Annals of Internal Medicine
May 15, 2012; 156 (10)
Ethical Dilemmas and Malfunctions in Clinical Trials Research
Stuart J. Pocock
Ann Intern Med May 15, 2012 156:746-747;
Two articles in this issue raise ethical concerns about the conduct and monitoring of clinical trials that are worthy of debate and future action.
Fan and colleagues (1) report a trial of a behavioral intervention strategy for patients with chronic obstructive pulmonary disease. The trial was stopped early primarily because of a statistically significant excess of deaths in the intervention group. This reminds us all that trials of behavioral strategies are no different from drug trials in their need for effective data monitoring for potential harm, even though some might argue that behavioral actions cannot conceivably harm patients. Appropriate structures (for example, planned interim analyses and an independent data monitoring committee) are an integral part of any major trial’s ethical armory in guarding against and acting on nasty surprises.
Having said that, just how convincing is the evidence of harm in this case? Twenty-eight deaths (17%) occurred in the comprehensive care management group compared with 10 deaths (7%) in the usual care group (P = 0.003). Trials that stop early because of an apparent treatment difference, whether for superiority or inferiority, are prone to exaggerate the true effect (2, 3). They stop on a “random high,” whereas the observed difference might well have regressed to the truth if they had continued longer. Inevitably, that unobservable truth will remain a mystery, but the idea of potential exaggeration is reinforced by the following 3 issues: The causes of death did not reveal any clear pattern consistent with a plausible reason for harm; the trial’s primary end point (time …