OBJECTIVE: Studies comparing medications adherence have become common yet they often do not account for differences in relative follow-up. Patient selection criteria may impact validity and comparability of these studies as well. METHODS: Adults with non-valvular atrial fibrillation, ≥1 rivaroxaban or apixaban dispensing (index date), and ≥1 year of pre-index eligibility were selected from IMS Health Real World Data Adjudicated Claims (IMS RWD Adjudicated Claims) and Truven Health MarketScan Research (Truven MarketScan) databases. Adherence was evaluated using proportion of days covered (PDC) ≥ 0.8 for treatment cohorts: (1) unmatched, with different follow-up, (2) propensity-score matched with similar follow-up, (3) matched, with similar follow-up and ≥2 rivaroxaban or apixaban dispensings, and (4) matched, with similar follow-up and chronic medication users only. Robustness was verified with PDC ≥0.9. RESULTS: In the IMS RWD Adjudicated Claims database, rivaroxaban users had a longer mean follow-up than apixaban users (408 versus 254 days, respectively; p < .01). While crude comparisons demonstrated lower adherence rates for rivaroxaban than apixaban (-12.4 percentage points [pp]; p < .05), these difference attenuated after matching and (1) balancing follow-up (-2.2 pp; p < .05), (2) excluding single-time medication users (0.2 pp; p > .05), and reversed after (3) excluding non-chronic medication users (5.0 pp; p < .05). Results obtained were consistent when these analyses were repeated within the Truven MarketScan databases and when using a PDC ≥0.9. CONCLUSION: Medication adherence comparisons need to account for differences in follow-up. Selection of chronic medication users may impact comparative adherence advantage between medications.
OBJECTIVE: Studies comparing medications adherence have become common yet they often do not account for differences in relative follow-up. Patient selection criteria may impact validity and comparability of these studies as well. METHODS: Adults with non-valvular atrial fibrillation, ≥1 rivaroxaban or apixaban dispensing (index date), and ≥1 year of pre-index eligibility were selected from IMS Health Real World Data Adjudicated Claims (IMS RWD Adjudicated Claims) and Truven Health MarketScan Research (Truven MarketScan) databases. Adherence was evaluated using proportion of days covered (PDC) ≥ 0.8 for treatment cohorts: (1) unmatched, with different follow-up, (2) propensity-score matched with similar follow-up, (3) matched, with similar follow-up and ≥2 rivaroxaban or apixaban dispensings, and (4) matched, with similar follow-up and chronic medication users only. Robustness was verified with PDC ≥0.9. RESULTS: In the IMS RWD Adjudicated Claims database, rivaroxaban users had a longer mean follow-up than apixaban users (408 versus 254 days, respectively; p < .01). While crude comparisons demonstrated lower adherence rates for rivaroxaban than apixaban (-12.4 percentage points [pp]; p < .05), these difference attenuated after matching and (1) balancing follow-up (-2.2 pp; p < .05), (2) excluding single-time medication users (0.2 pp; p > .05), and reversed after (3) excluding non-chronic medication users (5.0 pp; p < .05). Results obtained were consistent when these analyses were repeated within the Truven MarketScan databases and when using a PDC ≥0.9. CONCLUSION: Medication adherence comparisons need to account for differences in follow-up. Selection of chronic medication users may impact comparative adherence advantage between medications.