Comparing compliance patterns between randomized treatments

Abstract

When two equally efficacious drugs enter the market, the one with the better compliance is likely to be more widely used. Special management of the delivery may produce increased compliance. In this paper we analyze a trial of a single drug dosing prescription with patients randomized to either daily self monitoring of the outcome (blood pressure) or not. The study used Medication Event Monitoring Systems (MEMS) to record each exact time and date when a patient opened the pill container. No established method is available for comparing these high-dimensional compliance patterns between groups. This paper investigates several summary measures that highlight different dimensions of the pattern and the drug context in which they may be meaningful. Further, we examine conditional and marginal models that enable comparisons of the full pattern of daily dosing indicators for subjects between the groups. We found no simple difference in average compliance levels, but we found an interesting interaction between treatment and time: similar compliance existed initially among patients in both randomized groups, with a stronger decline over time for patients who did not monitor their blood pressure. We discuss how a balance between simplicity of interpretation and efficiency of data use may be sought in this case.