Assistant Professor University of Toronto University of Toronto Toronto, Canada
Background: Adherence to medications is critical for successful management of disease. While proportion of days covered “PDC” is a common metric for measuring adherence using claims data, it may be insufficient to distinguish clinically relevant differences in medication taking behavior over time. Group-based trajectory models (GBTM) have been used in addition to PDC to better represent medication taking behavior.
Objectives: This study aims to examine the adherence patterns one year after initiation among new-users of sodium-glucose cotransporter 2 inhibitors (SGLT2) using GBTM.
Methods: We conducted a cohort study of adults newly dispensed an SGLT2 from 2014-2018 in the United States using Truven MarketScan. Patients without continuous enrollment, pharmacy benefits, or who had a Type 1 diabetes diagnosis in the 12 months prior to the index date were excluded. We calculated 12-month PDC and categorized patients with PDC ≥ 80% as adherent. We performed multivariable logistic regression on adherence status conditioning on baseline demographics, comorbidities, and medication use. Estimates of associations were presented using odds ratios (OR) and 95% confidence intervals (CI). GBTM with three to six groups were performed to identify adherence patterns based on monthly PDC across twelve months after SGLT2 initiation. The final model was selected based on the lowest Bayesian information criterion value (BIC).
Results: In a cohort of 228,363 SGLT2 users, 45% were female and the mean age was 54.5 years. The mean PDC was 57% with 36% of the cohort being adherent. Overall, females (OR=0.73, 95%CI 0.71,0.74), and patients with anxiety (OR=0.93, 95%CI 0.90,0.96) or depression (OR=0.90, 95%CI 0.87,0.93) were less likely to be adherent. Older age (OR=1.01, 95%CI 1.00,1.02), and patients with antihypertensive dispensing (OR=1.16, 95%CI 1.12,1.20) or statin dispense (OR=1.16, 95%CI 1.14,1.19) were more likely to be adherent. GBTM identified six patterns of SGLT2 adherence: one-fill (PDC=0.08), consistently low (PDC=0.35), early discontinuation (PDC=0.22), moderate adherence (PDC=0.48), high adherence (PDC=0.79), and near-perfect (PDC=0.95). Near-perfect pattern has the highest average age while consistent low has the youngest (56.7 vs. 52.3). One-fill pattern has the highest percent of females while near-perfect has the lowest (53% vs. 37%).
Conclusions: We found a wide variation in adherence patterns among SGLT2 users in a national cohort. Importantly, the patient characteristics are varied across adherence status (PDC≥80%) and pattern.
