The published medical literature on antidepressants is missing a large
proportion of industry-sponsored clinical trials in which the active drug did
not beat placebo in a blinded, controlled trial, a study published in the
January 17 New England Journal of Medicine shows.
This publication bias hinders realistic assessment of the drug class and
could have exaggerated the efficacy of these medications.
The authors, led by Erick Turner, M.D., an assistant professor of
psychiatry and pharmacology at Oregon Health and Science University, compared
industry-sponsored clinical trials that have been published in biomedical
journals with those that have been submitted to the Food and Drug
Administration (FDA) as required for all currently marketed antidepressants.
About half (38) of the 74 clinical trials submitted to the FDA were deemed by
the agency to have positive results (that is, the primary endpoints, defined
before the trial initiation, were met). The authors classified study results
as positive or negative on the basis of the FDA reviewers' judgment. A
questionable study was one that the FDA judged to be neither positive nor
Almost all (37 of 38) of the positive studies were published in scientific
journals. In contrast, only three of 36 studies with results that were judged
negative or questionable were published as negative studies. Eleven of these
negative or questionable studies were published in a way that made the
outcomes, in the authors' opinion, appear positive, and the rest (22 studies)
were not published at all.
To measure the consequences of this publication bias, the authors
calculated the effect size—an estimation of clinical impact rather than
statistical significance—based on data from the 51 published trials and
from all 74 trials.FIG1
Not surprisingly, the effect size from the published trials came out larger
than that from all the trials, making the drugs' efficacy look greater than
the total data would support.
The authors acknowledged that their findings do not mean that
antidepressants are not efficacious. After all, the FDA had reviewed and
analyzed all the published and unpublished study data and approved each drug
for marketing. Meta-analysis by the authors also found each antidepressant to
be superior to placebo. Rather, the authors asserted that selective reporting
of trial data in the medical literature "deprives researchers of the
accurate data they need to estimate effect size realistically" and in
turn may "lead doctors to make inappropriate prescribing
Selective publication of industry-sponsored clinical trials is not a new
phenomenon or limited to psychiatry, APA President Carolyn Robinowitz, M.D.,
pointed out in a press release. She reiterated APA's official position of
supporting mandatory open access to all clinical-trial data.
"Organized psychiatry has actually been in the forefront of trying to
address this issue," said David Fassler, M.D., a clinical professor of
psychiatry at the University of Vermont. He cited APA's role in pushing the
AMA to issue a comprehensive report on publication bias and calling for a
centralized, publicly accessible registry of all clinical trials in 2003.
"Studies with negative outcomes are either less likely to be
submitted or they're rejected during the review process," observed
Fassler, who is an APA trustee-at-large. The International Committee of
Medical Journal Editors (ICMJE), representing 12 major international medical
journals, echoes this observation. Researchers and journal editors"
typically are less excited about trials that show that a new treatment
is inferior to standard treatment and even less interested in trials that are
neither clearly positive nor clearly negative, since inconclusive trials will
not in themselves change practice," the ICMJE editors wrote in September
In response to concerns about publication bias, ICMJE announced that,
beginning July 1, 2005, all clinical studies submitted for publication must
have been prospectively registered in a free, searchable public registry. This
requirement was soon adopted by many other journals, including the
American Journal of Psychiatry, rapidly spreading the practice of
clinical-trial registration among industry and academic researchers. The ICMJE
also urged journals to give more weight to negative studies when choosing
papers for publication.
In fall 2007, the U.S. Congress passed the Food and Drug Administration
Amendments Act, which contains general provisions for establishing a national,
publicly accessible, clinical-trial database. Pharmaceutical companies will be
required to register prospectively all phase 2 and phase 3 clinical trials and
report "basic results" from clinical trials in the database within
one year of trial completion. The results-reporting requirement is limited to
approved drugs. The Department of Health and Human Services will set specific
rules for this mandate, thus making the reporting of clinical trials a
In response to these policies, pharmaceutical companies have begun posting
clinical-trial results on the Internet routinely. It is conceivable that in
the near future all clinical-trial results of approved drugs will become
publicly accessible, even if they are not published in a biomedical
The disappointing performance of antidepressants in the unpublished,
industry-funded clinical trials reflects poor quality in conducting trials,
and the system of "drug approval, publication, and marketing provides
clear incentive for the publication of only positive results of clinical
trials," John March, M.D., M.P.H., told Psychiatric News."
The industry is simply following the rules of normal business
March is a professor of psychiatry and chief of the Child and Adolescent
Psychiatry Division at Duke University Medical Center and a principal
investigator in the Treatment for Adolescents With Depression Study (TADS),
one of several NIMH-sponsored large clinical trials that assessed the
comparative effectiveness of treatments for psychiatric disorders in clinical
The NIMH-sponsored clinical trials in children as well as adults provide a
sharp contrast to industry-sponsored studies in many areas: the publicly
funded studies directly compared medications with placebo and psychotherapy,
allowed enrollment of patients with comorbidities, studied vulnerable
populations including young children, applied more rigorous protocols and
procedures, and generated data that are open to researchers and the
"If you look at antidepressant trials in kids, the response rate to
active drug is essentially the same across all the trials at about 60
percent," said March. He blames high placebo-response rates for the lack
of efficacy in many of the unpublished industry trials. "In publicly
funded trials that were done at institutions with experienced
clinician-researchers, the active groups separated from placebo groups [that
is, became statistically significantly different] earlier, and the
effectiveness of active treatments was clearly positive," said
He pointed to the differences in the quality of study design and conduct
between industry- and NIMH-sponsored pediatric trials of medications for
several mental illnesses. The quality of some of these studies may have
prevented them from being published in medical journals, a possibility
acknowledged by Turner and colleagues.
The rules of regulatory approval and marketing-exclusivity extension reward
rapid completion of clinical trials, as long as the drug meets the minimum
requirements for statistically beating placebo in two trials. They do not
penalize marginally positive results or subpar trial design and conduct.
"A company can spend $40 million on a couple of quick pediatric
trials to satisfy the FDA requirement in exchange for a half to one billion
dollars of profit in the six-month patent extension," said March."
The incentives are all wrong."
March advocates a nationwide consortium of psychiatric clinicians and
researchers who would participate in a network of clinical trials to answer
key questions for patient care. "The industry is not likely to do trials
that are particularly beneficial to clinicians, particularly head-to-head
comparison trials and treatment augmentation trials for partial responders and
nonresponders," he said. "Public health is too important to leave
these questions entirely in the hands of the industry."
An abstract of "Selective Publication of Antidepressant Trials
and Its Influence on Apparent Efficacy" is posted at<content.nejm.org/cgi/content/abstract/358/3/252>.▪