Pharmaceutical Statistics has just published a special issue on estimands.

Editor Gerd Rosenkranz writes that ‘estimands offer the opportunity to define the effect of interest more clearly than in the current framework by considering post-randomization events that can interfere with the statistical analysis at the design stage of a clinical trial. Particularly, it avoids an approach that starts with defining analysis sets and analysis methods and lets the estimand turn out to become what is estimated by the analysis. Because not every estimand we can dream of may allow for an easily estimable effect estimator, the process of agreeing on estimands upfront can influence the definition of problem specific estimands to prevent difficulties on the analysis and interpretation side.

The papers in this issue of Pharmaceutical Statistics should be considered contributions to a vivid discussion rather than the final words. Taken seriously and pragmatically, estimands have the potential to become a new principle to improve the relevance of statistics for the interpretation of clinical trials data. We will see more in the time ahead, and hopefully, we have more clarity about the way forward.’