Background/Purpose:  Rheumatologists have many choices of medications to use for patients with rheumatoid arthritis (RA), but patients may not fill a prescription (primary non-adherence or PnA). Even for those who do initiate therapy, they may discontinue treatment shortly thereafter (secondary non-adherence or SnA). The factors that underlie never starting on therapy as distinct from early discontinuation are likely to be different but have not been well characterized. This study investigated predictors of PnA and SnA to to methotrexate (MTX), biologics/tofacitinib (B/T) in RA patients.

Methods:   A retrospective cohort analysis was conducted using a U.S. integrated claims and electronic health records (EHR) database. Patients were required to have a new prescription or treatment of MTX or B/T (index event) (12+ month clean period) with >=2 RA physician diagnoses (714.0, 714.2, or 714.81), no other autoimmune diseases or cancer before index, and continuous pharmacy and medical insurance coverage and EHR data availability before and after index. PnA was defined as a new written prescription of MTX or B/T (EHR data) not initiated within 90 days (linked pharmacy claims data). SnA was defined using only pharmacy data as treatment discontinuation (a gap >=90 days) within 12 months of initiation. Potential predictors (n=414 from EHR and claims data) were selected based on clinical judgment and included in separate lasso-regularized logistic regression.

Results: Of 1,117 and 744 patients with a new written prescription, 24% (MTX) and 27% (B/T) failed to initiate them. For those who started on treatment 12-month discontinuation rate was 50% for MTX and 40% for B/T. For PnA, multiple baseline treatment-based comorbidity markers positively and diagnosis-based comorbidity markers negatively impacted MTX primary adherence, whereas for B/T almost no comorbidity markers were significant. Model discrimination was good (AUC=0.81 for MTX, 0.71 for B/T). For SnA, some psychosocial issues such as depression and anxiety were associated with MTX discontinuation. For B/T, infection-related conditions at baseline were associated with discontinuation (Tables 1 and 2).

Conclusion:  Using novel linked EHR and pharmacy data, approximately 25% of new prescriptions for MTX, biologics or tofacitinib were not filled by patients within 90 days. Different factors were related to primary nonadherence (never filling first Rx) vs. secondary non-adherence (early discontinuation), implying the need for different interventions to improve adherence to MTX and B/T, both at the time of first fill and over time. Table 1: Logistic Regression Estimates of Primary Adherence: Odds ratio associated with filling first MTX and B/T prescription          

P < 0.05; **P < 0.01; ***P < 0.001. Table 2: Logistic Regression Estimates of Secondary Adherence: Odds ratio associated with 12-continuation with MTX and B/T 

P < 0.05; **P < 0.01; ***P < 0.001.