Abstract

We appreciate the comments raised by Subbaraman et al, highlighting valid concerns regarding the use of per-protocol (PP) rather than intention-to-treat (ITT) results in cost-effectiveness analyses. The authors appropriately pointed out that PP analyses are subject to potential selection bias. As such, using results from PP may result in misleading, and often overly optimistic, conclusions regarding the interventions’ cost-effectiveness. We fully agree with this premise. But blind reliance on ITT results is equally dangerous, especially in the context of pragmatic trials, which increasingly—and in our minds, appropriately—play an important role in informing decisions of whether to scale up interventions.