A simple method for calculating power based on a prior trial.

Abstract

OBJECTIVE: When an investigator wants to base the power of a planned clinical trial on the outcome of another trial, the latter study may not have been reported in sufficient detail to allow this. For example, when the outcome is a change from baseline, the power calculation requires the standard deviation of the difference, and it frequently happens that only the standard deviations of the baseline and the follow-up measurements are reported. Also when a complex analysis or an analysis with covariates is planned, the power calculation may be difficult or impossible. The objective was to develop a method to determine the power of a trial, based on minimal information from a previous (reference) trial. STUDY DESIGN AND SETTING: We investigated the power calculation for a range of statistical methods, including the t-test, analysis of covariance, analysis of variance, linear regression, logistic regression, Poisson regression, the Wilcoxon test, and the logrank test. RESULTS: A method to calculate the power of a trial solely based on the P-value or the confidence interval of the outcome of the reference study. CONCLUSION: A power calculation based on an earlier similar trial only requires its P-value.