(B-100) Bayesian Analysis of Factors Associated with Romosozumab Adherence Among Early Adopters in Medicare Beneficiaries

Background: Understanding adherence among early adopters of newly licensed drugs is crucial for management, but small study populations pose challenges. Romosozumab (Romo), approved in the U.S. in April 2019, is a parenteral osteoporosis (OP) treatment for postmenopausal women at high fracture risk. This study used Bayesian methods to explore factors associated with Romo adherence among its early adopters.

Objectives: To identify factors associated with Romo adherence using Bayesian regression.

Methods: Women aged ≥65 who initiated Romo in 2019 (as early adopters) with ≥4 years of continuous enrollment, including 3 years pre-initiation, were identified from 2006-2022 Medicare fee-for-service data. Women with baseline metastatic cancer or Paget’s disease were excluded. Based on prior research, we stratified the analyses by age (younger: ≤75; older: >75 years), and selected adherence predictors including race, geographic region, metropolitan residence (MR), Charlson Comorbidity Index (CCI), chronic kidney disease (CKD), COPD, antidepressant use (anti-DP), antidiabetics (anti-DM), hormone therapy (HT), steroids, and ER visits. Prior information of predictors was derived from previous literature. High Romo adherence was defined as proportional days covered (PDC) ≥80% over 3 months. Frequentist and Bayesian logistic regression models were evaluated Bayesian models reported odds ratios (ORs) with 95% credible intervals (CrI).

Results: Among 1,777 beneficiaries (younger: n=752; older: n=1,025), 77.8% of women had high adherence. Frequentist models found only that White race was associated with high adherence (OR=2.42, 95%CI: [1.35–4.36]) in younger patients and COPD was associated with decreased adherence in older patients (0.64[0.44–0.91]).
Bayesian analysis revealed additional associations. In younger patients, anti-DP (OR=0.81, 95% CrI: [0.74–0.89]), CKD (0.88 [0.81–0.97]), COPD (0.79 [0.74–0.83]), anti-DM (0.95 [0.91–0.99]), ER visits (0.90 [0.84–0.96]), and higher CCI (0.78 [0.70–0.88]) were associated with decreased adherence, while White race (2.83 [2.37–3.26]) , MR (1.29 [1.13–1.47]) and HT (1.05 [1.01–1.09]) were associated with high adherence. In older patients, CKD (0.89 [0.84–0.96]), COPD (0.81 [0.77–0.86]), steroids (0.78 [0.73–0.85]), and higher CCI (0.83 [0.77–0.90]) were associated with decreased adherence, while White race (2.05 [1.82–2.28]) and MR (1.34 [1.22–1.48]) were associated with high adherence.

Conclusions: Bayesian analysis identified additional adherence factors among early Romo adopters compared with frequentist models, including anti-DP and HT. Future research should consider Bayesian modeling when studying medication adherence, particularly when sample size may be limited.