250

Weight Disintegrant ➞ Physical Stability

Figure 12.

FORMULATION DESIGN DILUENT & DISINTEGRANT

80

Chair: Dr. Rinta Ibuki, Fujisawa Pharmaceu-tical Co., Ltd, Japan: Thank you very much, Dr.Kusai, for your presentation. You spoke in particularabout solid dispersion, and thank you for the tho-rough information you gave us. Any questions orcomments on this presentation, any discussion pointsto raise, please? We have 10 minutes left, so wehave time for discussion with all three lecturers. So ifyou have questions for any of them...

Dr. Akira Kusai, Sankyo Co. Ltd, Japan: Dr.Yamashita suggested earlier that one to one appro-priate method may exist to each API candidate. Withrespect to the poorly soluble APIs, we hope they areexcluded at the screening steps. When we formula-tion scientists are involved to handle them, it is betterto try a number of possible approaches such as soliddispersion, co-grinding, and various types of oily ba-sed dispersion instead of anticipating the most appro-priate method, for we do not have the precise mea-sures to assume. We have to be careful on thechemical and physical stability during storage even ifthe formulation exhibits good properties immediatelyafter preparation. If that is the case, we have to giveup.

As Dr. Kikuchi mentioned, a simple formulation isbetter. That is ideal, but we have to choose the realis-tic approach from the various options. I think this ishow we are managing it these days.

Dr. Shinji Yamashita, Professor, School ofPharmacy, Setsunan University: One furthercomment that's relevant to this point: we apply soliddispersion and micronization method and we can ex-pect to improve dissolution. But if we ignore the long-term stability issue and administer that type of pro-duct, we may find there is no efficacy. There aremany such examples.

So, depending upon the compound, a formulationstrategy may work, or it may not. The question is,how do we select from the options? We did makemistakes in the past. You at Sankyo, for example,tried with troglitazone to get succeeded. I think theone you described was a very good exsample. Weactually want to know the case of failure. How manymistakes have you made along the way in terms offormulation decisions?

Dr. Akira Kusai, Sankyo Co. Ltd, Japan: Well,in several cases we tried hard to improve the dissolu-tion behavior, but found that it did not improve ab-sorption. During applying various methods to improvedissolution, we faced with a number of issues. Forexample, we had to drop many substances whereformulation was not developed smoothly and the phy-sico-chemical stability was unfeasible. In a number ofcases, the dissolution behavior was improved butbioavailability was not.

And, as Dr. Kikuchi said, we conduct a lot of ex-periments using dogs before clinical stage, but arethese results truly held even in humans? Complexformulation and complicated manufacturing processare now developed to make dissolution behavior im-proved. We will conduct the human clinical studywith the drug products. However the formulation andmanufacturing process has never been justified bycomparing with the conventional dosage forms clini-cally. How can we make it? We have to accumulatesuch data in order to understand the relationship bet-ween dosage forms and clinical PK data and makethem use to promptly optimize the following formula-tion development study on poor soluble APIs withless effort.

I’m not sure this answer is appropriate to yourquestion, but this is my opinion. If any other manufac-

100

80

60

40

20

00 20 40 60 80 100 120

Time (min)

Dis

solu

tion

(%)

Troglitazone 200mg tablets using Solid Dispersion

Control tablets

Strength: 200mgApparatus: JP Method 2 (Paddle method)Medium: pH 9 Phosphate buffer, 900mLRotating speed: 75rpm

Figure 14.

DISSOLUTION PROFILE

5

4

3

2

1

00 2

Time (hr)

(Beagle Dogs)

Pla

sma

conc

entr

atio

n (µ

g/m

l)

4 6 8

Tablets using S.D.

Control tablets

AUC (0-8hr) Cmax Tmax (µg•hr/mL) (µg/mL) (hr)

Tablets 10.29±1.32 3.02±0.36 2.3±0.2 with S.D.

Control 3.72±0.76 1.02±0.15 3.2±0.3tablets

t-test p < 0.01 p < 0.01 –

Figure 15.

BIOAVAILABILITY (2)

81

turers here would like to give us some comments, Iwould be pleased to hear from them.

Dr. Soon-ih Kim, Ono Pharmaceutical Co.Ltd, Osaka, Japan: The work conducted by Dr. Ku-sai is very interesting. It focuses on what the industryis actually concerned with today. I am going to makea presentation on the selection of the compound la-ter. Can we apply these technologies at the selectionstage of candidates? Once you have got into a cer-tain phase, then there is left little freedom to pick op-tions and methodologies up. Now, it's better to makedosage forms with a conventional formulation: you willget their better stability. But when you have to im-prove bioavailability, the drug should be provided atan amorphous or liquid state. I think it is worthconducting that at the stage of screening API candi-dates instead of doing at the formulation laboratories.

Dr. Akira Kusai, Sankyo Co. Ltd, Japan: Tobe very frank with you, I think it would be desirable ifdrug substances were selected automatically basedon their physical and chemical properties before ap-plying those methodologies at the formulation stage.That's the area Dr. Kim is working today with sufficientexperience at the formulation field. You can choosethe most promising candidate before formulationstage. You need to get a certain amount of API candi-dates to ensure that you can select the most prefe-rable one among them. Also, as Dr. Lipinski mentio-ned, at the initial stage there's the issue of crystallinity.So it's a bit difficult to answer your question directly.

Dr. Shinji Yamashita, Professor, School ofPharmacy, Setsunan University: Listening to yourcomments from the chemist's point of view, what kindof improvement can the chemist expect at early for-mulation stage? Dr. Lipinski said that a difficult API isgoing to be two or three times more expensive, andrequire two or three times the effort – I imagine it asthree times. If that kind of cost is going to be incurred,then surely improvements should be made at formu-lation stage. If it's going to be expensive, very challen-ging and also time-consuming, would you give up? Isthis kind of technology going to be useful at earlystage or not? I think the final decision would be verymuch dependent on the applicability of the techno-logy at the early stages.

Chair: Dr. Rinta Ibuki, Fujisawa Pharmaceu-tical Co., Ltd, Japan: Perhaps Dr. Lipinski couldalso comment on this point, please? Maybe first fromDr. Kusai?

Dr. Akira Kusai, Sankyo Co. Ltd, Japan: Tra-ditionally, we would try to develop dosage forms wha-tever compounds have been selected as API candi-

dates. That was the case in the past. However, as Dr.Lipinski said, with poorly soluble APIs, they require alot of time to develop and also there exsist some in-herent risks. Are we ready to take those risks or not?It's true that we can discuss the risks when we pickthe API candidates up to be developed. These dayswe may contribute somehow to the decision-makingsteps. I don’t think it is too much to say, but in orderto keep the schedule, it's better to avoid developing acandidate if some of these properties are very difficultto handle during formulation.

On the other hand, if a candidate that might be-come as a blockbuster, whether to go or no go is amanagement decision. And if some kind of improve-ment can be attained, then we should do our best todevelop it. I think we need to take all those factorsinto consideration at well-balanced manner. I don'tknow if this could respond well to your question, butthe final decision whether or not to move on with anAPI candidate must be discussed within the organiza-tion.

Dr. Christopher A. Lipinski, Pfizer Inc., USA:There was a definite viewpoint – at least at the PfizerGroton Laboratories – and that is that at my lab andat other labs like ours, we actually completely discou-raged any early formulation work in discovery. Thereason was that we did not want to do anything thatwould hinder our chemists from changing the struc-ture. We say it's the chemist's job to try to improvethe structure, and to get better biological behaviorearlier with early formulation work just slows down theprocess. The chemist doesn't make the changes.

In fact, sometimes we even had to rescue it withformulation technology. We actually had to try to dis-courage the formulation scientists from presenting ex-citing information in internal poster sessions. We wereafraid that if our chemists saw examples of successstories then they would say, OK, there was no needfor them to change the chemical structure, the phar-maceutical scientist could eventually solve the pro-blem, and we did not want that message to getacross. We wanted the chemist to focus on trying tosolve the problem in chemistry, early on.

Chair: Dr. Rinta Ibuki, Fujisawa Pharmaceu-tical Co., Ltd, Japan: Thank you very much, Dr.Kusai, for your presentation.

Can I have one more minute? There were threepresentations from three speakers and there are twopoints I would like to take up at this juncture. Varioustechnologies will be discussed at the early develop-ment stage and it will be difficult to introduce newtechnology at that point. How can we improve this in

82

a simple way, how can we overcome the issue ofpoor solubility in a simple way? The answer is to fillnanocrystal or solubilized particles in capsules. Theseare very simple. Further, microemulsion is anothersimple approach. So the application of such techno-logy will be a very important factor in future.

My second point is that variability is also an issue,although poor solubility and absolute bioavailability areissues. How do we evaluate and assess that? That isour challenge, and frankly speaking we don't have theanswer unless we test it in humans. But how can wecollect the PK data from humans, how can we eva-luate the PK in humans? Of course there are regula-tory issues to overcome, but we would like to makeprogress here in Japan, so that we can have a sys-tem to evaluate the PK in humans.

So, thank you to all of the speakers.

83

Accelerating Discovery andDevelopment of Poorly Water-

soluble Actives by the AqueousSolubilizing System (ASS)

Dr. Soon-ih KIM

85

Chair: Roland Daumesnil, Capsugel Inc., NorthCarolina, USA: We are going to continue with thetheme of formulation strategy. I think it was extremelyinteresting and I'm sure we will have a debate on it after-wards during the panel discussion. As well as the pre-sentations on formulation strategy we have had so far,there will be additional ones from the industry and froma development center.

Before we start there is something I would like to talkabout. You are here facing a lot of top specialists, peopleyou are probably seeing for the first time and will neversee again, like Chris Lipinski. Don't be shy. Ask questions.There are no stupid questions. If you want to clarify some-thing, if there is something you do not understand, aska question. Do you think I know everything? I don't knoweverything, just like everybody else, so sometimes youneed to ask many questions. Also, we would be sopleased if you would ask questions before the professorialmachita, all the speakers, can get their oar in. They arehere for the purpose of talking, and Sugiyama loves tobe the first Beat them, be the first, and I will help you.

Our first speaker is Dr. Soon-ih Kim. I'm not evengoing to introduce him, he's extremely well-known inJapan. He has been involved in a lot of formulation,including the Capsugel expert system for capsules. Fora few years he has focused his activities on specifickinds of formulation for poorly soluble actives, and I'msure we will learn something through his presentation.Soon-ih, please, your turn.

Dr. Soon-ih Kim, Ono Pharmaceutical Co. Ltd,Osaka, Japan: Thank you very much, Chairman, foryour introduction, and I would like to thank Dr. Yamashita

and the staff of Capsugel for the opportunity to speakat this symposium.

Well, as you may know, at the discovery stage of thedrug development program there are many compoundsthat are poorly water-soluble. As a result, that affectsresearch efficiency. All pharmaceutical companies havenew chemical entities which, even though they are insol-uble in water, have been screened as promising can-didates because pharmacological potency takes priority.Yet, when such candidates enter the preclinical andclinical stages, they definitely fail and slow down devel-opment due to biopharmaceutical properties (forexample, they are poorly water-soluble). That happensin our company, too.

So, what is the level of solubility that does not causefailure or slow down discovery research? And howshould we solve insolubility problems? Those are theissues that I would like to discuss with you. It's my con-tribution to the symposium. Let me start, then.

I believe that you are familiar with the pie chart shownin Figure 1, which analyzes the reasons why compoundsfail. Lack of efficacy comes high up at 31 percent, as youcan see, while poor biopharmaceutical propertiesaccount for 41 percent of the failure rate. Looking atthe poor biopharmaceutical properties more closely,they encompass solubility; log D; chemical stability; per-meability; metabolism; protein binding; plasma stability;RBC binding, and in-vivo bioavailability.

I would now like to focus on the two critical issuesof solubility and permeability, and discuss the methodsof dealing with them in early-phase development.

Accelerating Discovery and Developmentof Poorly Water-soluble Activesby the Aqueous Solubilizing System (ASS)

Dr. Soon-ih Kim

Ono Pharmaceutical Co. Ltd, Osaka, Japan

86

Recently, there have been trials at early-stage dis-covery evaluating the performance of biopharmaceu-tical properties – solubility, permeability, metabolic stabilityand so forth (Figure 2) After selection of candidates,optimization of the bulk substance, due to crystal poly-morphism, is also considered to be important. How-ever, due to the fact that there are too many evaluationitems and that combinatorial chemistry, which leads toa deluge of promising candidates or compounds, is stilladvancing, a sophisticated screening flow and associ-ated criteria have yet to be put in place. So even if wehave the sophisticated screening flow, if it does notfunction well, then it is not meaningful. Therefore wehave to also address this critical issue.

Eight years ago, Dr. Hashida edited a publicationlooking at unfavorable properties for oral dosage formsin the Japanese pharmaceutical industry. In descendingorder of difficulty encountered, they were: poor bioavail-ability due to low permeability; poor bioavailability due tothe first-pass effect; chemical instability; low solubility, andpoor productivity due to a high dose.

Well, pharmaceutical scientists are one of the armsinvolved in the development program, and formulationresearchers have more than sufficient knowledge ofhow to improve solubility. Therefore I believe they canactively contribute to dealing with this issue. We shouldencourage the involvement of pharmaceutical scientistsand formulation specialists from the early discoverystage.

The graph in Figure 3 illustrates the predicted solu-bility distribution results from the commercial library runby in-silico Our company's library shows a similar dis-tribution, and in a recent project some compounds hada solubility of below 1 microgram per milliliter. I believe

Ono Pharmaceuticals is not the only company to expe-rience this type of phenomenon but, in any case, low sol-ubility is the major impediment in the discovery process.

So, as I mentioned earlier, if pharmaceutical scientistshelped to solubilize the compounds, that would facilitatetheir formulation processes in the later stage of devel-opment. I believe that this is one of the missions of phar-maceutical scientists or formulation specialists. Sincewe have such specialists in the audience, I would def-initely like to discuss the issue with you in today's sym-posium, to establish a more efficient discovery process.

For instance, we may come up with some lead com-pounds through in-vitro screening and if they are poorlysoluble, an aqueous solubilizing system (ASS) might beemployed in order to improve solubility (Figure 4) Themaximum potential of the compound can be evaluatedthrough in-vitro and in-vivo screening, and you cannarrow down the number of compounds. Then, youcan investigate the crystalline optimization for the selectedcompounds. I believe that this stepped process will turnout to be more efficient than the traditional screeningflow. It was recently initiated at Ono in Japan.

This morning, Dr. Yamashita suggested to us that ifyou look at the permeability of the compound, and ifthe permeability and absorption are good enough, wecan rescue that compound even if it is poorly soluble.As shown in Figure 4, even if a compound is poorly sol-uble we may be able to choose it as a candidate. Thedegree of solubility allowed would be chosen by phar-maceutical scientists. That is what the medicinal chemistswould like to propose to the formulation specialists.

We can cope with poor solubilility by micronization orASS.

Modern Drug Discovery, Jan/Feb, 1999

31 %Lack of Efficacy

22 %Toxicity

41 %Poor pharmaceuticalProperties

6 %Market Issues

Figure 1.

REASONS FOR COMPOUND FAILURE

Lead Candidate

Lead or Candidate

Preclinical Candidate

In vitroSpecificitySolubilityPermeabilityMetabolism Toxicity

Process chemistryCrystalline optimizationPurity

In vivoEfficacy, PharmacologyBiopharmaceuticsToxicity

Figure 2.

MODERN SCREENING FLOW

87

Based on these considerations, I came up with aportrait of the screening of preclinical candidates. Can-didates are classified as BCS Class I and II, or BCSClass II and IV. For those in BCS Class I and II, a stan-dard formulation is suitable, while for BCS Class II and IVit depends on the dosage – let's say a 100 mg dose asa standard dose. But I'm sure you have your own criteriafor the dose level at each of your companies. ForClass IV, a preformulation with solubilization andmicronization, for example, may be the way forward.But if what's thought to be a Class II drug turns out tobe a Class IV drug, then the use of a liquid or semi-solid solubilizing formulation may enable it to continue inthe development process.

What we have to be concerned about, however –and I would like to come back to this later – is that whenit comes to liquid and semi-solid products, we have totest compatibility, chemical stability and in-vivo perfor-mance Only candidates that pass these tests can beselected as preclinical candidates. That may be a viableapproach.

Now I would like to introduce a case study on howthis portrait can be applied. Compound X has very lowsolubility, although its other properties are not so bad.Since we did not have in-vivo data at this point, we didnot know whether it should be classified as BCS Class IIor BCS Class IV. But it was likely to be BCS Class IIfrom a pharmaceutical profile.

I organized a re-arrangement of the biopharmaceu-tical problems and strategy for Compound X. Oralbioavailability was 3 percent with monkeys and 20 per-cent with rats, but the toxic dose could not be suffi-ciently evaluated in rats because Compound X has anadverse action on rats, though not on humans and mon-keys. On the other hand, we could not make a confident

estimate of safety in monkeys after administration of thesuspension. Since the compound did not have verypoor permeability and had a low first-pass effect, wedecided to use the aqueous solubilizing system (ASS),in order to evaluate its maximum pharmacology andtoxicity potential.

The Table shows the bioavailability (BA), pharma-cology (ED50) and non-observed adverse level (non-toxic dose) of Compound X in suspension and in theASS system. I cannot go into the details of the phar-macological properties, but in monkeys, no matter howmuch we increased the suspension dose, the plasmaconcentration did not increase. Through ASS, the BAwas higher by about 10-fold, and the ED 50 was higherby about 10-fold. So it was possible to estimate thenon-toxic dose. As a result, Compound X was selectedas a preclinical candidate.

00 0.1 1 10 100 1 000

Predicted Solubility µg/mL

Poorly water-soluble

Num

ber

of C

ompo

unds

40

80

120

160

Figure 3.

IN SILICO PREDICTION OF SOLUBILITY

In vitro Screening

Poorly Water-soluble Actives

In Vitro and In Vivo Screening

Pre-Candidate

Aqueous Solubilizing System (ASS)

Optimization of Bulk

Figure 4.

PROPOSED DISCOVERY FOR POORLY WATER-SOLUBLE ACTIVES 1

Standardpreformulation

Dose

Liquid and Semisolid

Candidate Preformulated

Low (<100mg)or Class II High (>100mg) or Class IV

Candidate

Preclinical Candidate

BCS Class I, III BCS Class II, IV

Compatibility In vitro &in vivo Performance

Figure 5.

PROPOSED DISCOVERY FOR POORLY WATER-SOLUBLE ACTIVES 2

88

With Compound X, from early discovery we prepareda liquid or semi-solid formulation and we conductedpreformulation studies. The details of the ASS formula-tion are confidential, but if more than 120 mg/mL dis-solves, that should be considered sufficient forencapsulation with hard and soft capsules. As the mod-erator mentioned, we have the appropriate equipmentto conduct filling tests, and we looked at compatibility(Figure 5). The hard capsule filled liquid formulation ofCompound X can be prepared by the filling and sealingmachine (CFS 1000, produced by Capsugel), and thenthe capsules can be put through the compatibility test,and in-vitro and in-vivo profiles.

What was also checked at the same time – and whatwe had most difficulty with – was analysis. Experts inthis area will know how difficult that is. The decision onwhether to go on to preclinical or Phase 1 studies hadto wait until we had more stability data. As it turned out,we found that the formulation with ASS was unstablefor Compound X.

In parallel with liquid formulations, the amorphousstate and other possibilities were also examined, as Dr.Kusai suggested. But we found that this compound hadto maintain crystallinity to ensure stability, so once againwe went back to micronization as a possible formulationstrategy.

But how small should we make the particle size? Wehad to make a prediction. Using commercially availablesimulation software such as Gastro Plus (produced bySimulation Plus), we conducted an in-silico simulation ofdissolution and absorption (Figure 6) When we ran thesimulation, we found that with a dose level of 300 mgper person, the particle diameter size where we had nosaturation size problem in our prediction, or no satura-

tion absorption in our prediction, was below 10 microns.

I won't go into the details of the prediction approach,but we are trying out predictions for humans based onin-vivo rat PK profiling using ASS. How we make thatinference I would like to defer to another occasion,because competitors have introduced a similar product.And because 30 mg per person was predicted as theclinical effective dose, we came to the conclusion thatthis compound could be selected as a preclinical can-didate, and so it was entered into Phase 1 study.

I do not intend to be boastful, but pharmaceuticalscientists are able to make a large contribution to the dis-covery stage. Although we had to wait until the end ofour Phase I study to see whether Compound X wastruly successful, it means that in the future, when weencounter other poorly-soluble compounds, we wouldlike to take on the challenge.

Finally, with poorly-soluble compounds, their max-imum potentials (pharmacology, toxicity, pharmacoki-netics) should be evaluated by solubilization from earlydiscovery stage. After evaluation of the potentials, crys-talline optimization based on the relationship with solu-bility and absorption should be conducted on the trulypotent compounds These areas are the recent focusin pharmaceutical research. The Japanese industry hasto compete against mega-companies and survive in theworld. So we have to, I think, focus on pharmaceuticaltechnology such as solubilizing systems, physico-chemical technology, drug delivery systems and ana-lytical technologies. These technologies will supportJapanese-style discovery, and the Japanese pharma-ceutical industry will have to turn to these technologies.Pharmaceutical scientists therefore should be moreinvolved in discovery so that we have truly therapeuticdrugs. That is my firm belief. Thank you.

Chair: Roland Daumesnil, Capsugel Inc., NorthCarolina, USA: Let's try to apply the rule. We havethree or four minutes for questions, and I would likesomeone outside of the usual panel. I give you theproblem, I would like to have someone else. James,you are one of our speakers, so you will have the oppor-tunity, don't worry. But I would like to have someonefrom the floor. I think I am pushing a little, but I wouldlike to change the habit. So rather than have thespeakers and the chairs, if someone would like to askquestions before them… No? OK, James comes next.

Professor James Polli, University of MarylandSchool of Pharmacy, USA: Thank you very much foryour presentation. I know you spoke on discovery butI have a question that's maybe applicable to productson the market. We've heard a lot of discussion about def-initions for low solubility. Many drugs are ionizable; for

Almost absorbed

PredictedEfficacy Dose:30mg/person

Dose:300mg/person

Particle size:10µm

Preclinical & Phase I

Figure 6.

PREFORMULATION (2)

89

example, acids or bases, and their solubility can be verypH-dependent. I was just at a meeting in Portugal and,interestingly, the Swedish regulatory authorities areallowing biowaivers for some of the non-steroidal anti-inflammatories. These are acids which have very goodsolubility at a pH above 5, but below 5 they have verypoor solubility.

This group of compounds as a whole tends to bevery well absorbed in spite of its very low solubility at alower pH. In some sense they would appear to be low-solubility compounds; however, in other senses theyare high-solubility compounds, and in-vivo they certainlyperform as high-solubility compounds. In your studies,have you identified pH levels that are particularly impor-tant for acids or for bases in determining whether or notthey function effectively as highly soluble or lowly soluble?

Dr. Soon-ih Kim, Ono Pharmaceutical Co. Ltd,Osaka, Japan: With the ASS system we can preparea water-soluble formulation irrespective of pH. Our ASSsystem is pH-independent and poorly soluble com-pounds could be water-soluble at the whole pH range.I'm not addressing your question, right?

Professor James Polli, University of MarylandSchool of Pharmacy, USA: No, I don't think so. Imust admit, the reason I asked the question is becauseit's a very difficult one and I have not been able to find

anyone in North America that can answer it. It's a verychallenging question. But it certainly appears to be thecase that many drugs are ionizable and that for certaintypes of compounds, some regulatory authorities allowbiowaivers – for example, with non-steroid anti-inflam-matories, which they don't view as bio problems. Whyis it that some compounds, even though their entireprofile is not highly soluble, why is it that they perform asif they were highly soluble, and can we predict that inadvance?

Chair: Roland Daumesnil, Capsugel Inc., NorthCarolina, USA: I think, James, that that is a very inter-esting question. I would like to keep it for the panel dis-cussion, because you will have all the speakers andchairs challenging that. Thank you, Dr. Kim, for a nicepresentation.

The next speaker is Dr. Hassan Benameur, who ison the staff of Capsugel and is responsible for the devel-opment center at Capsugel. Hassan is a very well-knownspecialist in the formulation of poorly soluble activesusing microemulsion and self-emulsifying systems. Hehas patented a few formulations and I am sure that hewill be looking in detail at the pros and cons of thesekinds of formulation. Before joining Capsugel, Hassanwas responsible for formulation at Gattefossée, a com-pany specializing in lipid-based excipients. Hassan, it'syour turn.

91

Formulation of poorly solubleactives: how to make it

an industrial reality

Dr. Hassan BENAMEUR

93

Dr. Hassan Benameur, Development CenterDirector, Capsugel R&D Division Colmar France :Thank you. Yes, we will be focusing on lipid formula-tion, the lipid-based delivery system. In summary, andextrapolating on from what Dr. Lipinski said this morning,combinatorial chemistry and receptor-based screeninglead to two kinds of product: the brick dust and thegrease ball. The brick dust is a large, organic and polarmolecule with poor solubility and membrane transportproperties. The second category, that we call the greaseball, is a lipophilic molecule with poor aqueous solubilityand a very high log-P, or partition coefficient, which indi-cates good permeability.

It's currently impossible today to develop a good for-mulation approach for the brick dust molecule, whileconventional formulations do not work with the greaseball. As a formulator, if chemists ask us which kind ofdrug we want, or which one is easier to use in formu-lation, I think all of us would agree that it should belipophilic one’s or the so called grease ball As ProfessorYamashita said, scientists today, especially those inpharmaceutical development, have to use what I callinternally 'our two feet', to walk a way through the for-mulation. With these new molecular entities, we haveto use the physical factors but also be aware of the bio-logical factors in trying to find the solution (Figure 1).

We now know that the Biopharmaceuticals Classifi-cation System (BCS), which the US Food and DrugAdministration produced as guidance in August 2000,is very important. It means we can classify actives by sol-ubility and permeability and, to be simple, as either lowor high. Among the parameters that must be selected

at preformulation stage is the dose, defined as high(greater than 200 mg), medium (10 to 200 mg) and low(less than 10 mg). With a lipid-formulation organizedsystem, we normally use a dose of between 10 and200 mg, which is the classical range for this type of for-mulation; a low dosage works easily. Particle size ismainly important in terms of the pharmaceutical tech-nology, or production or characterization of the system,and is defined as high (greater than 100 micrometers),medium (1 to 100 micrometers), and low (less than0.2 to 1 micrometer).

A high partition coefficient (a log-P greater than 3) is,as I said, the best result at preformulation stage; amedium result is 1 to 3, while a low one is less thanzero. We need to aim for a lipophilic formulation, as thiswill integrate best with a lipid-based formulation and, of

Formulation of poorly soluble actives:how to make it an industrial reality

Dr. Hassan Benameur, Development Center Director

Capsugel R&D Division Colmar France

Major Impediments to Oral Drug Absorption Process

Physical Factors

SolubilityPartition

Coefficient

Polar/Non-PolarSurface

Area

Permea-bility

p-glyco-proteinefflux

system

cytochromep 450

metabolism

Biological Factors

Figure 1.

PHARMACEUTICAL DEVELOPMENT

94

course, we must be aware of the bioavailability param-eters: high is greater than 90 percent; medium, 20 to90 percent, and low, less than 20 percent.

I totally agree with Professor Lipinski that each experthas to do his job. Chemists have to provide the bestmolecule or the best information they have on newchemical entities (NCEs), and formulation scientists mustbring with them the best knowledge they have and usethe best tools they can get. So there is a lot of – let's say'homework' – to do on both sides, to collect all the infor-mation, and then the parties need to get together todiscuss and exchange information as scientists and tobuild bridges between disciplines.

So, to summarize several aspects of preformulationstudies for NCEs: achieving solid-state stability meansthat we need to look at the effect of temperature andhumidity, as well as the conditions laid down under theInternational Conference on Harmonization. Then thereis the effect of moisture, and the presence of a stablehydrate, as well as possible polymorphs – rememberthe story of Norvir, where the form that crystallized outfrom the hard gelatin capsule on storage was a differentpolymorph, with actives that were less water-soluble.There is also the effect of light.

In terms of the screening methods required, we haveto establish the pH solubility profile, the pH stability pro-file, and also the effect of oxidation. Based on this kindof information we can start choosing the non-aqueousexcipients or solvents best suited to a self-emulsifyingdelivery system (SEDDS) or a self-micronization drugdelivery system (SMEDDS). The precise selection willbe based on the NCE's solubility and stability.

How we use this information is our commitment tofollow a rational approach. Currently, there's a lot of dis-cussion in the literature about how to use SEDDS orrelated systems (microemulsion, emulsion…). The keypoint is the know how of the excipients which includetheir physico chemical characteristics of the excipientsand their blending. There are three main categories ofexcipients that can be blend together: the hydrophilictype, which is PEG (polyethylene glycol), PG (propyleneglycol), glycerol and so on; the lipophilic type, com-prising oils and MCTs (Medium chain triglycerides ), andthe amphiphilic surfactant type, the Gelucires, Tweens,Cremophors, and so forth.

Through our choice from this wide range of excipientswe can start combining them, in order to select the NCEwith the highest solubility and stability. If the drug isstable, we can begin formulation, using phase diagramanalysis.

Figure 2 is a hypothetical phase diagram in whichyou can see that these systems can organize them-selves when mixed and dispersed in water. The apex ofthe triangle shows the oil, while the base represents thesurfactant phase, using a surfactant co-surfactant(S/CoS), and the water phase. You can see that thevarious combinations possible on the surfactant/oil axisproduce different structures when diluted with water. Itis this kind of rearrangement that we want to develop andto formulate. Through specific combinations you will getdifferent types of arrangement (L1, L2, W/O; O/W…).

Let me run you through the diagram Because thesurfactant/oil composition is fixed, by adding a smallquantity of water you will get a water-in-oil micro-emul-sion. This kind of formulation can be valuable in somecases – let's say, when we want to include a peptide ora hydrophilic system in the capsule – because the con-tinuous phase will be lipophilic. By adding more waterwe come to a more complex situation, what we callbicontinuous micro-emulsion, which is a mixture of 50%water and 50%oil. If we go on diluting, we will end up witha coarse emulsion which is easy to recognize; it is milky.So you can see that, by blending the excipients, wecan get different organizations on the same axis.

The main objective with poorly water-soluble activesis to formulate an oil-in-water micro-emulsion (O/W)

SEDDS (clear, isotropic)or

oily dispersions

oil

coarseemulsion

w/o microemulsion(L2 Phase)

surfactant(s)

bicontinuousmicroemulsion

water

water

oil

o/w microemulsion

micelle(L1 Phase)

High HLB surfactantLow HLB surfactant or cosurfactant

water

oil

Figure 2.

HYPOTHETICAL PHASE DIAGRAM

95

straight away. The key point is to use the right combi-nation of excipients to avoid getting a different systemduring the dilution, which will crash out your active Thisstarts by using a clear, isotropic SEDDS system. In somecases an oily dispersion can work, but this is mainly forhydrophilic products because, by definition, water-insol-uble products will not be solubilised if they are not sol-ubilised at the beginning (isotropic solution). So thereare several ways of doing it.

Chemists have a range of approaches to micro-emul-sion, which has been widely used in classical chem-istry; the first micro-emulsions were developed in 1954.The approach we employ is based on titration. In otherwords what we do, as shown in Figure 3, is that we fixthe composition of the surfactant/co-surfactant – and Iwill explain the use of this later – with the lipidic phaseand we blend them with water to determine the area ofthe micro-emulsion.

Since solubility is very important, and as the activeshave to stay solubilised throughout the GI tract, theexcipient used will be non-ionic, because by using anon-ionic excipient we guarantee the solubility. On theother hand, a non-ionic surfactant and co-surfactantcome within what we call the phase inversion temper-ature: it means that their structure will change with thetemperature.

I admit it doesn't look obvious. But just to give you theeveryday story… We were involved in a study where allthe development had been done at 25 degrees Centi-grade, which is the lab temperature classically used,but when we repeated the dissolution activity at37 degrees, everything crashed out. It only happenedbecause we said that as 25 is close to 37, a 12-degreedifference would not be a big deal for such a complicatedformulation. But it was. So it's very important that thisparameter is fixed right from the beginning.

By making different blends, starting from a compo-sition with a low concentration of surfactant/co-surfac-tant and a high concentration of lipid, we can determinethe area of the micro-emulsion. And, as I have shownyou in the hypothetical diagram (Figure 2), you will getdifferent types of reorganization. But what we are lookingfor is to define precisely the point where we get theorganization that we want to use. In the case of Figure 4,which is an example of a water-in-oil micro-emulsionexistence field, the point corresponds to the inside con-tour of the area. This kind of system can be used forpeptides that we want to put into hard or soft capsules,because the continuous phase is lipophilic.

From this basis we now have to choose how to opti-mize the formula. Figure 5 gives an example of two lipidformulations with the same surfactant/co-surfactant oilyphase, and the same quantity of active. What happenswhen we add water is that one system will give you amicro-emulsion, the other will give you a milky dispersedsystem. It's clear that these two products will behave

S/CoS80

60

40

20

20

40

60

80

P

% of lipidic phase%

of S

/CoS

% of water phase

80 60 40 20Lipidicphase

Water

The ratio S/CoSmust be fixed.

The temperaturemust be fixed.

Figure 3.

MICROEMULSION TITRATION SYSTEM

S/CoS80

60

40

20

40

60

80

80 60 40 20Lipidicphase

Water

W/O Microemulsionexistence field

S/CoS andtemperature fixed.

20

Figure 4.

MICROEMULSION

Liquid formulationsfor HGC or SGC

When dilutedwith Water

Lipid formulation

Figure 5.

PHARMACEUTICAL DEVELOPMENT

96

totally differently from the point of view of absorption.It's well known that a small nano particle, as in clearsystem, will enhance bioavailability and that, secondly,it will be very stable. The question for the milky systemis the active into the organized system. If yes, it is onlyan increase of particles size which could be optimizedduring the formulation. If not it is the crash out of theactive that causes the milky dispersed system. In thiscase the formulation will be complex indeed surfactant/co-surfactant ratio should be redefined

So I drew the diagram in Figure 6. Normally, with thenon-visible system, because you have particles ofbetween 10 and 200 nanometers, it is clear to the nakedeye. When you have the water phase and the lipidicphase, which are two immiscible systems, what hap-pens first of all is that the surfactant forms an interfacialfilm. This is the classic approach formulate an emulsionor colloidal system On the other hand, you will also geta co-surfactant that reduces the interfacial tension, thusreducing the particle size and resulting in a micro-emul-sion system. So it is the combination of these three sys-tems – surfactant, lipophilic phase and co-surfactant – thatwill give you the right combination to do a micro-emulsion.

Now for the active. As I said, with this kind of systemthe more lipophilic the active, the better it is. This is whatformulators like, because then we can build up a systemin which the active is in the core of the micro-emulsion.It can also interact with the interface. But what we wantto avoid right from the beginning is getting an nonisotropic solution during the dilution, because it cancrash out the active from the system, and you end upwith two co-existing approaches: solid particles and themicro-emulsion.

The way to differentiate them quickly is to put themunder a polarized light. This also explains why we liketo keep the crystal form with this kind of formulation;because the particle is very small, you can see the crystaleven though you cannot see the particle. So a quickscreening that we can do during development is to takean isotropic solution and put it under the microscopeunder a polarized light, to check the absence or other-wise of the crystal. So, again, this is a key aspect Wehave to know in which power we are, and what all our for-mulators are. We have to draw the diagram in Figure 6for all of the constituents, all of the oily phases, all of thesurfactants, and this takes a lot of time and energy.

The way that we do the characterization is to carry outa rapid first screening, by eye. If it's clearly a liquid solu-tion we then check the isotropy under a polarized lightmicroscope, as I have explained. We also check the New-tonian behavior with a rheometer, as well as the thermo-dynamic stability, using the aging test and centrifugation.Finally, when we have the right formulation, we measurethe particle size by photon correlation spectroscopy.

So let's go through the Neoral success story. I'mmainly going to focus on the pharmaceutical aspects, theformulation aspects that make it into a success story.

WaterPhase

LipophilicPhase

Surfactants

ColloidalDispersion(10 to 200 nm)

Co-Surfactants

S: Forms the interfacial filmCoS: Reduces the interfacial tension, therefore reduces the droplet size

Figure 6.

SUPRAMOLECULAR STRUCTURE

x 50

Sandimmune® in water solution Classic Emulsion

Particle size distribution (nm) (log gra)Average Mean 1243.1 poly 8.561

Intensity Distribution

20

10

5

2

1 2 5 10 20 50 100 200 500 1000 5000

x 50

Neoral® in water solution Microemulsion

Particle size distribution determined by means of dynamic laser light scattering (PCS).

Particle size distribution (nm) (log gra)Average Mean 29.0 poly 0.038

Intensity Distribution

20

10

5

2

1 2 5 10 20 50 100 200 500 1000 5000

Figure 7.

PHYSICAL IN-VITRO EVALUATION

97

First of all, as I explained earlier, you get two kinds ofsolution from the physical in-vitro evaluation: the classicemulsion and the micro-emulsion (Figure 7). You cansee the classic emulsion via photon correlation spec-troscopy measurement, but the emulsion will have apoly-dispersed index and you will have large particlesizes distribution ranging between 500 and 5,000 nano-meters. On the other hand, with a micro-emulsion youwill get sizes of an order of magnitude lower (10 times),of between 20 and 50 nanometers, and you have asharp mono-dispersed system.

What is the in-vivo result of this? Figure 8 shows thebiological approach. First of all, the key point to notewith this reformulation is that the plasma profile is thesame. This was very important as a line extension ofthe existing Cyclosporine A formulation. Secondly, youcan see that the advantage of Neoral as a micro-emul-sion is enhanced bioavailability, because its 180 mgdose is now bioequivalent to Sandimmune, 300 mg.But on top of that, if we look at the results of the six-hour blood samples, you can see that there is reducedinter-subject variability with the microemulsion formula-tion.

When we do such formulations the question thatquickly arises is, what happens with this lipidic formulationin a fasted and non-fasted study? Figure 9 shows that

with the emulsion you get a clear differentiation betweenthe fed and non-fasted results. This is mainly due to thefact that because the particles are bigger in the emul-sion they interact with the meal, and they are more sen-sitive to lipid digestion On the other hand, with the smallnano particles of the Neoral micro-emulsion the formu-lation are bioequivalent (same Cmax and Tmax)..

Now to try to answer the question asked before theprevious talk. Unfortunately, we cannot use the studywe are currently running at the DC as example, so as ourmodel drug, we will use ketoprofen. With ketoprofen,can a micro-emulsion really be the answer? It is listed asa Class 2 molecule under the BCS classification. The sol-ubility in water is extremely low, 51 micrograms permilliliter, and we have a pKa of 4.45, which means thata pH higher than 4.45 will increase its solubility(Figure 10). The ketoprofen is well absorbed in the humansmall intestine at ph above 5.5.

Nevertheless it was a good challenge for us to showthat even with this product we can make a formulationto enhance the solubility. So we chose the formulationthat's on the market, which is ketoprofen, 50 mg, andmade a lipid formulation which was basically a semi-solid. We wanted to make it a semi-solid, rather thanthe more obvious liquid form, to show that we couldenhance the solubility of the product at acidic pH (inthe stomach). Formulation, consistent of a blend ofexcipient a hydrophilic (free PEG), the monostearate ofPEG as surfactant, monoditriglyceride as co-surfactant,

Sandimmune® 300 mg

Time [h]

Con

cent

ratio

n [n

g/m

L]

Interindividual comparison of Cyclosporine concentration-timeprofiles following single oral administration.

60

10

100

1 000 1 500

1 000

500

00 4 8

12 18 24 30 36 42 48

Neoral®180 mg

Time [h]

Con

cent

ratio

n [n

g/m

L]

60

10

100

1 000 1 500

1 000

500

00 4 8

12 18 24 30 36 42 48

Figure 8.

CLINICAL BENEFIT OF THE MICROEMULSION

Time [h]

Con

cent

ratio

n [n

g/m

L]

100

10

1 000 500

250

00 4 8

Sandimmune® 300 mg

Neoral®180 mg

Geometric mean whole-blood Cyclosporine concentration-time profiles followingsingle oral administrations under fasting conditions ( ) and with a fat-rich meal ( )to 24 healthy male volunteers.

100

8 16 24 32 40 480

1 000 1000

500

250

00 4 8

Figure 9.

UPTAKE BETWEEN FASTED AND NON-FASTED REGIMENS

98

with the as di and triglycerides as the lipidi. As a refer-ence formulation we used a powder which contain also50 mg of ketoprofen blended with the right powderexcipient lactose, colloidal silicon dioxide and magne-sium stearate.

What were our parameters? First of all, we used thepH of simulated gastric fluid without enzymes. That'spH 1.2, which means that the PK was below the PK ofketoprofen, to ensure that it had reduced solubility, andthat we really were dealing with a Class 2 molecule (lowwater solubility high permeability. On the other hand wedidn't want to use high speed stirring, so we used apaddle speed of 50 rpm , and we optimized the con-centration of the excipient by using a size 1 capsule.

Figure 11 shows the results of the dissolution profile,with the lower chart indicating our reformulation. It's clearthat by optimizing a lipid formulation you can solubilisethe active, whatever the pH. And this is typically theanswer: the system has to be independent of the pH,and this is the way we have proven it by increasing thewater solubility of ketoprofen in pH 1.2 by using orga-

nized lipid system in our case study water in oil micro-emulsion using non ionic excipients.

Let's now try to understand where ketoprofen will bepositioned in the micro-emulsion Here, it is very impor-tant to remember what we want to achieve as formula-tors. We can make a very good micro-emulsion, but ifthe active is no longer released from the system wehave fail in our formulation development We want toachieve a stable system that increases solubility, butthat will release the active for absorption and thusanswering the problem of class II molecules. To achievethis, we set up the study by using placebo microemul-sion and ketoprofen formulation in Figure 12 – we candiscuss this in the open panel if you like – and from it weknow that the formulation used will make an oil-in-watermicro-emulsion. By making measurements of particlessize of the microemulsion with and without ketoprofenwe obtained clear evidence that the ketoprofen is posi-tioned at the interface and not at the heart of the micro-emulsion. Indeed the particles size distribution increaselinearly with the increase of the ketoprofen concentration.

This was the part that we call the physical factorsand characterization.

We still need to answer the biological part of thedevelopment. For this we have developed a Caco-2cell formulation screening. Why? Because it's well-knownand its use is well-defined. Maybe it's not the best modeltoday, but it is the standard, and when you are dis-cussing its use with people from other disciplines, theyunderstand what you are talking about. Classically, themain objective is to put the formulation into the cultureand see if it can cross the barrier without damage tothe system (Figure 13).

Surfactants

Polar head

Without ketoprofen

Spontaneous curvature

LAS surfactant

With ketoprofen

W

Ketoprofen

W

Figure 12.

IN-VITRO DISSOLUTION TESTING STUDY

00 80

Time (min)

% D

isso

lved

20

40

60

80

100

120

604020 100 140120

SGF pH 1.2

Figure 11.

IN-VITRO DISSOLUTION TESTING STUDY

KetoprofenepKa = 4.45 Log P = 3.12

Solubility = 51µg/ml

LactoseSilice colloēdale

Magnesium Stearate

Formulation (50 mg ketoprofen)

(Semi-solid Solution)

pH = 1.2900 ml SGFpH1.2 (37°C)

Agitation paddleHGC Size 1

< pKa [ ]> Solubility

: 50 rpmOptimized [ ]of excipients

Lipid Formulation

Powder

Figure 10.

DISSOLUTION TESTING STUDY

99

We have already demonstrated that the product isa Class 2, which means that the solubility is the limitingfactor, not the permeability. What we now needed todemonstrate with the formulation, using CACO-2 bio-logical in-vitro evaluation, was that the results were inthe same area regardless of the approach (Figure 14).As reference we used the active solubilised in DMSO,and different formulations: we usually like to comparea liquid and a semi-solid against this reference. Theresults show that this kind of formulation gives the clearanswer that we have enhanced the absorption withoutmaking any modification.

So, to finish my talk, from the point of view of phar-maceutical development we really believe that a lipidformulation system can take us from a Class 2 moleculeto Class 1. It can be applied to poorly soluble actives,as I have shown, or to the related problems of poorlybioavailable actives, or to difficult-to-formulate products– remember to use the approach with peptides – andto actives that also have some inter-subject or intra-subject variability, as I demonstrated with the Neoralstudy.

I have reached the end of my talk, and I thank you foryour attention.

Chair: Roland Daumesnil, Capsugel Inc., NorthCarolina, USA: We have some time now for ques-tions. I'm not going to give up… Thank you, sir.

Kakashi, Toyo Capsule Co Ltd., Japan: In ourcompany we deal with many requests to increase thedissolution because we are manufacturers of soft gelatincapsules. I have a basic question. In the definition of amicro-emulsion that you talked about, what type of mea-surement method is used and what is the particle size?

Dr. Hassan Benameur, Development CenterDirector, Capsugel R&D Division Colmar France:If I understand correctly, you want the definition of amicro-emulsion and also the size of the droplets. I'lldemonstrate this by referring back to Figure 2, the hypo-thetical phase diagram. If you truly have a micro-emul-sion it will only be found in the area indicated in theFigure. So, by definition, a micro-emulsion is a systemwith a specific composition consisting of a surfactant, co-surfactant and oily phase and water phase. In the caseof an oil-in-water micro-emulsion, the continous phaseis water with oil inside; with a water-in-oil emulsion youwill get a continuous lipophilic phase with water inside,again with surfactant and co-surfactant. So there arealways four components.

From your question I can tell where the confusion iscoming from. The surfactant/oil axis in the Figure doesnot represent a micro-emulsion, but a self-emulsifying ora self-micro-emulsifying system. It means that whenwater is added, you get a more organized system. Witha micro-emulsion you always have four components:surfactant, co-surfactant, oily phase and water.

Caco-2 in-vitro evaluation

apical

B DB*

D*

hydrophilic

basolateral

glucose,ions, peptides

lipophilic lipophilic macromolecule

Figure 13.

BIOLOGICAL FACTOR EVALUATION

Biological In-Vitro evaluation

Test Percent Average PotentialArticle Recovery N=3

ACTIVE DMSO 17 3.79 High

Lipid form S1 68 1.74 High

Lipid form S2 65 2.14 High

Reference 3 ≤ 0.1 Low

Figure 14.

100

With regard to the particle size of a micro-emulsion,chemists would define it as around 20 to 150 nanome-ters maximum. We extend this to 10 to 200 nanometersin our own definition, because we include an assessmentby eye.

Kakashi, Toyo Capsule Co Ltd., Japan: In thatcase, what are the conditions of measurement, or theconcentration used in measuring the particle size? Couldyou elaborate on that point? What conditions do youapply?

Dr. Hassan Benameur, Development CenterDirector, Capsugel R&D Division Colmar France:The conditions to measure the particle size are clear. Ifyou are using an oil-in-water micro-emulsion you usethe diluability factor that you would use in your dissolu-tion method. It's mainly that you make a 900-millilitermicro-emulsion and you take your fraction and you do

the particle size measurement. If you are making a water-in-oil micro-emulsion, this time you have to do the mea-surement without any dilution and then you have yourparticle size measurement.

Referring to the Novartis formulation of cyclosporineA , the particles size giving in the figures is obtainedafter dilution in water for the Sandimune SEDDS andNeoral pre concentrated microemulsion. Indeed, in thepre concentrated microemulsion ethanol is used ashydrophilic phase. So when diluted with water the systemstay organized as microemulsion (O/W).

Chair: Roland Daumesnil, Capsugel Inc., NorthCarolina, USA: I think we have to stop. Thank you,Hassan. I'm sure there will be additional questions duringthe panel discussion. We are going to have a breaknow.

101

In-vitro dissolution:method considerationsand relation to in-vivo

Professor James POLLI

103

Noriko Yamanouchi, Capsugel Japan: Ladiesand gentlemen, please be seated. We are resumingthe session. We are very pleased to have ProfessorHashida of Kyoto University as chair of the session,Invited Lecture II-c. Thank you.

Chair, Professor Mitsuru Hashida, KyotoUniversity, Japan: This is the last of the Invited Lec-tures sessions, and we are very pleased to have twospeakers. So far, we have talked about formulationand formulation technology and various innovative ap-proaches. The last two speakers will be talking aboutregulatory issues and the approach to evaluate for-mulations. They will also be discussing issues aboutglobal standards and requirements.

The first speaker in this session is James Polli ofthe University of Maryland School of Pharmacy, wherehe is associate professor. Professor Polli acquired aPhD under Dr. Amidon at the University of Michiganand worked with Professor Amidon on oral dosageformulation and evaluation. He also frequently colla-borates with FDA and is active in the area of regula-tory affairs. He will be sharing with us his expertise inthose two areas. In relation to the regulations he willbe discussing in-vivo/in-vitro correlation. ProfessorPolli.

Professor James Polli, Associate Professor,University of Maryland School of Pharmacy:Thank you Professor Hashida for that kind introduc-tion. It is a great privilege for me to be here and I ap-preciate the fact that the organizers have invited meto this great conference. What I would like to speakwith you about in-vitro dissolution methodology, and

also how we can relate in-vitro dissolution to in-vivoplasma profiles.

I'll talk about dissolution method considerations,and really we'll be talking about three things. First,we'll talk about dose and sink-condition considera-tions; secondly, we'll talk about a biphasic systemdissolution system to provide sink-conditions and,thirdly, we'll discuss product-specific dissolution me-thods. After speaking about dissolution methods,we'll talk about in-vitro/in-vivo correlation methods andtwo particular IV/IVC methods: one that follows thespirit of the FDA guidance on IV/IVC and, secondly, adeconvolution approach to the IV/IVC form.

Dr. M. Pernarowski was a Canadian pharmaceuti-cal scientist who is credited with discovering the bas-ket dissolution method. Dr. Pernarowski has oftencommented on the role of dissolution, and I wouldjust like to highlight one particular comment from him.He says: 'A discriminating method for determining thedissolution characteristics is an excellent researchtool… Products with poor dissolution characteristicsare obviously poor candidates for the market place orfor clinical trials…'

Perhaps one of the most interesting things he hasto say, in the context of poorly soluble drugs, is: 'Labo-ratory tricks whose sole purpose is to increase the rateof solution… are not always a guarantee that the for-mulated product will be biologically available'. I will tryto refer back to this comment later in our discussion.

From Dr. Pernarowski's comments on dissolution,it appears there are two purposes of in-vitro dissolu-

In-vitro dissolution:method considerationsand relation to in-vivo

Pr. James Polli

Associate professor, University of Maryland.

104

tion testing. First, dissolution can serve as a qualitycontrol function for the manufacturing process and,secondly – perhaps a higher role – dissolution canserve as a surrogate, or a predictor, of in-vivo bioequi-valence.

To some extent, these can be very differing roles.These two different roles can be assessed by youranswer to the following question: should dissolutiondetect differences in formulation, or should dissolutiondetect differences in bioavailability? Is it that dissolu-tion should be discriminating with regard to diffe-rences in formulation, or do we want dissolution toactually reflect only those changes that manifest in in-vivo plasma profiles?

It appears that many misunderstandings in dissolu-tion may have to do with one person's perception ofthe role of dissolution versus another person's per-ception. I've been fortunate to speak with manypeople who have several years of experience in dis-solution testing, and here are some observations thatI've written down from speaking with them.

The first is that in early product development, the-re's a tendency for too much confidence in dissolutiontesting. The belief is that if we measure many physico-chemical parameters, perhaps we can use dissolutionto tell us whether formulation will be good or poor,even if we’ve made no effort to relate in-vitro dissolu-tion to in-vivo product performance. Then during lateformulation development, perhaps by being previouslyburnt or disappointed in early dissolution result, there'sa tendency for too little confidence in dissolution. Anda third observation is that unfortunately there is no onesimple universal dissolution test method for every for-mulation that may enter your laboratory.

This process is something you think about in yourlaboratory: the role of dissolution and the design of anin-vitro dissolution test, based upon drug solubilityand the formulation, including its excipients. Fromthese considerations, we start off with an initial ideaas to what the dissolution system should be. If wedecide that dissolution should serve as a surrogatefor in-vivo, then in-vitro dissolution testing should aimto mimic in-vivo dissolution (Figure 1). Then, basedupon underlying biopharmaceutical properties, wecan perhaps scale in-vivo dissolution – at least in ourminds – to how it will perform in-vivo.

Hopefully, we will have the opportunity to test theproduct in-vivo and then we assess whether the dis-solution test is performing as we hoped it would. And,of course, we actually need to go through this pro-cess many times before we have a good understan-ding of how in-vitro dissolution can actually mimic in-vivo dissolution and hence be a surrogate for in-vivoabsorption.

The same schematic is represented in Figure 2,as a time-line of the product development process.There are various stages in product development –an early stage, a middle stage and a late stage.Early in product development, we have to estimatewhat the role of dissolution will be in product perfor-mance.

After testing the early-phase product in early PKstudies, we learn something and typically revise ourunderstanding of how the formulation is performing in-vivo, and hence how we should tweak or change ourdissolution test method in order to serve as a bettertest. So there's a natural course of events where wehave to change our understanding of how we canperform dissolution in the best fashion.

initial effort

re-assess

prediction

in-vivo dissolution in-vivo performance

Figure 1.

DISSOLUTION IN FORMULATION DESIGN

Product Development Process

product concept/ assumed IVIVRtargetprototypepilot PK retrospective IVIVR

define formulationIVIVR PK study prospective IVIVRIVIVR definedprocess optimization

scale-up/pivitol PK/reg

SUPAC

from J. Devane and J. Butler

Figure 2.

stage 1

stage 2

stage 3

stage 4

105

What are the favorable properties of a dissolutiontest? Dissolution should be analytically and experi-mentally simple, since there's a lot of analysis to do.Arguably, one does in-vitro dissolution with the pur-pose of mimicking in-vivo dissolution. Obviously, dis-solution is not a permeability test. By itself, it's not atest that's able to predict bioavailability; that would re-quire additional information. But in-vitro dissolutionprovides the opportunity to mimic in-vivo dissolutionand, perhaps most importantly, assess for factors thatmay influence in-vivo dissolution. If there are formula-tion factors that are important in-vivo, we would likeour in-vitro test to be sensitive to those factors.

There are a couple of issues with regard to disso-lution testing. For instance, if we're trying to mimic in-vivo dissolution, what if a drug does not dissolvecompletely? There is the classic example of griseo-fulvin. Griseofulvin's absorption is incomplete be-cause of incomplete dissolution due to low solubility.An estimate is that the bioavailability of griseofulvin isabout 30 or 40 percent, due to incomplete dissolu-tion of griseofulvin in-vivo. So the natural question is,what should be the extent of dissolution in the in-vi-tro test for griseofulvin? Should it be 30 percent or40 percent? We typically assumption that in-vitrodissolution should go to completion, even if in-vivodissolution is not complete. This is an outstandingissue.

Similarly, we frequently design the dissolution testto provide sink conditions. This is most problematicfor poorly soluble drugs. To attain complete dissolu-tion of a poorly soluble drug, one frequently adds asolubilizing agent – for example, surfactant – and un-fortunately we know that surfactant can decrease thesensitivity of the dissolution test.

Another issue with regard to dissolution is that wewant everything to be the same across dosage

strengths. There could be a 10-fold range in dosestrength, including poorly soluble drugs. Yet we havethis historical tendency to prefer the same test me-thod and the same specifications, even though weknow that dose is an important fundamental barrier tocomplete dissolution, especially for poorly solubledrugs.

Figure 3 is an example where we have three diffe-rent products. It's really the same product, a poorlysoluble drug, at three different dose levels: level 1X;an intermediate level, 2X, where there's twice asmuch drug in the dosage form, and also a three-folddose level, 3X. This is a multi-particulate productwhere particulates are encapsulated, using a capsule,and that way there are three dose levels available.

The 3X dose is the slowest-dissolving. Nextcomes the intermediate, that is, the 2X dose Themost rapidly dissolving dosage form is the lowestdose. These are intended to be extended-releaseproducts so, for an extended-release product, thesevariations are really significant, particularly consideringthat the only difference between these products isreally just the number of particulates.

Why does dose slow dissolution? Based upon thesolubility of the drug and our test conditions, using900 ml, the two highest doses, the 2X dose and 3Xdose, do not provide sink conditions. Meanwhile, the1X dose provides sink conditions. I think we are all fa-miliar with the basic dissolution equation in Figure 4,where dissolution is a function of the solubility of thedrug, and also it considers the amount of drug that'salready in solution. So, by applying that model we canperhaps determine what's going on.

We took the data for the 2X product and the 3Xproduct and apply the dissolution model to fit thedata. Figure 5 is the observed data for the 3X productand the fitted profile. Figure 6 is the observed data forthe 2X product and the fitted profile. We're happy withthe fits.

Influence of Sink Conditions on Dissolution

Dose Sink Conditions ?

1 X Yes

2 X No

3 X No

dM =

AD (CS – C)

dt h

Figure 4.

100

80

60

40

20

00 4 12

Time (hr)

Per

cent

Dis

solv

ed

8

1 x2 x3 x

Figure 3.

INFLUENCE OF SINK CONDITIONS ON DISSOLUTION

106

Then we asked, what would be the predicted dis-solution profile for the low-dose product? The 1X pro-duct did not have problems with sink conditions.Using the model based upon the 2X and the 3Xdose, what we predicted is the dissolution profile inFigure 7, which pretty much matches the observedprofile. These results suggest that the lower dosesdissolve quicker since they are not limited by dissolvedrug.

And so the question we come back to is, whatshould we do? Since the formulations are essentiallyidentical but only differ in the amount of drug, it wouldbe reasonable to expect the lower-dose products toactually dissolve quicker, and allow specifications thatfollow that observation.

Figure 8 is another example. There is a fast disso-lution formulation version of a product, where the par-ticle size is relatively small. There's another formula-tion with larger drug particle size. We'll call them a fastformulation and a slow formulation. There plasma pro-files are reproducible, with the fast formulation yieldinga higher C-max. The fast formulation dissolves more

rapidly than the slow product, although the differenceis modest. So what was needed here was a highlysensitive dissolution method to discriminate betweenthese somewhat similar but reproducibly bio-inequiva-lent products.

Given Dr. Pernarowski's comments with regard tothe overuse of surfactants in dissolution testing, wedesigned a biphasic dissolution system (Figure 9),where we had 800 ml of buffer along with 100 ml ofoctanol, which functioned as a sink. The octanol layernegated the need for surfactant, yet provided sinkconditions for this poorly soluble drug.

We first wanted to understand the basic perfor-mance of this biphasic system. At time zero, drug so-lution was added to the buffer, and we monitoreddrug distribution between the phases. Over a shorttime, approximately an hour or two, about 15 percentof the drug remained in the buffer. Meanwhile, therest of the drug partitioned into the octanol. Figure 10shows disposition kinetics of drug within the biphasicsystem.

100

80

60

40

20

00 4 12

Time (hr)

Per

cent

Dis

solv

ed

8

2 x observed2 x fitted

Figure 6.

MEDIUM DOSE

100

80

60

40

20

00 4 12

Time (hr)

Per

cent

Dis

solv

ed

8

3 x observed3 x fitted

Figure 5.

HIGH DOSE

100

80

60

40

20

00 4 12

Time (hr)

Per

cent

Dis

solv

ed

8

1 x observed1 x predicted

Figure 7.

LOW DOSE

4

3

00 75 225

Time (hr)

Con

cent

ratio

n (µ

m/m

l)

2

1

150

SLOWFAST

Figure 8.

PLASMA PROFILES OF FAST AND SLOW FORMULATIONS

107

Since we anticipated that the differences betweenthe fast and the slow formulations were due to par-ticle size, we assessed if the system was sensitive toparticle-size. In evaluating fine particles and coarseparticles, Figure 11 shows their percent dissolvedversus the time profile in the biphasic system. Wewere happy with this result.

To assure that these results were not an artifact ofthe biphasic system, we essentially repeated thestudy in buffer only, using a small amount of drug, asmall amount of fine particles and also a smallamount of coarse particles, such that sink conditionswere not a problem. Buffer was able to discriminatethe fine particles from the coarse particles (Figure 12).We felt this biphasic system had some chance, al-though the use of octanol presents an environmentalchallenge.

Figure 13 shows the dissolution profile of the fastextended-release formulation in the biphasic system.We monitored the amount of drug in the buffer. Wealso monitored the amount of drug in the octanolphase,. We added them together. Figure 13 illustratesthe cumulative dissolution profile of the fast formula-tion. Very interestingly, we always saw about thesame constant level of drug in the buffer phase, re-gardless of how much drug had been dissolved up toany particular time.

Figure 14 is the dissolution profile from the slowformulation in the biphasic system, showing drug inbuffer, drug in the octanol phase and the total drug-release profile.

Then we compared fast and slow (Figure 15). Thefast formulation, which showed a slight but signifi-cantly higher C-max than the slow formulation in-vivo,

actually did perform as a fast formulation in this in-vi-tro test. A surfactant-based system was not able todiscriminate between the two formulations.

Just one last comment with regard to dissolutionmethods. Perhaps it's fair to say that dissolution isproduct-specific. This would be one point of view.

Octanol

Buffer

Screen

Dosage Form

Figure 9.

BIPHASIC DISSOLUTION MEDIUM

100

75

00 3 12

Time (hr)

Per

cent

dis

trib

utio

n

50

25

96

drug in octanoldrug in buffer

Figure 10.

DISTRIBUTION KINETICS IN BIPHASIC SYSTEM

120

90

00 3 12

Time (hr)

Per

cent

dis

solv

ed

60

30

96

coursefine

Figure 11.

DISSOLUTION PROFILES IN BIPHASIC SYSTEM:PARTICLE SIZE EFFECT

100

80

00 1 3

Time (hr)

Per

cent

dis

solv

ed

60

40

20

2

coursefine

Figure 12.

DISSOLUTION PROFILES IN BUFFER:PARTICLE SIZE EFFECT

108

Since the relevance of the dissolution method de-pends upon the drug substance as well as the actualdrug product itself (for example, the release mecha-nism), dissolution testing will generally be consideredto be drug-product specific. The experimental evi-dence for such a broad statement is that unfortuna-tely we still have never found a universal dissolutiontest that is applicable to a wide range of dosageforms.

And the implication of the idea that dissolution isproduct-specific is that perhaps it would be best todo some dissolution early on – maybe not too much,but focus on getting some in-vivo data as early aspossible. This approach will allow for an early evalua-tion of what dissolution is able to tell us about in-vivo.

To further give evidence for the notion of product-specific dissolution, there are many marketed ver-sions of theophylline extended-release capsules. TheUnited States Pharmacopeia (USP) accepts 10 diffe-rent dissolution tests. These 10 tests have eight diffe-rent dissolution media. They range from simulatedgastric fluid (SGF) for one hour, then pH 6 phosphate;

through pH 3 phosphate for 3_ hours, then pH 7.5phosphate; to pH 4.5 phosphate with octoxynol 9;and pH 7.5 simulated intestinal fluid (SIF), without en-zyme. These 10 different dissolution tests have 10different dissolution criteria, 10 different sample timesand acceptance intervals.

One dramatic aspect is that two of the formula-tions have the same test method and media, so eve-rything's experimentally identical. Clinically, they arealso identical in that they are both intended to be gi-ven every 12 hours. So everything is similar aboutthese products in terms of their dissolution method,dissolution media and clinical utilization.

However, they differ dramatically in their specifica-tions. Specifications for one product allow 85 to 115percent at five hours, while the other product is muchslower-dissolving and is allowed 50 to 80 percent atfive hours. Clearly, these two products are different.Even though they perform the same in-vivo, they per-form very differently in in-vitro tests, but meanwhileare still safe and effective products.

So those were some comments with regard to dis-solution methodology.

What I would like to talk about next is in-vitro/in-vivocorrelation. In particular, we will talk about an importantdocument that the FDA issues. In 1997, the FDA is-sued its in-vitro/in-vivo correlation guidance, which isapplicable to both new drugs and generic drugs.

The guidance describes what data is necessaryfor an IV/IVC, and formulation requirements. It dis-cusses predictability evaluation – the IV/IVC has toshow predictability – and the guidance says there aretwo areas of application. First, an IV/IVC will allow forbiowaivers; that is, waivers of in-vivo bioequivalencestudies. Also, an IV/IVC will allow one to justify disso-lution specifications.

100

75

00 12 48

Time (hr)

Per

cent

dis

solv

ed

50

25

24 36


total drug

Figure 13.

DISSOLUTION PROFILE OF FAST FORMULATIONIN BIPHASIC SYSTEM

100

80

00 12 48

Time (hr)

Per

cent

dis

solv

ed

60

40

20

24 36

SLOWFAST

Figure 15.

DISSOLUTION PROFILES OF FAST AND SLOWIN BIPHASIC SYSTEM

100

75

00 12 48

Time (hr)

Per

cent

dis

solv

ed

50

25

24 36


total drug

Figure 14.

DISSOLUTION PROFILE OF SLOW FORMULATIONIN BIPHASIC SYSTEM

109

So, briefly, an overview of IV/IVC in the context ofthis guidance. It says that one can take in-vitro disso-lution release, couple it with a model for IV/IVC whichessentially is a pharmacokinetic model, and predictin-vivo plasma profiles. If one is able to do that withaccuracy, one is then allowed to use dissolution as areplacement for in-vivo bioavailability studies.

The term in-vitro/in-vivo correlation refers to a pre-dictive mathematical model describing the relationshipbetween an in-vitro property of an extended-releasedosage form, and also a relevant in-vivo response.

This IV/IVC guidance talks about various levels ofIV/IVC. There's an FDA Level A correlation definition,based on convolution. I would just caution you thatthe Level A definition in the FDA guidance is extre-mely different from the USP Level A definition. For theFDA, they say a Level A correlation is a predictive ma-thematical model for the relationship between the en-tire in-vitro dissolution profile and the entire in-vivoresponse profile, that is the plasma profile. So it hasto be a predictive model that maps all dissolution datato all PK plasma data.

The guidance addresses methodology as far asdata formulation requirements are concerned. As Iread it, essentially three formulations are needed, andthese formulations need to be different. The scope ofthis guidance has to do with trying to identify the im-portant formulation variables, so there needs to besome effort to vary the important formulation va-riables, and they presumably will have different in-vitrorelease profiles.

There are dissolution data requirements. Formula-tions have to employ the same dissolution system,which preferably does not have a pH greater than6.8. There has to be a sample size of at least 12 andthere needs to be a fair difference, at least 10 per-cent, between the fast, medium and slow formula-tions. Plasma data is needed in humans and, prefera-bly, it would be advantageous to have crossover PKdata, and a reference dosage form.

An important component is predictability evaluation(Figure 16). The IV/IVC has to demonstrate that it canpredict plasma profiles. There are two terms descri-bed: internal predictability and external predictability.There are circumstances when one can do internalpredictability; for example, if your drug is a broad the-rapeutic index drug. If your drug is a narrow therapeu-tic index drug, then you need to do external predicta-bility.

Internal predictability concerns demonstrating thatyou can adequately fit the data used in your IV/IVCmodel building. External predictability is more challen-ging, in that you need to show that you can predict aformulation PK profile, when that formulation was notused to build the IV/IVC model. The measure that isused to assess prediction is the percent predictionerror, and that's the observed value minus the predic-ted value over the observed value, so it's a percent.

Internal predictability is the only requirement if youhave a drug with a wide therapeutic index, providedthree formulations are studied. If only two formulationsare studied, a limited benefit is given without externalpredictability. As far as the specifications are concer-ned, if you have three formulations – fast, mediumand slow – the average predicted error has to be lessthan 10 percent across all three formulations for theC-max as well as the AUC. Additionally, the predictionerror for each individual formulation must be less than15 percent.

If the drug possesses a narrow therapeutic index,if internal predictions are inconclusive, or if only twodifferent formulations were studied in IV/IVC develop-ment, you need to subject your IV/IVC to external pre-dictability. It's preferred that the formulation subjectedto external predictability is one that has a different re-lease rate than the formulas that contributed towardsbuilding the model in the first place. It may or may not

Predictability evaluation

• internal predictabilityvs. external predictability- dependent on amount of data, therapeutic index,

range of release rates studied

% PE =observed value – predicted value

• 100%observed value

Figure 16.

100

75

00 4 16

Time (hr)

Per

cent

dis

solv

ed

50

25

8 12

MEDIUM

FAST

SLOWSLOW fit

MEDIUM fit

FAST fit

Figure 17.

DISSOLUTION

110

be acceptable if the formulation for external predicta-bility analysis has the same release but different ma-nufacturing aspects, or is the same product but froma different lot.

The passing criterion is 10 percent. You are allo-wed a 10 percent prediction error for the C-max andAUC in order to demonstrate external predictability. Aprediction evaluation of between 10 and 20 per centis inconclusive, and more data is needed. A resultgreater than 20 percent fails.

Figure 17 is an example of IV/IVC. Following theFDA guidance, there are three formulations: slow, me-dium and fast. Figure 18 shows the plasma profiles ofthese slow, medium and fast formulations. The goal isto take the dissolution data and come up with a phar-macokinetic model, an IV/IVC model, that fits theplasma profiles.

Figure 19 gives the pharmacokinetic metrics, theC-max and AUC. Since the slow and fast formula-tions are used to build an IV/IVC model, we'll performinternal predictability analysis on those two formula-tions. The third formulation in the middle, the mediumformulation, will be use to evaluate external predicta-bility. Hopefully the slow and fast will fit, and if so, we'llsee if the model can predict the medium formulation.

The slow formulation was one of our two formula-tions used to build the model. Figure 20 shows theslow observed plasma profiles, and the model wasable to match the profile. The other formulation thatwas used to build the IV/IVC was the fast formulation(Figure 21). Again, we have observed data and theIV/IVC is trying to match the observed data.

We need to calculate prediction errors. We haveobserved C-max values for our two formulations, andwe have predicted C-max values from the IV/IVC. Wecan determine the C-max prediction error: what is therelative difference between our observed and predic-ted values? These values are each less than 10 per-cent, so it passes (Figure 22).

We calculated the internal predictability AUC forthe two formulations that went into the model-buil-ding, slow and fast (Figure 23). Again, we are lookingat the percent prediction error. They are both under10 percent, so it passes the acceptance criteria.

Since the model was able to pass internal predic-tability for the two formulations, we then took the thirdformulation and put its dissolution profile into theIV/IVC model, yielding the predicted profile in Fi-gure 24.

Figure 25 provides the percent prediction errors.Errors are unusually low. In this regard, this not a typi-cal example, since results are better than normal,that's for sure. However, these results exemplify anapproach to carry out the FDA guidance on IV/IVC,which is a convolution approach emphasizing plasmaprofiles.

There is a second way to do an IV/IVC. Early inproduct development, we're interested in drug ab-sorption per se, rather than plasma profiles, since alot of our laboratory testing is focused on drug ab-sorption itself. With an interest in drug absorption, de-

1.5

1

00 8 12

Time (hr)

drug

con

cent

ratio

n (n

g/m

l)

0.5

642 10

MEDIUM

FAST

SLOW

Figure 18.

PLASMA PROFILES

0.8

0.6

00 12 48

Time (hr)

drug

con

cent

ratio

n (n

g/m

l)

0.4

0.2

24 36

SLOW pred.SLOW obs.

Figure 20.

INTERNAL PREDICTABILITY

Predictability evaluation

Formulation Cmax AUC48hr(ng/mL) (ng hr/mL)

SLOW 0.680 8.61

MEDIUM 1.38 9.27

FAST 1.50 9.46

Figure 19.

111

convolution approach to IV/IVC is an approach to em-phasize.

This deconvolution approach is built on a verysimple model for oral absorption. There is a dissolu-tion step and a permeation step. From this simplemodel, we derive an equation where fraction absor-bed is a function of the fraction dissolved (Figure 26).As more drug dissolves, more drug is absorbed.

However, the nature of the relationship betweenabsorption and dissolution is affected by a couple ofparameters. In particular, this relationship is modula-ted by the term 'alpha'. Alpha is simply a ratio of thetwo rate constants (permeation and dissolution). Inthis two-step model, we have a rate constant for per-meability as well as a rate constant for dissolution.

Large alpha values indicate permeability is muchfaster than dissolution, which is to say that dissolutionis rate-limiting (Figure 27). At the other extreme, verysmall alpha values indicate permeability is very slowcompared to dissolution, so that would be a permea-tion rate-limited absorption product. We can take theequation and vary alpha, in a simulation-wise fashion.

2

1.5

00 12 48

Time (hr)

drug

con

cent

ratio

n (n

g/m

l)

1

0.5

24 36

FAST pred.FAST obs.

Figure 21.

INTERNAL PREDICTABILITY

Internal Predictability: Cmax

Formulation Observed Predicted %PECmax Cmax(ng/ml) (ng/ml)

SLOW 0.680 0.643 5.4

FAST 1.50 1.43 4.7

Figure 22.

External Predictability

Metric Observed Predicted %PE

Cmax 1.38 1.40 – 1.4(ng/ml)

AUC48hr 9.27 9.19 0.8(ng hr/mL)

Figure 25.

Internal Predictability: AUC

Formulation Observed Predicted %PEAUC48hr AUC48hr(ng hr/mL) (ng hr/mL)

SLOW 8.61 9.20 – 6.9

FAST 9.46 9.17 3.1

Figure 23.

1.5

00 12 48

Time (hr)

drug

con

cent

ratio

n (n

g/m

l)

1

0.5

24 36

MEDIUM pred.MEDIUM obs.

Figure 24.

EXTERNAL PREDICTABILITY

Model

• Fa is the fraction of the total amount of drug absorbed at time t,

• fa is the fraction of the dose absorbed at t = infinity,

• alpha is the ratio of the first-order apparent permeationrate coefficient (kp

app) to the first-order dissolution ratecoefficient (kd), and

• Fd is the fraction of drug dose dissolved at time t.

Polli, J.E., Crison, J.R., and Amidon, G.L. (1996): A novelapproach to the analysis of in-vitro-in-vivo relationships. J. Pharm.Sci. 85:753-760

Figure 26.

112

Figure 28 plots the family of curves that result fromthe deconvolution IV/IVC equation. Since this modelencompasses the classic USP Level A model, theupper line or curve may look very familiar to you. Thisline of unity is the classic USP Level A IV/IVC. Astraight-line relationship is obtained only when alphais very large. The only time we see this linear relation-ship between absorption and dissolution is when dis-solution is rate-limiting. As the drug is dissolving, it'sbeing immediately absorbed, because the permeabi-lity is several-fold faster than dissolution.

I suggest that scientists have been frequently di-sappointed with in-vitro/in-vivo correlation, particularlyfor immediate-release products, in part since an in-correct model is applied. The classic model has beenthe USP model. Unfortunately, the USP model LevelA is only useful for products that are dissolution rate-limited. Many products, particularly immediate-releaseproducts, may not be dissolution rate-limited.

I would like to thank several graduate students atthe University of Maryland, and my collaborators in

dissolution, Dr. Shah and Lawrence Yu at the FDA.Most importantly, I appreciate being here with youand being invited. Shinji and Roland, thank you verymuch for inviting me. I appreciate your hospitality.Thank you.

Chair, Professor Mitsuru Hashida, KyotoUniversity, Japan: Thank you very much, ProfessorPolli. In the first half of his presentation he gave usefulgeneral considerations on dissolution methodologies,and in the second part he talked about in-vitro/in-vitrocorrelation and examined several perspectives in ex-plaining a unique concept of the correlation. Thankyou very much. Since it is a very rare opportunity, Iwould like to open the floor for discussion. Yes?

Hamaura, Sankyo Pharmaceuticals: Profes-sor Polli, thank you very much for the insightful dis-cussion. I have two questions. The first is about thedefinition of sink conditions. The definition of sinkconditions may vary, so what is your definition of sinkconditions when you describe them in your biophasicsystem?

Professor James Polli, Associate Professor,University of Maryland School of Pharmacy: Iam under the impression that there are at least twodefinitions. I think most pharmaceutical scientists sayit's a factor of 10, minimally. If the concentration of so-lution reaches one-tenth that of the solubility (ormore), then you no longer have sink conditions. I be-lieve the British Pharmacopoeia uses a value of three,or one-third, and here I use the one-third value. Whenthe concentration of the dissolving drug exceededone-third of the drug’s solubility, then it was deemedto be non-sink.

Hamaura, Sankyo Pharmaceuticals: Thankyou very much. I have a second question. In your talk,you mentioned that IV/IVC is difficult to undertake foran immediate-release formulation. In Figure 2, whereyou showed the IV/IV relationship with PK studies, thePK study was mentioned as part of an IV/IVC. Well, ifthere's no problem in formulation then the formulationmay not be changed from Phase 1 to the end. Andyet we are expected to change the formulation, andwe have to see the PK in humans to look at theIV/IVC. Do we waive that study or not?

Professor James Polli, Associate Professor,University of Maryland School of Pharmacy:Your question has to do with when should one do anIV/IVC? I think I understand your question, and I hopeI don't put regulators on the spot, but broadly spea-king it seems as if we have things backwards. It's in-teresting that the FDA has an IV/IVC guidance on ex-tended-release formulations.

Alpha

k app

α = p

kd

• large alpha: dissolution rate-limited absorption

• small alpha: permeation rate-limited absorption

• alpha = 1: mixed rate-limited absorption

Figure 27.

1.0

0.00 0.25 1

Fraction dissolved

Frac

tion

abso

rbed

0.5

0.5 0.75

100

10

5

2

1.01

0.5

0.2

0.1

0.01

Figure 28.

THEORETICAL IVIVRS

113

You say you have had the fortunate opportunity tohave formulations that are not very sensitive to formu-lation changes. Arguably, that's a very good positionto be in. If one had to give a biowaiver based upondissolution, wouldn't it make sense to do it for formu-lations that represent low risk?

Meanwhile, we have an IV/IVC guidance for relati-vely high-risk products, extended-release products.While IV/IVCs for extended-release products can bevery appropriate, it would appear unusual that imme-diate-release products are not provided a biowaiveroption via IV/IVCs since immediate-release productsare intrinsically lower risk products, compared to ex-tended-release products. Maybe we have arrived atthis circumstance since whenever we've subjectedimmediate-release products to IV/IVC, we always getbioequivalence.

There was a large collaborative study of immediaterelease formulations performed at the University ofMaryland. This was headed by Larry Augsburger andthe FDA, and in all cases there was bioequivalencebetween a fast, medium and slow formulation, in spiteof real formulation differences. At some level, such re-sults are disappointing, since there are differences in-vitro but they're all the same in-vivo. But should wereally be disappointed in that? Many immediate-re-lease products are water soluble, and I suggest thattheir dissolution kinetics are not overtly important. Ki-netically, dissolution is often rapid, such that you canchange dissolution and yet it does not have a big im-pact in-vivo. In that sense, it's good not to have a cor-relation between dissolution and bio.

Hamaura, Sankyo Pharmaceuticals: If I maytake a different perspective… From the QC perspec-tive, quality control, of course the specification of thedissolution has to be determined, and there, the for-mulation may be changed. Do you believe that the PKstudy is needed, to determine the PK in the specifica-tion?

Professor James Polli, Associate Professor,University of Maryland School of Pharmacy: It'smy opinion that, for may products, we frequently don'thave a good understanding of the role of dissolutionin absorption kinetics. I do think that dissolution isvery quick for some products, such that slowing dis-solution has no ill effect.

The analogy that comes to my mind is that in theUS there are two very wealthy people. There's BillGates and the financier Warren Buffett from Nebraska.Bill Gates is twice as rich as Warren Buffett. But as faras I'm concerned, they're both very wealthy.

At some level we can't simply concern ourselveswith the fact that there's change, but we need to bemore sensitive to the importance of that change. If aproduct is not dissolution rate-limited and dissolutionis slowed, that change certainly may not matter. Ithink that because we often want dissolution to doeverything all the time, we're excluding other impor-tant considerations, such as the role of gastric emp-tying and intestinal permeability kinetics.

Chair, Professor Mitsuru Hashida, KyotoUniversity, Japan: Thank you very much. Can Pro-fessor Sugiyama please be quick?

Professor Yuichi Sugiyama, University ofTokyo: I'm not so familiar with this field, so I'm sorryto ask such a very, very basic question. You mentio-ned that when people are interested in drug absorp-tion they use the deconvolution method, and whenthey are interested in overall pharmacokinetics theyuse the convolution method. That's what I do not ne-cessarily understand because I am not a specialist inthis field, although I know pharmacokinetics very well.So, when you use the deconvolution method to as-sess drug absorption, what is the rate function, whatis the output function? And when you apply theconvolution method, what is the input function, whatis the rate function? That's my very, very basic ques-tion.

Professor James Polli, Associate Professor,University of Maryland School of Pharmacy: Ithink I understand the nature of your question and Iwould agree with you that both of those proceduresare more similar than they are different.

Professor Yuichi Sugiyama, University of To-kyo: Yes, yes, convolution is just the mirror…

Professor James Polli, Associate Professor,University of Maryland School of Pharmacy:But I will say this. Sometimes, in the laboratory, we'regenerating dissolution data. We're also generatingpermeability data, and we're interested in drug ab-sorption. We are less concerned with distribution andelimination. We want to learn about how dissolution isaffecting drug absorption, how permeability is affec-ting drug absorption. So our emphasis is really ondrug absorption.

Meanwhile, perhaps very much later in drug deve-lopment, maybe at the registration stage, we play agame. We want to match plasma profiles; I want AUCand C-max to be the same. That's a different circum-stance than maybe early on in product developmentwhere we're collecting permeability data and dissolu-tion solubility data. However, the underlying pharma-

114

cokinetic methods are essentially the same. But theyemphasize two different things.

Professor Yuichi Sugiyama, University of To-kyo: I don't necessarily understand what you said,but I can discuss the data with you. What I want toknow is, what is the input function, and what is therate function for the convolution method? And, for thedeconvolution method, what is the output functionand what is the rate function? That is my question.But I can talk to you later.

Professor James Polli, Associate Professor,University of Maryland School of Pharmacy:Thank you. We could have a whole symposium onIV/IVC methods…

Chair, Professor Mitsuru Hashida, KyotoUniversity, Japan: Thank you very much. I'd like toclose the lecture now. Thank you, Professor Polli.

The next speaker is Dr. Lawrence Yu. Let me intro-duce Dr. Yu. His details are included in your handout.Currently, he works as Director for Science in the Of-fice of Generic Drugs at CDER, part of the FDA.There are a lot of very active people working in theOffice of Generic Drugs and Dr. Yu is a leader of thatOffice. Recently, his name has been frequently citedin various society meetings, or we often hear from himdirectly at these meetings. He will be talking aboutWater-insoluble Drugs: scientific issues in drug deve-lopment and drug regulation.

115

Water-insoluble drugs:scientific issues

in drug developmentand drug regulation

Dr. Lawrence YU

117

Dr. Lawrence Yu, Director for Science, Of-fice of Generic Drugs, CDER, Food and DrugAdministration, USA: Thank you. Good afternoon,everyone. Thank you, Dr. Hashida, for your kind intro-duction. I'd also like to thank the organizers for invitingme. I want to thank Capsugel for your continued effortin developing formulation science. It is my pleasureand privilege to be here to give a talk entitled, Water-insoluble drugs: scientific issues in drug developmentand regulatory evaluation. Since I am from the FDA Iam required to provide a disclaimer: 'Opinions ex-pressed in this presentation are those of the speakersand do not necessarily reflect the view or policies ofthe US FDA'. What this specifically means is that nei-ther the FDA nor myself are responsible for what I amgoing to say for the next 45 minutes!

My talk will cover four aspects, namely: the Bio-pharmaceutical Classification System (BCS): the sci-entific basis for biowaiver or biowaiver extensions; lim-its to oral drug absorption, to understand whatcauses poor oral drug absorption; examples of thedelivery of poorly water-soluble drugs and, finally, avery brief discussion of the challenges to regulatoryevaluation.

Let's start with the BCS scientific basis forbiowaiver or biowaiver extensions (Figure 1). What isthe BCS? The BCS is a scientific framework for clas-sifying drugs, based on their aqueous solubility andintestinal permeability. Basically, there are two param-eters, solubility and permeability. Each parameter hastwo levels, high and low. The combination of two pa-rameters and two levels creates four Classes, namelyClasses I, II, III and IV. Class I is high solubility/highpermeability, Class II is low solubility/high permeability,Class III is high solubility/low permeability and, finally,Class IV is low solubility and low permeability.

In August 2000 the FDA issued a guidance en-titled, Waiver of in-vivo Bioavailability and Bioequiva-lence Studies for Immediate-release Solid OralDosage Forms, Based on a Biopharmaceutical Clas-sification System, or BCS. Please note that the waiverrefers to the studies, not the requirements of bioe-quivalence or bioavailability; it simply is a waiver of in-vivo bioavailability/bioequivalence studies. What hasthis guidance specifically discussed? If the solid oraldosage form meets certain criteria, in-vivo bioavail-ablity or bioequivalence studies can be waived(Figure 2). First, rapid dissolution; namely, if more than

Water-insoluble drugs: scientific issuesin drug development and drug regulation

Dr. Lawrence Yu

Director for Science, Office of Generic Drugs, CDER, Food and Drug Administration, USA.

What Is the BCS?

• The BCS is a scientific framework for classifying drugs based on their aqueous solubility and intestinal permeability.

Biopharmaceutics Class Solubility Permeability

I High High

II Low High

III High Low

IV Low Low

Figure 1.

118

85 percent dissolves in 30 minutes at the three dis-solution media, pH 1.2, pH 4.5 and pH 6.8. Second,high solubility; namely, that the highest strength of thedrug dissolves in less than 250 ml aqueous media ina pH range of 1 to 7.5. Third, high permeability;namely that the drug's extent for intestinal absorptionis equal to or more than 90 percent. Finally, the FDAalways uses a risk management approach. As a re-sult, only drugs with a wide therapeutic index are in-cluded. Although at this point we have not reached aconsensus on the definition of a wide therapeutic win-dow drug, we are still investigating. So this is theguidance.

Where or when can it be used? It can be usedanywhere, from pre-clinical to Phase IV and in multi-source drugs or generic drug applications (Figure 3).The BCS has received tremendous support from thescientific community, especially from the FDA Advi-

sory Committee for Pharmaceutical Science, otherexperts, FDA staff and public workshops. We've runat least four or five workshops on this topic.

Concerns expressed at these public workshops,along with comments on the BCS guidance, suggestthat the approach is overly conservative. As you cansee, this is an initial step towards biowaivers, so it's nat-ural to be conservative. Therefore we are making a con-tinued effort to relax the solubility class boundary condi-tions, or maybe extend biowaivers to BCS Class IIIdrugs, namely, high solubility/poor permeability drugs(Figure 4). I can hear you asking why is that? Why not aClass II? Why not a Class IV, instead of Class III?

At this point we want to examine Class III drugs(Figure 5). Let's review the scientific basis for bio-waiver extension for Class III drugs. The oral absorp-tion of Class III drugs is likely limited by their perme-ability, and less dependent upon the formulation. Ifdosed in rapidly dissolving solid oral dosage forms,

RAPID DISSOLUTION

BIOWAIVER

HIGH SOLUBILITY

HIGH PERMEABILITY WIDE THERAPEUTICINDEX*

* Definition under discussion at this time

Figure 2.

WAIVER OF IN-VIVO BE STUDIES OF PHARM.EQUIVALENT IR PRODUCTS

Initial Classification

Class Confirms to BCS Criteria

Equivalence In Vitro

Equivalence In Vitro - Level 3

Equivalence In Vitro

Pre-clinicalClinical

Phase I-IIIFormulation Changes

Manu. Changes….

MarketedFormulation

Post-ApprovalChanges

Multi-SourceProduct

Figure 3.

BCS APPLICATIONS

Solid

=

=

IN VITRO

STOMACH

INTESTINE

COLON

Solution

DissolutionComplete

GastricEmptying

IntestinalAbsorption

IntestinalAbsorption

ColonicAbsorption

ColonicAbsorption

MouthSalivaryglands

Esophagus

Liver

Gallbladder

Stomach

Pancreas

Smallintestine

Largeintestine

Rectum

Figure 5.

IN-VIVO ABSORPTION OF A BCS CLASS III DRUG:SOLID VERSUS SOLUTION

10

1

0.1

0.010 100 000

Volume (ml) of water required to dissolve the highest dosestrength at the lowest solubility over pH 1.2-7.5 range

Class IModel DrugsAbsorption

“Not limited”

Class IISolubilityLimited

Absorption

Class IIIPermeability

LimitedAbsorption

Class IVProblem Drugs

PoorAbsorption

Jeju

nal P

erm

eabi

lity

Pef

f (x1

0 -4

) cm

/sec

10 100 1 000 10 000

Figure 4.

BIOWAIVER EXTENSION: BCS CLASS III DRUGS?

119

the dissolution can be complete in the stomach.Thus, whether it's a tablet or a capsule, once it entersthe stomach, dissolution essentially takes place there.But a solution will simply go through mixing in thestomach. What left the stomach is a drug solutionwhether it is dosed in solid dosage forms or in solu-tions. Since absorption mainly occurs in the small in-testine, if excipients have no effect on absorption orGI motility, intestinal absorption is expected to be thesame, whether it is dosed from a solid oral dosageform or a solution. Therefore there's no scientific rea-son to believe that rapidly dissolving immediately re-lease products with highly soluble/poorly permeabledrugs could show bio-inequivalence, unless exci-pients utilized in the solid dosage forms strongly affectthe permeability or GI motility.

So we have the following hypothesis for biowaiverextension for Class III drugs. If two solid immediate-release dosage forms of BCS Class III drugs dis-solve rapidly at all physiologically relevant conditionsand contain no excipients that may affect oral drugabsorption, then the bioequivalence of these twosolid oral dosage forms is assured. Therefore we be-lieve that, scientifically, a biowaiver can be granted.At this time we are still collecting data to providemore evidence, so that hopefully some day we canconvince the scientific committee and the public thatfor those BCS Class III drugs, a biowaiver can begranted.

Now let's move on to the next topic – the limits tooral drug absorption. Today we have discussed poororal drug absorption, from perspective of permeability,and from perspective of solubility. I want to go onestep further. With respect to solubility, what are actuallythe causes of poor oral drug absorption? In order to do

that, please allow me to go back to review mechanisticmodels for predicting oral drug absorption.

This is basically a brief description of the absorp-tion process. When administered to the patient, asolid drug such as a tablet or capsule will disintegrateor dissolve in the stomach. Dissolved and undis-solved drug empties from the stomach into the smallintestine, where dissolution continues. The dissolveddrug will cross the intestinal membrane, passesthrough the liver and reaches the systemic circulation.We use the rate and extent of absorption to charac-terize the efficacy of these processes. So the funda-mental processes of oral drug absorption are transit,dissolution, absorption and metabolism (Figure 6).

In order to predict oral drug absorption, we use theso-called a reduction approach, which models andsimulates one process at a time (Figure 7). How dowe go about developing a mathematical processmodel? First, we develop a transit model. We assumethe drug molecules are like a plastic ball, so there'sno absorption/intestinal permeability, there's nometabolism, there is no dissolution. It's simply thedrug flowing through the intestine. Once the transitmodel is designed and validated, you then add ab-sorption, and at this point you assume dissolution isvery fast and therefore it's not a limiting step. We callit the absorption and transit model, or CAT, for com-partmental absorption and transit model. Next youadd the dissolution model for poorly soluble drugsand we then have an advanced compartment ab-sorption and transit model (ACAT). Finally, we includethe last step, hepatic metabolism, the ACAT-BA, foran advanced dissolution and transit model, with BAstanding for bioavailability.

MouthSalivaryglands

Esophagus

Liver

Gallbladder

Stomach

Pancreas

Smallintestine

Largeintestine

Rectum

Metabolism

TransitDissolution

Absorption

GastricEmptying

Figure 6.

ABSORPTION PROCESSES

Transit model

+ Absorption / permeation

+ Dissolution

+ Hepatic metabolism

Absorption and transit model(CAT)

Abs., diss., and transit model(ACAT)

ACAT - BAmodel

Figure 7.

PREDICTING ORAL DRUG ABSORPTION

120

The final ACAT-BA model consists of seven com-partments (Figure 8). The top row in the Figure standsfor the solid drug and the lower row stands for the liq-uid drug, and we have the liver and the systemic cir-culation. Now, I'm now giving you equations to de-scribe the ACAT-BA model. If you write down themathematical equation it can consist of over 16 ordi-nary differential equations. It appears complicated,but with a computer it's a piece of cake.

Figure 9 is the compartmental transit model (CAT),showing one compartment, nine compartments andseven compartments. Based on human experimentaldata for small intestine transit time we found thatseven compartments best represented the data, andso we regard the seven-compartment model as thetransit model that models the drug flowing through thesmall intestine.

The model can certainly predict transit time. Bytaking the next step, and adding absorption, you can

now also predict the extent of oral drug absorption, orwhat I call the extent of intestinal absorption (Fig-ure 10). At this point, dissolution is very fast, and wecan show you a couple of compounds – cefatrizineand ranitidine – where prediction works reasonablywell. However, even though the figures appear verygood, if you look at low permeability drugs, we havelarge differences, so it's not perfect.

Not only can we predict the extent of intestinal ab-sorption, we can also predict the rate of absorptionfor saturable drug absorption, as in the case of cefa-trizine in Figure 11. As you probably know, cefatrizineis a carrier-mediated transporter. For saturable ab-sorption the extent of bioavailability at 250 mg and500 mg is about 75 percent, and at 1,000 mg it'sabout 50 percent. So not only can you predict therate of passive absorption, you can predict the sat-urable absorption.

Small Intestinal Tract

Kt

MexitM0

Ka Ka

Ka

Dissolution

Liver Plasma

Figure 8.

ACAT - BA MODEL

100

80

00 400 600

Time (min.)

Yu et al. Int. J. Pharm. 140:111-118 (1996)

Per

cent

of D

ose

in C

olon

60

40

20

300200100 500

9 Compartment1 Compartment

Experimental7 Compartment

Figure 9.

COMPARTMENTAL TRANSIT MODEL (CAT)

1.0

0.8

0.00.0 2.0 3.5

Human Permeability (cm/hr)

Yu et al. Int. J. Pharm. 186:119-125 (1999)

Frac

tion

of D

ose

Abs

orbe

d

0.6

0.4

0.2

1.51.00.5 3.02.5

ExperimentalPredicted

Figure 10.

CAT MODEL: EXTENT OF INTESTINAL DRUG ABSORPTION(PERMEABILITY-LIMITED)

12

10

0.00 8 12

Time (hr)

Yu et al. Eur. J. Pharm. Biopharm. 45:199-203 (1998)

Pla

sma

Con

c. (

µg/m

L)

8

6

4

2

642 10

Dose = 250 mgDose = 500 mgDose = 1000 mg

Figure 11.

CAT MODEL: RATE OF ABSORPTION SATURABLEABSORPTION (CEFATRIZINE)

121

Because it is made up of seven compartments,we can assign different absorption rate constants toeach compartment; for example, the jejunal, duodenalor ileal. Eventually this leads to being able to predictthe regional drug absorption, as you can see in thecase of ranitidine (Figure 12).

As I said, these prediction models consist of over16 ordinary differential equations. Certainly if I pre-sented this to chemists, as Dr. Lipinski pointed outthis morning, they would be likely puzzled and wouldshake their heads.

Therefore I am trying to develop a very simplescheme, to help our friends in the formulation depart-ment, in the pre-formulation department and in themedicinal chemistry department to understand whatthis specifically means. In order to do that I want todefine so-called dissolution-limited absorption andsolubility-limited absorption.

If we look at the dissolution rate equation inFigure 13 you have a diffusion coefficient, D, dissolu-tion surface area, S, aqueous boundary thickness, h,solubility, Cs, and the concentration in the dissolutionmedia, Cl. If we look at this equation carefully, we willquickly find there are two key parameters that canbring about significant change. The aqueous bound-ary thickness or the diffusion coefficient does notusually change that much from drug to drug. So thetwo fundamental parameters are the dissolution sur-face area – as you know, reduced particle size in-creases the surface area – and the solubility. So ifslow dissolution is caused by the surface area, I callthis dissolution-limited absorption. If slow dissolutionis caused by slow solubility, I call this solubility-limitedabsorption.

Now let's look at the potential utility, taking the defi-nition one step further. In order to do that, we firsthave to define some parameters which we can use tosee whether it is a dissolution-limited absorption, apermeability-limited absorption, or a solubility-limitedabsorption. In order to do that, we have to definethree parameters: dose volume, dissolution time andthe maximum absorbable dose Figure 14. Dose vol-ume is essentially the same definition as in the Bio-pharmaceutical Classification System. Dissolutiontime represents the minimum time required to dis-solve a particle. Finally, the maximum absorbabledose is based on the permeability, solubility, absorp-tion, effective absorption surface area, and transittime.

Figure 15 is basically a very simple schema that Ihave put together. The main transit time of 199 min-utes represents the mean human small intestinal tran-sit averaged from 440 data points collected from theliterature. The 2 x 10-4 centimeters per second is ba-sically the human permeability and represents over90 percent fraction dose absorbed, or a 90 percentextent of intestinal absorption. Finally, we have the ab-

0.5

0.4

0.00 8 12

Time (hours)

Pla

sma

Con

cent

ratio

n (µ

g/m

L)

0.3

0.2

0.1

642 10

ExperimentalTheoretical

Figure 12.

REGIONAL DEPENDENT ABSORPTION (RANITIDINE)

DissolutionPermeation

• DissolutionDissolution rate = D *S/h (Cs - Cl)• D - diffusion coefficient• S - dissolution surface area :

Drug product/Drug substance• h - Aqueous boundary thickness• Cs - Solubility: Drug substance• Cl - Concentration in dissolution media

• Permeability: Drug substance

Figure 13.

ABSORPTION PROCESSES

Quantitative Estimation of Absorption

• Dose VolumeVdose =

M0Cs

• Dissolution TimeTdiss =

phr03DCs

• Absorbable Dose Dabs = Peff Cs A < Tsi >

Figure 14.

122

sorbable dose over the actual dose. So, under thisscenario we can define the dissolution-limited ab-sorption, permeability-limited absorption, and solubil-ity-limited absorption.

Let’s review the comments included in the Figure,especially when applying the schema to preclinicaltoxicity evaluation. For dissolution-limited absorptionor permeability-limited absorption, the absoluteamount of drug absorbed increases with increaseddoses. However, for solubility-limited absorption – I'mtalking about conventional dosage forms here – theabsolute amount of drug absorbed does not increasewith increased dosing.

I want to give you some examples to show howyou can try out the schema. Case Study 1 is a reallive case. The dose is relatively low, 5 milligrams. Sol-ubility is 4 micrograms per milliliter, while permeabilityis high, at 8 x 10-4 centimeters per second. Obviously,this is a very lipophilic compound, poorly soluble. Inorder to improve the bioavailability, over 20 formula-tions were designed and manufactured, with bioavail-abilities varying from 20 to 30 percent.

Sixteen formulations were evaluated in animalsover a year; you can imagine how long it takes toevaluate 16 formulations in three to six dogs. Andover the year it was found that that all these formula-tions had a similar animal bioavailability, ranging from20 to 35 percent. In other words, at the end of over ayear's effort, during which numerous studies wereconducted, the outcome was basically the same. So,as a formulation scientist, you ask yourself the ques-tions, why, what, and how?

Now in this case, even though the solubility is ex-tremely low, only 4 micrograms per milliliter, in fact ab-sorption is complete, believe it or not, because of thefairly low dose. Because the solubility here is aqueous

solubility, solubility in-vivo might be tremendously in-creased. What happened here to cause the lowbioavailability is simply hepatic metabolism. The for-mulation has a limited effect on bioavailability, and alimited number of animals may not show not signifi-cant bioavailability improvement (43 percent CV (coef-ficient of variation)). In other words, whatever kind offormulation you use, unless you inhibit hepaticmetabolism, you cannot improve the bioavailability.

Now let's look at Case Study 2. In this case, solu-bility is even lower, 2 micrograms per milliliter, the doseis 400 milligrams, and permeability is 10 x 10-4 cen-timeters per second. So we have a high dose and, aswe discussed this afternoon, the first approach youwould try is to reduce particle size, because it's lowsolubility. It's very natural. It's a simple approach, prob-ably the first approach option all of us would normallytake, formulation scientists not excepted.

The particle size was reduced from 20 microme-ters to 0.5. As you know, to reach 0.5 micrometers isnot a trivial task, and when you conduct in-vitro disso-lution and find a 25-fold increase you are extremelyhappy because 25-fold is a very significant improve-ment. However, after they are given to dogs the re-sults came back, saying, sorry, there's no improve-ment at all. So again you have to ask yourself thequestions why, what and how?

In this case, solubility-limited absorption is thecause; particle size has no significant effect on absorp-tion. What this specifically means is that because it is ahigh dose, you are simply saturating the small intestine.Therefore, with the reduced particle size you can prettymuch reach maximum solubility in the intestine, andthat's the most you can do. So you see a significantimprovement in-vitro because of the sink conditionsthere. However, you cannot see actual improvement in-vivo, because there are no sink conditions.

Limits to Oral Drug Absorption

Rate-limiting steps Conditions Comments

Dissolution limiting Tdiss > 199 minutes The absolute amount of drug absorbedPeff > 2 x 10 -4 cm/sec increases with the increasing of the dose.Dabs >> Dose

Permeability limiting Tdiss < 50 minutes The absolute amount of drug absorbedPeff < 2 x 10 -4 cm/sec increases with the increasing of the dose.Dabs >> Dose

Solubility limiting Tdiss < 50 minutes The absolute amount of drug absorbed doesPeff > 2 x 10 -4 cm/sec not increase with the increasing of the dose.Dabs < Dose

Figure 15.

123

Figure 16 is the classic example of digoxin, show-ing dissolution-limited absorption. In this case, there-fore, a reduced particle size improves bioavailability,or the fraction dose absorbed.

The next example is griseofulvin (Figure 17). Athigh dose we basically do not see much improve-ment, because absorption is limited by the solubility.However, at low dose, absorption is limited by bothsolubility and dissolution, therefore we can see limitedimprovement. But unlike with digoxin, where it canpretty much reach 100 percent absorption, it is notso in this case.

So far, we have discussed the BiopharmaceuticalClassification System, and we have discussed thelimits of oral drug absorption. Now let me discuss thedelivery of poorly water-soluble drugs.

This afternoon we've heard about a variety of ap-proaches: lipid-based delivery systems, solid disper-sion, co-solvents, nanocrystals, prodrugs, salts, andso on. Now I'd like to ask you to pay attention to whatwe can learn from the absorption of cholesterol (Fig-ure 18). Even though its solubility is extremely low, Iam sure all of us would like to have a low absorptionfor cholesterol. If you look at the solubility, it is 12nanograms per milliliter. How low should it be? It's lowenough, isn't it? It has a C-log of at least 9; that's ex-tremely high. However, for whatever reasons, naturedesigned in such a way that all of us can absorb anaverage of 3 grams of cholesterol a day. So even ifyou eat a lot of large pizzas, they can pretty much getabsorbed.

Now this is 3 grams, where normally you wouldsay that a dose of 1 or 2 grams for poorly solubledrugs is extremely high. So we can learn from nature.She gives us an indication for the possible delivery of

poorly soluble drugs. We are still at the trial and errorstage. Hopefully, in the future, formulation science willdevelop to where, based on drug properties, we cansee what kind of approach we need to take to deliverthese problem drugs.

The example I want to give now is one that I was in-volved with before I joined the FDA – amprenavir (APV).This is an HIV protease-inhibitor, and it has a low solu-bility and is poorly wetted. Conventional oral dosageforms, including the powder-in-capsule form, basicallyhave no detectable plasma levels. So you can see thatbioavailability is very close to zero and the dose is ex-tremely high, 1,200 milligrams, twice daily.

Now, with this situation, we often use a salt ap-proach. With a PK of around 4 or 5, for example, thesalt approach works very well. But in this case the PKof 1.9 (Figure 19), it is difficult to make it.

So the approach to take is the self-emulsifyingdrug delivery system. In fact, a micro-emulsion includ-ing vitamin E-TPGS, or tocopheryl polyethylene glycol1,000 succinate, PEG 400, and polyethylene glycol.

1.0

0.00 80 100

Particle diameter (µm)

Yu. Pharm. Res. 16:1884-1888 (1999)

Frac

tion

of D

ose

Ab

sorb

ed 0.8

0.6

0.4

0.2

604020

Figure 16.

DIGOXIN: DISSOLUTION-LIMITED

1.0

0.00 50 60

Particle diameter (µm)

Yu. Pharm. Res. 16:1884-1888 (1999)

Frac

tion

of D

ose

Ab

sorb

ed 0.8

0.6

0.4

0.2

40302010

Dose = 500 mgDose = 1000 mg

Dose = 250 mg

Figure 17.

GRISEOFULVIN: SOLUBILITY-LIMITED

• Humans can effectively absorb extremely insolubleand extremely lipophilic substances• Fat-soluble Vitamin A (S ≤ 1 µg/ml, CLogP = 6.5)

Vitamin E (S ≤ 1 µg/ml, CLogP = 12.8)• Cholesterol (S ~12 ng/ml, CLog P= 9): ~ 3g/day

• Absorption mechanism and implicationsfor delivery of insoluble drugs?

After Dr. Ping Gao

Figure 18.

LESSONS FROM NATURE

124

The initial clinical studies, which utilized amprenavirdissolved in E-TPGS in hard gelatin capsules, hadgood bioavailability. Because we did not do any IVpharmacokinetics, we would never know the actualbioavailability.

Figure 20 is just to give you an idea of pharma-cokinetics in dogs with a variety of formulations. Thedry-filled drug capsule gives zero bioavailability. But atthe 50 percent point of Vitamin E-TPGS, we prettymuch reach the high limit. Now, you might ask, whynot use 50 percent of Vitamin E-TPGS? There are allkinds of reasons, one of which is that it is way too ex-pensive. On the hand, 20, 30 or 40 percent of Vita-

min E-TPGS give reasonable levels of plasma levels;they have a reasonable bioavailability. So this com-pound, which is actually water insoluble, the self-emulsifying delivery system provides a reasonablebioavailability and can be developed in a product.

Scientifically, I want to discuss the role of TPGS inthe improvement of bioavailability of amprenavir. In thiscase, the solubility of amprenavir was significantly im-proved in the presence of TPGS through micellar solu-bilization, and TPGS also enhanced the absorption orpermeability of amprenavir in-vitro. So, overall, TPGSenhanced the absorption flux by increasing solubility,and probably also enhanced the permeability.

pH-Solubility Profile / Amprenavir Crystal Form A

0.010 1 2 3 4 5 6 7 8 9 10 11 12 13 14

0.1

1

10

0.01

0.1

1

10

Sol

ubili

ty (m

g/m

L)

pH

pH-stat

pKa = 1.97

NH2

H3C

CH3

HN

O

O

O

O

SN

OH

O

Figure 19.

PH-SOLUBILITY PROFILE

Pharmacokinetics in Dogs

Formulation Cmax Tmax t1/2 AUC (0 to 24h) Bioequivalencea

(µg x hr/mL) (µg/mL) (hr) (h x µg/mL) (%)

Dry Fill Capsule 0 0 0 0 0

Peg 400 Solution 3.85 ± 1.25 1.1 ± 0.9 4.2 ± 1.7 12.2 ± 1.46 47 ± 11

20% Vit E-TPGS 5.41 ± 0.69 1.7 ± 0.6 3.6 ± 0.8 22.1 ± 4.52 86 ± 10

25% Vit E-TPGS 5.03 ± 0.44 1.7 ± 0.6 2.0 ± 0.8 20.6 ± 4.85 84 ± 13

30% Vit E-TPGS 8.24 ± 0.12 1.3 ± 0.6 2.0 ± 0.7 23.5 ± 4.97 93 ± 12

40% Vit E-TPGS 6.92 ± 0.94 1.7 ± 0.6 1.9 ± 0.6 24.4 ± 4.55 98 ± 16

50% Vit E-TPGS 7.63 ± 1.46 1.7 ± 0.6 2.5 ± 1.3 26.8 ± 8.27

(Hard gelatin capsule)

Figure 20.

125

Figure 21 has been shown before. Basically, theaddition of Vitamin E-TPGS improved solubility byaround four to 10-fold, and permeability improvedtwo-fold, from five to around 10.

So, in this very brief overview, we have discussedthe Biopharmaceutical Classification System, we havediscussed limits to oral drug absorption, and we havealso used examples to discuss the delivery of poorlywater-soluble drugs. Now I would like to talk aboutthe challenges to regulatory evaluation.

In the development of self-emulsifying delivery sys-tems, we face a number of issues. One is the lack ofconsensus on the appropriate in-vitro dissolutionmethods that are predictive of in-vivo absorption.Here, you can look at the dissolution method for Neo-ral in the USP; in fact, for Neoral, the USP require-ment changes to a disruption test instead of dissolu-tion. In other words, what is specifically meant is thetest of disruption of the capsule, instead of dissolutionof the capsule.

Then there is the lack of scientific rationale for ex-cipient selection. We have been the same levels for avery long period of time. We try and see what hap-pens, but we cannot predict what might be happen-ing. We sometimes have some kind of idea – for ex-ample, if we looked at a micro-emulsion versus anemulsion, we probably would know the micro-emul-sion is better than the emulsion with the absorptioncharacteristics. However, if we used the same micro-emulsion formulation and changed the excipients, youreally wouldn't know what might happen.

Certainly, the number of surfactants and solventsis very limited, and at early-stage formulation develop-ment it's pretty much that we let animals lead humanbeings; we depend upon the in-vivo study for screen-

ing. We're testing in dogs, we're testing in rats andwe see what happens, even though we know animalscan not always have good correlations with humans.But this is the best approach available to us rightnow. Finally, there is a poor understanding of the in-vivo drug absorption mechanism via self-emulsifyingdelivery systems.

So because of those challenges, we really don'tknow if this is the right or wrong dissolution test. But weknow this dissolution test is sufficient for quality control,and that is the usual objective of the dissolution test.We know that, but we do not know whether this disso-lution test is sufficient or correlates to the in-vivo perfor-mance of specific self-emulsifying delivery products.

I want at this point to thank the people who havehelped me in this presentation: Dr. Ajaz Hussein ofthe FDA, Professor Gordon Amidon of the Universityof Michigan, Professor James Polli of the University ofMaryland, and Dr. Arup Roy of Eli Lilly.

My final wish is that some day we will know whatkind of formulation, what kind of excipients we shouldchoose, based on drug properties, to give us theright answer to save our time and development costsand to help human beings. But we have not yetreached this stage with respect to self-emulsifyingdelivery systems. We are at the stage of the art of trialand error, we have not reached the stage that sci-ence asks us to do. It depends on you, it depends onall of us to get there.

This concludes my presentation. I welcome anycomments and criticism. Thank you for your attention.

Chair, Professor Mitsuru Hashida, KyotoUniversity, Japan: Thank you very much, Dr. Yu.You mentioned biowaiver extensions to the BCS. Youalso clarified the limits to oral drug absorption. Youtalked about modeling, and based upon the modelingyou talked about various analyses of the kinetics ofabsorption and also various delivery systems. Youalso talked about the challenges to regulatory evalua-tion. So you covered a broad range of topics in youroverview. I would like to invite any comments or ques-tions about your presentation.

Hassan Benameur, Capsugel Inc., NorthCarolina, USA: Thank you for the excellent presen-tation you gave us. Let me challenge you. I have twoquestions for you. One is regarding your Case Study1, where you said that from the formulation point ofview, when an active has a metabolic effect, formula-tion cannot enhance bioavailability. Is that right?

What we know today is that hepatic metabolismcertainly occurs. But it is known that before that, dur-

1.5

0.00.0 1.5 2.0

Vitamin E-TPGS (mg/mL)

Yu et al. Pharm. Res. 16:1812-1817 (1999)

Sol

ubili

ty (

mg

/mL)

1.2

0.9

0.6

0.3

1.00.5

CMC = 0.2 mg/mL

Figure 21.

EFFECT OF VITAMIN E-TPGS ON SOLUBILITY OF AMPRENAVIR

126

ing absorption, there is some intestinal metabolism,and we can develop more individual models by usinghuman microsomes to see if we first have intestinalmetabolism. And on the art of formulation, if we canshow by using human intestinal microsomes thatthere is a first metabolism at the intestine, can wechoose this as a formulation strategy and so enhancethe bioavailability? So this is some information aboutthe approach, and I totally agree with you. We are farfrom the end of these studies, we are still at the be-ginning.

Dr. Lawrence Yu, Director for Science, Of-fice of Generic Drugs, CDER, Food and DrugAdministration, USA: Well, for Case Study 1, in-vestigators found a very limited effect on bioavailabilityfor all the formulations that had been evalauted. I amwell aware that some polymers, for example, are ableto inhibit gut metabolism, at least in-vitro. At this point,in the case of all the 16 formulations that were evalu-ated, a significantly improvement in-vivo was notdemonstrated. I do not know whether this is actually anormal occurrence or a very limited occurrence. Butbased on the data, it suggests that any contributionfrom gut metabolism is very limited.

Hassan Benameur, Capsugel Inc., NorthCarolina, USA: Did you check if you got an intestinalmetabolism off your active?

Dr. Lawrence Yu, Director for Science, Officeof Generic Drugs, CDER, Food and Drug Ad-ministration, USA: That certainly is another topic ofsymposium. Based on my understanding of some ofthe compounds, in general the intestinal contributionis not that significant for the compounds evaluated.

Chair, Professor Mitsuru Hashida, KyotoUniversity, Japan: Thank you very much. Any otherquestions, please?

Walt Walters, Simulations Plus, USA: Thecompounds that you dealt with apparently do nothave the kind of characteristics that perhaps Dr. Be-nameur's question related to. But there are certainly alarge number of compounds on the market that are3A4 substrates, and certainly potentially there will bemore. It's possible that by delaying release of thecompound, so that release occurs beyond regions ofhigh 3A4 expression in the upper small intestine, youmight actually affect bioavailability by avoiding thoseregions.

Dr. Lawrence Yu, Director for Science, Of-fice of Generic Drugs, CDER, Food and DrugAdministration, USA: That's correct.

Prof. Yuichi Sugiyama: Related to that question,I would like to add a comment. Recently, in my labo-ratory, based on the literature data, we tried to under-stand first- pass GI metabolism, particularly for theCIP 3A4 substrate. We used about 12 compounds.Then, from the clinical pharmacokinetic data, after oraland IV administration, we calculated the so-calledFa/Fg as well as the hepatic clearance. Then we plot-ted the Fa/Fg intrinsic hepatic clearance passes, be-cause in the 3A4 it should be the same in the liverand in the intestine; that is the same iceline.

So what did we find? We are now writing that pa-per, it's going to appear in a journal very soon. Underconventional microsomal stability studies, 1 milligramof microsomal protein per milliliter is a liver micro-some. Our data indicates that if the half-life of micro-somal stability is less than five minutes, then you haveto worry about first-pass GI metabolism. That is ourconclusion, after summarizing all of the information. Inthat case, according to our findings, whether thatcompound is a substrate of PGP or not, it does notaffect our conclusion. That is just our summary.

Walt Walters, Simulations Plus, USA: It's greatinformation, thank you. So its natural passing is…What is the average half-life of the compound? Ifwe're lasting five minutes, the gut metabolism is sig-nificant.

Prof. Yuichi Sugiyama: Less than five minutesmeans the in-vitro microsomal stability studies, usingin-vivo microsomes…

Dr. Lawrence Yu, Director for Science, Of-fice of Generic Drugs, CDER, Food and DrugAdministration, USA: For the majority of com-pounds which you have investigated, what kind ofpercentage show that from in-vitro they have a half-life lasting five minutes?

Prof. Yuichi Sugiyama: Not so many; I wouldsay less than 30 percent of the cultured 3A4 sub-strate. That's our result.

Dr. Lawrence Yu, Director for Science, Of-fice of Generic Drugs, CDER, Food and DrugAdministration, USA: Pgp has no impact on yourconclusion. That is interesting!

Chair, Professor Mitsuru Hashida, KyotoUniversity, Japan: Thank you very much onceagain, Dr. Yu. I would now like to close the session.

127

Open discussionCase studies/

Marketed products

Roland DAUMESNIL

129

Chair, Professor Mitsuru Hashida, KyotoUniversity, Japan: Ladies and gentlemen, we arevery pleased to open the final session. I would like toask Professor Yamashita and our colleague, RolandDaumesnil, to chair the open discussion session.

Thank you very much for staying with us overmany hours. Many topics have been discussed andperhaps your brain is saturated with information. ButI'm asking you to stay on because there are some is-sues that haven't yet been fully discussed or ad-dressed, and we would like to cover those in thissession. I would first l ike to ask my coach here,Roland Daumesnil, to summarize the issues and ap-proaches regarding new and marketed products, andlater he will also give us a summary of today's discus-sions. He says that his presentation will take 20 min-utes and I am sure that you will enjoy it. After that, thefloor will be open for discussion, with the chance forquestions or comments to any of the speakers in to-day's symposium.

Roland Daumesnil, Capsugel Inc., NorthCarolina, USA: Let's examine some numbers first. Ifyou look at the value in billions of dollars of the globalpharmaceutical market in 2002, total OTC sales plusRx prescriptions were US $420 billion, split as $53billion for OTC, and $367 billion for Rx. Not bad. Wetalk a lot about drug delivery: the drug delivery busi-ness represents $50 billion, or 12 percent of the total.What we talked about today, the poorly soluble ac-tives that are on the market, already represent $110billion; in other words 26 percent of the total value iscreated by poorly soluble actives.

We have been dealing with poorly soluble activesfor years, but there is certainly room for improvement.Remember what Chris Lipinski said? When high-throughput screening started, his team was generat-ing a lot of poorly soluble actives. Of the newmolecules launched last year, in 2002, 58 percentwere poorly soluble actives. There must be a goodreason for this, and I'm sure you got it today. The cur-rent pipeline is no different. People continue to ap-proach their work using the techniques we havetalked about, and 40 to 50 percent of new moleculesunder development – even when the Rule of 5 is ap-plied – are poorly soluble.

Going back to the marketed products that accountfor 26 percent of the total market value, there areabout 100 candidates for reformulation among allthese poorly soluble actives. Maybe your companyhas one. Certainly, they can be reviewed with an eyeto increasing the bioavailability, improving the PK pro-file, improving the stability, creating line extensionswith functionality and, last but not least, to extendingpatent protection.

Some of these products are blockbusters.

Let's look at the 12 most important poorly solubleactive products already on the market (Tables 1and 2). Let's start with Pfizer. The first one is atorvas-tatin, the famous Lipitor. Look at the solubility,0.11 mg/ml. Also it's not because you have a poorlysoluble active that you don't have a blockbuster. Thesame for simvastatin. These two products, onepoorly soluble, the other insoluble with high first-passmetabolites, already accounted for US $13 billion

Open discussionCase studies/Marketed products

Roland Daumesnil

Capsugel Inc.,North Carolina, USA

130

in 2001. And so the list goes on. All these productscould be improved and are going to be improved.Celecoxib is one of them, loratidine is another.Something can be done to give a second life to aproduct.

Paroxetine is the same (Table 2). Cyclosporine –we'll be talking a lot about this product, because forme it represents a unique example. Despite all the is-sues that surrounded its launch, this product hassaved the lives of thousands of people. Plavix(clopidrogel) and the anti-cancer products, Taxol andTaxotere, are also insoluble. They have great valueand deserve to be improved, if possible, to give thema patent extension. So with 12 products already onthe market which belong to the blockbuster category– meaning they are worth more than $1 billion – this isalready saying something important about the poten-tial of insoluble drugs.

We already talked about how to formulate Class IIdrugs. In reality, we're talking about a lot of systems:creating salts or a prodrug, reducing particle sizes, in-cluding the active in a polymer complex, and creatinga micro-emulsion or SEDDS. But how many suchproducts are on the market, and using which technol-ogy? I checked the Physicians' Desk Reference, thePDR, plus other compendia, and I came up withsome interesting results: there are not very many.

What you will indeed see more and more are prod-ucts based on micro-emulsions, emulsions or self-emulsifying systems, such as cyclosporine, ritonavirand isotretinoin capsules. So let's look at them, plus afew others. I will certainly be talking about amprenavir– I'm not sure I totally agree with Larry Yu's viewpoint.But we are here to be challenged.

Sandimmune. I love this product (Figure 1). A lot ofpeople around the world love this product, not be-cause it was a big challenge, a big issue in terms offormulation, but because, as I said before, this pro-duct has saved thousands of lives. Without this pro-

Marketed products. Candidates

Generic name Trade name Solubility Bioavailability Worlwide salesin mg/ml in % 2001 in Billion $

Atorvastatin Lipitor 0,11 HPFM 20-40 6,5Simvastatin Zocor Insoluble < 5 6,7Loratidine Claritin Insoluble 0-20 3,2Celecoxib Celebrex Insoluble ? 3,2Olanzapine Zyprexa Insoluble 40 3,1Sertraline Zoloft < 10 20-40 2,4

Table 1.

Marketed products. Candidates

Generic name Trade name Solubility Bioavailability Worlwide salesin mg/ml in % 2001 in Billion $

Paroxetine Paxil 5,4 HPFM ? 2,5Cyclosporine Neoral 0,04 20-40 1,4Clopidrogel Plavix Insoluble 50 1,4Paclitoxel Taxol Insoluble ? 1,2Clarithromycin Biaxin Insoluble 50 1,2Docetaxel Taxotere Insoluble ? 1,0

Table 2.

• Formulation• Cyclosporine 25/100mg• Alcohol• Labrafil M 2125 CS• SEDDS ( Droplet size: 864 nm)

• Product Characteristics• Poor/Variable absorption• Absolute Bioavailability = 30%• High inter/intra patient variability• P-gp inhibitor• Extensively metabolized

by Cytochrome P.450 CYP 3A4• Inhibitors: Ketoconazole, Grapefruit...• Inducers: Rifampin, St John's Wort

Figure 1.

SANDIMMUNE CYCLOSPORINE CAPSULES

131

duct, it is impossible to achieve successful trans-plants. Novartis developed an emulsion which was,by the way, a self-emulsifying system, not a micro-emulsion, and Dr. Okamoto told us that the dropletsize is around 864 nanometers. It has poor and vari-able absorption, everybody knows that, as well aslow bioavailability and high inter/intra patient variability.It is a P-gp inhibitor – extensively metabolized by cy-tochrome CYP 3A4.

What does all this mean? It means that, yet again,these concepts are sometimes the terrible reality wehave to deal with.

What do inhibitors mean? If you inhibit the P-gp orthe cytochrome CYP 3A4, you will increase the con-centration in the blood. You can do that with keto-conazole and grapefruit. But what if it is the reverse, ifyou use an inducer of P-gp like rifampin or St John'sWort? You simply reduce the quantity, the concentra-tion of cyclosporine in the blood, up to the pointwhere you could have a graft loss, a rejection of thetransplant. You simply kill the patient.

So, talking about P-gp inhibitors, talking about cy-tochrome CYP 3A4, and so on, has great value forthese kinds of product. They are life-saving products,and looking at the contra-indications and drug inter-actions is extremely important in better understandingthe role of transporters, inhibitors or inducers.

That is why Neoral was developed (Figure 2). It is amicro-emulsion with a small droplet size of 39 nano-meters. It is not equivalent to Sandimmune because ithas between 20 and 50 percent more AUC, and40 to 106 percent more in terms of the C-max. Thereis no bioequivalence, and less inter/intra patient vari-ability. Nothing else changes.

What is interesting about Figure 3 is that Gengraft isa generic version of cyclosporine targeted at Sand-immune and Neoral. It is a self-emulsifying system andyou can see they use quite different excipients to beable to circumvent Novartis' patent and come up witha formulation bioequivalent to Neoral. Bear in mind thatfor this kind of product you also have to undertake clin-ical studies. You will never get approval from the FDAor any ministry of health without doing clinical trials.

Chris Lipinski was talking about the fact that solu-bility is very dependent on product synthesis, onchemistry synthesis. For some products, this is criti-cal. Norvir in Figure 4 is an interesting example. Let'stalk about it.

Abbott was obliged to change the formulation be-cause it changed the synthesis. The product startedto precipitate and the crystals were simply insoluble.So they had to withdraw it from the market. Also, thisproduct is extremely sensitive to the fed condition.

• Formulation• Cyclosporine 25/100mg• Alcohol• Corn oil-mono-di-triglycerides• Polyoxyl 40 hydrogenated castor oil• MICROEMULSION (Droplet size: 39 nm)

• Product Characteristics• Not bioequivalent to Sandimmune

• IAUC = + 20- 50%• Cmax = + 40-106%

• Less inter/intra patient variability• P-gp inhibitor• Extensively metabolized

by Cytochrome P.450 CYP 3A4• Inhibitors: Ketoconazole, Protease inhibitors,

Grapefruit...• Inducers: Rifampin, St John's Wort

Figure 2.

NEORAL CYCLOSPORINE CAPSULES

• Formulation• Ritonavir 100mg• Alcohol• Oleic acid• Polyoxyl 35 castor oil• SEDDS

• Product Characteristics• Absolute Bioavailability = ?• Solubility very dependent

on chemistry synthesis• Dose 600mg

• AUC= 120 ± 54 Ķ/ml• Cmax= 11 ± 3,6 Ķ/ml• fed conditions

• Metabolized by Cytochrome P.450 CYP 3A4• Inhibitors: Ketoconazole, quinidine, Cisapride• Inducers: Rifampin, St John's Wort

Figure 4.

NORVIR CAPSULES

• Formulation• Cyclosporine 25/100mg• Polyoxyl 35 castor oil• PEG 400• Polysorbate 80• Sorbitan mono oleate• SEDDS

• Product Characteristics• Improved absorption• Absolute Bioavailability = 30%• Reduced inter/intra patient variability• P-gp inhibitor• Extensively metabolized

by Cytochrome P.450 CYP 3A4• Inhibitors: Ketoconazole, Grapefruit...• Inducers: Rifampin, St John's Wort

Figure 3.

GENGRAFT CYCLOSPORINE CAPSULES

132

You can increase the blood concentration by up to 30percent if you take it with a heavy meal. Once again,cytochrome CYP 3A4 is involved, and the inhibitorsand inducers are exactly the same as the ones wediscussed before.

Another good example is Agenerase, or ampre-navir (Figure 5). Dr. Yu, I believe that TGPS is a P-gpinhibitor and that the inclusion of TPGS simplychanges the permeability. You are able to increasebioavailability by including TPGS. The active is alsometabolized by CYP 3A4. But in this case the in-hibitor, ritonavir, and the inducer, saquinavir, are differ-ent. They are protease inhibitors.

Rapamune (sirolimus), 1 milligram, is another im-munosuppressant used for transplantation (Figure 6).The product was launched as a solution. A solution isnever the perfect dosage form, because you neverknow if the patient is going to take a little more or a lit-tle less. So they used nanocrytals to include 1mg ac-tive into a tablet. This is an interesting example wherethe nanocrystals really bring something to the prod-uct. But remember, the dose is extremely low.

Food effect is important with this drug. Sirolimus isalso a substrate of cytochrome CYP 3A4 and P-gp.Professor Sugiyama was talking this morning about

their synergistic effects, and here you have a productwhich is sensitive to both. Once again there are in-hibitors that increase the concentration in the blood,while inducers decrease it. In this case, rememberthat if you decrease the blood concentration, as withcyclosporine, you risk graft loss.

Isoretinoin is a suspension (Figure 7). What is im-portant here is not the dissolution – this is a point forall of you to take in, especially those involved in for-mulation. Sometimes, you have to forget the dissolu-tion. What discriminates batches is the particle size.

I will go even further. If you are interested there is adissolution method in the British Pharmacopoeia. Itdoesn't work. It simply doesn't work. Try it, and callme if it works. This is a typical case where the disso-lution method doesn't work. It must be taken withfood. It's also very sensitive to oxidation, so the for-mulation is quite difficult. But if you look at the differ-ence between fed and fasted, bioavailability is roughlyfive times greater.

Figure 8 (slide 15 on the presentation) is a genericversion of isotretinoin. I haven't given the name of thecompany, because the product has not beenlaunched. As it's not yet on the market, the companydid not want me to include its name. It's a genericscompany in Europe. To develop it they used exactlythe same formulation, which is a suspension with sur-factant. They got bioequivalence with Accutane, bythe way, but at a lower dose, which could be impor-tant in terms of cost because isotretinoin is a very ex-pensive active.

What about OTC? You can also do something withan OTC product.

Solufen G is a formulation of 200 mg of ibuprofen,an OTC product (Figure 9). It uses Gelucire 44/14,

• Formulation• Amprenavir 50/150mg• PEG 400• Propylene Glycol• TPGS• SEDDS

• Product Characteristics• TPGS = P-gp inhibitor• Absolute Bioavailability = ?• Metabolized by Cytochrome P.450 CYP 3A4

• Inhibitors: Ritonavir...• Inducers: Saquinavir...

Figure 5.

AGENERASE AMPRENAVIR CAPSULES

• Formulation• Sirolimus 1 mg• PEG 8000• Microcrystalline cellulose• Poloxamer 188• NANOCRYSTALS

• Product Characteristics• Food effect• Substrate of Cytochrome

P.450, CYP 3A4 and P-gp.• Inhibitors: Rifabutin, Carbamazepine...• Inducers: Nicardipine, Metoclopramide...

Figure 6.

RAPAMUNE SIROLIMUS TABLETS

• Formulation• Isotretinoin 10/20/40mg• Hydrogenated soybean• Hydrogenated vegetable oil• Soybean oil• SUSPENSION

• Product Characteristics• Mean Particle size = 90-100 Ķm• Bioavailability = 20%• High inter and intra individual variability• Very sensitive to oxidation• Must be taken with food

• Fed/Fasted• Cmax= 60/300• Tmax= 860/300• AUC= 6080/2400

Figure 7.

ACCUTANE ISOTRETINOIN CAPSULES

133

and you've got a self-emulsifying system with one ex-cipient. I think they were lucky there, because youcannot put more than 200 milligrams into this capsulesize. It has excellent stability, a rapid onset of actionand it is differentiated from the competition, which isquite tough in this segment.

Among the new approaches still under develop-ment, nitric acid-releasing NSAIDs (non-steroidal ant-inflammatories), are simply naproxen linked with nitricoxide. Using a SEDDs system increases the solubilityand bioavailability.

For celecoxib, Pharmacia used a SEDDS with tran-scutol and what they got – and this is important forthis kind of product – was a rapid onset of action.Don't ask me if Pharmacia/Pfizer is going to launchthis product. Even if I knew, I wouldn't tell you, butthey say it works very well. If you look at the patent,you will find something very interesting. The patent de-scribes the formulation strategy and how they came toa SEDDS using a ternary diagram. The selection ofexcipients is also explained in the patent. It is very welldone, it's an excellent rationale.

In the case of Paclitaxel, they decided to put PVPinto a SEDDS formulation, and the result was excel-lent. They increased the solubility and the bioavailabil-ity five times.

Last, but not least, elitriptan is an excellent productfrom Pfizer. This product is very unstable and to havea rapid onset of action you need to create a formula-tion which I would say is different from a normaltablet. So they developed a SEDDS, and what hap-pened was a rapid onset of action with excellentchemical stability, which was not the case with a lot ofthe formulations they had investigated.

Including the active in a polymer complex is alsoan alternative to formulate poorly soluble active. Herewe need to mention cyclodextrin. Cyclodextrin isused a lot for injectables but, orally, there is only onemarketed product, an oral solution of itraconazole inwhich you use cyclodextrin to complex the active.

The poorly soluble active represents a superb mar-ket. There are 100 marketed active compounds withpoor solubility: candidates for reformulation. Look atyour portfolio of products, maybe you have one inthere. The level of 40 to 50 percent poorly solublenew actives is here to stay. And I think Chris Lipinski'smessage was very clear; even if you use the Rule of5 and other filters, it will stay.

There are new technologies available to achieveacceptable absorption: nanoparticles, polymer com-plexes, micro-emulsions and SEDDS.

But don't just take my word for it. Do the same ex-ercise as I did. Look at the patents, look at all the liter-ature. You will see that self-emulsifying micro-emul-sion systems have become an industry reality. We'renot talking about the past, but today reality. Why? Be-cause companies dealing with poorly soluble activesunderstand more and more that they need to investi-gate alternative dosage forms.

Chair, Professor Mitsuru Hashida, KyotoUniversity, Japan: That was Roland's presentation.Now we would like to have an overall questions andanswers discussion, including his presentation andcovering the whole of today. So please raise anyquestions that you have. We won't restrict any topics,so feel free to direct your questions to anybody,please.

Perhaps it will help if we review the scope of dis-cussion a little? A number of topics were covered to-day. We discussed various things, but formulationstudies were the focus of our presentations and dis-cussions. In the morning Dr. Lipinski made a presen-tation, questioning the type of activity chemists are

• Formulation• Isotretinoin 10/20/40mg• Stearoyl macroglyceride (Gelucire 50/13)• Sorbitane oleate (Span 80)l• Soybean oil• SUSPENSION WITH SURFACTANT

• Product Characteristics• Mean Particle size= 90-100 Ķm• Bioequivalence• Generic 16mg= Accutane 20mg• High inter and intra individual variability• Very sensitive to oxidation• Must be taken with food

• Fed/Fasted• Cmax= 60/300• Tmax= 860/300• AUC= 6080/2400

Figure 8.

GENERIC ISOTRETINOIN CAPSULES

• Formulation• Ibuprofen 200 mg• Gelucire 44/14• SEDDS with one excipient!

• Product Characteristics• Excellent stability• Rapid onset• Differentiation

Figure 9.

SOLUFEN G IBUPROFEN CAPSULES

134

engaged in during pharmaceutical development, andasking how chemists, formulation people and alsodevelopment scientists can collaborate with eachother. And to take this idea further, as Dr. Yu did, towhat extent will the formulation study of the future beinvolved in the drug development process? Whatcontribution can the formulation study make to thediscovery stage, for example?

So could anybody with any opinions, any experi-ence to share, please break the ice? Of course, it canbe any other topic.

Dr. Soon-ih Kim, Ono Pharmaceutical Co.Ltd, Osaka, Japan: Earlier, I put a question to Dr.Kusai. Now I would like to ask a question to Dr. Lipin-ski. He said that early formulation should be discour-aged as it makes the chemist's work too relaxed. Ithink that in Japanese companies the specialists informulation studies become rather confused with thattype of statement. So we would like to have Dr. Lipin-ski confirm that statement again, please.

Dr. Christopher A. Lipinski, Pfizer Inc., USA:Yes. I'll state what I said again. At least in the PfizerGroton Laboratories, we actually discourage veryearly discovery formulation work, because there's adanger it will prevent the chemists from doing what istheir job, which is to change the chemistry structureso as to make the compounds better absorbed, moresoluble and more permeable. I recognize that this iscontroversial; I know that there are some companiesthat in fact have early discovery pre-formulation work.

I must say, I influenced what the Groton Laborato-ries did. But it's my opinion that if you have com-pounds that are so insoluble and so impermeable thatyou have to do formulation work in early discovery toget biological activity on those compounds, then per-haps you should consider that you should not evenwork on those kinds of compounds.

Roland Daumesnil, Capsugel Inc., NorthCarolina, USA: I have a question for a group ofspeakers, this afternoon's formulators. If you takeonly the log-P as a criterion, do you think the strat-egy in terms of formulation with a log-P of 3 or a log-P of 6 is the same, or would you approach it differ-ently?

Dr. Lawrence Yu, Director for Science, Of-fice of Generic Drugs, CDER, Food and DrugAdministration, USA: I'll talk about formulation here,as a regulatory scientist. I think that when you look atthe solubility, C log-P, you need to look at how muchdrug is needed to be effective. So what I'm trying tosay is that the question is not that simple. Based on

one parameter, you're saying, here you are, here isone simple criterion you may use. But you have tolook at the overall picture, you have to look at otherparameters such as total dose to make reasonablestrategic decisions.

Now I'd like to comment about something that re-ally worries me very much. Between 1995 or 1997and last year, the cost of drug research and develop-ment increased drastically, at least in the USA. I don'tknow the figures. Meanwhile, the number of ap-provals by the FDA has drastically decreased.

Basically, things are going in opposite directions.On the one hand, expenditure has gone up, on theother hand the number of new molecules approvedby the FDA has declined drastically. I think last yearcosts rose by as much as 40 percent, while newmolecule approvals dropped by over 10 percent. Idon't know the precise figures. I think companiesshould be prepared to do something about it, be-cause we all know these trends cannot continue.

Roland Daumesnil, Capsugel Inc., NorthCarolina, USA: On average, Dr. Yu, over the last fiveor six years the overall number of new molecules peryear has been around 40, so it's quite low. I'm talkingworldwide, not just the USA.

Dr. Lawrence Yu, Director for Science, Of-fice of Generic Drugs, CDER, Food and DrugAdministration, USA: I think there are the sametrends in Japan and Europe. The number of newmolecules approved by regulatory agencies has de-creased drastically. The cause is most anybody'sguess, and I'm sure that people are looking into it. ButI'm also sure that if all of us take a responsible atti-tude, those trends cannot, and will not, continue.

Sototo Iso, Japan: So what is the cause of sucha situation in clinical studies and new molecules? Is itbecause of ADME, or something else?

Dr. Lawrence Yu, Director for Science, Of-fice of Generic Drugs, CDER, Food and DrugAdministration, USA: I really don't know. I thinkADME's contribution has drastically decreased lately. Ithink there have been some break right-off or yellowduster/molecules, but not that many. You know, I'vespent eight years in the industry, and I've only seenone molecule which cannot be delivered.

Question from the audience: To follow up onthat… I was very pleased to hear your remarks thismorning, Yamashita San, about the need for breadthand depth of knowledge, and the need for a multidisci-plinary approach to the problems that we have. Liste-ning to Roland's presentation, we saw things where

135

you have to think about solubility, permeability, P-gp,CYP 3A4 formulation. We have to put all of this toge-ther.

Perhaps one of the reasons that the number of ap-provals is down is that we know more things that cango wrong now than we did some years ago. So weknow what to look for that limits the compound frombecoming approved.

I think that as an industry the multidisciplinary ap-proach, where we start thinking about all of thesethings together, rather than one person working in acubicle over here on toxicity, and another person overthere on solubility, is perhaps one of the answers. Weneed to, I believe, train and educate more generalistswho can consolidate the information from the special-ists and apply it. Of course, I'm a bit biased, but I be-lieve simulation and modeling is about the only way topull all of that together and make sense out of it.

Chair, Professor Mitsuru Hashida, KyotoUniversity, Japan: I completely agree with yourcomments. Maybe somebody from the Japanesecompanies can comment on this issue?

Yano, Ono Pharmaceutical Co., Ltd, Osaka,Japan: I'm one of the chemists working under Dr.Kim. We have to look at both the overall aim and thecosts at the same time. When we are doing synthe-sis, we have to try hard to bring the molecules to thevery end of development. We also have to look at theformulation, and we have to minimize cost and riskboth in the synthesis stage and the formulation stage.I think that is the best method. However, we have notyet found the right approach. It's difficult to find theright approach, and we don't yet know the limitationsof what can be successfully developed in the synthe-sis stage. So those are the constraints. We have toknow what the limit is.

Chair, Professor Mitsuru Hashida, KyotoUniversity, Japan: Thank you very much for yourvaluable contribution. Anyone else – preferably notfrom Ono Pharmaceuticals? Ono Pharmaceuticals'strategy may be somewhat unique. Dr. Kikuchi, youdiscussed the dose number and chemists, for exam-ple. I think you have to explain to chemists about thedose number and how such knowledge is used ob-jectively in synthesis, and generate consensus.

Dr. Hiroshi Kikuchi, Principal Investigator,Daiichi Pharmaceutical Co. Ltd, Japan: Thecompound that I used in the presentation was fromfive years ago, and people handling the synthesis ofcourse became aware of the dose number and theRule of 5. But on the other hand, we had some diffi-

culties – I am doing some soul-searching here – be-cause we sweated so much over solubility that weforgot the importance of membrane permeability. Ifmembrane permeability is high, then formulation maybe successful even with a relatively low solubility. Forquite some time, there was too much stress on solu-bility. But dose number and the Rule of 5 are very wellunderstood by chemists and in early screening weonly have alerts.

In the ninth reformulation approach, I looked atchanging the salt of the crystal. Today, we canchange the crystal salt and if the salt had been HCL itwould have been much easier. But under theJapanese regulations prevailing at the time, it was noteasy to switch from one salt to another. In discoverywe were working to the standards of the Research In-stitute of Material Properties which meant that notonly the chemical stability but the absorption of a saltto be used in crystalline form had to be checked inmonkeys or dogs, and an overall evaluation made be-fore moving ahead. So we made a change to thesystem, but because it was five years ago we had anolder system where we could not change the salt inthe crystal.

If we can cope with the problem by changing thesalt or doing something with the crystalline form, thatshould be tried first before reformulation. The salt is-sue still remains today so far as regulations are con-cerned. Once a certain type of salt is adopted youcan't change it. Is that the case at the FDA, Dr. Yu?

Dr. Lawrence Yu, Director for Science, Of-fice of Generic Drugs, CDER, Food and DrugAdministration, USA: Based on current regulations,a salt is a new molecular entity. In other words, I con-sider a chlorine salt and a free-based one two differ-ent molecular entities. Actually, you can gain market-ing advantage from this sometimes.

Roland Daumesnil, Capsugel Inc., NorthCarolina, USA: Professor Sugiyama, you gave a su-perb presentation as usual. But I'm scared, becauseyou keep on discovering more and more influx/effluxsystems or transporter. Do you know what you aredoing? You're increasing the number of possible druginteractions. Take your time, take your compendium,take the PDR, and look at all these critical molecules.You have pages of, 'don't take this drug if you belongto a particular patient group', or 'don't take that drugwith this product'.

If it continues like this I think that for all products, allmolecules with an efflux system, or where a cy-tochrome metabolism is involved, maybe you willhave to say, I don't want to launch it. I'm sure that

136

there are some pharma companies which are thinkingabout that, because you limit the product's potentialon the market tremendously. I don't know if I'm exag-gerating, but that was the way I understood your su-perb presentation.

Professor Yuichi Sugiyama, University ofTokyo: I don't know whether I should answer thisquestion in English or Japanese. Let me answer inJapanese. Suppose the number of transporters is go-ing to increase in the intestinal tract or the liver… Weneed to go back to the history of cytochrome discov-ery 20 years ago, when more and more cytochromesubtypes were being discovered. Today, this is a to-tally matured research area and our understanding ofit is quite extensive. For instance, if there is a drug-drug interaction, CYP3A4 is the important enzymeand if there is genetic polymorphism then 2D6 and2C19 are the risky CYP isozymes.

Therefore, we are continuing the effort to find newtransporters. This is important because, if we can doit, then in five years' time, even if there are many,many transporters, we will be able to identify and nar-row down the difficult ones. We will be able to statewhich ones incur a drug interaction risk and pinpointthe ones that carry a risk of genetic polymorphism.What's important is to clarify the truth. The most im-portant thing is the scientific value. That is my answer.

This question needs to be posed in Japanese andI hope that the translator will do a good job. I'll speakvery slowly. Well, we have talked with other pharmacompanies and they have confirmed what I am aboutto say. Using excipients, the efflux transporter such asP-glycoprotein is inhibited, and so absorption is en-hanced. This is a good strategy.

However, there is the patent, inclusive of the con-cept of the strategy. What I don't know is the circum-stances in which we would be infringing this patentprotection. If inhibiting efflux enhances absorption, ofcourse to support or to demonstrate that would benot so easy. Are there any occasions when pharmacompanies would be hesitant in doing further re-search on that compound, because of that patentprotection? Since I am a layman in the patent area Iam putting this question to either the Japanese or theoverseas-based pharma companies.

Chair, Professor Mitsuru Hashida, KyotoUniversity, Japan: Thank you very much. Has any-one had the experience Professor Sugiyama de-scribes?

Dr. Soon-ih Kim, Ono Pharmaceutical Co.Ltd, Osaka, Japan: In this company, three years

ago, 3A4 inhibitors were used to increase the plasmaconcentration. We achieved our aim, but found thatthere was an established US patent for that applica-tion. We asked the CRO about the licensing fee andthe royalty was so huge that we dropped the product.That's our experience.

Question from the audience: Well, Dr. Kim, ifthere's good evidence for an activity then of courseyou may be infringing a patent. But it's not that clear.You know, there are many, many excipients and thereare many, many enzymes that inhibit the influx systemand if we say that this function constitutes a patent in-fringement in itself, then maybe they cannot be usedas a substrate. I really do not know how a patent canbe applied to the very wide range of reactions andphenomena in-vivo. We use TPGS.

Professor Yuichi Sugiyama, University ofTokyo: May I just make one final comment? This is aquestion to people working in Japanese and foreignpharmaceutical companies. I have a lot of opportunityto discuss with people in companies about the drugdiscovery process, and they have a good under-standing of PK. Still, I also encounter some othercases that I would like to talk about now, as I wouldlike you to be aware of them.

With the development of LC/MS spectrometrybioavailability screening is carried out extensively inrats, using cassette dosing. Once the IV and oral ad-ministration data is available, we get the BA datathrough that. If you conduct that kind of experimentwith rodents to obtain the BA(bioavailability) values,then if the BA is below 10 percent, how can we inter-pret the results?

Actually such low BA in rodents is not due to thepoor GI absorption in many cases, but rather due tofirst-pass hepatic metabolism or possibly first-passhepatic eliminations, though we have many excep-tions.

There are thus a lot of cases where the BA is verylow due to the extensive hepatic metabolism in ro-dents. Nevertheless, based only on the low BA val-ues, they have tendency to consider that it is merelydue to poor intestinal absorption. However, there existmany data available in which the low BA comes fromthe extensive first-pass hepatic clearance, so youshould also conduct other investigations to know theexact mechanism for the low BA values.

In my experience, even for the drugs with low BAin rodents which come from the extensive first passhepatic metabolism, if you apply that to humans ,then the metabolic activity is very often less than one-

137

tenth of rodents and the BA is high enough. In hu-mans, in other words, hepatic first-pass metabolismis often minimal. But to come to such conclusion, youshould have your IV and oral administration data inyour hands, so I think you should make good use ofthem and make good accurate readings of the datathat is available.

By doing so, you can know the low BA may comefrom the poor GI absorption or from the extensivefirst-pass hepatic metabolism. If you are already doingthat, then please overlook my remarks, but listening tosome of the comments, I have noticed that there arepeople who are rather ignorant about this situation.That's the reason why I wanted to bring it up. For ex-ample, once permeability and solubility have beencorrectly evaluated then you will be able to identify theproblem from that kind of data , as Dr Yu indicated inCase Study 1.

Question from the audience: Yes, I think that'scorrect. There are some predetermined PK analysismethodologies available. So whether the reason is in-solubility or whether it is the permeability in the GItract, when the GI absorption is poor, if you apply it tohumans you cannot expect the extent of absorption

to be higher. But where the BA is low due to an ex-tensive first-pass hepatic metabolism, then in a lot ofcases it is only a very small problem when applied tohumans.

Lawrence Yu, Director for Science, Office ofGeneric Drugs, CDER, Food and Drug Admin-istration, USA: In answer to Dr. Sugiyama's com-ments, I never published this data. The reason I didn'tdo it is because I was trying to see whether a com-puter model is better than animal testing or not, andbecause of other responsibilities I never got round topublishing. I collected about 22 compounds, from2001 and 2002; because I am at the FDA I havethese advantages.

When I speak about correlation from the rat, I'monly talking about bioavailability, I'm not talking aboutdrug disposition. Certainly, animal studies are utilizedfor drug disposition, to understand where the drug islocated. But for absolute bioavailability, the predictivecorrelation from rats to humans is R2, or 0.19, whileit's also about 0.2 from dogs. Actually it's in my com-puter. So what these figures mean is that absolutebioavailability in animals has no predictive value at allfor humans.

139

Closing remarks

HANKS TO ALL OF YOU. I think it was a superb symposium with a great deal of learnings.We know better about the importance of the influx / efflux system and the mutation of the P-gp which can result in substantial changes in PK profiles. We also learnt that even if youapply Lipinski's rule of five properly, the poorly soluble actives are here to stay.

Another interesting input came from the facts that reduction in particle size doesn'tnecessary mean increased absorption. Good in-vitro profile doesn't mean good in-vivoprofile. Good in-vivo results in rats or dogs doesn't mean good in-vivo results on humans.

In other words until you have performed the human clinical trials, do not limit your in-vivo testson one formulation. Test also the alternatives which didn't give you a good profile on animals.

We also learnt that new industrial technologies are now available to achieve an acceptableabsorption.

The last but not the least, considering the dissolution test as a predictor of in-vivo absorptionis wrong. This QC dissolution test doesn't correlate with the in-vivo performance. New testsmust be developed which will become good in-vivo predictor noticeably for class II.

So, thanks to all speakers who made excellent presentations. Thanks to all of you who madethis symposium a very interactive one. Thanks to the translators who did a good job andthanks to the Capsugel team who perfectly organized such a superb day.

I'm going to finish with the chairman, my friend Professor Shinji Yamashita who spent a lot oftime to prepare this symposium and who as usual efficiently chaired it. He also made sure,with elegance that everybody spoke within the allocated time. He deserves a special thankson behalf of all of us.

Roland Daumesnil

T

®

AmericasGreenwood, U.S.A.Capsugel Americas535 North Emerald RoadGreenwood, South Carolina 29646Tel: +1-864-223-2270

+1-800-845-6973Fax: +1-864-942-3848

Puebla, MEXICOPfizer Consumer Health Care México, S. de R.L. de C.V.Division CapsugelPaseo de los Tamarindos no. 40Col. Bosques de las Lomas05120 México, D.F.Tel: +(525) 91 77 31 18

1-800-010 04 63Fax: +(525) 52 57 51 74

Rio de Janeiro, BRAZILCapsugelLaboratórios Pfizer Ltda. Av. das Américas, nº 8.445 / 1.202-1.207 Barra da TijucaRio de Janeiro - RJ Cep : 22793-081 - Brazil Tel: +55-21- 2432-5264

+0800704-2445 Fax: +55-21-2432-5270

InternationalHeadquartersMorris Plains, U.S.A.Pfizer Inc.Capsugel Division201 Tabor RoadMorris Plain, New Jersey 07950Tel: +1-973-385-2000Fax: +1-973-385-2317

AsiaBangkok, THAILANDCapsule (Thailand) Co., Ltd.1/82 M005, Rojana Industrial ParkRojana Road, Tambol KhanhamAmphur U-Thai, Ayutthaya 13210Tel: +6635-334000Fax +6635-227088

Jakarta, INDONESIACapsugelJl Raya Bogor Km. 42P.O. Box 15/CBICibinong 16916, Jawa BaratTel: +62-21-875-2226Fax: +62-21-875-2227

Sagamihara, JAPAN Capsugel3-36 Minamihashimoto4-Chome Sagamihara CityJapan 229-1133Tel: +81-42-700-6700Fax: +81-42-700-6719

Suzhou, CHINASuzhou Capsugel Ltd.Jinjihu RoadSuzhou Industrial ParkSuzhou, Jiangsu Province, 215021Tel: +86-512-6761-6666Fax: +86-512-6761-8807

EuropeBornem, BELGIUMCapsugelRijksweg 11B-2880 BornemTel: +32-3-890-05-11Fax: +32-3-889-26-22

Colmar, FRANCECapsugel10, rue TimkenF-68027 Colmar CedexTel: +33-3-89-20-57-09Fax: +33-3-89-41-48-11

Ploërmel, FRANCECapsugel Z.I.de CamagnonB.P.320F-56803 Ploërmel Cedex Tel: +33-(0)2-97-72-10-00Fax: +33-(0)2-97-72-15-00

AustraliaSydney, AUSTRALIACapsugel32 Cawarra RoadCaringbah, SydneyNSW 2229 AustraliaTel: +61-2-9710-6642Fax: +61-2-9710-6283

Ora

lDel

iver

yof

Poor

lySo

lubl

eA

ctiv

es–

From

Dru

gD

isco

very

tom

arke

ted

Prod

ucts

®

BAS 267 E 2004

oral delivery of poorly soluble actives – from drug discovery to marketed products ·...

Documents