Appropriate analysis of trial data is vital, but often far from straightforward. It must respect the chosen design, provide a meaningful and unbiased summary of the treatment effect, whilst having sufficient robustness to unforeseen issues. As a practical CTU we are able to maximize the impact of our work by developing, testing and applying our proposed solutions within our own trials.
Missing outcome data
There is a rising awareness of the need to take account of missing outcome data in trials, but the range and novelty of methods is a barrier to many trialists. We focus on practical, accessible methods for exploring the assumptions underlying missing-data methodology, and on frameworks for analysis and reporting.
Treatment switching (which does not reflect routine clinical practice) complicates the interpretation of a trial. For example, control arm patients may receive an unlicensed and/or unfunded experimental treatment on disease progression. Again, our focus is on developing practical methods and frameworks for analysis and interpretation of trials in the presence of treatment changes.
The MRC CTU at UCL is leading the way in developing and implementing best practice for meta-analysis (MA). For example, we show that treatment-covariate interactions, on which stratified medicine relies, can and should be corrected for so-called “aggregation bias” whether with Individual Participant Data (IPD) or with published data. We also clarify when so-called “one-stage” and “two-stage” models are likely to give similar results and when not; and how complex novel trial designs should best be included in both standard and Network MA.