g.l. drusano, m.d. co-director ordway research institute professor and director

Convincing the Pharmaceutical Industry to Use Surrogates for Antibiotic Development:

What is Gained and What is Lost

G.L. Drusano, M.D.Co-Director

Ordway Research InstituteProfessor and Director

Division of Clinical PharmacologyAlbany Medical College

Research PhysicianWadsworth Center, David Axelrod Institute

New York State Department of Health

Surrogates & Antibiotic Development• First, let me state my understanding of “surrogates”• It is important to acknowledge the very nice slide deck

of Dr. Arthur Atkinson posted on the WEB• For antibiotic trials, we can have a surrogate endpoint

as well as a biomarker (or surrogate marker, if you do not mind bucking the following guys, who will be happy to meet you out back)

BIOMARKERS DEFINITIONS WORKING GROUP: BIOMARKERS AND SURROGATE ENDPOINTS: PREFERRED DEFINITIONS AND CONCEPTUAL FRAMEWORK. CLIN PHARMACOL THER 2001;69:89-95.

The surrogate endpoint can be microbiological outcome instead of clinical outcome. The biomarker can be a transformation of the measured drug concentration (e.g. free drug AUC/MIC ratio or free drug Time > MIC)

All of the above is definition driven by statisticians

Surrogates & Antibiotic Development• So, for antibiotic studies, we can measure a

biomarker and we can use a surrogate endpoint• While measuring the biomarker is

straightforward, there are issues about the surrogate endpoint (discussed shortly)

• Why do we do reasonably well with identifying exposure-response relationships in anti-infective trials?

• The answer is that the MIC gives us a normalizing measure for docking the same drug into different receptors (different bacteria)

Surrogates & Antibiotic Development

• For antibiotics, we try to develop relationships between some measure of drug exposure and some measure of effect

• Most often, these are clinical or microbiological outcome, both of which are dichotomous outcome variables


• We have seen Dr. Forrest present a large number of patient studies, BUT were one to perform a literature search, I would expect that the numbers of such studies would differ from the numbers of in vitro and animal studies by several orders of magnitude

• Why the difference?

• First, the former measure colony counts (a continuous variable) while the latter deal with dichotomous outcome variables


• Second, it is MUCH cheaper and faster to perform in vitro and animal studies

• So, what can we gain from the former and what from the latter?

• It is important to remember that the FDA, IDSA and ISAP recently held a workshop that addressed much of this

• Most of the outcome can be addressed with one slide, after which I will guild the lilly

“Closing the Loop”

In-Vitro/ Animal Models

Phase 1 Studies

Phase 2 Studies

Phase 3 Studies

Pre-Clinical Clinical Development

Office of Clinical Pharmacology and Biopharmaceutics Office of Clinical Pharmacology and Biopharmaceutics IDSA/ISAP/FDA Workshop 4/16/04 – IDSA/ISAP/FDA Workshop 4/16/04 – Dr. Charles BonapaceDr. Charles Bonapace

This is what the FDA expects from sponsors

Surrogates & Antibiotic Development• Let us first examine preclinical studies

• At the point of entry into Phase I trials, we are trying to choose a safe and effective dose for Phase II and Phase III trials

• So, what do we need?♦ An exposure target (hence the preclinical studies♦ Population pharmacokinetic information♦ Protein binding data from animals & man♦ Target organism MIC distribution

Journal of Clinical Investigation 2003;112:275-285 &Nature Reviews Microbiology 2004;2:289-300

Drusano GL. Nat Rev Microbiol 2004;2:289-300


Jumbe et al J Clin Invest 2003;112:275-285

0 mg/kg

90 mg/kg

215 mg/kg 600 mg/kg

Peripheral (thigh)Compartment (Cp)

Central Blood Compartment (Cc)IP

injection

kcp kpc

+ Bacteria(XT/R)

f(c)

dCc= kaCa+kpcCp-kcpCc-keCc

dt

ke

dXS=KGS x XS x L - fKS(CcH ) x XS

dtdXR= KGR x XR x L- fKR(Cc

H ) x XR

dt

Kmax CcH

C H

50+CcH

f(CcH)=

Y1=XT=XS+XR

Y2=XR

[4]

[5]

[6]

[7]

[8]

, =K and = S,R

[2]

L = (1- (XR + XS)/POPMAX)

[9]

dCp = kcpCc - kpc Cp

dt

[3]

dCa= -kaCa

dt[1]

KmaxGS

0.117

KmaxGR

0.163

KmaxKS

94.01

KmaxKR

12.16

HKS

6.26

HKR

2.37

C50KS

123.5

C50KR

129.8

KmaxG -maximum growth rate (hr-1) in the presence of drug

KmaxK -maximum kill rate (hr-1)

C50K -drug concentration (g/mL) to decrease kill rate by half

HK -rate of concentration dependent kill

Popmax -maximal population size

Mean Parameter Estimates of the Model.

Popmax = 3.6 x 1010

Jumbe et al J Clin Invest 2003;112:275-285Drusano GL. Nat Rev Microbiol 2004;2:289-300


Pre-Clinical to Phase I/II: What about Resistance?Can this be done in vitro?


The hollow fiber model was described by Blaser and Zinner and employedextensively by Dudley


Tam V et al. Bacterial-population responses to drug selective pressure: Examination of garenoxacin’s effect on Pseudomonas aeruginosa. J Infect Dis 2005;192:420-428

Surrogates & Antibiotic Development P. aeruginosa - Prevention of Amplification of Resistant

Subpopulation• The amplification of the

resistant sub-population is a function of the AUC/MIC ratio

• The response curve is an inverted “U”.

• The AUC/MIC ratio for resistant organism stasis is circa 185/1

Resistant organismsat baseline

All other data points representresistant organism counts at48 hours of therapy

Surrogates & Antibiotic DevelopmentPropspective Validation Study

Tam V et al. Bacterial-population responses to drug selective pressure: Examination of garenoxacin’s effect on Pseudomonas aeruginosa. J Infect Dis 2005;192:420-428


• We can bridge preclinically to Phase II/III through the use of Monte Carlo simulation

• This is the plot of probability for target attainment (resistance-suppression exposure) for a FQ as derived from the preclinical model


Surrogates & Antibiotic Development• The previous data was

for resistance, but straightforward cell kill is also possible

• Later we will see how much kill correlates with what happens in clinical trials

• BUT, please remember that a total drug AUC/MIC = 88 gives a 2 log kill



• We can also perform such studies in Phase II/III, as Dr. Forrest has already shown you

• As an example, a trial of nosocomial pneumonia with the same FQ as in the mouse study will be presented

Population mean pharmacokinetic parameter values derived from 58 Patients with Nosocomial Pneumonia Receiving Levofloxacin as a 1.5 Hour Constant Rate, Intravenous Infusion

Vol Kcp Kpc CLUnits L hr-1 hr-1 L/hr

Means 34.4 7.65 6.07 7.24

Medians 23.3 2.66 0.924 6.24

S.D. 33.5 9.59 12.0 4.36

Vol = Volume of the central compartment; Kcp and Kpc are first order ntercompartmental transfer rate constants connecting the central and peripheral compartments; CL = Total clearance of Levofloxacin

Drusano GL, SL Preston, C Fowler, M Corrado, B Weisinger, J KahnJ Infect Dis. 2004;189:1590-1597.

Final model for microbiological outcome for patients with nosocomial pneumonia receiving levofloxacin daily

Final Model for Microbiological OutcomeConstant Parameter Odds Ratio 95% Confidence Interval for

Odds Ratio (AUC/MIC > 87)

-2.1971.374 3.952 11.596 – 1.347

(Age)0.067 1.069 1.138 - 1.004

p = 0.001; McFadden’s 2 = 0.31



Can we bridge from mouse to man?

Again, the answer is YES, using Monte Carlo Simulation

Surrogates & Antibiotic Development• The previous data was

for resistance, but straightforward cell kill is also possible

• Later we will see how much kill correlates with what happens in clinical trials

• BUT, please remember that a total drug AUC/MIC = 88 gives a 2 log kill


The Reminder Slide

Surrogates & Antibiotic Development• So, the exposure target (AUC/MIC) mediating a

2 log10 cfu/g drop in the mouse is identified as the exposure needed to drive a high probability of a good microbiological outcome in patients with nosocomial pneumonia

• BUT the clinical study was a pneumonia study & the mouse study was mouse thigh infection

• Andes and Craig showed for a FQ that Mouse thigh and mouse lung targets are virtually identical (AAC 2002;46:1665-1670)

• How often does a fixed dose of drug achieve this target?


• So, what do we gain?♦ Information to choose the best dose♦ A high likelihood that the Phase III trial will work (at the cost, that is

worth something!)

• What is there to lose?♦ Nothing that I can see


• Let us examine the clinical trial result just presented and speculate about the future

• We looked at a continuous microbiological variable in the preclinical study and a dichotomous microbiological variable in the clinical study (eradication vs. persistence)

• Is the microbiological outcome from the clinical trial a surrogate outcome?

• According to the working group the answer is yes

EXAMPLES OF LABORATORY ENDPOINTS

THERAPEUTIC BIOMARKER/ CLINICAL CLASS SURROGATE_ OUTCOME ANTIBIOTICS NEG. CULTURE CLINICAL CURE

ANTI-DIABETIC BLOOD GLUCOSE MORBIDITY

LIPID LOWERING DRUGS CHOLESTEROL CAD

DRUGS FOR PROSTATE CA PSA TUMOR RESPONSE

ANTI-HIV DRUGS CD4; VIRAL RNA DELAY AIDS

From Arthur Atkinson, M.D.

Surrogates & Antibiotic Development• I know it is “the law” that approvals are only given on

clinical outcomes, but is it reasonable?• Subpart H does exist

TITLE 21, PART 314, SUBPART H SEC. 314.510 APPROVAL BASED ON A SURROGATE ENDPOINT OR ON AN EFFECT ON A CLINICAL ENDPOINT OTHER THAN SURVIVAL OR IRREVERSIBLE MORBIDITY “FDA MAY GRANT MARKETING APPROVAL FOR A NEW DRUG PRODUCT ON THE BASIS OF ADEQUATE AND WELL-CONTROLLED CLINICAL TRIALS ESTABLISHING THAT THE DRUG PRODUCT HAS AN EFFECT ON A SURROGATE ENDPOINT THAT IS REASONABLY LIKELY … TO PREDICT CLINICAL BENEFIT…”

• AND, the outcome for pharyngitis is COMPLETELY microbiological in nature

• So, can trials for approval be smaller, faster, more informative because they are based on microbiological outcomes and still count for NDA submission/approval

The answer may be addressed in the next talk• But, let us examine the issue


• Yes, I am a true statistical believer and following to a survivorship endpoint may have the most robustness for being sure that the intervention is doing somethingBUT, antibiotics (and antivirals) are a bit different from other drugs

• We are NOT docking the drug with a human receptor, but rather the receptor in the pathogen


• BIOLOGICAL PLAUSIBILITY• EPIDMIOLOGIC EVIDENCE THAT MARKER IS A RISK FACTOR • MARKER MUST BE CONSISTENT WITH PATHOPHYSIOLOGY • MARKER MUST BE ON INTERVENTION PATHWAY • CHANGES IN MARKER REFLECT CHANGES IN PROGNOSIS• ADVERSE DRUG EFFECTS MUST NOT BE CONFOUNDING (again, thanks

to Dr. Atkinson)

• So, let us look at microbiological outcome as a surrogate endpoint

growing organisms in sterile areas is a riskthe pathophysiology has the organism at the center of itthe organism IS the target of the interventionmaking it go away improves prognosisadverse drug effects are confounding TO CLINICAL OUTCOME

• Are there instances where the microbiological effects do not drive outcome clinically?


• Sure – nothing is perfect – macrolides and their effects on inflammation may make things better faster clinically in otitis media

• BUT clinical outcome may also be skewed

• In pneumonia, the impact of infection on gas exchange may be so far advanced that the drug does its job perfectly (eradicate organisms) but the patient dies of ventilatory failure

• This is a physiological, NOT drug failure


• If the drug does what it is supposed to do can we scientifically impute failure?

• Currently, this is not THAT big a deal, as clinical outcome and microbiological outcome are tightly intertwined because of the way they are measured

• Eradication is frequently assessed as a good clinical outcome combined with no chance of primary infection site resampling

• BUT, it may not always be so

Surrogates & Antibiotic Development• PCR tests, antigen tests, quantitative imaging

testing or other methodologies have the possibility of having a quantitative resampling microbiological outcome

• Dr. Forrest has published a study in which the lower respiratory tract was resampled sequentially

• Dr. Paul Ambrose has published a quantitative resampling study of sinusitis

• The future may be now

Surrogates & Antibiotic Development• Look at the results in HIV disease

• MAJOR advances came about only AFTER quantitative RNA PCR was identified as a valid surrogate endpoint

• The conversation needs to begin for antibacterials


• The insistence on clinical endpoints instead of microbiological endpoints is driven by two things:1) Many statisticians are nihilists – see the silliness of intent-to-treat2) There is a confounding of safety with effectiveness


• I believe that we CAN safely use prior information to conduct our clinical trials (In this I am a Bayesian)

• Nihilistic ITT statisticians say that we need to be ULTRA sure of the validity of the results and, so biological plausibility and prior knowledge need to go out the window

• The question is: How many times do we reject the true outcome because of statistical nihilism?


• What causes more harm?

• The other issue is confounding endpoints of safety and effect

• It is true that in HIV disease, patients with MAC had a higher death rate with high dose clarithromycin

• WHAT IS THE TAKE HOME FROM THIS?

• In my view, the outcomes should be viewed separately


• First, the dose response should be evaluated to ascertain which clarithromycin dose is most effective

• Second, a safety assessment should be done

• Here, we would conclude that there were some, but marginal differences in the microbiological activity of high dose clarithromycin

• However, on the safety evaluation, it had more deaths


• Conclusion: The loss of safety with high dose clarithromycin outweighed the minor increase in effect

• Should a microbiological endpoint ALWAYS be used to measure effect?ANSWER: NO!

• There are certain instances where there is a high cure rate with no therapy (Polyanna Effect)

• Use of Time-To-Event analysis with measures of “How do you feel?” endpoints are reasonable


• We all know about sinusitis and AEBCB and (in kids) otitis media

• Sujata Bhavnani, Paul Ambrose and I were recently VERY surprised by the analysis of a CAP trial


Note 70% probability of cure at zero drug exposure

These patients were all Fine 1’s and 2’s


• In such circumstances, “is your cough or sinus pain or ear ache better?” is a perfectly reasonable approach tied to a time-to-event analysis

• On the other hand, in patients with VAP with destroyed lung from Pseudomonas exotoxin A getting Ceph du jour, asking about feelings may be impeded by the ET tube and asking the antibiotic to regrow the lungs will be like Waiting for Godot AND IS NOT REASONABLE

Surrogates & Antibiotic Development• We have new tools and better understanding

that allows better, more scientific inferences to be drawn without exaggerating risk

• There are times when evaluation of “how do you feel?” is the most appropriate approach (and makes a heck of a lot more sense than waiting 10-14 days after the sentinel event takes place to make the evaluation)

• In sick patients, the microbiological endpoint, particularly with resampling makes more sense

Surrogates & Antibiotic Development• In ALL instances, the safety evaluation remains

key, but should be done separately

• Ultimate inferences about the drug, dose and schedule for the indication need to balance safety and effectiveness

• Let the discussion begin with the regulatory agencies

g.l. drusano, m.d. co-director ordway research institute professor and director

Documents

antibiotic studies

studies phase

clinical outcome

measure of drug exposure

webfor antibiotic trials

microbiological outcome

surrogate endpoints

surrogate endpointwhile