Each week, we will be publishing layman’s abstracts of new articles from our prestigious portfolio of journals in statistics. The aim is to highlight the latest research to a broader audience in an accessible format.
The article featured today is an Open Access Tutorial in Biostatistics from Statistics in Medicine, with the full article now available to read here.
, , , . Sensitivity analysis for clinical trials with missing continuous outcome data using controlled multiple imputation: A practical guide. Statistics in Medicine. 2020; 39: 2815– 2842. https://doi.org/10.1002/sim.8569
Clinical trials are vital to test the effectiveness and safety of new medical treatments. Unfortunately however, it is most typical that some participant outcomes will not be collected. This may be due to missed participant visits, trial withdrawals or various other unavoidable events. Thus often some required data will be missing from the analysis. Missing data creates a problem since any method of statistical analysis will make an untestable assumption about what the unobserved data looks like. If the wrong assumption is made, then misleading conclusions may be drawn about the treatment under study, which could have disastrous clinical implications. When there are missing data, it is therefore important that primary analysis be conducted under the most plausible assumption for the missing data. Sensitivity analysis under a range of different credible assumptions should then be undertaken to assess how robust the trial results are.
One method which readily enables contextually relevant sensitivity analysis, and has recently seen increased discussion and developments in the statistical literature, is Controlled Multiple Imputation. Controlled multiple imputation procedures include delta-based multiple imputation, which enables investigators to explore the impact of a worse or better outcome for the unobserved, than that predicted based on the outcomes of the observed. An alternative example is reference-based multiple imputation, which enables the impact of individuals with missing data behaving like a specified reference group in the trial to be assessed. For example, in a two-group placebo controlled trial, data can be imputed for all individuals missing data in the active treatment group as if they followed the behaviour of the placebo group.
In this new tutorial article, the authors provide an accessible overview of controlled multiple imputation procedures for missing data sensitivity analysis in clinical trials, and a practical guide to their use for a continuous outcome. Worked examples and Stata code are included to facilitate adoption of such methods, to enable robust evaluation of clinical trial results.