1 statistical perspective acamprosate experience sue-jane wang, ph.d. statistics leader alcoholism...
TRANSCRIPT
1
Statistical PerspectiveAcamprosate Experience
Sue-Jane Wang, Ph.D.
Statistics Leader
Alcoholism Treatment Clinical Trials
May 10, 2002
Drug Abuse Advisory Committee Meeting
DACCADP/DB2/OB/CDER/FDA
2
OUTLINE
• The Three European Pivotal Trials– Pelc II, Paille, PRAMA– Differential treatment discontinuation
• The US 96.1 Trial– Acamprosate effect (2000 mg/day) observed in only
the sponsor defined Post-Hoc Analysis but not supported by many other analyses
• Difference between European and US Trials– Analytic Issues
3
• Placebo (pbo)
• Acamprosate 1332 mg/day (low dose)– European: two 333 mg tablets bid
• Acamprosate 1998 mg/day (medium dose)– European: two 333 mg tablets tid– US: two 500 mg tablets bid
• Acamprosate 3000 mg/day (high dose)– US: three 500 mg tablets bid
Dosages of Treatment
4
Pelc II• Study Design: A multicenter, double blind,
randomized, placebo-controlled 3-arm study
• Study Objective: Effectiveness and
tolerance of acamprosate in helping to maintain abstinence in weaned alcoholic
• Main criterion of judgement:
The consumption of alcohol
• Trial duration: 3 months
• Trial period: June 1990 to April 1992
5
Pelc II(3-mons)
Pbo Acamp1332 mg/d
L-dose
Acamp1998 mg/d
M-dose
Nominalp-value
N 62 63 63
% dropouts1 48% 30% 32% 0.065
Time on treatmentDiscontinuation
84 d 85 d 85 d 0.146
% no relapse-Dropout as is-as relapse
21%15%
51%41%
44%41%
0.0010.002
Time to 1st relapse 17 d 53 d 53 d <0.0011Dropout rates differ, primarily in ‘lost-to-follow-up’, No death
Pelc II
6
Paille• Study Design: A multicenter, double blind,
randomized, placebo-controlled 3-arm study
• Main Objective:
– maintenance of abstinence with acamprosate 1332 mg/day (low dose) in alcoholic patients who were followed as outpatients after withdrawal
• Trial duration: 360 days
• Trial Period: April 1989 to November 1992
7
PaillePaille
(360 days)Placebo Acamp
1332 mg/dL-dose
Acamp1998 mg/d
M-dose
Nominalp-value
N 177 188 173
% dropouts1 65% 55% 48% 0.006
median time ontreatment period
8-mon 10.5 11.8 0.051
% abstinence-Dropout as is-use 340d cutoff-as relapse
23%11%7%
27%18%14%
30%19%15%
0.2850.0960.044
Time to 1st relapse 1-mon ~1 mon 2-mon .01-.071Dropout rates differ, primarily in ‘relapse’, ‘refusal’; 2-deaths per arm
8
PRAMA• Study Design: A multicenter, double blind,
randomized, placebo-controlled 2-arm study– PBO vs. Acamprosate
• Study Objective:– Effectiveness and tolerance of acamprosate,
which helps to maintain abstinence after detoxification in the alcoholic patients
• Trial Duration: 48 weeks
• Trial Period: Oct. 1990 to December 1992
9
PRAMA
• Primary efficacy: time to 1st relapse
Definition of relapse• short-term relapse: alcohol consumption for
up to 24 hours • long-term relapse: alcohol consumption for more
than 24 hours with/without the need for hospitalization
• continuous relapse: constant alcohol consumption
10
PRAMA(48 weeks)
Placebo Acamp1332 mg/d&1998 mg/d
L- & M-dose
Nominalp-value
N 136 136
% dropouts 60% 42% <0.001median time ontreatment period
169 days 337 days 0.005
% abstinence-Dropout as is-as relapse
40%12%
51%29%
0.052<0.001
Time to 1st relapse 3-mon 8.4-mon 0.005Time to 1st relapse 1.5-mon 4.5-mon <0.001
PRAMA
11
The European Pivotal Trials
• Drinking Data was retrospectively collected
• Dropout rate higher in pbo than in acamprosate
• Effect of acamprosate 1998 mg/d (r.t. pbo) was shown in “% complete abstinence”
• Effect of acamprosate 1332 mg/day (low dose) was not shown in Paille Trial
• Trials were conducted in late 1980s to early 1990s
12
US 96.1 Trial• Patient Population: Alcohol dependence who
had been withdrawn from alcohol or who had completed medicated detoxification within 2 to 10 days of study entry
• Study Design: a multicenter (21 centers), double-blind, randomized, placebo controlled
• Randomization: well-balanced among 3-arms: PBO, M-dose (2 g/d), H-dose (3 g/d)
• Data: rigorous TLFB drinking measurement
13
• Primary Objective– to confirm efficacy & safety of medium dose (2000
mg/day) acamprosate in association with standardized but minimal psychosocial support, guided by a protocol-specific manual
• Secondary Objective– explore high dose (3000 mg/day) acamprosate
efficacy & safety
• Trial Duration: 24 weeks
• Trial Period: May 1997 to January 1999
US 96.1 Trial
14
US 96.1 Trial
US 96.1(24 weeks)
Placebo Acamp2000 mg/d
M-dose
Acamp3000 mg/d
H-dose
Nominalp-value
N 260 258 84
% dropouts 45% 59% 48% 0.005
Time to treatmentdiscontinuation
168d 136d 164d 0.110
15
• Protocol specified primary efficacy endpoints
– Time to 1st day of any drinking
– Time to 1st day of heavy drinking
(6 drinks for men, 4 drinks for women)
– Cumulative abstinence duration (CAD)
– CCAD (=percent days abstinence PDA)
– Rate of complete abstinence
US 96.1 Trial
16
Originally Planned Efficacy Analyses(ITT popn)
Parameter Statistic PlaceboACAMP2 g/day
ACAMP3 g/day
% Patients whorelapsed to drinking
% 89% 92% 90%
Time to 1st drink (d) Median 4 4 4Time to heavy drink Median 12 14 17CAD (d) N
Mean (SE)Median
25683.3 (3.9)
78
25372.3 (3.7)
56
8281.5 (7.1)
75 CCAD (%) N
Mean (SE)Median
25651.2 (2.2)
53
25345.5 (2.2)
38
8249.9 (4.1)
47
17
• Medium dose acamprosate 2g/d failed to show a superior effect on pre-specified endpoints
• Exploratory analysis pre-specified– CAD or CCAD adjusted for T, C, Detoxification
• Supportive analysis pre-specified– adjusted for T,C, amount of psychosocial therapy– adjusted for T,C, illicit drug use
US 96.1 Trial
18
• New primary efficacy endpoint
– Cumulative Abstinence Duration (CAD)
(post-hoc definition)
• Endpoint considered: CCAD (=PDA)
• Endpoint actually used: ALCCAD (PDA adjusted for treatment discontinuation)
US 96.1 Trial
19
Chosen by the sponsor
Use the following covariates for adjustments:– treatment exposure
– pooled site ()
– baseline CGI-severity
– stage of readiness to change
– psychological antecedent
– addiction index – goal of abstinence
Post-hoc ANCOVA Model #1
20
- use 6 covariates (excluding treatment exposure) from model #1
Treatment Exposure: defined astreatment compliance*treatment duration/100
-potentially treatment related -due to differential time to discontinuation
-and differential dropout
Post-hoc ANCOVA Model #2
21
• Modified CCAD: No Statistically significant findings on Model #1 or Model #2 or unadjusted analysis
• ALCCAD: endpoint actually used (CCAD adjusted for treatment discontinuation)
Versions of Primary Efficacy Outcome
22
ALCCAD: Treatment Group Comparisons and Adjusted (Least-Square) Means ITTPopulation
Plbo Aca 2 g/d(M-dose)
Aca 3 g/d(H-dose)
p-valM-dose: P
p-valH-dose:P
N 256 253 82
Unadjustedmean
54.3 56.1 60.7
Unadjustedmedian
59 59 63 0.703 0.205LS Means #1 52.3 58.2 62.7 0.044 0.009LS Means #2 53.4 56.8 63.1 0.296 0.021
23
ALCCAD: Treatment Group Comparisons and Adjusted (Least-Square) Means
Population StatisticACAMP2 g/day Placebo P-Val
Intent-to-Treat nLSMean (1)LSMean (2)
25358.256.8
25652.353.4
0.044*0.296
Efficacy Evaluable nLSMean (1)LSMean (2)
17762.360.6
19854.855.8
0.023*0.157
Motivated Intent-to-Treat
nLSMean (1)LSMean (2)
10070.068.3
11558.159.0
0.021*0.100
Motivated EfficacyEvaluable
nLSMean (1)LSMean (2)
7175.573.0
8659.461.3
0.008**0.068
24
Why post-hoc ANCOVA model #1?Seven covariates + Trt
Why post-hoc ANCOVA model #2?Excluding ‘treatment exposure’
US 96.1 Trial
25
Reviewer’s exploratory analyses -
to examine how the results of the sponsor’s post-hoc ANCOVA depend on which covariate(s) to include
- center always included
- each covariate included one at a time
- other combinations of covariates included with or
without the ‘abstinence goal’
Post-Hoc Model #1:Why 7-covariates
26
US 96.1 Trial (ITT Pops)Covariates
Placebo(n=256)
Aca 2 g/d(n=253)
(M-dose)
Aca 3 g/d(n=82)
(H-dose)
Treatment ExposureMean (T related?)Median;p=.0172
16.521
14.215
15.720
Abstinence Goal(p=0.033 trend) 45% 40% 32%Slip is Possible(p=0.072 trend) 28% 31% 39%Abstinence Goal + Slipis Possible 73% 71% 71%
27
Figure 2. Observed Mean Heavy Drinking Days at Each VisitUS 96.1 Trial
0
1
2
3
4
5
6
Weeks on Study
Hea
vy D
rinki
ng D
ays
pbo
medium(2g/d) dose
high (3g/d) dose
Time% n_pbo%n_2000%n_3000
wk24wk20wk16wk12wk8wk4wk2wk10.570.610.670.720.800.890.940.99
0.560.590.630.660.760.880.991.000.420.470.560.660.770.890.941.00
28
Figure 3. Observed Mean Any Drinking Days at Each VisitUS 96.1 Trial
0
5
10
15
20
25
Weeks on Study
Any
Dri
nkin
g D
ays
Pbo Aca 2g/d Aca 3 g/d
wk1 wk24wk20wk16wk12wk8wk4wk20.99 0.94 0.89 0.80 0.72 0.67 0.61 0.57
%n_3000%n_2000% n_pbo
Time
0.420.470.560.660.770.890.941.001.00 0.99 0.88 0.76 0.66 0.63 0.59 0.56
29
Medium dose (2 g/d) effective?
US trial was sufficiently powered for efficacy evaluation of this dose
=> No evidence of effect, after adjusting for any
one covariate alone
=> Excluding treatment exposure from ANCOVA,
No evidence of the effect
=> Numerically worse than placebo in mean
heavy drinking days
30
Medium dose (2 g/d) effective?
Whether this dose appears to show efficacy depends on post-hoc selection of covariates for adjustment
e.g., including “abstinence goal” + “trt exposure” or all 7 covariates (multiplicity!!)
31
Exploration for High Dose???
Not sufficiently powered (1/3 sample size)
Insufficient safety data in the US trial
Numerically superior to placebo in mean heavy
drinking days
The abstinence goal seemed prognostic of high dose
(3 g/d) effect, but only if using the ALCCAD
Effect was not seen if adjustments did not include
abstinence goal
32
Summary - US 96.1 Trial
• Dropout rate was significantly higher and treatment exposure was shorter in medium 2g/d dose acamprosate than in high dose or placebo
• Effect of 2 g/day (r.t. pbo) not established on protocol specified primary efficacy outcome or post-hoc defined primary endpoint (CCAD = percent days abstinence)
33
Summary - US 96.1 Trial
• Post-hoc chosen model #1 or #2 on ALCCAD can be problematic:
- The significant results for 2 g/d acamprosate depend on which post-hoc independent covariate(s) to include for adjustments; no effect after multiplicity adjustments
- Analysis for 3 g/d acamprosate was exploratory:
time to 1st heavy drinking, mean heavy drinking over
time; but, small ‘n’
34
Difference between European vs. US
• Acamprosate in Europe: 3-mon, 360d, 48-wk– less dropouts – longer treatment exposure
• Acamprosate 2g/d in US: 24-wk– more dropouts – shorter treatment exposure
• Analytic Issues– need of well-thought pre-specified algorithm for
handling dropout patterns
35
Differences between European vs. US• Drinking data: European: retrospective collection from clinician US: TLFB diary• Criteria of total abstinence at study entry Europe: explicitly required US: not explicitly required• Medicated detoxification European: 100% US:10%• Patients’ baseline characteristics European: 18 to 65 years of age US: no upper age limit
36
Differences between European vs. US• Psychosocial support
European: non-structured psychosocial therapy US: standardized, manual-guided psychosocial support• Dosage form
European: 333 mg tablets US: 500 mg tablets• Other design features of the US study were not
typical in the European studies e.g., mandatory follow-up algorithms for missed visits or missed phone contacts