ABSTRACT
Chapter 4 discusses, and offers examples of, a multitude of trial types, including phase I,II, and III, parallel arm, paired, crossover, cluster-randomized, superiority, noninferiority, and multi-armed trials. We show through the balanced block principle that cluster-randomized trials are inefficient and should be avoided unless the question cannot be answered any other way. Additional design topics include blinding, appropriate controls, choice of primary endpoint, and baseline variables, the latter topic leading naturally to the intention-to-treat (ITT) principle, principal stratification, and regression to the mean. We offer the Cardiac Arrhythmia Suppression Trial (CAST) as a testament to the fact that the effect of a treatment on a surrogate endpoint need not translate to an effect on the clinical outcome of interest, and the Coronary Drug Project as a testament to the importance of the ITT principle. We also illustrate how to reduce variability through replication and averaging, differencing, regression, and stratification. The variability ratio principle implies that reduction of within-arm variability improves power. We revisit the ‘O minus E except after V’ principle and show that it leads to the Mantel-Haenszel test and a related estimator for stratified analysis. We also discuss the measurement and replication principles, which have important implications on the design of clinical trials.
