Use of Bayesian net benefit regression model to examine the impact of generic drug entry on the cost effectiveness of selective serotonin reuptake inhibitors in elderly depressed patients.

INTRODUCTION: Since their invention in the late 1980s and early 1990s, selective serotonin reuptake inhibitors (SSRIs) have become the primary form of pharmaceutical treatment for depression. As the patents of several top-selling SSRIs have expired or are soon to be expired, the SSRI market is expected to witness an increasing share of generic SSRIs. We explored the impact of generic drug entry on the cost effectiveness of SSRIs.
METHOD: Using Medicare MarketScan claims data, we compared the cost effectiveness of sertraline, citalopram, escitalopram and fluoxetine with paroxetine in elderly depressed patients, before and after the entry of generic paroxetine. We followed users of SSRIs for 6 months, starting from the date of their first prescription of an SSRI. For each patient, we measured costs (C(i)) as total medical costs and quantified effectiveness (E(i)) as the avoidance of treatment failure, which was defined as having a break exceeding 45 days in the use of antidepressants. We then calculated individual net benefit as lambda x E(i)- C(i) and employed both net benefit and Bayesian net benefit regression models to examine the impact of generic paroxetine on the cost effectiveness of the other four SSRIs compared with paroxetine, while controlling for patients' sociodemographic characteristics, co-morbidities and patterns of medication switch.
RESULTS: Deterministic analysis showed that paroxetine was dominated by most SSRIs prior to the availability of generic paroxetine, and that, after the entry of generic paroxetine, citalopram and escitalopram were dominated by paroxetine. Net benefit regression analysis found that, at a number of lambda values ($US1000, $US5000 and $US10,000), sertraline and escitalopram were more cost effective than paroxetine in the pre-generic-entry period but not in the post-entry period, although the difference in net benefit between the two SSRIs and paroxetine was not statistically significant in both periods. The Bayesian net benefit regression analysis reached similar conclusions. At lambda = $US5000, the probability that sertraline, citalopram, escitalopram or fluoxetine was more cost effective than paroxetine was 96.7%, 77.6%, 96.3% and 97.0%, respectively, in the pre-entry period in the pooled analysis. These probabilities reduced to 36.7%, 62.7%, 33.0% and 60.1%, respectively, in the post-entry period. The probabilities became 94.1%, 71.9%, 89.1% and 92.1% in analysis using the pre-entry data as a prior to update the post-entry data rather than using the pooled data.
CONCLUSION: Using generic drug entry as an example, our study demonstrated the importance of including the economic life cycle of pharmaceuticals in cost-effectiveness analyses. Additionally, the proposed Bayesian framework not only preserves the advantages of the net benefit regression framework, but more importantly, it introduces the possibility of conducting probabilistic cost-effectiveness analyses with claims data.