Objective: To examine adherence to adalimumab (ADA) and etanercept (ETA) and health care costs in rheumatoid arthritis (RA) patients, and to explore the association between adherence, utilization and costs.
Research design and methods: Using administrative claims data from a large managed health care plan, RA patients treated with etanercept or adalimumab during the period from 01/01/2005 through 12/31/2005 were identified. The first dispensing date was defined as the index date. Patient adherence and costs were assessed during the 1 year post-index period.
Main outcome measures: Nonadherence (medication possession ratio <80%) was modeled using logistic regression. Hazard ratios (HR) comparing time to discontinuation were estimated using Cox proportional hazard (PH) models. Propensity score matching with multivariate generalized linear modeling adjustment was done to assess cost difference between ADA and ETA.
Results: Of 3829 eligible RA patients, 1292 (765 existing, 527 naïve) and 2537 (1834 existing, 703 naïve) patients used ADA and ETA, respectively. Compared with ADA users, ETA users had longer average treatment duration (316 vs. 291 days; p < 0.0001). Unadjusted adherence rates for naïve and existing users were 63% and 70% (ADA), and 65% and 73% (ETA). Logistic regression analysis indicated that compared with ETA users, ADA users were more likely to be nonadherent (OR, naïve 1.24; existing; 1.25). Cox PH models indicated that existing ADA users were more likely to discontinue (HR = 1.11; p = 0.06) their medication than existing ETA users. Compared with ADA users, ETA users had significantly lower RA-related pharmacy costs (naïve: $10,892 vs. $12,534, p < 0.01; existing: $12,192 vs. $13,752, p < 0.01) and RA-related total costs (naïve: $11,976.42 vs. $13,511.99, p < 0.05; existing: $14,031 vs. $15,454, p < 0.05).
Conclusions: ETA users had longer treatment duration, were more likely to adhere to their medication regimen and had lower RA-related pharmacy and RA-related total costs compared with ADA users. These findings must be considered within the limitations of this database analysis.