Fig. 1: Cohorts construction from the main study populations.

First, to identify the adverse event profile associated with CML, we matched (1:1) CML patients in the H12O, EMEA, and U.S. networks to non-CML patients from the same network using propensity score matching based on age and sex. Second, for the analysis that assesses the effect of CML on the survival and infection (ICD-10-CM diagnosis branch A00-B99) incidence of patients, we compared non-oncologic patients from the H12O, EMEA, and U.S. networks to the subset of CML patients from the same network without any other neoplasm, i.e., ICD-10-CM diagnosis in the branch C00-D49 (Neoplasms) other than C92.1 using propensity score matching (1:1) based on age and sex. Third, sub-cohorts from the original CML cohorts from the H12O, EMEA, and U.S. networks, based on the first TKI record in the EMR, were created to compare survival and incidence of second malignancy between patients who received imatinib as their first TKI versus those who received dasatinib or nilotinib. Fourth, we segmented the original CML cohorts from the United States and Global networks into two cohorts based on whether they received their first TKI between 2001 and 2010 or between 2011 and 2020.