To the Editor: Dr. Correa raises three main concerns with our paper. First, he believes that it does not make sense that branded antidepressants outperform generic medications. Our study does not test the effectiveness of branded versus generic medications; rather, the focus is the net effect of step therapy. The comment may reflect a common misunderstanding of step therapy. Step therapy does not merely require that a generic be substituted for a branded form of the same medication. Rather, step therapy for antidepressants requires that a limited list of antidepressants be tried first before other types of antidepressants can be prescribed. We find that this type of formulary design results in patients receiving less antidepressant medication, which we believe is the main explanation for the negative effect of step therapy on outcomes. We hypothesize that the reason step therapy results in less antidepressant utilization is that it creates administrative and financial barriers to receiving prescribed drugs. This explanation is consistent with other empirical evidence, such as the results of a survey of patients who were subject to step therapy that found that 11% subject to step therapy for SSRIs never received the medication, and 24% paid out of pocket for the brand medication (1).

Dr. Correa raises other concerns as well. He points out that there are differences between the baseline characteristics and pre-period utilization of the step therapy and comparison populations. In our study, differences between the two populations were addressed by two methods: (a) outcomes are examined before and after the implementation of step therapy relative to two comparison groups, so that all time-invariant differences are netted out, and (b) multivariate regression adjusts for time-varying differences in between the two populations. We agree that differences in the pre-period trends may have influenced the results; however, pre-post observational designs are always challenged by finding a comparison group with similar trends in the pre-period, and by using two different comparisons, we believe we largely mitigated this threat to internal validity.

Third, Dr. Correa points out that patients often receive multiple psychotropic medications and the study does not account for polypharmacy. While polypharmacy is common, we do not believe that changes in polypharmacy between the two comparison and two intervention populations could be an explanation for why the groups under step therapy would experience a change in antidepressant utilization and medical care utilization after the implementation of step therapy.

Finally, we agree that evaluations of step therapy are complex undertakings and would encourage other researchers to examine the potential unintended consequences of this widely used formulary design.