A Critique of NIMH’s Major Study of Anti-Depressant Effectiveness

by Ed Pigott | April 26th, 2006

On March 23rd the Washington Post reported in a front-page article on the findings just released from NIMH’s 35-million dollar “Sequenced Treatment Alternatives to Relieve Depression” study. The study results were met with mixed reviews. The url:
The results from this study were published in three articles and discussed in an editorial in the New England Journal of Medicine (NEJM):

  1. The 1st article appeared in the January 2006 American Journal of Psychiatry (AJP) and focused on the initial response to citalopram (an SSRI) by 2,876 patients presenting with major depression in either a psychiatric clinic or primary care setting.
  2. The 2nd article appeared in last week’s NEJM and focused on the effectiveness of bupropion, sertraline, or venlaffaxine for patients who failed to respond to citalopram.
  3. The 3rd article appeared in last week’s NEJM and focused on medication “augmentation” by adding either bupropion or buspirone to citalopram for those patients who failed to respond to citalopram alone.

The study’s key design and methodology characteristics were:

  1. The subjects had a score of 14 or higher on the 17-item Hamilton Depression Rating Scale (HAM-D).
  2. Subjects were excluded who had a history of non-response or intolerance to antidepressant medication in their current depression episode.
  3. Nearly 80% of the subjects had recurrent depression with a mean duration of 16 years.
  4. Approximately 50% of the subjects were unemployed (for many this was due to their depression since in 1990 unipolar depression was the fourth leading cause of permanent disability with the projections even worse, by 2020–see reference below).
  5. Neither the subjects, the physicians, NOR the raters were blind to the treatment condition in the first phase of the study.
  6. The raters were blind to the subjects’ treatment condition during phase two of the study.
  7. The patients had follow-up visits on weeks 2, 4, 6, 9, and 12 with a physician who was a “depression specialist” and could have an optional week 14 visit during both phases of the study.
  8. The physicians were assisted by clinical research coordinators in the application of measurement-based care, which included the routine measurement of symptoms and side effects at each treatment visit and the use of a treatment manual that described when and how to modify and/or augment medication doses based on these measures.
  9. Concomitant medications were allowed to be used to treat the associated symptoms of depression (e.g., sleep, anxiety, and agitation) and to treat the side-effects from the study’s antidepressant medications (e.g., primarily sexual dysfunction) based on each “depression specialists’” clinical judgment during both phases of the study.
  10. Once the subjects achieved the main criterion for “remission” as defined by a score of 7 or less on the HAM-D and then this score was maintained for a minimum of two weeks (maximum of four weeks); they were then moved into the study’s 12-month “natural” follow-up phase.
  11. The 16-item Quick Inventory of Depressive Symptomatology, Self-Report (QIDS-SR) was the study’s secondary outcome measure.
  12. The study also used the SF-12’s Mental Component Score (MCS). The SF-12’s MCS scale is considered the gold-standard for easily assessing the extent to which someone’s mental health problems impacts his/her activities of daily living with over 10-million administrations per year.
  13. The study only reports the subjects’ baseline measure on this scale, NOT their score on it when the subjects’ had achieved “remission” and were transferred into the 12-month follow-up stage.
  14. These articles do not report on the relapse data during the study’s follow-up stage nor even on the subjects’ HAM-D scores after the 14 weeks in phase one–and for those subjects that needed it–phase two.
  15. The study did not have a placebo control condition or any other control condition.
  16. Thankfully for us, psychotherapy was explicitly banned for subjects during the study.

The study’s key findings were:

  1. The subjects’ MCS score was 25.6 (SD=7.9) indicating that on average the subjects were almost 2.5 standard deviations below the mean. This score is indicative of people whose mental health problems are significantly impacting their ability to perform normal activities of daily living.
  2. During phase one, 28% of the 2,876 subjects had a two-to-four week remission during their up to 14-weeks of flexible dosing using citalopram and then were moved to the study’s 12-month follow-up stage.
  3. During phase two, one group of 727 citalopram non-responders were randomly assigned between bupropion, sertraline, and venlaffaxine.
  4. In this group, 21.3% achieved a two-to-four week remission on bupropion, 17.6% on sertraline, and 24.8% on venlaffaxine.
  5. Despite an N of 727 subjects, there were no significant differences between medications.
  6. Averaging across these medications, 21.2% of the subjects were able to obtain a two-to-four week remission after switching to a different SSRI medication and then they were moved to the study’s follow-up stage.
  7. The combined depression “remission” rate between those subjects who achieved a two-to-four week remission on citalopram (28%) and those subjects who achieved such a remission after switching to a different SSRI (21.2%) was only 43% (28% + (72% x 21.2%) during the 28 weeks of phases one and two.
  8. During phase two, one group of 565 citalopram-alone non-responders were randomly assigned to one of two medication “augmentation” groups in which either bupropion or buspirone was used to “augment” citalopram. Note to psychologists: This “augmentation” practice is common in psychiatry even though it has no empirical basis in published peer-reviewed studies that include a placebo control group. Unfortunately, this study does not include such a control group either.
  9. In this group, 29.7% of bupropion augmentation subjects achieved a two-to-four week remission compared to 30.1% for buspirone augmentation subjects.
  10. Averaging across these medications, 29.9% of medication augmentation subjects were able to obtain a two-to-four week remission and then were moved to the study’s follow-up stage.
  11. The combined depression remission rate between those subjects who achieved a two-to-four week remission on citalopram alone (28%) and those subjects who achieved remission after augmenting citalopram with either bupropion or buspirone (29.9%) was only 49.5% (28% + (72% x 29.9%) during the 28 weeks of phases one and two.
  12. In addition to study medications, 17.2% of the subjects also took trazodone; 11.9% also took an anxiolytic medication; and 16.7% also took either a sedative or hypnotic medication (page 1238, NEJM). The authors do not report what percent of men were prescribed Viagra or another medication for sexual dysfunction nor what percent, and which medications, were prescribed to treat the other side-effects from the study’s various antidepressant drugs that were administered either alone or in combination with said drugs.
  13. The study’s combined overall depression “remission” rate was 46% during the 28 weeks of phases one and two. Interestingly, everybody keeps reporting it as half when given even the researchers’ own unorthodox statistical methodology it was actually four percentage points less than half. That is a difference of 115 subjects who did not obtain their watered down definition of “remission” (2,876 x .04).
  14. When evaluated from an “intent-to-treat” perspective–the heretofore accepted standard in drug studies–the remission rate was only 39%. That is a difference of 316 subjects who did not even obtain a two-to-four week “remission” from their symptoms of depression during 28 weeks of the highest quality of drug-care that 35 million dollars can buy.

There were 2,876 subjects enrolled into this study 28 percent of whom (or 805 subjects) had a “remission” during phase one. That leaves 2071 subjects that were encouraged to participate in phase two but only 1292 (727 + 565) chose to do so with the remaining 779 subjects having lost hope in this form of care and choosing to no longer be experimented on. In the “let’s try a 2nd SSRI” trial, 154 subjects had a remission (727 x 21.2%) while in the “let’s add more mind-altering drugs” trial, 169 subjects (565 x 29.9%) did so.

In total, at the end of the 28 weeks 1128 (805 + 154 + 169) subjects had achieved a two-to-four week remission and then were moved to the 12-month follow-up stage while 1,748 of the original subjects failed to ever achieve even a mere two-to-four week “remission” from their perhaps drug-induced elongated episode of major depression.

NEJM editorial and Washington Post reporting:

  1. In the NEJM editorial written by David Rubin, he laments psychiatry’s 20th century drift away from general medicine and writes, “this change led to a definition of the cause of depression in behavioral terms–anger turned inward or abnormal reaction types–and suggested treatment that focused on insight or behavioral modification in the context of talk therapy. Depression thus was divorced from other medical illnesses. It was the search for new treatments (i.e., new psychopharmacological ones) that initiated the reconciliation between psychiatry and medicine, and the current questions surrounding the treatment of depression bring fully into focus our abilities and limitations in reducing the ravages of this disorder.”
  2. The Post points out that the study used 4 different antidepressants (Celexa, Wellbutrin, Zoloft and Effexor) that each had their own unique mode of action yet even though every patient who didn’t respond to citalopram got exposed to another one, there was essentially no difference in effectiveness. The Post quotes Dr. Rubinow, “this suggests that the underlying brain mechanisms of depression are far more complicated than simple notions of a single chemical imbalance.”
  3. Despite the pitiful results under optimal circumstances to maximize the percentage of “remissions,” Dr. Rubin’s editorial clearly states his belief in biological psychiatry’s eventual success in designing more specific and individualized antidepressant drug therapies. We’re talking serious “clinical” denial here on Dr. Rubin’s part.
  4. The Post cites the researchers’ acknowledgement that the intensive patient monitoring and “high-quality” depression care evaluation system found in this study occurs at most in only 10% of medical practices, something that Augustus Rush, the study’s lead investigator has a financial incentive to see improved upon (please see point # 9 in the “My Observations” section).
  5. The Post cites one researcher who states that in actual practice the remission rate for depression using antidepressant drugs is likely “in single digits.”
  6. The Post quotes Thomas Insel, NIMH’s director, as stating “the glass is half full from our perspective but the glass is half empty in that we need to come up with better treatments in the future.”
  7. The Post again quotes Insel as stating “this (study) involved depression-care specialists who made sure there was every careful monitoring of side effects and a relentless effort to optimize the dose.”
  8. Dr. Rubin’s NEJM editorial (page 1306) states that “about 25%” of subjects have a remission after switching to a different SSRI after the failure of the first SSRI medication. The actual rate was only 21.2% or, if we are rounding this statistic like Dr. Rubin seems to prefer, “about 20%.”
  9. The Post article correctly points out how pharmaceutical companies cherry pick/have extensive exclusion criteria such that their drug trial results simply don’t reflect the variety of comorbid problems that present in real life.
  10. The Post quotes “Augustus” John Rush, the lead investigator as stating, “a 50% remission rate is extraordinarily good, given the nature of these disorders.” Why is it that these biological psychiatrists (BP’s) feel such a need to consistently round way up their success percentages???

My observations:

  1. I hate to point out to Augustus Rush but the actual “remission” rate was only 39% from an intent-to-treat perspective over a 28-week time period and all that we know is that the “remission” lasted for between two and four weeks and then the subjects were “rushed” into the follow-up stage so as of yet we have no information regarding how long these alleged “remissions” actually lasted.
  2. Neither the Post article nor the NEJM’s editorial points out that that this was a “tag-you-are-healed” research design such that shortly after a subject achieves “remission” he/she are then rushed out of the data pool. Personally, I’ve never seen this type of research design before and it seems utterly bizarre to me. Has anybody else seen such a thing before???
  3. Neither the Post article nor the NEJM’s editorial points out that depression has a “spontaneous” remission rate of around 20% or so within any two-to-three month time period (if I remember right) and I assume that it is substantially higher in the six-to-seven month time period that this study lasted particularly when using their biweekly measurement methodology and counting a “remission” as being for as little time as only two weeks.
  4. Neither the Post article nor the NEJM’s editorial points out that one-third or more of the subjects were taking one or more “non-study” medications to treat their mental state including trazodone, anxiolytic medications, sedatives, hypnotic medications, and Viagra or similar medications for sexual dysfunction as well as the unreported number of other mind-altering medications to treat the other side-effects from the study’s antidepressant medications.
  5. The Post article did not point out that this study included no placebo or no-treatment control condition. Given the extensive amount of physician time, numerous medication dosing adjustments, initiating of new medications (both study, and non-study, medications), and the use of a highly sophisticated measurement-based care monitoring system including a telephonic voice-response system for subjects, the placebo effect in this study should be far more substantial than that found in traditional placebo-controlled studies.
  6. Neither the Post nor the NEJM’s editorial points out how interpreting this study’s 39% “remission” rate is scientifically impossible. There are simply too many uncontrolled-for “non-specific” factors that permeate throughout this study that—based on previous research—are known to spuriously increase the number of depression “remissions.”All that is known is that four different “antidepressant” drugs that were provided to depressed people either alone, or in combination with a variety of other mind-altering drugs in a study designed to inflate the subjects’ “remission” rate. This resulted in a paltry two-to-four week remission rate of only 39% over a 28-week time period before the subjects were rushed out of the data pool. That is all that is known from this study…that, and how intellectually dishonest these researchers are.
  7. Neither the Post nor the NEJM’s editorial points out that from an intent-to-treat perspective (which is the accepted standard in drug studies), the “remission” rate was only 39% and this was only for a two-to-four week window in a 28-week study.Instead, the Post headlines and psychiatry’s spin is that “the glass is half full” when in fact the “glass” is far closer to being only one-third full and we do not know how many holes are in this glass causing even this “remission” rate to quickly dissipate.
  8. Neither the Post nor the NEJM’s editorial points out that eight of the thirteen authors report receiving funding from multiple pharmaceutical companies with the average # of reported “potential conflicts of interests” being 10.25 per author (page 1241, NEJM). The NEJM did not require all of the 134 researchers participating in this study to report their “potential conflicts of interests.” Instead, NEJM wisely committed itself to a “save the trees” effort.
  9. Neither the Post nor the NEJM’s editorial points out that two of Augustus’ 19 reported “potential conflicts of interest” are advisory board payments, and more insidiously, royalty payments that he receives from HealthCare Technology Systems based on sales (http://www.healthtechsys.com/index.html). This company provided the depression “disease management” measurement-based care monitoring and telephonic voice-response system that was used in this study.While not specifically naming the company, Augustus advocates for such systems’ widespread adoption in the article for which he was the lead author.
  10. Fundamentally, this study appears designed to render the largest “remission” rate possible. No wonder that SEVEN pharmaceutical companies provided medication at no cost to this study even though there were only FOUR SSRI medications being “evaluated.” In a witches brew, the more numerous and powerful the mind-altering ingredients you have, the better the brew. In the marriage between the pharmaceutical companies and biological psychiatry, we have the self-deluded and aggrandizing makings of the “perfect brew.” Even pharmaceutical companies who do not make any antidepressant drugs still make money from this market through the sales of their drugs that help to treat the side-effects of antidepressant drugs.
  11. What a racket for pharmaceutical companies and BP’s. BP’s give desperate people powerful mind-altering drugs for depression, none of which have ever been shown to be any better than any other neuro-active medication and, as Dr. Rubin was forced to acknowledge, the BP’s don’t really even know how or why such mind-altering drugs occasionally work. When one mind-altering drug doesn’t work, BP’s then give the now even more desperate person another mind-altering drug or “augment” the first drug by combining a new mind-altering drug with the old one. BP’s then often times have to give the desperate and now side-effect ridden person a new mind-altering drugs to counteract all of the side-effects of the current mind-altering drugs that the desperate person is now taking. I would have loved to have seen what the remission rate would have been if a simple time-released version of No-doz had been included in this study. I bet that it would have been superior to anything that the BP’s cooked up while resulting in far fewer side-effects.This racket has been foisted onto society by the collusion between pharmaceutical companies and BP’s generating a perpetual money machine for both parties. Pharmaceutical companies get ever increasing sales’ revenue while BP’s consulting revenue is also ever increasing due to the rising demands for their presumed expertise.
  12. The fact is that despite at each point rigging the study design and methodology so as to render the largest possible remission rate that these “researchers” could plausibly defend, the remission rate was still only 39% and even this meager result is fleeting at best, symbolized by Augustus Rush’s decision to rush subjects out of the data pool once they’ve received as little as a two-week “remission.”
  13. I reanalyzed the 15-week data that I had on a similar group of subjects as were included in this study using Augustus’ methodology (see below). This analysis showed that by rushing subjects out of the data pool after only a two-to-four week period of time–versus at the end of the treatment period–resulted in a 19% overstatement of the “remission” rate in a 15-week time periodIf this same overstatement rate held for the two 14-week phases in this study, phase one’s more accurately reported remission rate would have been 22.7% versus 28% (.28 — (.28 x .19)). In phase two, the “let’s try a 2nd SSRI’s ” remission rate would have been 17.2% versus 21.2% (.212 – (.212 x .19)) while the “let’s add more mind-altering drugs” remission rate would have been 24.2% versus 29.9% (.299 — (.299 x .19)). The remission rate, if it were measured and reported on at the end of the 28-week study using the commonly accepted “intent-to-treat” methodology, would likely be significantly less than 31.6% (.39 x (.39 x .19)). The reason that it is likely significantly less than 31.6% is due to the longer 28-week time frame, and the well-known fact that people taking antidepressant drugs cease doing so the longer that they have taken them due to these mind-altering drugs’ innumerable side-effects while rendering only limited effectiveness for the vast majority of sufferers.

Biological Psychiatry’s Waterloo:

No matter how hard Augustus and Rubin try to spin it, this is biological psychiatry meeting its Waterloo. Biological psychiatry’s clinical practices are worse than medieval medicine. At least medieval medical practitioners typically used benign substances, thereby rendering a purer placebo effect without the known and unknown long-term risks associated with taking the various combinations of mind-altering drugs that BP’s regularly prescribe on an ongoing basis.

Treatments Based On Psychological Science Can, and Are, Doing Far Better:

Fundamentally, I believe that treatment programs and methodologies based on psychological science are doing a far better job than biological psychiatry in helping this needy group of people…our profession just lacks pharmaceutical companies’ deep pockets to highlight and widely disseminate our expertise, and for the most part, both our leadership, and ourselves, are utterly clueless in how to maximize the benefit to all concerned that our superior science makes possible.
As a profession, we can’t do anything about biological psychiatry’s deep pocket advantage nor BP’s inherent networking advantage by having gone through the same “medical” training as other physicians, but I relish being the underdog with far superior science.
Ed Pigott, Ph.D.
Maryland Licensed Psychologist
Principal, NeuroAdvantage, LLC
Email: ed@neuro-advantage.com
Phone: 443.812.9497


Murray, C.J. & Lopez, A.D. (1997). Alternative projections of mortality and disability by cause 1990-2020: Global Burden of Disease Study. Lancet, 349: 1498-1504.

Madhukar H. Trivedi, M.D., A. John Rush, M.D., Stephen R. Wisniewski, Ph.D., Andrew A. Nierenberg, M.D., Diane Warden, Ph.D., M.B.A., Louise Ritz, M.B.A., Grayson Norquist, M.D., Robert H. Howland, M.D., Barry Lebowitz, Ph.D., Patrick J. McGrath, M.D., Kathy Shores-Wilson, Ph.D., Melanie M. Biggs, Ph.D., G. K. Balasubramani, Ph.D., Maurizio Fava, M.D. and STAR*D Study Team, (2006). Evaluation of Outcomes With Citalopram for Depression Using Measurement-Based Care in STAR*D: Implications for Clinical Practice, Am J Psychiatry, 163, 28-40

Rush, A.J, et al (2006). Buproprion-SR, Sertraline, or Venlafaxine-XR after Failure of SSRIs for Depression, N Engl J Med, 354(12), 1231-1242

Trivedi, M.H., et al (2006). Medication Augmentation after the Failure of SSRIs for Depression, N Engl J. Med, 354 (12), 1243-1252

Rubinow, David R. (2006). Treatment Strategies after SSRI Failaure — Good News and Bad News, Editorial, N Engl J Med, 354(12), 1305-1307

Research on Anti-Depressants

One can readily critique the studies on anti-depressant efficacy, as Ed Pigott has done, but I would like to take another perspective. This development could also be seen as an approach to greater realism in research. Realistically, people don’t take just antidepressants, and often they don’t take just one. Ideally also, there is supportive care available as well. In some ways the studies therefore replicated a more realistic clinical situation, whereas in others it represented an ideal—with respect to access to supportive care, for example. It would have been essentially impossible to institute a valid placebo control arm when there were so many uncontrolled elements in the treatment.

We have reason to applaud the attempt to give latitude to clinical realities in design research. We may in time claim the same latitude in our neurofeedback studies. This $35M study is now the second major such study (the $6M ADHD Multi-Site study was the first) in which a placebo arm of the design was excluded. The question of fundamental efficacy of stimulants in one case, and of anti-depressants in the other, was no longer considered to be on the table. We should take the same liberty, after lo these many years, with regard to neurofeedback. The raw question of efficacy is simply no longer at issue.

All that said, it remains true that the study design in its details could hardly have been constructed more favorably to the desired outcome, nor could researcher conflict of interest have been any more blatant. To think that this industry had the gall to accuse British researcher Wakefield of bias in having his studies of vaccine mercury hazard funded by those who shared his concern. Oh, horrors! Who can trust the results?

So my own response to this work is two-fold. While flogging the details of the design we can still hold it up as an attempt to move toward clinical realism, one on which the NIH saw fit to lavish $35M. In particular, we ought to take advantage of the fact that the insistence on a placebo arm is not inviolable.

The placebo issue is actually somewhat problematic here. Given the fact that the kind of depression being researched was variable in presentation over time, the issue of the “natural course of the disease” is actually inseparable from the discussion. Implicitly, the quotation of medication efficacy of nearly fifty percent made the assumption that there would have been no change at all in the absence of the intervention. Further, the declaration of remission after only two weeks of symptom relief hides from view any subsequent deterioration. The latter will be compartmentalized in a separate publication on the follow-up phase, the impact of which can be managed as necessary. Few will make the connection with the claim of 50% response currently being publicized.
The jury shall disregard….

By Ed Pigott, Ph.D.

Leave a Reply