Purpose: To investigate whether demographic, clinical and treatment-related risk factors known at treatment initiation can be used to reliably predict future hormonal therapy non-persistence in women with breast cancer, and to inform intervention development.
Methods: Women with stage I-III breast cancer diagnosed 2000-2012 and prescribed hormonal therapy were identified from the National Cancer Registry Ireland (NCRI) and linked to pharmacy claims data from Ireland's Primary Care Reimbursement Services (PCRS). Non-persistence was defined as a treatment gap of ≥180 days within 5 years of initiation. Seventeen demographic, clinical and treatment-related risk factors, identified from a systematic review, were abstracted from the NCRI-PCRS dataset. Multivariate binomial models were used to estimate relative risks (RR) and risk differences (RD) for associations between risk factors and non-persistence. Calibration and discriminative performance of the models were assessed. The analysis was repeated for early non-persistence (<1 year of initiation).
Results: Within 5 years of treatment initiation 680 women (19.9%) were non-persistent. Women aged <50 years (adjusted RR 1.41, 95% CI 1.16-1.70) and those prescribed antidepressants (RR 1.22, 95% CI 1.04-1.45) had increased risk of non-persistence. Married women (RR 0.82 95% CI 0.71-0.94) and those with prior medication use (RR 0.62 95% CI 0.51-0.75) had reduced risk of non-persistence. The area under the receiver-operating characteristic (ROC) curve for non-persistence was 0.61. Findings were similar for early non-persistence.
Conclusion: The risk prediction model did not discriminate well between women at higher and lower risk of non-persistence at treatment initiation. Future studies should consider other factors, such as psychological characteristics and experience of side-effects.
Keywords: Adherence; Breast cancer; Hormonal therapy; Medication taking behaviour; Persistence.