Parallel-Group and Crossover Designs for Drug Evaluation in the Various Forms of Angina Pectoris

The choice of a parallel-group or crossover design depends on a number of considerations, mainly related to the characteristics and prevalence of the disease. Crossover designs have the advantage of greater statistical efficiency (which may be particularly useful when patient recruitment is a problem); however, they may lead to equivocal results for a number of practical and statistical reasons. In addition, the small number of patients involved in crossover trials precludes any detailed evaluation of the efficacy/risk ratio. The statistical efficiency of parallelgroup designs may be increased by incorporating baseling measurements in the analysis and by using factorial designs.