Conditional power for assessing population interventions

AIM: To calculate conditional power in comparative two-period studies with previously observed baseline data. METHOD: Isolate the variability attributable to the yet-to-observed data and modify the standard power formulae. RESULTS: For illustration, we examine rates of opioid overdose before and after a reformulation of one opioid product. The null hypothesis posited no impact of the reformulation, alternative hypotheses posited possible impacts, and ancillary hypotheses posited different secular pre-post changes directly observable in comparators. Conditional power varied with the size of the comparator population and with the assumed pre-post change for the comparator. CONCLUSION: Pre-post designs can be initiated after the baseline period is over. Power calculations that are conditioned on observed baseline data account differently for variability in the baseline and follow-up periods.