carvedilol and the food and drug administration approval process: an introduction
TRANSCRIPT
Carvedilol and the Food and DrugAdministration Approval Process:An Introduction
Lloyd D. Fisher, PhD, and Lemuel A. Moye, MD, PhDDepartment of Biostatistics, University of Washington, Seattle, Washington; and Universityof Texas School of Public Health, Houston, Texas
ABSTRACT: We discuss briefly the new drug carvedilol (Coregt), a beta-blocker, alpha-blocker,and antioxidant. This drug was developed for congestive heart failure in a series of trials,four in the United States and one in Australia and New Zealand, briefly summarized inthis document. We also summarize the classical paradigm of the U.S. Food and DrugAdministration (FDA) for drug approval and the FDA’s use of advisory committees.This document serves as background to the discussion of carvedilol’s approval. Con-trolled Clin Trials 1999;20:1–15 Elsevier Science Inc. 1999
KEY WORDS: FDA, advisory committee, carvedilol, beta-blocker, alpha-blocker, antioxidant, randomizedclinical trial
INTRODUCTION
Carvedilol is a drug that recently received approval from the U.S. Food andDrug Administration (FDA). The data, approval process, and issues raisedwere relatively unusual. They brought starkly to the forefront fundamentalissues in the application of hypothesis testing and strength of evidence. Weparticipated in this process and had strong opposite points of view which wegive in the accompanying articles [1, 2]; this joint introduction gives backgroundfor those articles. There were many issues involved in the discussions, and weomit from this introduction some issues not in the following articles. Thesearticles have several purposes: to debate the appropriateness of a rigid adher-ence to the hypothesis-testing paradigm; to give simplified background aboutthe usual FDA approval process; and to illustrate the fundamental role thatbiostatistical considerations can play in evaluating new drugs, biologics, anddevices. To begin with relevant background, we first discuss the illness treated,congestive heart failure (CHF); second, the drug under consideration (knownas carvedilol generically or by its brand name Coregw); and third, the usual FDA
Address reprint requests to: Dr. Fisher, Dept. of Biostatistics, Box 357232, University of Washington,Seattle, WA 98195-7232.
Received April 9, 1998; accepted September 16, 1998.
Controlled Clinical Trials 20:1–15 (1999) Elsevier Science Inc. 1999 0197-2456/99/$–see front matter655 Avenue of the Americas, New York, NY 10010 PII S0197-2456(98)00052-X
2 L.D. Fisher and L.A. Moye
paradigm for drug approval. A discussion of the use of Advisory Committees bythe FDA follows.
This background information sets the stage for a description of the clinicaldevelopment program for the compound and the FDA approval. There weretwo advisory committee meetings that ended with recommendations for theFDA. We must also state for the record that Lloyd Fisher was a paid consultantfor the sponsor, SmithKline Beecham Pharmaceuticals. He had previouslyserved two terms (8 years) as the biostatistician on the Cardiovascular andRenal Drugs Advisory Committee and also reviewed new drug applications(NDA), on an outside appointment to the FDA, for 10 years before this experi-ence. Dr. Moye was the biostatistician on the committee during these delibera-tions (and still serves in that role).
CONGESTIVE HEART FAILURE
CHF results when the heart cannot produce enough blood circulation tomaintain the body in its normal state. Several factors can cause it. For example,a heart attack that kills a portion of the heart muscle diminishes the heart’sability to pump blood. This failure can lead to the buildup of fluid in the bodyand extremities; one standard treatment is thus with diuretic drugs that helpthe body eliminate fluid. Several drugs have been proven useful in heart failure.Angiotensin-converting enzyme inhibitors (ACE inhibitors) have been shownto prolong life, and digoxin reduces the combined incidence of death andhospitalization. Thus, patients are often treated with the triple therapy of diuret-ics, digitalis, and ACE inhibitors. The amount of impairment due to CHF rangesfrom none after appropriate compensation by drugs to the patient’s beingtotally bedridden and incapable of any normal functioning. The most commonsymptoms are shortness of breath (dyspnea) and fatigue. One standard measureof the disease’s severity is the subject’s ability to exercise.
CHF is a problem of tremendous societal importance. There is a prevalenceof approximately 4 million patients in the United States and 6% of individualsover the age of 65. In 1992 it was the single most costly medical problem inthe United States, using 5.4% of the federal health care budget.
CARVEDILOL, OR COREG
Carvedilol is unusual in that it has a number of activities that are possiblybeneficial. The rationale for its use is given in the quote below from the briefingdocument that the advisory committee received. It relates the drug’s activityto the hormones that trigger our “fight or flight” reaction and increase theactivity of our cardiovascular system:
Activation of the sympathetic nervous system is one of the most importantpathophysiologic abnormalities in chronic heart failure. Circulating levelsof catecholamines are increased in this disorder in proportion to the severityof the disease, and patients with the highest plasma levels of norepinephrinehave the most unfavorable prognosis. These observations have lead to thehypothesis that sympathetic activation has a major influence on the clinicalstatus of patients with heart failure and plays an important role in determin-ing the rate of progression of the underlying disease. Norepinephrine can
Carvedilol and the FDA Process: Introduction 3
exert adverse effects on the circulation, both directly and indirectly, andinterference with its actions can ameliorate the pathophysiologic abnormali-ties of heart failure and can retard progression of disease in experimentalmodels of left ventricular function.
Drugs that interfere with the sympathetic nervous system, by acting ona-, b1, or b2 receptors, or preferably all three receptors, might be expectedto block these deleterious effects. Alpha-receptor blockers can exert periph-eral and coronary vasodilator effects and may have favorable effects onmyocardial hypertrophy and arrhythmias. Beta-receptors (particularly non-selective agents) can interfere with the toxic effects of catecholamines, mayprevent and reverse the structural changes that occur during the progressionof heart failure, and may reduce the contribution of arrhythmias.
These observations have led investigators to propose the beta-adrenergicblocking drug might be useful in the management of heart failure. Beta-blockers were previously considered to be contraindicated in the disorderbecause of their propensity to produce short-term effects, but studies initi-ated in Sweden more than 20 years ago raised the possibility that long-termtherapy with these drugs might produce hemodynamic and clinical benefits.Controlled trials with several different beta-blockers have shown that thesedrugs can be beneficial in patients with heart failure.
As one would suspect by the rationale above, carvedilol is an a-, b1, and b2
blocker. In addition to the mechanisms of perceived potential benefit, carvedilolis one of the most potent antioxidants ever found. Antioxidant activity (suchas possessed by vitamins C, E, and beta-carotene) may have long-term benefiton the development and progression of coronary artery disease. This is, how-ever, not thought likely to be an important mechanism of action in the settingof CHF.
USUAL FDA PARADIGM FOR DRUG APPROVAL
To have some measure of predictability and order in drug approval, sponsorsof new drugs, biologics, or devices and the FDA must both have a mutualunderstanding of what is expected for approval. (A biologic is a substance thatoccurs naturally in the human body, such as insulin). Because we are studyingthe drug carvedilol, our further discussion refers to drugs; however, the princi-ples are the same for biologics and devices. Consider a situation in which theend point or outcome variable for a favorable drug effect ethically allowsreplication of the results. For approval in this case, a sponsor normally needsto have two randomized clinical trials, double-blind where possible, each ofwhich shows statistically significant benefit of the drug being studied. Thistwo-study paradigm is interpreted in the total experiment context of the drugdevelopment program. If there are more than two studies, then the totalityand consistency of the data are examined. Further, the downside of the drug(usually adverse events or “side-effects”) is examined to ensure a favorablerisk-to-benefit ratio. For rational use of the drug, one needs to provide a labelfor the drug giving physicians information concerning the correct dose ordoses, possible interactions with other drugs commonly used in the illness orphysiologic state (called the indication for use) being studied, and a carefulprecise description of which patients may reasonably be expected to benefitfrom the drug. Still, all in all, one first looks for two positive trials as minimal
4 L.D. Fisher and L.A. Moye
evidence of drug effectiveness for a regulatory decision. Note that the one-sided p value associated with two such trials is 0.025 3 0.025 5 0.000625, andthe corresponding two-sided p value is 0.00125.
As stated above, we need objective and commonly understood standardsto allow industry to plan drug development and decision making with a reason-able prospect of success if all works out as hoped. Having standards understoodby both sides also allows the agency (the FDA) not to approve a drug submittedto it. This introduces a fairness into the process.
FDA ADVISORY COMMITTEES
The FDA often uses committees to advise them on the approvability anduse of specific drugs, biologics, and devices and on conceptual issues associatedwith such approval. The agency does not bring all compounds to advisorycommittees, but the division evaluating a drug decides whether or not to solicitinput from one. Different divisions of the FDA use their committees in differentways, but among the more common issues on which the committees meet arethe introduction of the first drug in a new drug class, a new clinical indicationfor an existing drug if it is the first such approval in a drug class, when thereis a “close call” or disagreement within the agency, and the introduction intothe public domain of reasonable standards and methods for developing certainclasses of drugs.
Before considering a drug at such a meeting, the committee members typi-cally receive a briefing document prepared by the sponsoring drug company(in this case, SmithKline Beecham); FDA review documents, both clinical andbiostatistical, or one document integrating both types of review; and questionsto be asked of the committee at the meeting. The questions direct the committeeto issues and questions considered important by the agency. The carvedilolmeeting questions are given in Appendices A and B as further background tothe considerations at the two advisory committee meetings for carvedilol. Bydesign, the committee members deliberate in public and are not allowed totalk to one another about the upcoming review before the meeting. The agencystaff makes this clear and certainly do not allow such discussions in theirpresence (e.g., when a review is to take place the second day of the meeting).If a committee member has questions that the sponsor could answer, she orhe can contact the sponsor through the executive secretary of the committee;the sponsor usually answers promptly. The agency requires all committeemembers to report conflicts of interest and can, and does as deemed appropriate,disqualify members either from voting or from voting and participating in thedeliberations. Conflicts of interest can arise not only from involvement withthe sponsor but also from involvement with a company with a competing drug.More minor conflicts of interest can be announced but still allow a committeemember full participation in deliberations and voting if it is deemed in thepublic interest. Public announcements about the conflicts are made at the begin-ning of the meetings; the meetings are announced in the Federal Register andare open to all interested parties. In addition to representatives of patientgroups and medical and commercial members of the audience, stock analystsoften attend.
Carvedilol and the FDA Process: Introduction 5
Tab
le1
End
Poin
tsfo
rth
eC
arve
dilo
lSt
udie
sE
ndpo
int
Des
crip
tion
Com
men
ts
Mor
talit
yA
llca
use
mor
talit
y.6-
min
corr
idor
wal
kd
ista
nce
Rec
ord
edis
the
dis
tanc
eth
atca
nbe
wal
ked
As
CH
Fim
pair
sth
eab
ility
toex
erci
se,a
nin
crea
sein
6m
in.
inex
erci
seab
ility
isco
nsid
ered
bene
fici
alto
pati
ents
;atl
east
inth
esh
ortt
erm
,bet
a-bl
ocki
ngd
rugs
low
erth
em
axim
alex
erci
sere
spon
se.
9-m
intr
ead
mill
dis
tanc
eT
hed
ista
nce
wal
ked
ata
cons
tant
spee
don
atr
ead
mill
in9
min
.M
axim
alex
erci
seto
lera
nce
Tim
eon
atr
ead
mill
,ac
cord
ing
toa
def
ined
prot
ocol
,unt
ilth
epa
tien
tca
nnot
cont
inue
any
long
er.
New
Yor
kH
eart
Ass
ocia
tion
CH
FC
lass
I—N
olim
itia
tion
:or
din
ary
phys
ical
Thi
scl
assi
fica
tion
sche
me,
alth
ough
rela
tive
lyfu
ncti
onal
clas
s(N
YH
Acl
ass)
acti
vity
doe
sno
tca
use
und
uefa
tigu
e,cr
ude,
cont
ains
cons
ider
able
info
rmat
ion
dys
pnea
,or
palp
itat
ion.
abou
tth
epa
tien
t’ssu
rviv
alst
atus
and
the
Cla
ssII
—Sl
ight
limit
atio
nof
phys
ical
acti
vity
:ab
ility
tod
oph
ysic
alac
tivi
ty.
Itis
the
mos
tsu
chpa
tien
tsar
eco
mfo
rtab
leat
rest
.co
mm
only
used
clas
sifi
cati
onsc
hem
efo
rC
HF.
Ord
inar
yph
ysic
alac
tivi
tyre
sult
sin
fati
gue,
dys
pnea
,or
angi
na.
Cla
ssII
I—M
arke
dlim
itat
ion
ofph
ysic
alac
tivi
ty:
alth
ough
pati
ents
are
com
fort
able
atre
st,
less
than
ord
inar
yac
tivi
tyw
illle
adto
sym
ptom
s.C
lass
IV—
Inab
ility
toca
rry
onan
yph
ysic
alac
tivi
tyw
itho
utdi
scom
fort
:sy
mpt
oms
ofC
HF
are
pres
ent
even
atre
st.
Wit
han
yph
ysic
alac
tivi
ty,
pati
ent
expe
rien
ces
incr
ease
dd
isco
mfo
rt.
Car
dio
vasc
ular
hosp
ital
izat
ions
Hos
pita
lizat
ion
for
aca
rdio
vasc
ular
caus
e.Pa
tien
t’sgl
obal
asse
ssm
ent
Poss
ible
answ
ers:
mar
ked
lyim
prov
ed,
Such
‘‘glo
bal
asse
ssm
ents
’’in
man
yse
ttin
gsar
em
oder
atel
yim
prov
ed,m
ildly
impr
oved
,no
inse
nsit
ive
mea
sure
s.U
sing
them
,it
isof
ten
chan
ge,
slig
htly
wor
se,
mod
erat
ely
wor
se,
dif
ficu
ltto
show
trea
tmen
tch
ange
s.an
dm
arke
dly
wor
se.
(con
tinu
ed)
6 L.D. Fisher and L.A. Moye
Tab
le1
(con
tinu
ed)
End
poin
tD
escr
ipti
onC
omm
ents
Phys
icia
n’s
glob
alas
sess
men
tSa
me
asab
ove.
Sam
eas
abov
e.L
eft
vent
ricu
lar
ejec
tion
frac
tion
,L
VE
FT
hees
tim
ated
prop
orti
onof
bloo
dpu
mpe
dL
VE
Fan
dN
YH
Acl
ass
are
extr
emel
ypr
ogno
stic
orE
Fou
tof
the
left
vent
rica
l(t
hepu
mpi
ngin
dic
ator
sfo
rsu
rviv
al;
how
ever
,d
rugs
that
cham
ber
ofth
ehe
art)
dur
ing
ahe
art
beat
.ch
ange
LV
EF
wit
ha
resu
ltin
gch
ange
insu
rviv
alha
veno
tbe
enad
equa
tely
stud
ied
toal
low
itto
bem
ore
than
asu
rrog
ate
outc
ome
vari
able
.L
VE
Fis
anex
trem
ely
impo
rtan
tsu
rrog
ate
end
poin
tat
this
tim
e.M
inne
sota
Liv
ing
wit
hH
eart
Failu
reA
ques
tion
nair
esp
ecif
ical
lyd
esig
ned
tosh
owQ
uest
ionn
aire
chan
gein
CH
Fsy
mpt
omat
olog
y.It
had
not
prev
ious
lybe
enus
edas
apr
imar
yen
dpo
int
befo
reth
eca
rved
ilol
prog
ram
.
Carvedilol and the FDA Process: Introduction 7
Tab
le2
Sum
mar
yof
Car
ved
ilol
Stud
ies
Plan
ned
Pati
ent
Follo
w-u
pSt
udy
Sele
cted
Pati
ent
Elig
ibili
tyC
rite
ria
Prim
ary
End
Poin
t(m
o)
U.S
.Pro
gram
220
LV
EF
<0.
35N
YH
Acl
ass
II–I
VC
hang
ein
exer
cise
dis
tanc
efr
omba
selin
e.B
oth
6-m
inco
rrid
or6
CH
F6-
min
corr
idor
wal
kw
alk
dis
tanc
ean
d9-
min
trea
dm
illte
st.
dis
tanc
e15
0–45
0m
221
LV
EF
<0.
35N
YH
Acl
ass
II–I
VC
hang
ein
exer
cise
dis
tanc
efr
omba
selin
e.B
oth
6-m
inco
rrid
or6
CH
F6-
min
corr
idor
wal
kw
alk
dis
tanc
ean
d9-
min
trea
dm
illte
st.
dis
tanc
e15
0–45
0m
239
LV
EF
<0.
35N
YH
Acl
ass
II–I
VC
hang
ein
exer
cise
dis
tanc
efr
omba
selin
ean
dM
inne
sota
qual
ity
6C
HF
6-m
inco
rrid
orw
alk
oflif
esc
ore.
dis
tanc
e,
150
m24
0L
VE
F<
0.35
NY
HA
clas
sII
–IV
Clin
ical
prog
ress
ion
ofhe
art
failu
re,
def
ined
asd
eath
due
to12
CH
F6-
min
corr
idor
wal
kC
HF,
orho
spit
aliz
atio
nfo
rC
HF,
orw
orse
ning
CH
Fre
quir
ing
dis
tanc
e.
450–
550
min
crea
sing
back
grou
ndm
edic
atio
nsby
50%
for
atle
ast3
0d
ays.
Aus
tral
iaN
ewZ
eala
ndSt
udy
223 E
arly
phas
eH
isto
ryof
dys
pnea
atre
stor
Max
imal
exer
cise
tole
ranc
eon
trea
dm
illpl
ustw
ohe
mod
ynam
ic6
fati
gue
atre
stw
ith
EF
,0.
45m
easu
rem
ents
(LV
EF
and
the
LV
inte
rnal
dim
ensi
ons)
.N
YH
Acl
ass
IVw
asex
clud
edL
ater
phas
eSa
me
asab
ove.
All
caus
ed
eath
and
all
caus
eho
spit
aliz
atio
n.(T
his
was
only
18–2
4cl
arif
ied
late
rin
the
stud
y.)
(con
tinu
ed)
8 L.D. Fisher and L.A. Moye
Tab
le2
(con
tinu
ed)
Num
ber
Ran
dom
ized
and
Impo
rtan
tSe
cond
ary
End
Poin
tsD
rug
Dos
age
Use
d
U.S
.Pro
gram
Phys
icia
nan
dpa
tien
tgl
obal
asse
ssm
ent;
NY
HA
func
tion
alcl
ass;
CH
Fsy
mpt
omp
584
;c6.
25m
g5
83;
220
scor
e;ca
rdio
vasc
ular
hosp
ital
izat
ions
;EF;
9-m
inw
alk
dis
tanc
e;qu
alit
yof
life
c12
.5m
g5
89;
scor
es.
c25
mg
589
221
Phys
icia
nan
dpa
tien
tgl
obal
asse
ssm
ent;
NY
HA
func
tion
alcl
ass;
CH
Fsy
mpt
omp
514
5sc
ore;
EF;
qual
ity
oflif
esc
ores
.c
25m
g5
133
239
Min
neso
taL
ivin
gw
ith
Hea
rtFa
ilure
Que
stio
nnai
reas
am
easu
reof
the
Qua
lity
p5
35of
Lif
e.c
25m
g5
7024
0Ph
ysic
ian
and
pati
ent
glob
alas
sess
men
t;N
YH
Afu
ncti
onal
clas
s;C
HF
sym
ptom
p5
134
scor
e;ca
rdio
vasc
ular
hosp
ital
izat
ions
;EF;
9-m
inw
alk
dis
tanc
e;qu
alit
yof
life
c25
mg
523
2sc
ores
.
Aus
tral
iaN
ewZ
eala
ndSt
udy
223 E
arly
phas
eN
YH
Afu
ncti
onal
clas
sp
520
8C
linic
alst
atus
(doi
ngve
ryw
ell,
wel
l,fa
iror
poor
)c
520
7L
ater
phas
eN
/A
(All
dat
aea
rly
and
late
incl
uded
inlo
nger
term
follo
w-u
pan
alys
es.)
p5
208
c5
207
p,pl
aceb
o;c,
carv
edilo
l.
Carvedilol and the FDA Process: Introduction 9
COREG (CARVEDILOL) STUDIES
We do not discuss here all of the carvedilol data but only the primary trialsthe sponsor offered as proof of efficacy and a satisfactory risk-to-benefit ratio.At the time the carvedilol trials were designed, the standard end point fordrugs being studied in CHF was improvement in ability to exercise. Onestandard test is to see how far the subjects can walk during a prespecified timeinterval. The sponsor considered mortality as an end point, but sample-sizecalculations suggested studies that were too large and expensive. Further,although some persons viewed the potential for a survival benefit as possibleor even likely, others viewed it as fairly remote. Beta-blocking drugs will bluntexercise response in the short term (because the body is less able to stimulatethe cardiovascular system), and therefore ability to exercise would not seemthe most promising end point for such drugs, at least in the short term. Never-theless, the main thrust of the drug development program was to use exerciseas the primary end point with one exception in the United States. Very briefdescriptions of the end points (Table 1) and of the important studies (Table 2)follow. The program in the United States had four component studies identifiedas studies 240, 221, 220, and 239. Table 2 summarizes the design characteristicsand other features and uses a number of secondary end points that we describeand summarize in Table 1.
REFERENCES1. Fisher LD. Carvedilol and the Food and Drug Administration (FDA) approval
process: the FDA paradigm and reflections upon hypothesis testing. Controlled ClinTrials 1999;20:16–39.
2. Moye L. Endpoint interpretation in clinical trials: the case for discipline. ControlledClin Trials 1999;20:40–49.
APPENDIX A
FDA Questions, 2 May 1996, for the Cardiovascular and Renal DrugsAdvisory Committee (with slight editorial changes)
COREGe (carvedilol), an approved agent for the treatment of hypertension,has been investigated as a treatment for moderate (90% New York Heart Associ-ation, NYHA, II–III) chronic congestive heart failure. In support of this indica-tion, the sponsor has submitted the results of four well-controlled U.S. multicen-ter studies plus a well-controlled Australia–New Zealand multicenter study ina population with less severe disease. In addition, there are three shorter termand smaller double-blind, placebo-controlled, single-center studies. If carvedilolwas to be approved for improvement of morbidity in congestive heart failure,it would be the first agent with that claim.
The Advisory Committee has received summaries of these data in the formof a briefing package prepared by the sponsor and a medical and statisticalreview. The two are in substantial agreement with regard to what happenedin these trials; the issues relate to how the results should be interpreted.
In discussion of the findings from these studies, it will be important todistinguish which analyses attributed worst rank to subjects who died or with-drew and which analyses propagated the last measured value. One of thequestions asks for a comparison of the two approaches.
10 L.D. Fisher and L.A. Moye
The carvedilol development program collected much clinically relevant data,and much of that data, viewed broadly, is supportive to benefit. The problemis the identification of specific benefits a patient with congestive heart failurecan expect to receive. If it is only in the global gestalt that the benefits aremanifest (i.e., if no single benefit is established to conventional standards ofproof), then what should the indications be for use?
1. Was there a mortality effect in study 220? If so,1.1. Describe the statistical analysis that leads to that conclusion, including
1.1.1. Correction of the estimated p value for multiple interimanalyses,
1.1.2. Correction of the estimated p value for multiple prespecifiedprimary and secondary endpoints, and
1.1.3. Reasons for ignoring at least one potential analysis (retrospec-tively conducted by the FDA: page 21 of the combined medical/statistical review) that denies a treatment effect (p . 0.05 fora random-effects model analysis of mortality plus near-mor-tal events).
1.2. Was this mortality effect confirmed by other studies individually(particularly study 221) or in combination?
1.3. Was such an effect expected on the basis of known effects of otheragents in this pharmacologic class or this agent in other settings?
1.4. If carvedilol were approved for the treatment of congestive heartfailure, should its approval include a claim for prolonging life? If not,what should be done to obtain such a claim?
Unlike mortality, worsening heart failure (a combined end point of sudden-or heart failure-related death, heart failure-related hospitalization, or increasedheart failure medications) was a primary hypothesis tested during the develop-ment program in study 240. The sponsor and reviewers performed very differ-ent analyses, neither of which were exactly what was prespecified.
2. Was there an effect of treatment on worsening heart failure in study 240?2.1. The sponsor’s analysis of study 240 indicated a treatment effect that
was highly statistically significant (p 5 0.003).2.1.1. Was the analysis plan a reasonable one?2.1.2. Was the analysis amenable to verification?2.1.3. How were missing medications data handled?
2.2. The reviewers analysis of study 240 (page 97 of the combined medi-cal/statistical review) indicated a treatment effect that was marginallystatistically significant (p 5 0.04). However, missing medications dataled to exclusion of 60% of subjects from the analysis. Was this areasonable analysis plan? If so, what weight should be given anapparent treatment effect when so large a fraction of the study popula-tion was excluded from the analysis?
2.3. Were the findings in study 240 confirmed by other studies? Cite theconfirming studies and analyses.
2.4. If carvedilol were approved for the treatment of congestive heartfailure, should its approval allow a claim for reducing the rate of
Carvedilol and the FDA Process: Introduction 11
progression of chronic moderate heart failure? If not, what shouldbe done to obtain such an indication?
3. The following are other prospectively defined primary or secondary endpoints for the five multicenter efficacy studies:
• Exercise tolerance by 6-min corridor walk test• Exercise tolerance by 9-min treadmill test• Quality of life (21-question set)• NYHA classification• Ejection fraction• Subjects’ global assessment (single question)• Investigators’ global assessment (single question)• Total cardiovascular hospitalizations• Hospitalizations for heart failure• Heart failure symptoms (7-question set)• Cardithoracic ratio
3.1. For each, address the relative merits of analyses based on3.1.1. Last value carried forward;3.1.2. Assignment of a low score to deaths and withdrawals.
3.2. For which end points are the data to support a benefit of treatmentsufficiently compelling to warrant inclusion?3.2.1. As part of the indications for use?3.2.2. As part of the clinical pharmacology?
3.3. If any of the secondary end points were thought to show a compellingtreatment benefit,3.3.1. Was any adjustment necessary because of failure to show an
effect of treatment on the cited studies’ primary end points?3.3.2. Was any adjustment necessary because of the multiplicity of
secondary end points?3.3.3. Was any adjustment necessary because the blind was compro-
mised by treatment effects on vital signs?4. If carvedilol were to be approved for use in heart failure,
4.1. What initial dose should be recommended? Cite the rationale anddata to support a choice.
4.2. What target dose should be recommended? Cite the rationale anddata to support a choice.
5. Should carvedilol be approved for the treatment of congestive heart fail-ure? If so,5.1. Should its use be restricted to patients with symptomatic heart
failure?5.2. What should be said about its use in patients with NYHA class IV
heart failure?5.3. Are there other characteristics of populations that should be specifi-
cally included or excluded?6. Comment on the following as strategies for drug development programs
in chronic congestive heart failure. For each, what are the implicationsfor the certainty with which one will likely know a real benefit exits?
12 L.D. Fisher and L.A. Moye
• Prespecify primary and secondary end points. Use secondary endpoints only in interpreting treatment effects on primary end points.
• In round one, prespecify every end point of potential interest. Use theend points for which a treatment effect appears evident as primaryend points in confirming studies.
• Prespecify every end point of potential interest. Seek approval for thosedemonstrating a benefit of treatment.
APPENDIX B
FDA Questions, 27 February 1997, for the Cardiovascular and Renal DrugsAdvisory Committee (with slight editorial changes)
In May 1996, after consideration of the data and analyses that were availableat that time, the Advisory Committee recommended nonapproval of carvedilolfor the treatment of heart failure. Since that time, long-term follow-up fromone study (study 223) and new analyses of all multicenter studies have becomeavailable. These new data are all, to some degree, supportive of benefit fromtreatment with carvedilol. This matter is once again brought to the AdvisoryCommittee, so that all available data can be brought to bear on the final recom-mendation of the Committee.
The Advisory Committee’s initial decision was based in part on the positionthat one cannot reach definitive conclusions about secondary end points froma study that fails to demonstrate effectiveness using its primary end point. Thisposition requires careful consideration, as it is consistent neither with pastAgency actions nor past Advisory Committee recommendations. For example,in 1993, the Advisory Committee recommended approval of enalapril for thetreatment of asymptomatic left ventricular dysfunction on the basis of a singletrial in which there was little evidence for support of benefit on the primaryend point, mortality (p 5 0.3), but fairly strong evidence for a benefit of oneof eight prespecified secondary end points, time to first hospitalization forcongestive heart failure (p , 0.001, with no statistically significant differencefor all-cause hospitalizations). The Advisory Committee’s recommendation ofapproval with regard to enalapril was somewhat controversial, but it wassustained by Agency action; prevention of hospitalization for congestive heartfailure was added to the Indications section of enalapril’s labeling.
In reconsidering carvedilol, the Committee is reminded that federal lawpertaining to approval simply calls for evidence of effectiveness, from adequateand well-controlled clinical trials, that is convincing to experts. Regulations donot specify two-sided p , 0.05 on the primary end points in each of two studies,and some Agency approval decisions involving end points of irreversible harmor rare conditions have not had two such studies. Regulations also do notspecify that the end points in such studies be the same. One effect of relyingon 2 (out of 2) studies with p , 0.05 on their primary end points is that thelikelihood of incorrectly identifying a treatment benefit is ,0.0025 [sic]. Thedegree of clinical confidence in the treatment benefit is often further enhancedby the observation of apparent treatment effects on prespecified secondary endpoints and other study data, where these observations support a mechanism
Carvedilol and the FDA Process: Introduction 13
of action or are otherwise expected in association with favorable effects on theprimary end point.
The questions that follow assume that the Committee does not wish torecommend a change in the standard of approval, in the sense of specifying adifferent level of type I error rate. Instead, the questions are intended to solicitthe Committee’s judgment about how the data from the carvedilol developmentprogram might support a regulatory decision with a degree of confidenceequivalent to the usual standard. The first three questions suggest alternativestrategies by which specific benefits might be said to have been established bythe carvedilol development program. Failing that, the Committee is asked, inquestion 4, if nonspecific benefit should be the basis for approval.
1. Most approvals are made on the basis of p , 0.05 for primary end pointsrepresenting clinical benefit in each of two adequate and well-controlledstudies. A case can be made that carvedilol meets that standard withstudies 240 and 223.1.1. Study 240 had a primary end point of time to the first event of sudden
death, death from progression of heart failure, hospitalization forworsening heart failure, or sustained increase in a specified groupof heart failure drugs. Elimination of the medications component ofthe end point from either the sponsor’s analysis (which includedcause-specific mortality and hospitalization) or the reviewer’s analy-sis (which included all-cause mortality and hospitalization) greatlyincreased the p value (from 0.003 to 0.029 and 0.04 to 0.378, respec-tively), suggesting that most of the statistical power lies in the medica-tions component. What effect does this observation have on the clini-cal interpretation of the results of study 240?
1.2. With regard to Study 223,1.2.1. What clinical benefit was the primary end point in the short-
term phase?1.2.1.1. Identify the words in the protocol leading to that con-
clusion.1.2.1.2. What analysis leads one to conclude there was a treat-
ment benefit?1.2.2. What clinical benefit was the primary end point in the long-
term phase?1.2.2.1. Identify the words in the protocol leading to that con-
clusion.1.2.2.2. What analysis leads one to conclude there was a treat-
ment benefit?1.2.3. For what baseline prognostic factors should one adjust?1.2.4. What adjustment in p values is indicated for multiplicity of
end points?1.3. With appropriate consideration of the supporting evidence from pri-
mary and secondary end points of these and other clinical trials,should carvedilol be approved for the treatment of heart failure onthe basis of p , 0.05 on the primary end points in each of two adequateand well-controlled studies?
14 L.D. Fisher and L.A. Moye
2. If not, carvedilol might be approvable on the basis of compelling evidence,from prespecified secondary end points, of a specific benefit. The promo-tion of a secondary end point to the status of a primary end pointand a potential bias of approval carries with it some implications withregard to the overall type I error rate. Below is a list of the prespecifiedsecondary end points of clinical benefit for which the sponsor’s analysesyielded a nominal p , 0.05 in at least one adequate and well-con-trolled study:
• Physician’s global assessment• Patient’s global assessment• NYHA class• Hospitalization for cardiovascular causes• Heart failure signs and symptoms
2.1. For each study end point, how should the nominal p value be ad-justed for?2.1.1. The number of the study’s primary end points?2.1.2. The number of the studies prespecified secondary end points?
2.2. For which secondary end points and studies, if any, is the evidenceof benefit as convincing as the observation of p , 0.05 for a primaryend point analysis?
2.3. With appropriate consideration of the supporting evidence from pri-mary and other secondary end points of these and other clinical trials,should carvedilol be approved on the basis of compelling evidenceof clinical benefit in some specific secondary end point or some combi-nation of such end points?
3. If not, carvedilol might be approvable on the basis of compelling evidence,from retrospectively defined end points, of a specific benefit. The promotionof a retrospective end point to the status of a primary end point anda potential basis of approval carries with it some presumably more seriousimplications with regard to the overall type I error rate. Below is a listof some retrospective end points of clinical benefit for which analysesyielded a nominal p , 0.05 in at least one adequate and well-con-trolled study:
• Heart failure mortality and hospitalization for heart failure• All-cause mortality• All-cause mortality and hospitalization for heart failure• All-cause mortality and hospitalization for cardiovascular causes• All-cause mortality and all-cause hospitalization
3.1. For each such study and end point, how should the nominal p valuebe adjusted for?3.1.1. The number of the study’s primary end points?3.1.2. The number of the study’s prespecified secondary end points?3.1.3. The number of the study’s other retrospective end points?
3.2. For which retrospective end points and studies is there evidence ofbenefit at least as convincing as the observation of p , 0.05 for aprimary end point analysis?
Carvedilol and the FDA Process: Introduction 15
3.3. With appropriate consideration of the supporting evidence from pri-mary, secondary, and other retrospective end points of these andother clinical trials, should carvedilol be approved on the basis ofcompelling evidence of specific clinical benefit with respect to somerespectively defined end point or some combination of such endpoints?
4. If not, carvedilol might be approvable on the basis of compelling evidenceof clinical benefit, without naming the specific benefit. For example, onemight conclude that, overall, some set of measurements pertaining tosymptomatic heart failure was indicative of benefit, but one might beunable to conclude that any specific benefit measurement met standardsthat would permit it to be named as the expected benefit of treatment.4.1. What analysis pertains to the assessment of overall benefit?
4.1.1. What weight was given to the primary end points?4.1.2. How did that analysis adjust for each study’s multiple prespeci-
fied end points?4.1.3. How did that analysis include retrospectively defined end
points?4.2. Should carvedilol be approved on the basis of compelling evidence
of an overall clinical benefit?