Evaluating a predictive model of tyrosine kinase inhibitor therapy failure in a European-type cohort: a step towards population-specific tools.

Predicting therapeutic failure in patients with chronic phase-chronic myeloid leukemia (CP-CML) treated with tyrosine kinase inhibitors (TKI) remains a major challenge for personalized care management. The Sokal and EUTOS long-term survival scores were designed to predict CML-related mortality, but are also used to guide therapeutic choices, despite their poor performance for this purpose. A recent study proposed a refined predictive model of therapy failure specifically tailored for patients treated with imatinib and second-generation TKIs that showed promising results in a Chinese cohort. The present study evaluated the performance and applicability of this predictive model in a real-world, multicenter cohort from the French CML Observatory. The key differences identified between the Chinese and French cohorts (age, baseline hemoglobin levels, and treatment regimens) likely influenced the model performance. Specifically, the new model did not allow for discriminating risk groups effectively in the French cohort. However, the model reconstruction using this cohort identified other predictive variables (sex, leukocytosis, comorbidities, high-risk additional chromosomal abnormalities) that better stratified patients at risk of therapy failure. Our findings highlight the influence of demographic and clinical differences on predictive models and emphasize the need for local or population-specific tools to optimize risk stratification and therapeutic decision-making in CP-CML.