I give a lecture regarding the statistical methodology employed in the 2001 Paxil (paroxetine) Study 329: Efficacy of paroxetine in the treatment of adolescent major depression: a randomized, controlled trial. Keller MB, Ryan ND, Strober M, Klein RG, Kutcher SP, Birmaher B, Hagino OR, Koplewicz H, Carlson GA, Clarke GN, Emslie GJ, Feinberg D, Geller B, Kusumakar V, Papatheodorou G, Sack WH, Sweeney M, Wagner KD, Weller EB, Winters NC, Oakes R, McCafferty JP. J Am Acad Child Adolesc Psychiatry. 2001 Jul;40(7):762-72.
Paxil Study 329 Retracted: A Critical Statistical Analysis
1. Efficacy of Paroxetine (Paxil) in the
Treatment of Adolescent Major
Depression: A Randomized,
Controlled Trial
CASE STUDY PRESENTED BY: CARLO CARANDANG, MD
MCDA 5520, NOVEMBER 27, 2017
2. Paroxetine Study- Background (circa
2001)
To compare paroxetine with placebo and imipramine with placebo for the treatment
of adolescent depression
1988- Prozac- dawn of SSRI blockbuster drugs (multi-billion dollars)
Push to find new indications for SSRIs, given their profitability
I chose this study, as this study was eventually debunked for misleading statistical
analysis
Big Pharma and Psychiatry: SSRIs were lucrative for both sides
Big Pharma overseas and sponsors the majority of the studies submitted to FDA for
approval
Perfect Storm of bias and greed affecting the scientific process
3. Justification For Selecting This Study
I chose this study as it shows how statistical analysis can be gravely affected by human
intervention due to bias and the pressures of:
Financial gain
Academic fame
Pressure on researchers to produce positive studies (“publish or perish”)
Pressure on journals to only produce positive studies, and not publish negative ones
For business statistics, this study emphasizes how important integrity and sound methodology
are to statistical analysis and not participating in cherry-picking variables based on best response
for financial gain and recognition
4. Paroxetine Study: Demographics
275 adolescents (12-18, males and females) with major depression began 8 weeks of double-
blind paroxetine (20–40 mg), imipramine (gradual upward titration to 200–300 mg), or placebo
10 centers in the United States and 2 in Canada
425 subjects were screened, and 275 were randomly assigned to one of the treatment arms
Study and investigators funded by GSK (GlaxoSmithKline)
Keller et al., 2001
5. Paroxetine Study: Outcome Measures
2 primary outcome measures were endpoint response (Hamilton Rating Scale for Depression
[HAM-D] score ≤8 or ≥50% reduction in baseline HAM-D) and change from baseline HAM-D
score
Secondary outcome measures:
HAM-D depressed mood item
depression item of the Schedule for Affective Disorders and Schizophrenia for Adolescents-Lifetime
version (K-SADS-L)
Clinical Global Impression (CGI) improvement scores of 1 or 2
nine-item depression subscale of K-SADS-L
mean CGI improvement scores
6. Paroxetine Study: Statistical Methods
Power Analysis: a sample size of 90 patients per arm was required to provide approximately
80% power to detect an effect size of 0.4 between an active regimen and placebo with an α level
of 5% (two-tailed)
Continuous variables, such as changes from baseline to endpoint in the HAM-D total score, CGI
improvement scale, and K-SADS-L, were analyzed by a two-factor analysis of variance using the
general linear model procedure of the Statistical Analysis System (SAS)
Categorical variables, such as percentage of subjects responding to treatment, were analyzed
with logistic analysis implemented in the categorical modeling procedure (CATMOD) of the SAS
Pairwise comparisons between each active treatment and placebo were two-tailed and
performed at an α level of .05. Data are reported as least square means (±SD or SE)
7. Paroxetine Study: Results
A total of 190 subjects (69% of 275)
completed the 8- week study
Premature withdrawal rates were 24% for
placebo, 28% for paroxetine (p = .60 versus
placebo), and 40% for imipramine (p = .02
versus placebo)
Significant for primary outcome variable of
HAM-D < 8
Problem: in the completers, much less than
the power analysis requirement of 90 for each
treatment arm
This may have consequences in the outcome
variables, as study is underpowered
8. Paroxetine Study: Results
High placebo response
Paroxetine does not really separate from
placebo
Not significant for primary outcome
variable of HAM-D > 50% decrease from
baseline
9. Paroxetine Study: Weaknesses
High placebo response (actually, why not just give placebo for depression?)
Underpowered
High drop-out rate (most due to side effects)
One of the primary outcome variables (HAM-D < 8) was not significant (could be from being
underpowered, or could be from lack of treatment effect)
Emphasis on secondary outcome variables when presenting the results (amateurish)
So one of the primary outcome variables is significant, but most likely not clinically significant
(low effect size)
Not enough to say statistically significant…need to calculate and report effect size (Big Pharma
probably suppressed this)
10. Paroxetine Study Retracted- Reanalysis
A recent reanalysis of a 2001 paroxetine (Paxil) study has put into serious doubt the whole field
of psychiatry and how it conducts clinical trials (Le Noury et al., 2015; Keller et al., 2001)
The reanalysis showed that Paxil (paroxetine) is not effective for adolescents with depression
and also found that there was an increase in harm on Paxil compared to a placebo, including
suicidal ideations and behaviors and other serious side effects (Le Noury et al., 2015)
These findings are in stark contrast to the original 2001 Paxil study 329, which published its
findings in the leading journal for child and adolescent psychiatrists, that Paxil was both safe and
effective for adolescent depression (Keller et al., 2001)