2019/06/26 by John C. Whitehead, Whitehead, John, Yasin Desai +3 · 1 citation
Economics, Econometrics and Finance · Mathematics · #Advanced Causal Inference Techniques #FOS: Computer and information sciences #Health Systems, Economic Evaluations, Quality of Life #Methodology (stat.ME) #Statistical Methods in Clinical Trials
paper · pdf · doi:10.48550/arxiv.1906.11324
openalex publication_date 2019/06/26 · openalex created_date 2022/07/22 · openalex updated_date 2026/07/28
When a clinical trial is subject to a series of interim analyses as a result\nof which the study may be terminated or modified, final frequentist analyses\nneed to take account of the design used. Failure to do so may result in\noverstated levels of significance, biased effect estimates and confidence\nintervals with inadequate coverage probabilities. A wide variety of valid\nmethods of frequentist analysis have been devised for sequential designs\ncomparing a single experimental treatment with a single control treatment. It\nis less clear how to perform the final analysis of a sequential or adaptive\ndesign applied in a more complex setting, for example to determine which\ntreatment or set of treatments amongst several candidates should be\nrecommended.\n This paper has been motivated by consideration of a trial in which four\ntreatments for sepsis are to be compared, with interim analyses allowing the\ndropping of treatments or termination of the trial to declare a single winner\nor to conclude that there is little difference between the treatments that\nremain. The approach taken is based on the method of Rao-Blackwellisation which\nenhances the accuracy of unbiased estimates available from the first interim\nanalysis by taking their conditional expectations given final sufficient\nstatistics. Analytic approaches to determine such expectations are difficult\nand specific to the details of the design, and instead "reverse simulations"\nare conducted to construct replicate realisations of the first interim analysis\nfrom the final test statistics. The method also provides approximate confidence\nintervals for the differences between treatments.\n