Rewarding the Companies That Cheated the Most in Antidepressant Trials


It is well known that we cannot trust the data the drug companies publish,1 and it seems that, in psychiatric drug trials, the manipulations with the data are particularly pronounced.2 As just one example, half of the deaths and half of the suicides that occur in randomised trials are not published.3

When the FDA in 2006 published its meta-analysis of 100,000 patients who had received depression pills or placebo in randomised trials, after having asked the companies how many suicides had occurred in their trials without checking the veracity of the information, the suicide rate on pills was 1 per 10,000 patients.4 However, five years earlier, Thomas Laughren, who chaired the large FDA meta-analysis, published his own meta-analysis of the drugs, based on data in FDA’s possession, and this time the suicide rate on pills was 10 per 10,000 patients, or 10 times as many.5 It is difficult to comprehend discrepancies of this magnitude, but what is abundantly clear — and which has been demonstrated by many researchers — is that the companies have deliberately concealed many cases of suicide and suicide attempts in their trials and in their reports to the drug regulators. In many cases, this has amounted to fraud.

When, very rarely, independent researchers get the possibility of analysing the trial data themselves, the results are often markedly different to those the companies have published. This was, for example, the case in the re-analysis of GlaxoSmithKline trial 329 in children and adolescents.6 The company had reported that paroxetine given to children and adolescents was effective and well tolerated, but none of this was true. Paroxetine was ineffective and harmful, and it is also harmful in adults.

The Lancet 2018 network meta-analysis by Cipriani et al.

Fraud and selective reporting are of course not limited to the most serious outcomes but also affect other trial outcomes. Several of the authors of a 2018 network meta-analysis in the Lancet are well aware that published trial reports of depression pills cannot be trusted. I therefore do not understand why they are authors on this paper. Erick Turner, for example, has been a reviewer for the FDA and he showed in 2008 that the effect of depression pills was 32% larger in published trials than in all trials in FDA’s possession.7 And John Ioannidis published a paper in 2008 called “Effectiveness of antidepressants: an evidence myth constructed from a thousand randomized trials?” where he mentions in the abstract that “selective and distorted reporting of results has built and nourished a seemingly evidence-based myth on antidepressant effectiveness.8

Yet the authors included 421 trials from the database search, 86 unpublished studies from trial registries and pharmaceutical company websites, and 15 from personal communication or hand-searching other review articles.9 By far most of the data came from published trial reports, which we know are seriously unreliable for depression trials. The authors’ meta-analytic exercise is academic, with no clinical value, and they drown the many biases in the trials in statistics that are so complicated that it is impossible to know what all this leads to. But we do know that statistical maneuvers cannot make unreliable trials reliable.

The authors included both head-to-head comparisons of drugs and comparisons of drugs with placebo. They found an effect compared to placebo of an SMD of 0.30, which is very similar to numerous earlier meta-analyses.10

However, they went much further than this. Despite the doubtful effect, which is far below what is clinically relevant, they ranked the drugs according to their effect and acceptability (drop-out for any reason).

This is a futile exercise, and when I first saw this network meta-analysis, my thought was that the authors had rewarded those companies that had cheated the most with their trials. My suspicion was strengthened when I looked at the results in their abstract. The authors claim, for example, that in head-to-head trials, agomelatine, escitalopram, and vortioxetine were more effective than other antidepressants and that the same three drugs were also more tolerable than other antidepressants. One doesn’t need to be a clinical pharmacologist to know that this seems too good to be true. Drugs that are more effective than others (which is often a matter of giving them in higher, non-equipotent doses), will usually also be more poorly tolerated. It is highly unlikely that some depression pills are both more effective and better tolerated than others. I therefore took a closer look at these three drugs.

I cannot know of course whether the companies behind these drugs were worse than others or whether the odd findings just reflected fundamental errors in the network meta-analysis. I do not accuse anyone, I just give some facts below.


Agomelatine is marketed by Servier. It was touted in 2011 in the Lancet as being an outstanding drug by two authors,11 including leading Australian psychiatrist Ian Hickie, who had numerous financial conflicts of interest. The authors claimed that fewer patients on agomelatine relapsed (24%) than do those on placebo (50%), but a systematic review by other psychiatrists found no effect on relapse prevention, no effect as evaluated on the Hamilton depression scale, and that none of the negative trials had been published.12 Three pages of letters — which is extraordinarily many — to the editor in Lancet (21 January 2012) pointed out the many flaws in Hickie’s review.

Escilatopram and vortioxetine

These drugs are marketed by Lundbeck. It is really far-fetched to believe that escitalopram can be better than citalopram because the active substance is the same. Citalopram is a stereoisomer consisting of an active part and an inactive mirror molecule, and escitalopram only contains the active substance. When studied by Lundbeck in its own head-to-head trials, the active molecule is better than itself.

However, when independent researchers made a meta-analysis based on indirect comparisons, comparing escitalopram with placebo, and citalopram with placebo, there was no difference.13 Their results are very telling. In a meta-analysis of seven head-to-head trials (2,174 patients), efficacy was significantly better for escitalopram than citalopram (odds ratio 1.60; 95% confidence interval 1.05 to 2.46). For the adjusted indirect comparison of 10 citalopram and 12 escitalopram placebo-controlled trials (2,984 and 3,777 patients respectively), escitalopram wasn’t any better than citalopram (indirect OR 1.03; 0.82 to 1.30). A similar discrepancy was found for treatment acceptability. Such results cast serious doubts about the reliability of network meta-analyses of depression pills.

Vortioxetine seems to be a very poor drug. When independent researchers compared vortioxetine with duloxetine and venlafaxine in meta-analyses, these drugs were significantly more effective than vortioxetine at three of the four dose levels tested.14 It is of note that every author in all of the published short-term trials had significant commercial ties to Lundbeck. This is a sure way of controlling that what gets published supports the company’s marketing ambitions.

Such ties were also apparent in Lundbeck’s meta-analysis that compared escitalopram with citalopram. All three authors worked for Forest, Lundbeck’s US partner, one as a consultant and the other two in the company. What are we supposed to make out of a paper published in a bought supplement to a journal edited by a person who is also bought by the company?15 Nothing.

Network meta-analyses of published trial data are not reliable

In a study of network meta-analyses (NMA), the authors used data from 74 FDA-registered placebo-controlled trials of 12 depression pills and their 51 matching publications.16 For each dataset, NMA was used to estimate the effect sizes for 66 possible pair-wise comparisons of these drugs. To assess how reporting bias affecting only one drug may affect the ranking of all drugs, they performed 12 different NMAs for hypothetical analysis. For each of these NMAs, they used published data for one drug and FDA data for the 11 other drugs. They found that pair-wise effect sizes for drugs derived from the NMA of published data and those from the NMA of FDA data differed in absolute value by at least 100% in 30 of 66 pair-wise comparisons (45%).

Extreme media hype

The Lancet network meta-analysis contains nothing new and what it claims to be new, is so unreliable that we should ignore it. These facts did not prevent the first author of the network meta-analysis, Andrea Cipriani, to hype the paper to the extreme, e.g. in BBC News:17

Lead researcher Dr Andrea Cipriani, from the University of Oxford, told the BBC: ‘This study is the final answer to a long-standing controversy about whether anti-depressants work for depression’ … Scientists say they have settled one of medicine’s biggest debates after a huge study found that anti-depressants work. The study … showed big differences in how effective each drug is.”

The authors of the report, published in the Lancet, said it showed many more people could benefit from the drugs … The Royal College of Psychiatrists said the study ‘finally puts to bed the controversy on anti-depressants’.”

Researchers added … At least one million more people in the UK would benefit from treatments, including anti-depressants.”

What is the reality?

It is still the reality that, despite serious flaws in depression trials — of which the most important ones are lack of blinding because of the conspicuous adverse effects of the pills, cold turkey in the placebo group because people were already on depression pills before they were randomised, industry-funding, selective reporting and data massage — the average effect is considerably below what is clinically relevant. In the Lancet network meta-analysis, the mean baseline severity score on the Hamilton Depression Rating Scale was 25.7, which is considered very severe depression according to the American Psychiatric Association’s Handbook of Psychiatric Measures.18 The oft-heard claim that these drugs work for very severe depression is wrong. Furthermore, when it is sometimes found in meta-analyses that the effect seems to be larger in severe than in moderate depression, this is likely just a mathematical artefact: the higher the depression score at baseline, the more the unblinding bias will distort the result.19

If the balance between benefits and harms of depression pills was positive, fewer people would drop out while on drug than while on placebo. The network meta-analysis did not report anywhere the average drop-out rate for the drugs, but it seemed to be very close to 1, which means that the drugs are no better than placebo. But it is worse than this. We have access to clinical study reports from the European drug regulators for depression pills, which are more reliable than what the companies publish. They also allowed us to include patients the drug companies had excluded from analysis. Despite the many biases in the trials, which included introducing withdrawal effects in the placebo group, we found significantly more drop-outs on drug than on placebo (Tarang Sharma, personal communication, submitted for publication). This means that placebo is the better drug.

Use psychotherapy for depression, not pills, and help people come off the pills

My conclusion is that patients should not be treated with depression pills. I no longer call them antidepressants, as they are ineffective and increase the risk of suicide and violence, which in the worst case can lead to homicide, with no upper age limit.2021 Further, they have numerous other serious harms, e.g. sexual dysfunction in about half of the patients who had a normal sex life before they started on the pills.

The patients should be treated with psychotherapy, which halves the risk of a new suicide attempt in those who have been admitted after a suicide attempt. Those who are currently on depression pills should be offered help to taper off them slowly and safely, and we should all focus on offering withdrawal courses (see and One of my PhD students and I lecture at such courses and we have an approved title and have submitted a protocol for a Cochrane review of studies of withdrawal of depression pills.

Show 21 footnotes

  1.  Gøtzsche PC. Deadly medicines and organised crime: How big pharma has corrupted health care. London: Radcliffe Publishing; 2013.
  2.  Gøtzsche PC. Deadly psychiatry and organised denial. Copenhagen: People’s Press; 2015.
  3.  Hughes S, Cohen D, Jaggi R. Differences in reporting serious adverse events in industry sponsored clinical trial registries and journal articles on antidepressant and antipsychotic drugs: a cross-sectional study. BMJ Open 2014;4:e005535.
  4.  Laughren TP. Overview for December 13 Meeting of Psychopharmacologic Drugs Advisory Committee (PDAC). 2006 Nov 16.
  5.  Laughren TP. The scientific and ethical basis for placebo-controlled trials in depression and schizophrenia: an FDA perspective. Eur Psychiatry 2001;16:418-23.
  6.  Noury JL, Nardo JM, Healy D, et al. Restoring Study 329: efficacy and harms of paroxetine and imipramine in treatment of major depression in adolescence. BMJ 2015;351:h4320.
  7.  Turner EH, Matthews AM, Linardatos E, et al. Selective publication of antidepressant trials and its influence on apparent efficacy. N Engl J Med 2008;358:252–60.
  8.  Ioannidis JPA. Effectiveness of antidepressants: an evidence myth constructed from a thousand randomized trials? Philosophy, Ethics, and Humanities in Medicine 2008;3:14.
  9.  Cipriani A, Furukawa TA, Salanti G, et al. Comparative efficacy and acceptability of 21 antidepressant drugs for the acute treatment of adults with major depressive disorder: a systematic review and network meta-analysis. Lancet. Published Online February 21, 2018.
  10.  Jakobsen JC, Katakam KK, Schou A, et al. Selective serotonin reuptake inhibitors versus placebo in patients with major depressive disorder. A systematic review with meta-analysis and Trial Sequential Analysis. BMC Psychiatry 2017;17:58.
  11.  Hickie IB, Rogers NL. Novel melatonin-based therapies: potential advances in the treatment of major depression. Lancet 2011;378:621-31.
  12.  Koesters M, Guaiana G, Cipriani A, et al. Agomelatine efficacy and acceptability revisited: systematic review and meta-analysis of published and unpublished randomised trials. Br J Psychiatry 2013;203:179-87.
  13.  Alkhafaji AA, Trinquart L, Baron G, et al. Impact of evergreening on patients and health insurance: a meta analysis and reimbursement cost analysis of citalopram/escitalopram antidepressants. BMC Med 2012;10:142.
  14.  Cosgrove L, Vannoy S, Mintzes B, et al. Under the influence: the interplay among industry, publishing, and drug regulation. Account Res 2017;24:99-115.
  15.  Gorman JM, Korotzer A, Su G. Effi cacy comparison of escitalopram and citalopram in the treatment of major depressive disorder: pooled analysis of placebo- controlled trials. CNS Spectr 2002;7(4 Suppl. 1):40–4.
  16.  Trinquart L, Abbé A, Ravaud P. Impact of reporting bias in network meta-analysis of antidepressant placebo-controlled trials. PLoS ONE 2012;7(4): e35219.
  17.  Therrien A. Anti-depressants: Major study finds they work. BBC News 2018; 22 Feb.
  18.  Fournier JC, DeRubeis RJ, Hollon SD, et al. Antidepressant drug effects and depression severity: a patient-level meta-analysis. JAMA 2010;303:47–53.
  19.  Gøtzsche PC, Gøtzsche PK. Cognitive behavioural therapy halves the risk of repeated suicide attempts: systematic review. J R Soc Med 2017;110:404-10.
  20.  Bielefeldt AØ, Danborg PB, Gøtzsche PC. Precursors to suicidality and violence on antidepressants: systematic review of trials in adult healthy volunteers. J R Soc Med 2016;109:381-92.
  21.  Maund E, Guski LS, Gøtzsche PC. Considering benefits and harms of duloxetine for treatment of stress urinary incontinence: a meta-analysis of clinical study reports. CMAJ 2017;189:E194-203.


Mad in America hosts blogs by a diverse group of writers. These posts are designed to serve as a public forum for a discussion—broadly speaking—of psychiatry and its treatments. The opinions expressed are the writers’ own.


Mad in America has made some changes to the commenting process. You no longer need to login or create an account on our site to comment. The only information needed is your name, email and comment text. Comments made with an account prior to this change will remain visible on the site.


  1. I like to mention to people that have mental health challenges in my community that they are strong. I like to say from time to time so they know.

    Simply calling to issue that antidepressants and psychotropics prescription rates have more than doubled over years and saying we need to curtail this would be helpful. I mean that’s pretty well known and people can determine if they want to try or not. That’s how I see it. I’d feel ultra insensitive if I pushed one way or another on people in recovery. That’s a bias too.

    With me I like to go home after long hard day and listen to good music. See this video message to illustrate.

    Report comment

    • A problem with this study, and most of the individual studies it is based on, is that patients who take part in these drug trials have been on an antidepressant before the trial . They are then put on placebo for 10 days , a so-called washout. Then half the group, now in cold turkey witdrawal , is now put back on a similar drug to what they had 10 days earlier, and the other group gets to continue their Cold turkey withdrawal.

      That means that all these studies are testing how good it is for patients to go back on a similar drug compared to being in withdrawal with a lot of abstinence symptoms. It is then actually quite surprising that the placebo patients are doing so well. The studies actually show most patients can stop their antidepressants and do quite well.

      The fact that these studies are just testing relief from abstinence symptoms by taking a similar drug, could explain why there is no effect in children and young adults. Many of them may not have taken any drug before they try the antidepressant.

      This testing of relief from abstinence, may also be the reason why the results are better with very serious depression. People who have very serious depression will most probably be more addicted to higher doses of antidepressants already, and have quite bad abstinence symptoms when they are put in cold turkey withdrawal.

      What is actually quite amazing is that the placebo groups reach the same level of reduction of depression as the drug group, only one week later. Why not wait?

      Report comment

      • Those are really interesting points. The totally different results in under-18’s baffled me, leading me to believe that what they were suffering from was maybe different. But drug naivety in that group explains it completely, and they may just be measuring relief from abstinence in many of the adults in the trials.

        By the way, the psychiatrist euphemism for “feeling crap in withdrawal” is “realising how well the antidepressant was working”. Lol.

        So, maybe its in under-18’s that the best data lie? And in that group, whats different about the under-18 fluoxetine trials? I wonder if there is a story there.

        Report comment

        • Just to follow up on fluoxetine. Robert’s book “Under The Influence” describes the original under-18 studies from the early 2000’s, and what was striking to me was the the small SMD effect size (0.2-0.5 depending on the scale). But 2 studies from 2014, one by the original author Emslie, and one by Atkinson, which were principally done to look at Duloxetine also, basically said there’s nothing there.

          Report comment

        • I think we all need to keep in mind that 18 years is an arbitrary legal (not biological) cutoff point. Puberty is a process not an event.
          So treating >18 and <18 on a yes/no basis will miss a lot of nuance. It might be the best we can do with existing data, but we need to make this caveat more explicit.

          Report comment

  2. “Half of the deaths and half of the suicides that occur in randomised trials are not published.” Let that sink in for a moment.

    “I no longer call them antidepressants…” Great. This is good. Dr. Breggin has calls these drugs what they are, namely neurotoxins. They might even be called “depressants,” but neurotoxins is perhaps the most accurate way to refer to these drugs that are euphemistically called “medications.”

    And yes, Paroxetine is too terrible to describe in words.

    Report comment

    • This article could be an example of how to manipulate study so that it seems to produce very positive results. If you have enough patients in the study, you can find so-called significant results of almost any difference. The difference can be totally insignificant, but it will show up as statistically significant, and be called significant in the text. Most people don’t realize that we are talking about statistical significance, and not clinical significance. In the appendix of the study they published a standard mean difference between placebo and the drug And this result was in the very weak range, actually weaker than the Kirsch et al study, that claimed that the antidepressants don’t work. The so-called significant difference between drug and placebo is approximately two points on the Hamilton depression scale. The difference has to be at least three for either patient or therapist to notice a difference. So what this study has confirmed, is that antidepressants can create a totally insignificant difference compared to a placebo pill.

      The placebo pill is often combined with attention and close follow up with a professional, and this has a very positive effect. So actually what we should do is give a lot of professional attention, testing and preferably some kind of therapy. Antidepressants don’t add anything, and many people say that it’s hell on earth to discontinue them. And 70% lose their sex life with these pills, for nothing.

      Report comment

    • Selection criteria are extremely important in any meta-analysis. You can get almost any results you want if you choose the studies you want to include. Here the researchers asked the pharmaceutical companies to provide unpublished trails to their liking. It is quite obvious that the companies would then be able to supply information that was to their advantage, that didn’t have to go through the rigorous process of peer review to be published. So including whatever the pharmaceutical companies wanted to provide, even if it sounds like this is being very thorough and searching for unpublished data, this is not science.

      Report comment

  3. A close look at the widely used HAM-D explains a lot. You can complete it in such a way as to score high enough to qualify as moderately depressed and get into a drug trial, without having indicated to any normal person who reads your answers that you have any problems of import other than insomnia. Then you can show some changes during the course of the trial that cut your HAM-D depression score in half, especially if they put you on benzos or sleeping pills, ruining your life, but actually look sicker than you did when you started the trial, based once again on a read of the 17 questions and your new answers. That reduction by 50% was what Cirpriani used as a marker of successful treatment.

    I’m afraid to post a link to the graphic lest my comment be censored.

    Report comment

  4. Excellent work Dr Gøtzsche and thank-you for this information, which we can now reference.

    “sexual dysfunction in about half of the patients who had a normal sex life before they started on the pills.”

    My view…this is what it is about, reducing people breeding, so, from their perspective SSRI’s work, they do not care about depression nor the suicide/homicide because it can always be put down to ‘mental illness’, so they get away with it. We need a situation where the coroners courts have to take into consideration pharmacogenetics tests, then we will get the truth of the matter, which will be closer to this:

    Excellent work Brian, hat off to you !

    This could be the answer, but the UK authorities have control on it. People need to – independently – do their own pharmacogenetics test, no authority can in any way be trusted, especially if you have a MH label :

    Report comment

      • I have had “mental health” underlings pressure me to date guys I found unattractive if not repulsive. They would get mad at me, accuse me of being stuck up and judgmental.

        They would remind me I was no better than the others. And our “mental illnesses” meant we had so much in common–only another “bipolar” could understand me. A therapist actually told me that. We’re all alike after all! 😛

        Then they say, “You’re not your illness.” Hmm. Someone can’t keep the narrative straight.

        Report comment

  5. Dear Peter Gotzsche,
    leaving aside the specific case of antidepressants, I’d be interested in your wider perspective on what should count as reliable evidence from clinical trials.
    I’m aware that bias and distortions are widespread, having authored this study for TranspariMED:
    Is your perspective that none of the currently available sources of information on drug safety and efficacy are reliable? If so, what does that imply for clinicians and patients in areas of medicine beyond psychiatry? Should we dismiss all published clinical evidence and go by doctors’ and patients’ personal experiences alone?
    Genuinely interested in your view, RSVP.
    Thank you!

    Report comment

    • I believe that the medical field, including all medical specialties and not just psychiatry, have allowed themselves to be poisoned by the drug companies to the extent that we must be very careful what we accept as gospel truth these days. Just because a doctor says something is true doesn’t mean that it is true. I believe that we must investigate things on our own. A lot of times doctors only know what drug reps have told them about a particular drug. If I want to find out reliable information about a drug, how to use it and how to get off of it, I go to my pharmacist and have her or him explain things to me. I’ve found them to be much more reliable than doctors in this regard. And in fact, when I decided to get off the “antidepressant” I was put on when held in the “hospital” I went to my pharmacist to find out how to do that safely.

      Report comment

    • Science is self evident. If the drug works, you know immediately and the condition is halted or the illness eliminated. The problem is that a two study approval with no independent researchers and the lure of a great job in pharmaceutical after a stint with the FDA is too much of an opportunity to give up.

      Unfortunately a lot of the drugs being launched are being questioned with the amount of years we are currently losing off the average life span. We’ve lost many years off the last two years of data.

      Report comment

  6. Dr. Gøtzsche,

    Thank you for this review. Can you please comment on how you think Cipriani failed to detect any effect of sponsorship, considering Kelly found huge effects of sponsorship?
    I asked Cipriani, but did not get a clear answer. I am startled by the dramatic discrepancy. Does the network meta-analysis method somewhere dilute or conceal this effect?


    p.s. I am a novice researcher wanting to turn the following take-down of antidepressants into a peer-reviewed article:
    I hope to meet some researchers on this site to help guide me in this, though I may sit on this article for a year or so before doing much about it.

    Report comment