Rates of Adherence to Colorectal Cancer Medications and Predictors of Non-Adherence: A Systematic Review and Meta-Analysis

PURPOSE: To synthesize adherence rates to colorectal cancer medications and identify predictors of nonadherence. METHODS: Following PRISMA, we searched PubMed, Embase, PsycINFO, and Web of Science through August 13, 2024. Observational studies reporting adherence or predictors were eligible. Two reviewers independently screened and extracted data and assessed risk of bias using JBI Checklist for Cohort Studies. Adherence was grouped by measurement approach: claims-based PDC/MPR, chart/clinical record review, or patient-reported. Random-effects meta-analyses were performed within clinically and methodologically homogeneous subgroups. RESULTS: Thirteen studies (n = 13) met inclusion, with adherence ranging from 33% to 100%. In claims-based analyses using PDC/MPR thresholds, pooled adherence was about 40% [95% CI, 0.36-0.44] with substantial heterogeneity (I-squared = 84.6%). Pooled adherence was about 83% in both chart/record [95% CI, 0.44-0.97] (I-squared = 71.4%) or patient-reported measures [95% CI, 0.68-0.92] (I-squared = 93.8%), also with substantial heterogeneity. Nonadherence was more likely with advanced stage, ECOG greater than or equal to 1, multiple prior regimens, female sex, and treatment-related adverse events. The overall risk of bias was low, although some included studies lacked complete follow-up or strategies to address it. CONCLUSION: We synthesized adherence to CRC medications and identified consistent predictors of nonadherence. Adherence was lowest with claims-based PDC/MPR and higher with chart or patient-reported measures. These findings support targeted interventions for patients at higher risk of non-adherence, including those in the advanced stage of the disease, those with multiple regimens, and those experiencing adverse events. Future work should use standardized adherence definitions and metric-specific reporting to enable valid pooling.