Bob Nakagawa and Rick Hudson suggest that captopril was the ACE inhibitor of choice in hypertension at the time of our study and that patients in hospitals might have been preferentially prescribed captopril. Our data, however, show that among first-time users of ACE inhibitors, the use of captopril was considerably lower than that of the other 2 agents; this does not indicate preferential use of captopril initially. Moreover, our study included only prescriptions dispensed on an outpatient basis, and availability of the drug on the hospital formularies should not have direct relevance to our study. Thus, we disagree with the claim that we present "little evidence" of the presence of therapeutic differences among ACE inhibitors on that basis.
Maurice McGregor contends that our conclusions are premature. He refers to a study by Caro and colleagues1 that showed very low rates of persistence with ACE inhibitors in a similar cohort. We agree entirely that one should take such changes in drug use into account when trying to infer causality between drug use and subsequent use of health services, which our study did not. Our intent-to-treat analysis was a first step in using population-level data to assess whether agents belonging to the same therapeutic class differ in respects other than simply their chemical structures, such as the way they are prescribed to different patients and their impact on health services utilization. Additional studies accounting for complex patterns of drug use would be welcome.
Reference-based pricing policies aim to ensure that the more cost-effective medication is used. Although we are advocates of this approach, we believe that such policies should be carefully evaluated, not only in terms of health-related spending but also in terms of population health. McGregor's statement that "this is a provocative study that merits clarification" is certainly true. Indeed, our study had several methodological limitations, as Paul Grootendorst and Anne Holbrook correctly pointed out in an accompanying editorial,2 and there may be other plausible explanations for the observed differences. However, population-based studies are essential in evaluating whether policies aimed at reducing costs may not in fact increase long-term costs and, more important, negatively affect public health.