NDA 20-358 |
Page 5 |
(2)
Adjusted
Results
Given the two dose groups in
this study, it was necessary to make an adjustment for multiple
comparisons. One approach was to use
Dunnett’s test, which yielded a critical p-value of 0.025 for declaring any
particular finding positive. Using this
criterion p-value, the positive findings on the HAMD-21 and MADRS scores for
the 150 mg dose group generally prevailed, but again, only for LOCF analyses.
Impression: We consider
this to be a negative study that cannot provide support for the antidepressant
efficacy of either the 150 or 300 mg/day bupropion SR doses.
Overall Conclusions Regarding Efficacy Data for Bupropion SR
In summary, none of these 3 studies provided
evidence for the antidepressant efficacy of bupropion SR in the dose range
being studied. The sample sizes for
both studies should have been adequate, and on the basis of HAMD total scores
at baseline, the study populations had depressive symptoms of sufficient
severity to expect they might be responsive to drug treatment. While even apparently adequately designed
studies of antidepressants often fail, there was, unfortunately, no active
control arm to test the sensitivity of any of these trials for detecting a drug
effect. The failure of the 300 and 400
mg doses to show clear effectiveness, despite their bid equivalency to IR doses
of 300 or 400 mg suggests that it may be the study assay sensitivity that is
the problem. Unfortunately, whatever
the explanation is, the studies do not support the lower dose range for
bupropion.
We note your conduct of a NONMEM analyses for
studies 203, 205, and 212 combined, which you cite as providing supplementary
evidence for the effectiveness of bupropion SR. As an alternative to your exploratory analysis, Dr. Hoberman from
the Division of Biometrics performed a simple meta-analysis using all three
studies to investigate whether or not insufficient power might in part be an
explanation for the weak results for the individual trials. His analysis focused on the HAMD-28 total
score, CGI-Severity, and HAMD-Depressed Mood Item at week 8 for the 300 mg vs
placebo comparison. Given the greatly
increased sample size, it is perhaps not surprising that significance was
achieved for two of the three variables, i.e., HAMD-28 total score and
CGI-Severity, both for LOCF and OC analyses, but importantly not for HAMD-Depressed
Mood.
This analysis, showing effectiveness of an approved
daily dose level, provides some support for the view that these studies were
underpowered to detect the response to this formulation in this dose range and
for this population being studied.
While such an analysis has some explanatory value, it cannot support the
effectiveness of the lower doses: (1) Whatever the outcome of the meta-analysis
or the NONMEM analysis, neither was in the original analytical plan for this
program, and thus, neither can be considered definitive in assessing the
success or failure of the program. (2)
The sample sizes involved in the meta-analysis, i.e., almost 400 for the 300 mg
dose group and almost 500 for placebo, raise a concern about the possibility of
having a sample size large enough to be able to achieve statistical significance
for a treatment effect that is of marginal clinical significance. The point estimates of the effect sizes
(Tables 4-6) seen in these studies are very small. Although similar estimates have been seen in some studies of
active drugs, active drugs usually have larger estimated effects in some
studies.
Back a Page
Next Page
Back to Wellbutrin SR® NDA Index Page
Back to Main
Index