The Editorial that Wasn't: Evidence for Systematic Research Manipulation Undetectable by Critical Review

Woe to those of us who have been advocates for evidence-based medicine.  A short description of the evidence-based medicine is medicine whose practice is informed by critical, rigorous review of the best available evidence from clinical research as revealed by systematic search of the published research literature, as well as by the clinician's understanding of biology and the medical and biopsychosocial context, and by the patient's own values.

Evidence-based medicine is based on some key assumptions.  One is that a systematic review will reveal all the results of research studies that are relevant to the issue at hand.  A second is that while the research studies may be flawed and imperfect, they are reported honestly.

Unfortunately, as we have repeatedly discussed, there is more and more evidence that a systematic review will not reveal all relevant results, because research studies may be suppressed, perhaps often, when their results are unfavorable to vested interests.  (Look here for further discussion.)  There is also considerable anecdotal evidence that the design, implementation, and analysis of research studies may be manipulated to make the results more likely to favor vested interests.  (Look here for further discussion.)

An article by Vedula et al just published in the renowned New England Journal of Medicine suggests that such manipulation might be systemic, and that the reporting of manipulated studies may not clearly show what manipulation was done. (1)

To summarize, the authors got access through legal proceedings to internal study protocols and research reports from clinical trials sponsored by Parke-Davis (later merged into Pfizer Inc) of gabapentin (Neurontin) for a variety of clinical problems other than seizures.  Gabapentin was originally marketed as an anti-seizure drug, but Pfizer later "admitted guilt for off-label marketing."  (We discussed the stealth marketing campaign for Neurontin here.)  Vedula et al compared the primary outcome variables specified in the original research protocols, internal research reports, and any publications of the trials' results.  They identified 21 trials, 13 of which were published.  For 12 of the 13, the authors had access to the internal protocol, report, or both.  The main results were:

For 8 of the 12 published trials, there was a disagreement between the definition of the primary outcome in the protocol and that in the published report.... Sources of disagreement included the introduction of an entirely new primary outcome in the published report (in the case of 6 trials); failure to distinguish between primary and secondary outcomes in the published report, even though the protocol did distinguish between them (2 trials); relegation of a primary outcome in the protocol to a secondary outcome in the published report (2 trials); and failure to include in the published report one or more primary outcomes specified in the protocol (5 trials).

Furthermore, it appeared that failure of published articles to clearly and fully report results in terms of the original, pre-specified primary outcome variables occurred when these comparisons were not favorable to gabapentin. As the authors summarized:

Thus, trials with findings that were not statistically significant (P≥0.05) for the protocol-defined primary outcome, according to the internal documents, either were not published in full or were published with a changed primary outcome.

As shown in Figure 3, all the changes that took place between what was specified in the protocol, what was known before publication (as presented in the internal company research reports), and what was reported to the public led to a more favorable presentation in the medical literature of gabapentin's efficacy for unapproved indications.

They concluded:

We are concerned that the reporting practices observed in our analysis do not meet the ethical standards for clinical research or maintain the integrity of scientific knowledge. Fair and honest treatment of patients enrolled in clinical trials of any kind requires full, open, and unbiased reporting. Journal publication, a formalized platform for scientific discourse and dissemination of knowledge, should not be used as a marketing tool for off-label drug use.

Reporting biases such as those we describe here increase the likelihood that interventions will appear to be effective when they are not. Such biases can lead to the omission of negative findings in systematic reviews of intervention effectiveness and in evidence-based guidelines. For example, the 2005 Cochrane systematic review regarding the effectiveness of gabapentin for acute and chronic pain concluded that it is effective on the basis of published findings and should now be updated with the inclusion of unpublished information made available through litigation.

I believe that the article by Vedula et al is particularly important because it shows what appears to be systematic manipulation of the analysis and reporting of multiple clinical trials of the same drug (but for different indications) that had the effect of making the drug appear efficacious when it likely was not. Furthermore, the manipulation was concealed. The published research articles did not completely describe what the intended primary efficacy outcome variables were, did not provide results in terms of these variables, and instead provided results only in terms of variables that were chosen post-hoc as new primary outcome variables.

To address that latter point, I independently reviewed three of the research publications cited by Vedula et al.(2,3,4) All three were noted to have reported significant results favoring gabapentin in terms of a primary efficacy outcome variable that was not identified as such in the corresponding studies' original research protocols or reports (see the supplementary data provided with the New England Journal article.) Per my review, none of the three published articles offered any hint that what they reported as primary outcome variables were not the variables originally chosen in that capacity, nor did they identify what those original primary outcome variables were, or how comparisons made using them turned out.

The implications of the article by Vedula et al are very important. Hence, I was surprised that the article appeared without an accompanying editorial to discuss these implications, and that its publication did not generate much media interest.

So that gives me an opportunity to comment further.

One could start with the implications for evidence-based medicine. As noted above, a short description of the evidence-based medicine is medicine whose practice is informed by critical, rigorous review of the best available evidence from clinical research as revealed by systematic search of the published research literature, as well as by the clinician's understanding of biology and the medical and biopsychosocial context, and by the patient's own values. Evidence-based medicine depends on critical, rigorous review, but the review process is generally done under the assumptions that research publications honestly describe what was done and what its results were. The review process was never designed to detect dishonest reporting or find information that was deliberately concealed. Manipulation of research (design, analysis, and implementation), concealment of that manipulation, and outright suppression of research threaten the foundations of evidence-based medicine. Yet the article by Vedula et al is part of a growing body of evidence that such manipulation, concealment and suppression are widespread, and done to serve vested interests, often commercial.

That is a huge problem for proponents of evidence-based medicine, but also for physicians who want their practices to be based on science, for patients who want their care to be based on science, and for all those in society who see the advancement of medical science as a way to improve peoples' lives.

As senior author Professor Kay Dickersin noted in an interview with Bloomberg "The trouble is, as a scientist, the publication has always been held up to me as the truth. It's the scientific record. What this study indicated is we can't believe that record."

Furthermore, to be a bit more concrete, most physicians, patients, and policy makers depend on what appears to be honest clinical research to make decisions about individual care and health policy. Deliberate and deceitful manipulation of clinical research to favor sponsors' products has likely lead to excessive use of and payments for drugs and devices that are less effective than advertised, if not useless or dangerous. Thus, it is likely that such manipulation is partially responsible for ever increasing health costs and poor health outcomes.

Finally, we need to start thinking about how we can detect and compensate for manipulation of clinical research in the past, and deter such manipulation in the future. One possible deterrent would be, as was noted by Vedula et al, detailed clinical trials registries that contain complete information about trial protocols.  For this to be effective, there need to be mechanisms to assure compliance, and penalties for non-compliance.  Moreover, since clinical research is now global, the registries must have global scope, and enforced assurance of compliance must also be global.

Registries might decrease future manipulation and suppression of research.   No one has suggested, as far as I know, a systematic way to detect and correct for previous manipulation. It would require a major, global investigative effort to uncover manipulation, and it would be a major scientific and policy endeavor to reveal most suppressed research and correct most manipulations.

However, before anything is done, patients, physicians and policy makers must acknowledge and understand the problem.  Yet it seems that even discussing these may be topics that are very uncomfortable for some of us. The longer we shrink from addressing them, however, the worse will be the results for patients, physicians, science and society.

References


1. Vedula SS, Bero L, Scherer RW, Dickersin K. Outcome reporting in industry-sponsored tirals of gabapentin for off-label use. N Engl J Med 2009; 361: 1963-1971. Link here.


2. Mathew NT, Rapoport A, Saper J et al. Efficacy of gabapentin in migraine prophylaxis. Headache 2001; 41: 119-128. Link here.


3. Vieta E, Goikolea JM, Martinea-Aran A et al. A double-blind, randomized, placebo-controlled, prophylaxis study of adjunctive gabapentin for bipolar disorder. J Clin Psychiatr 2006; 67: 473-477.  Link here.


4. Caraceni A, Zecca E, Bonezzi C et al. Gabapentin for neuropathic cancer pain: a randomized controlled trial from the gabapentin cancer pain study group. J Clin Oncol 2004; 22: 2909-2917. Link here.