OBJECTIVES: Currently, there are several tyrosine kinase inhibitors (TKI) approved for the treatment of chronic myeloid leukemia (CML). The aim of our study was to evaluate the long-term cost-effectiveness of different therapy regimens for CML focusing on the evaluation of the uncertainty using probabilistic sensitivity analysis (PSA).   METHODS: We performed a cost-effectiveness analysis using a state-transition Markov model. The model evaluates seven treatment strategies including different combinations of TKIs as well as chemotherapy or stem cell transplantation. For model parameters, we used published trial data, and Austrian clinical, epidemiological, and economic data from the Austrian CML registry, statistical and economic databases. We performed a cohort simulation over a lifelong time horizon, adopted a societal perspective with an annual 3% discount rate. We conducted extensive uncertainty analyses and contrasted different methodological approaches to define parameter uncertainty distributions from our source data. We compared the base-line derived from the mean parameter values with the mean outcomes of all PSA scenarios. RESULTS: In the base-case efficiency frontier, nilotinib without second-line TKI resulted in an ICUR (1) of 118,600 €/QALY compared to the baseline strategy imatinib without second-line TKI. Imatinib followed by nilotinib after failure yielded an ICUR (2) of 123,900 €/QALY compared to nilotinib without second-line TKI. Nilotinib followed by dasatinib after failure resulted in an ICUR (3) of 149,400 €/QALY compared to imatinib followed by nilotinib after failure. The remaining strategies were excluded due to absolute or extended dominance. The PSA resulted in ICURs of (1) 116,200 €/QALY (2) 130,300 €/QALY and (3) 130,300 €/QALY.    CONCLUSIONS: Based on our analysis, we recommend imatinib followed by nilotinib as the most cost-effective treatment strategy including a second-line TKI. Mean results from PSA show only small deviation from the base-case analysis. When new path to cure data are available, results will need to be updated.