CORRUPTION OF CLINICAL TRIALS REPORTS:
There is a disconnection between the FDA’s drug approval process and the reports we see in medical journals. Pharmaceutical corporations exploit this gap through adulterated, self-serving analyses, and the FDA sits on its hands. I suggest we need a new mechanism to fix the problem – by independent analyses of clinical trials data.
When they analyze and publish their clinical trials in medical journals, pharmaceutical corporations have free rein to shape the analyses. The FDA conducts independent analyses of the data submitted by the corporations, and it may deny or delay approval. But the FDA does not challenge the reports that flood our medical journals, both before and after FDA approval. It is no secret that these publications are routinely biased for marketing effect, but the FDA averts its gaze. That failure of the FDA – a posture known as enforcement discretion – has been well documented. The question is why? At the same time, exposing the biases has been difficult for outsiders because the data are considered proprietary secrets.
A Case Study
Now, a detailed example of deliberate corporate bias has finally been documented, through materials released in litigation. This exposé was reported by Drs. Jon Jureidini, Jay Amsterdam, and Leemon McHenry. Their findings were recently published, and their article is freely available on-line. This example concerned a clinical trial of an antidepressant drug in children and adolescents. The drug, citalopram, was already approved for use in adults, and its off-label use in children would spread if there was published supportive evidence. An Investigational New Drug (IND) protocol and plan of analysis were filed by Forest Laboratories with the FDA in 1999. The trial was completed in 2002, and the results were published in American Journal of Psychiatry in 2004 – but the FDA did not accept the results as sufficient to approve this drug for use in children or adolescent patients. By that time the patent on citalopram had expired and Forest Laboratories introduced a virtual twin drug, escitalopram (single active enantiomer). That more expensive version of citalopram was heavily promoted, and it was approved in 2009 for use in children, but even then the FDA specifically noted that safety and efficacy were not established in children under age 12. Since then, new analyses suggest that most antidepressant drugs have little evidence of efficacy even in older children.
Tricks of the Trade
In service of a positive report, the statistical analyses performed by Forest Laboratories deviated from the IND plan of analysis, and negative results were edited out. The biases now documented by Dr. Jureidini and his colleagues for that 2004 sponsored report in American Journal of Psychiatry included:
· Inflating the main measure of the drug's effect by reporting an incorrect and clearly exaggerated effect size. On being challenged, the authors later explained their misinformed computation without actually acknowledging the error.
· Failure to report secondary measures of response because they were negative. Those measures had been stipulated in the IND protocol to serve as cross-checks on the main result. These negative findings were airbrushed out of the publication by corporate marketing.
· Unplanned, new secondary measures of response were inserted ex post facto because they were positive (that is a real no-no).
· Violations of the IND protocol were not reported and were then fudged (patients who had properly been excluded per protocol were put back in for analysis, which made a nonsignificant primary outcome analysis turn positive).
· Adverse events were analyzed and summarized in a misleading way.
· The finding that the drug had no effect on depression in children under age 12 was not reported, even though an age-effect interaction analysis had been specifically projected in the IND protocol. This strategic omission left the impression that off label use of citalopram in younger children could be clinically reasonable.
· The corporation knew that another, unpublished, trial in children, conducted by their European partner Lundbeck was negative, and that it raised concerns about suicide risk, but that information was withheld. The authors later were challenged in the journal about this concealment. Their response was utterly disingenuous.
· The published article failed to acknowledge that it was authored by a non-medical ghostwriter, who took direction from marketing executives – the 2004 publication was a marketing product purporting to be an objective scientific report.
· Academic authors were recruited only after the manuscript was written, reviewed, and approved in-house – these nominal academic authors were signed on to front for the corporate narrative.
· The perfunctory role of the academic (ahem) authors is clear from the fact that they failed to recognize the wildly inflated effect size claimed for the drug – something that was instantly obvious to several groups of competent readers.
These changes created the appearance of a positive result, and the publication drew wide attention. According to Thomson Reuters Web of Science, it has been cited over 160 times, placing it in the top 5% of cited articles in clinical medicine from 2004. This early publication gave plausible justification for off-label use of citalopram/escitalopram in children, even with FDA approval having been denied, and even though the trial was actually negative. The FDA has reported that between 2005 and 2010 well over 750,000 patients up to age 17 received escitalopram, including almost 160,000 under age 12. Thus, the sleight of hand about failure to show even fudged efficacy in younger children is especially deplorable. Internal memoranda reproduced in the exposé by Dr. Jureidini and his colleagues give a clear picture of the corporate manipulation of the scientific publication process. Now we know – in black and white – just how bad the bias can be. This kind of data manipulation, with ad hoc cherry picking and moving of goalposts, is unacceptable, but it is entrenched. Indeed, it is business as usual – and the FDA looks away.
A Specific Proposal
Our primary defense against such perversions of scientific reporting is fidelity to the registered IND protocol and plan of statistical analysis. The solution is not hard to see: We need independent analyses of clinical trials because we cannot trust the corporate analyses. In effect, we need something like the Underwriters Laboratory to verify the statistical analyses of clinical trials. Nobody takes the manufacturing corporation’s word for it concerning the safety and performance of X-ray machines or cardiac defibrillators. Why treat the statistical analysis of drug trials any differently? It’s highly technical work.
Who should assume that responsibility? Why not the FDA? After all, they alone see all the data. My specific proposal is for Congress to mandate that the FDA analyze all clinical trials data strictly according to the registered protocols and analysis plans. That requirement should apply to new drugs or to approved drugs being tested for new indications. It should apply also to publications reporting new trials of approved drugs. Corporations and investigators should be prohibited from publishing their own in-house statistical analyses unless verified by FDA oversight.
There are three good reasons for prohibiting in-house corporate analyses of clinical trials data. First, as the present example illustrates, the inherent conflict of interest is simply too great to be ignored. Second, when corporate statisticians who answer to marketing executives get “creative” in the ways exposed here, then the conditions for valid statistical analyses no longer apply – the statisticians are then on a fishing expedition and they are no longer testing the defined study question with fidelity to the methods specified in the IND protocol. In that case, any nominally significant statistical findings are just exploratory, not actionable – not good enough to justify off-label use of the drug, especially when properly evaluated alternatives are available.
Third, there can be no justification for treating the production of influential publications in medical journals any differently than we treat the production of potent drugs. Our FDA continuously inspects production facilities for evidence of physical adulteration, even as far away as China. They now need to monitor the adulteration of clinical trials reports in medical journals. The harms of adulterated analyses can be just as serious as the harms of adulterated products.
Push Back from Pharma?
We can expect the pharmaceutical industry to mount a First Amendment challenge to this proposal. It will fail, because the public health is too important. Just as there is no First Amendment right to shout fire in a crowded theater, so also corporations have no First Amendment right to say a drug is safe and effective when they know it isn’t. That is a betrayal of patients.
The corporations will also claim piously that their publications undergo peer review. Sadly, that is no barrier to this pervasive corporate bias because the peer reviewers for medical journals don’t see all the real data – they see only the data the corporation wants them to see. Only the FDA sees all the data. We can no longer cling to the myth of informed and unbiased peer review of clinical trials reports. The corporations rely on that myth as a fig leaf to support their First Amendment claims and to defend their practice of in-house statistical analyses. Moreover, medical journals also are subject to bias and conflict of interest. We could note that the Associate Editor of American Journal of Psychiatry in 2004 was also a major U.S. key opinion leader for Forest Laboratories. According to one of the released depositions, he was instrumental in securing acceptance of the report by the journal.
Business as Usual?
The present example is not an isolated case. Dr. Jureidini and his co-authors described several similar, recent examples. One of those was the reanalysis by Jureidini and others of an infamous trial of paroxetine for pediatric depression. And still, fresh exposés keep appearing. The latest is from Lisa Cosgrove at the University of Massachusetts in Boston and her colleagues, involving “ghost management of the information delivery process” for another new antidepressant drug, vortioxetine – available on-line here. (What is it with the antidepressants, anyway?) On this Health Care Renewal blog, Roy Poses has called attention to these issues. As recently as June 8, 2016 he discussed the Transparency International report on corruption in the pharmaceutical sector.
Eric Topol, who helped to expose the Vioxx scandal, made similar points recently in a BMJ commentary: “The bad science in clinical trials has been well documented and includes selective publication of positive results, data dredging, P hacking, HARKing, and changing the outcomes that were prespecified at the beginning of the study…. Furthermore, the disparity between what appears in peer reviewed journals and what has been filed with regulatory agencies is long standing and unacceptable.”
It’s No Time for Old Solutions
As the eye-popping numbers of children treated with escitalopram show, even off-label use of an undistinguished drug in a niche population can be highly profitable. That is why I am proposing that the statistical analysis of clinical trials data can no longer be entrusted to pharmaceutical corporations, on account of their massive inherent conflict of interest. Open access to patient level data, as well as pre-registration of protocols and of data analysis plans, have been actively promoted for some years now to clean up the corporate bias in clinical trials. These are positive developments, but they will not close the disconnection highlighted just above by Dr. Topol. The once idealistic world of clinical trials has changed irreversibly in the past 30 years. As one observer has noted, “… in the course of time the coordinated actions of industry, government, and the biomedical research community have degraded the basic rules of empirical science…” We would do well to acknowledge this fact, and to recognize with Abramson and Starfield that “The first step is to give up the illusion that the primary purpose of modern medical research is to improve Americans’ health most effectively and efficiently. In our opinion, the primary purpose of commercially funded clinical research is to maximize financial return on investment, not health.”
When corporations are involved, there is no point in prolonging the myth of noble and dispassionate clinical scientists searching for truth in clinical trials. It’s over. We would do better to stop pretending that corporate articles in medical journals are anything but marketing messages disguised with the fig leafs of coöpted academic authors and of so-called peer review. The case study reported out by Drs. John Jureidini, Jay Amsterdam, and Leemon McHenry shows us the real face of business as usual in commercial clinical trials. That being the case, it makes no sense to expect corporations and academic key opinion leaders suddenly to reform their biased and conflicted behavior. Only a structural change from the outside like I propose here has any chance of succeeding. The statistical analysis of clinical trials is too important to be entrusted to the sponsoring corporations.
It is time for Congress to grasp this nettle. The time for enforcement discretion is past, and we need Congress either to direct the FDA to act or to create a new mechanism of oversight. To do nothing would be unthinkable.
Bernard J. Carroll
Professor and Chairman Emeritus,
Department of Psychiatry,
Duke University Medical Center.
The writer is a former chairman of the Psychopharmacologic Drugs Advisory Committee, Food and Drug Administration, U.S Public Health Service.
Acknowledgment: Several colleagues commented and made suggestions on drafts of this post – in particular John M. Nardo, MD, Donald F. Klein, M.D., and Patrick Skerrett from STAT News.
Update 06-23-2016: This post was cross-posted on Naked Capitalism, where some interesting comments can be found.