Identifying Tyrosine Kinase Inhibitor Nonadherence in Chronic Myeloid Leukemia: Subanalysis of TAKE-IT Pilot Study

Identifying nonadherence (NA) in chronic myeloid leukemia (CML) remains a challenge. Tyrosine kinase inhibitor adherence was measured by electronic monitoring and Basel Assessment of Adherence to Immunosuppressive Medications Scale (BAASIS) self-report in 55 CML patients over 4 months. The BAASIS had 67% sensitivity and 71% specificity for diagnosing NA. The BAASIS and the risk factors for NA found in this study provide a basis for identifying nonadherent CML patients. Background: There are inconsistencies in reports on correlates for nonadherence (NA) to tyrosine kinase inhibitors (TKIs) in chronic myeloid leukemia (CML). The diagnostic accuracy of subjective adherence measures using electronic monitoring (EM) as the reference standard is yet to be determined. This study aimed to evaluate correlates of TKI NA using EM and test the diagnostic accuracy of subjective adherence measures. Patients and Methods: CML patients receiving a TKI for any duration were enrolled at 4 hematology institutes, and adherence was measured for 4 months. EM adherence was the reference adherence measure, expressed as the percentage of days with the drug taken as prescribed. Subjective adherence was measured using the Basel Assessment of Adherence to Immunosuppressive Medications Scale (BAASIS) self-report and clinician-reported visual analog scale (VAS) at 2 time points. Baseline t heory-derived correlates of NA were identified using single and multiple regression analysis. The diagnostic accuracy of BAASIS and clinician-reported VAS was tested against an exploratory EM NA cutoff of < 95%. Results: The median EM adherence (n = 55) was 97.5% (range, 48-100%), while the 25th percentile was 92.1%. Lack of membership in a CML patient support group, living alone, and third-line treatment were associated with EM NA on multiple regression analysis. The BAASIS self-report (n = 94) had a sensitivity of 67% and a specificity of 71% for diagnosing NA, while clinician-reported VAS (n = 89) had a sensitivity of 78% and specificity of 42%. Conclusion: A quarter of patients had potentially clinically meaningful NA. These NA correlates and the BAASIS provide a basis for identifying nonadherent patients who can be targeted by interventions.