-
Views
-
Cite
Cite
Pamela M. Murnane, R. Yates Coley, Jared M. Baeten, Murnane, Coley, and Baeten Respond to “Every Good Randomization Deserves Observation”, American Journal of Epidemiology, Volume 182, Issue 10, 15 November 2015, Pages 861–862, https://doi.org/10.1093/aje/kwv201
Close - Share Icon Share
Extract
We thank Westreich and Edwards for their insightful commentary (1) on our paper about estimating the efficacy of preexposure prophylaxis (PrEP) for human immunodeficiency virus (HIV) prevention in a randomized trial with imperfect adherence (2). Westreich and Edwards address the important question of how we might generalize or transport efficacy estimates from study samples to real-world populations. We all agree that intention-to-treat efficacy estimates from randomized trials, despite being study-population average estimates and unbiased in their design, do not necessarily represent real-world effectiveness.
A common assumption is that persons who enroll in randomized trials are highly motivated and, barring significant side effects or other unacceptability of the intervention, will adhere to the study protocol. Results from PrEP trials have demonstrated that such protocol compliance does not always occur. Poor adherence in these trials may be attributable to qualities of the intervention itself (e.g., the burden of daily pill-taking, stigma around using medicines commonly prescribed for treatment of HIV infection) as well as the clinical trial process (e.g., uncertainty about the efficacy of an unproven product and use of placebo) (3, 4). Ultimately, adherence likely reflected participants’ personal motivation about the research question and/or their commitment to doing what researchers requested of them, perhaps more so than the real-world direct desire to prevent HIV acquisition. Indeed, the results of PrEP trials are a stark reminder regarding any intervention requiring ongoing adherence by trial subjects, including interventions for health conditions other than HIV, that intention-to-treat efficacy might significantly underestimate true biological effects and potentially real-world effects as well.
