TL21: The Prevention and Treatment of Missing Data in Clinical trials – How far have we come?
*Gosford Aki Sawyerr, Janssen Pharmaceuticals 

Keywords: missing data, prevention, treatment, sensitivity, multiple imputation, pattern mixture

The National Research Council issued the final report on The Prevention and Treatment of Missing Data in Clinical Trials in December of 2012. The panel favored “estimating-equation methods and methods that are based on a statistical model for the data. In particular, weighted estimating equations and multiple-imputation models have an advantage in that they can be used to incorporate auxiliary information about the missing data into the final analysis, and they give standard errors and P values that incorporate missing-data uncertainty. Analyses that are performed with such methods often assume that missing data are missing at random, and such an assumption often makes sense for the primary analysis. However, the observed data can never verify whether this assumption is correct. Therefore, to assess robustness, sensitivity analyses are recommended. We advocate sensitivity analyses that are easy to interpret by clinicians.” (NEJM 2012, Oct. 4, 367(14); 1355-60.)

Between the time of issuance of the draft NAS Report (July 2010), it’s finalization in December 2012, and since then, several submissions/filings have been made to regulatory agencies, in a wide range of therapeutic areas. Several presentations have been made in a wide variety of settings where the theoretical framework has been discussed. Terms such as pattern mixture, MMRM, and multiple imputation, have become more commonplace as the recommendations are being implemented. However, key issues remain regarding choice of estimands, primary endpoint, what sensitivity analyses are to be done and how the basic inference is to be carried out in the presence of these multiple analyses.

In this round table or panel discussion, we would like to discuss regulatory, academic and industry experiences in implementing the recommendations, and key lessons learned. Of additional interest are development of computing resources and decision-making within a multi-disciplinary environment.