What Should Be Done About Bias and Misconduct in Clinical Trials?

Abstract

• A 2007 analysis of promotional detailing visits for gabapentin from 1995-1999 found that 46% led to physician intention to prescribe or recommend the drug, and 38% of those visits involved discussions in which the "main message" was at least 1 off-label use. Various components of the health care system, including government regulators, journal editors, and pharmaceutical manufacturers, are cobbling together a patchwork of solutions to the problem of inaccurate "evidence" about the efficacy and safety of prescription drugs. But will their efforts be sufficient to encourage Turner et al. found a strong association between favorability of results for the study drug and likelihood of publication. Overall, of 74 FDA-registered studies conducted in 12,564 patients, 51 (68.9%) were published, including 37 of 38 studies (97.4%) that the FDA had judged as indicating positive findings for the study drug. In contrast, of 36 studies judged by the FDA to have produced negative (n = 24) or questionable (n = 12) results, 3 (8.3%) were published with the negative findings presented, 11 (30.6%) were published with a representation that the findings were positive, and 22 (61.1%) were not published. The researchers concluded that "the efficacy of [the antidepressant] drug class is less than would be gleaned from an examination of the published literature alone . . . . As a result of selective reporting, the published literature conveyed an effect size nearly one third larger than the effect size derived from the FDA data," leading physicians to make "inappropriate prescribing decisions."6 Whittington et al., studying trials of the use of selective serotonin reuptake inhibitors in children and adolescents with depression, had produced similar findings in 2004, concluding that when the unpublished data were considered, "risks could outweigh benefits of these drugs (except fluoxetine) to treat depression in children and young people . . . . Non-publication of trials, for whatever reason, or the omission of important data from published trials, can lead to erroneous recommendations for treatment." In an analysis that examined the reporting of clinical trial data used in new drug applications submitted to the FDA, Lee et al. (September 2008) identified 909 trials supporting 90 new drugs approved between 1998 and 2000, using FDA review documents and the approved drug label. The authors categorized trials as "pivotal" if they were described in both the summary document for drug approval and the "clinical studies" section of the E D I T O R I A L

Drug–drug interaction trials incompletely described drug interventions in ClinicalTrials.gov and published articles: an observational study

“…The expanded FDAAA mandated the inclusion of the ID element to trial data, as specified in the 2016 Final Rule for Clinical Trials Registration and Results Information Submission, released by the NIH [33], and effective for trials initiated on or after January 18, 2017. This data element was formed to provide timely insight into the methodology of a trial and to possibly prevent biased results [33,34]. Nevertheless, we are mindful that 88% of analyzed trials were excluded from the FDAAA legislation and ICMJE requirements, mostly because of phase 1 status, which could be a potential source of bias in our study.…”

Selling Real-World Health Care Research to Reluctant Buyers-Evidence-Based Education or Marketing a Defective Product?

“…As we and others have observed previously, the practices of selective reporting of study findings, publication planning, and other forms of misconduct are, sadly, reportedly endemic in health care research. 2,3 Studies that use observational or "real-world" data, particularly pharmacoeconomic modeling and retrospective analyses of administrative databases, are particularly vulnerable to manipulation; it is especially easy to make post hoc changes to a planned protocol behind closed doors when only claims data and hypothetical patient populations, not prospectively studied human subjects, are involved. 4 Thus, to the extant problem presented by Brixner et al in their commentary on use of real-world data in this issue of JMCP 5 -that decision makers are sometimes reluctant to rely on analyses of real-world data-one reasonable response is that the most reluctant "buyers" of research may well be the best informed.…”