Responder Analyses—A PhRMA Position Paper

Abstract

Ideally, a clinical trial should be able to demonstrate not only a statistically significant improvement in the primary efficacy endpoint, but also that the magnitude of the effect is clinically relevant. One approach to address this question, often proposed by clinical societies and regulatory guidance, is a responder analysis, in which a continuous primary efficacy measure is dichotomized into "responders" and "nonresponders." This article represents a Pharmaceutical Research and Manufacturers of America (PhRMA) position on responder analyses. With respect to demonstration of the existence of a treatment effect, we find that the well-known loss of statistical power associated with a responder analysis outweighs any real or perceived benefits of this approach. However, between-group comparisons of the percentages of "responders" can play a role in the assessment and reporting of the clinical meaningfulness of the treatment effect.

Personalized medicine: Time for one-person trials

“…Researchers typically get disappointing results with a drug in large, population-based trials. This leads them to conduct ad hoc post-trial analyses, to try to identify the factors that cause some of the people in the trial to seem to be responsive 3 .…”

Interpretation of patient-reported outcomes

“…8,50 Consider a situation where lower change or more negative scores are better or more favorable ( Figure 6). In Figure 6, 70% of the subjects in the experimental group had scores of 10 or less (that is, 10 or better) compared with 55% of the subjects in the control group.…”

Incorporation of Clinical Meaningfulness Into the Analysis of a Continuous Variable: A More Powerful Alternative to the Responder Analysis

“…One well-known disadvantage of the responder analysis is reduced power relative to an analysis of the original continuous variable (Deyi, Kosinski, and Snapinn 1998;Fedorov, Mannino, and Zhang 2009;Senn and Julious 2009;Uryniak et al 2011). In this section we present the results of a brief simulation study to evaluate the power associated with the four analysis variables described above in order to get a sense of the degree to which analyses of the two alternative transformed variables, labeled Transformation 1 and Transformation 2, preserve the power of the analysis of the original continuous variable.…”

“…While in common use in several therapeutic areas, this approach has been criticized on multiple grounds (Snapinn and Jiang 2007;Uryniak et al 2011), most notably for a substantial cost in statistical power relative to an analysis of the original continuous variable (Deyi, Kosinski, and Snapinn 1998;Altman and Royston 2006;Fedorov, Mannino, and Zhang 2009;Senn and Julious 2009). The responder analysis does attempt to address an important issue: ensuring that the effect of the treatment is not only statistically significant, but also clinically meaningful.…”