BACKGROUND: Chronic myeloid leukemia (CML) treatment is lifelong, and while it is important for patients to remain adherent to treatment, there are conflicting findings with respect to differences in adherence and persistence with dasatinib or nilotinib during second-line treatment. OBJECTIVE: To compare the rates of adherence, persistence, and discontinuation of 2 oral second-generation tyrosine kinase inhibitors (TKI), dasatinib and nilotinib, in CML patients during their first 12 months of second-line treatment. METHODS: Adult CML patients (ICD-9-CM: 205.1x) with 2 consecutive dasatinib or nilotinib prescription claims within 12 months were identified from the Truven Health MarketScan Databases (January 1, 2006-September 30, 2011). Patients were excluded if they had FDA-approved non-CML indications for imatinib, had less than  6 months continuous enrollment, or had a stem cell/bone marrow transplant in the 6 months pre-index. Patients were followed until the first occurrence of index TKI discontinuation/switch; enrollment end; December 31, 2011; or 12 months follow-up post-index. Index treatment (dasatinib ≤ 100 mg or nilotinib) was categorized as second-line if there was evidence of only 1 alternative TKI exposure (e.g., imatinib, dasatinib, or nilotinib) anytime during the patient's available claims history. When comparing adherence, persistence, and discontinuation, inverse probability treatment weighting (IPTW) was used. Adherence and persistence measures were calculated as specified by the International Society for Pharmacoeconomics and Outcomes Research Medication Compliance and Persistence Special Interest Group. Treatment adherence was calculated using medication possession ratio (MPR) and was reported as both continuous and binary measures (i.e., high adherence = MPR ≥ 85%). Persistence was reported as the proportion of days covered (PDC) and estimated level of persistence (ELPT). Finally, discontinuation was defined as a treatment gap of greater than 90 days and absence of index TKI during the remainder of the follow-up period. Time to discontinuation and high adherence of index TKI were compared using weighted Cox proportional hazards and logistic regression models, respectively. RESULTS: After propensity weighting, the 219 second-line dasatinib patients and the 158 second-line nilotinib patients were similar in mean age, gender, cancer complexity, and comorbidity burden at baseline. Age as a categorical measure, population density, and index year remained imbalanced and were, therefore, included as covariates in the multivariate analysis of adherence. In the bivariate analyses, mean MPR (88.2% vs. 84.4%, P = 0.036); proportion of patients with high adherence (72.7% vs. 63.3%, P = 0.006); and ELPT (70.4% vs. 62.7%, P = 0.026) were significantly higher among dasatinib patients than nilotinib patients. Mean PDC was not significantly different between dasatinib and nilotinib patients (0.79 vs. 0.77, P = 0.328) after propensity weighting. In addition, a significantly lower proportion of second-line dasatinib patients discontinued their index therapy compared with second-line nilotinib patients (4.4% vs. 8.6%, P = 0.020). With a hazard ratio (HR) of 0.50 (95% CI = 0.27-0.93, P = 0.029), dasatinib patients had half the possibility of discontinuing treatment compared with nilotinib patients at any point in time. After accounting for the baseline factors remaining imbalanced and controlling for cancer complexity and number of concomitant medications at baseline, second-line dasatinib patients were 1.7 times (95% CI = 1.2-2.4) more likely to be highly adherent than second-line nilotinib patients (P = 0.0016). CONCLUSIONS: Among second-line TKI-treated CML patients, dasatinib patients had significantly higher adherence and lower discontinuation rates compared with patients receiving second-line nilotinib.