Methods to increase efficiency in clinical trials with restricted sample size

[摘要] Efficiency is a perennial motivation of statistical analysis and clinical trials.This is most pertinent when sample size is constrained. When trials and their analyses are more efficient, results can be more precise, can be disseminated quicker, and impact the clinical pathway faster. This thesis describes methods developed and investigated by the author in three trials at the Cancer Research UK Clinical Trials Unit.Methods for seamless phase I/II trials that conduct dose-finding by efficacy and toxicity outcomes are studied. A repeated measures analysis in an ultra-rare disease yields a feasible trial where standard approaches do not.Finally, this thesis develops methods for a phase II trial with co-primary outcomes and predictive covariate information. We conclude that two common goals to increase efficiency are: i) use more outcomes to answer trial questions; and ii) use all available information. In our examples, analysing efficacy and toxicity in dose-finding lets these trials simultaneously achieve phase I and II objectives.However, this thesis highlights operational issues that can impair efficiency. We show that statistical performance is improved by analysing the information in repeated measures and predictive baseline covariates. Methods developed herein help to achieve conventional error rates without prohibitive increases in sample size.

[发布机构] University:University of Birmingham;Department:Institute of Cancer and Genomic Sciences

[关键词] R Medicine;R Medicine (General)