Interaction in Cardiovascular Crossover Studies: The Standard and the Clinical Analysis

Abstract

The crossover trial, which is considered the most powerful means of deter mining the efficacy of cardiovascular drugs, is frequently biased by treatment- by-period interactions. A second problem is that the standard analysis (Hills-Armitage analysis) for these biases is unreliable because of a low sensitiv ity. The author recently introduced an alternative method for the detection of interaction bias entitled the clinical analysis, because it looks at the clinical performance of the separate treatment groups in a trial and not, as in the stand ard analysis, at the means of the groups. In a mathematical model this analysis was capable of detecting interaction at a 30% lower level than the standard analysis. The present study was undertaken to show not only that the clinical analysis performs better than the standard but also that it enables the investiga tor to differentiate among different types of interactions, eg, physical carryover effect, rebound phenomenon, and psychological carryover effect.