To the Editor:
Ian Shrier1 argues that the nature of blinding in the traditional randomized controlled trial (RCT) yields treatment-effect estimates that do not reflect the way in which treatments are administered in the clinical context. It is claimed that “deception” RCTs, in which all participants are told they are receiving an active treatment while some receive placebo, may sometimes be unethical but more closely reflect the clinical context. I present a brief defense of the traditional RCT by arguing that it remains an excellent tool for estimating the causal treatment effect of interest in most blinded trials.
In the clinical context, providers must often choose between administering an active treatment or not. When active treatment is administered, patients will be told they are receiving it. However, when no treatment is administered, patients will presumably not be told they are receiving an active treatment. The deception RCT does not match this clinical context exactly because placebo recipients are deceived; in fact, it is the nonblinded RCT that most closely corresponds to this clinical context.
If the nonblinded RCT reflects clinical reality, why go through the trouble of blinding? The main reason is that in many cases the target of estimation in the RCT is not the total effect of treatment (ie, treatment effectiveness) but rather the pure biological effect of the treatment absent any effects due to what patients are told (ie, treatment efficacy). The arguments for focusing on treatment efficacy rather than effectiveness boil down to generalizability: the biological activity of a treatment is more likely than a placebo effect to be retained across populations.
While it would generally be regarded as unethical, a “covert” RCT, in which patients are randomized to active treatment or placebo without their knowledge, would provide reliable estimates of treatment efficacy. Hence, the appeal of traditional RCTs for estimating efficacy: they are a close but ethical relative of the covert design, revealing as little as possible about treatment assignment. Deception RCTs are a more distant relation and will reliably estimate treatment efficacy only if the deception has the same effect on those receiving active treatment and placebo (an assumption that Dr. Shrier rightly points out is often tenuous).
Although estimating treatment efficacy may not always be of primary interest, when it is, the traditional RCT design remains a worthy choice.
Division of Biostatistics, University of Minnesota School of Public Health, Minneapolis, MN, firstname.lastname@example.org
1. Shrier I. Estimating causal effect with randomized controlled trial. Epidemiology. 2013;24:779–781