A Servier Drug And A ‘Flawed’ Paper In The Lancet

Last May, an article in The Lancet touted the attributes of melatonin analogs, which have generally been used to treat sleep disorders, for combating major depression. In particular, the piece singled out agomelatine, an antidepressant that is sold by Servier, the French drugmaker, and was approved by European regulators in 2009. Known commercially as Valdoxan, the pill is not available in the US.

The article, which reviewed existing literature, touted the advantages the drug may have over other antidepressants, including Prozac and Paxil, such as improved sleeping, reduced waking after onset, fewer relapses and the potential for fewer side effects. One piece of data that was trumpeted – fewer patients on agomelatine relapse – 23.9 percent – than those who were given a placebo, or 50 percent.

While agomelatine was compared with other melatonin analogs, the authors wrote that, “importantly, only agemelatine has been reported to have clinically significant antidepressant effects… This drug might occupy a unique place in the management of some patients with severe depression and other major mood disorders.” (here is the abstract).

Since the article appeared, however, several scientists have raised objections to its publication. This week, in fact, The Lancet published no fewer than a half dozen letters that harshly criticize the methodology and conclusions, and maintain that the extent of the connections held by the authors – Ian Hickie and Naomi Rogers – to Servier was not fully disclosed.

The episode is the second such instance to arise this month in which a medical journal has been chided over a decision to publish an article in which the main thesis and motivations of the authors have been challenged. The other recent example concerned the Reviews in Cardiovascular Medicine, which is a peer-reviewed journal, and ran a review article about the use of fibrates in treating high levels of triglycerides. The article was paid for by Abbott Laboratories (read more here).

The circumstances involving the paper in The Lancet are slightly different, but the outcry from scientists from different corners of the earth has underscored growing concern about the vetting process used by medical journals prior to publication. Although The Lancet editor declined to comment, he did note – by way of a Tweet the other night on @RichardHorton1 – that “we are very heavily criticised for publishing a review on melatonin-based drugs for depression. Biased and overstated, say many.”

[UPDATE: Horton also Tweeted these remarks: "It is this kind of complicity that damages any hopes of a positive partnership between medicine and industry... As troubling is the fact that one author took part in speaking engagements for the company making one of these drugs... this paper purported to be an unbiased review of a new drug class. Peer review improved it, yet not enough... The bias in this paper is very disturbing - it might be fine to argue your case in a Viewpoint or letter. But...]

The publication, in fact, is threatening to grow into a scandal. A paper published this week in Evidence-Based Mental Health focuses on recent papers about agomelatine and cites these as an example of declining standards. “Most medical journals require adherence to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses, or PRISMA). It is an evidence-based minimum set of items for reporting in systematic reviews and meta-analyses,” write Corrado Barbui and Andrea Cipriani of the Department of Public Health and Community at the University of Verona in Italy.

“Adherence to PRISMA is not required in review articles dealing with basic science issues as these articles are not focused on clinical trials. In practice, however, the agomelatine case indicates that clinical data are regularly included and reviewed with no reference to the rigorous requirements of the PRISMA approach. These articles have this way become a modern Trojan horse for reintroducing the brave old world of narrative-based medicine into medical journals” (read the abstract).

Here is a sampling of the criticism that The Lancet has published in its correspondence section… The first letter, by the way, is from Barbui and Cipriano of the University of Verona, who write that “agomelatine has clinically significant antidepressant properties. However, this claim is not based on a systematic review of available evidence nor supported by the efficacy data that Hickie and Rogers themselves present.”

“In terms of tolerability, Hickie (in photo to the right) and Rogers do not mention the potential relation between agomelatine and hepatic problems,” they continue. “For long-term data, they included three studies, two negative and one positive comparison between agomelatine and placebo, but in the text of the article and in the summary only the positive study is mentioned” (read here).

Then, Robert Howland of the Department of Psychiatry at the University of Pittsburgh Medical Center wrote: “their assertion that agomelatine has similar efficacy to (Effexor, Prozac and Zoloft) is based on studies not designed a priori to test the antidepressant efficacy of these drugs.” He also noted that “the only placebo-controlled study in elderly patients found no benefit for agomelatine” (read here).

The next missive was from Celia Lloret-Linares, Jean-François Bergmann and Stéphane Mouly, who work at various hospitals in Paris. They noted that, of 10 placebo-controlled trials involving various doses of agomelatine, half were negative and one gave inconclusive results in terms of reaching primary endpoints. “These results do not support the statement that agomelatine is ‘unique’ and displays a ‘significant antidepressant effect’ and a ‘favorable safety profile’ ” they wrote.

They also point out that four other melatonin analogs are briefly presented on one page, while agomelatine is described in four pages and three tables. The authors, they add, have “numerous conflicts of interest with Servier. Clinical trials and audit sponsored by Servier, unrestricted educational grants, consultancy fees, and honoraria for lectures might explain the subjective nature and inappropriateness of Hickie and Rogers’s conclusions” (see this).

There’s more. “Where is a pooled analysis of all the listed placebo-controlled trials? Excluding those data from analyses is disingenuous. Where are the data for categorical outcome (response vs nonresponse) in all the trials, not just the three registration trials? The hidden data are needed to estimate the composite number needed to treat. And where are categorical outcomes (relapse vs non-relapse) to support the claim of prevention? Why are these data hidden?” asks Bernard Carroll of the Pacific Behavioral Research Foundation in California.

He also complains that “the financial arrangements are not transparent. Was this review commissioned? Was it proposed by Hickie and Rogers to Servier? Was the unrestricted educational grant from Servier Laboratories to Rogers effectively an honorarium to them? Who paid how much to whom, and where did the money go? Who paid the three acknowledged assistants? Was the paper drafted by the assistants? Did the corporation supply content?

“Did employees of Servier Laboratories participate in shaping the misleading manner of reporting the clinical trials data? Did they review the manuscript or request changes? This paper seems to break new ground for sponsored writing in medical journals, with conflicts of interest hidden in plain sight while bias continues,” he concludes (here is his letter).

Then, a letter from Jon Jureidini of the University of Adelaide and Melissa Raven of Flinders University charges that there is a “serious discrepancy between the summary and the body of the paper.” Why? The summary claims fewer patients on agomelatine relapse than those on placebo, but that was just one trial. Elsewhere, the paper mentions two other trials that did not find lower relapse rates. “Unfortunately, many will only read the summary and will be misled,” they write.

They also complain that the authors “selectively emphasize” that agomelatine has a lower risk of causing common side effects, but do not cite authoritative sources. And they maintain there are “several examples of citation misrepresentation.” They conclude that “publication of this flawed paper will undoubtedly validate marketing of Valdoxan, and we are curious to see how many paid Valdoxan advertisements will be published in Elsevier journals.” Elsevier, which publishes The Lancet, declined to tell them how many reprints Servier may have purchased, citing confidentiality (see this).

In response, Hickie and Rogers write the complaints “focus rather narrowly” on efficacy, clinical significance of using the drug to manage depression, side effects and their relationships with Servier. They argue antidepressants should be assessed by “other important outcomes,” including cases of suicide, reduced co-prescribing of other meds for anxiety and sleep disorders, and improved economic and social participation. “These issues represent a major challenge to the current ways in which the efficacy and clinical significance of antidepressant interventions are assessed,” they write. And they maintain that potential changes in liver function tests were previously notes in clinical trials.

As to conflicts, they say the paper was commissioned by The Lancet and “developed solely by us. It was not initiated or supported financially by Servier Laboratories. The additional research assistants who assisted with the paper’s preparation are long-standing employees of our institution who have worked previously on many similar datasets via our own financial resources.” Fact checking, however, did involve Servier, but they insist that “the opinions expressed and the conclusions drawn are those of the authors.”