Fumisato Takagi, Takaaki Ono, Naoto Takahashi, Yosuke Minami, Chiaki Nakaseko, Noriyoshi Iriyama, Katsumichi Fujimaki, Kazuhiko Kakihana, Yoji Ogasawara, Masaya Okada, Tetsuzo Tauchi, Yoshiko Atsuta, Ritsuro Suzuki, Kazuhito Yamamoto, Shigeki Ohtake, Toshihiro Miyamoto, Kazunori Ohnishi, Emiko Sakaida, Shin Fujisawa, Yukio Kobayashi, Tatsuya Kawaguchi, Masahiro Kizaki, Hitoshi Kiyoi, Yasushi Miyazaki, Itaru Matsumura
Risk stratification to predict treatment failure in chronic-phase chronic myeloid leukemia (CP-CML) remains an important challenge in the tyrosine kinase inhibitor (TKI) era. We conducted a pooled analysis of 710 newly diagnosed CP-CML patients enrolled in the prospective JALSG CML212 and New TARGET studies, approximately 80% of whom received second-generation TKIs (2G-TKIs) as first-line therapy, to identify predictors of failure-free survival (FFS). In this study, we incorporated the ELN 2025-defined "Warning" category as an initial adverse state in the definition of FFS events and evaluated transitions to "Unfavorable". Within this framework, the EUTOS long-term survival (ELTS) score consistently stratified FFS whether treatment failure included both "Warning" and "Unfavorable" events or was restricted to "Unfavorable". Among patients treated with 2G-TKIs, ELTS demonstrated superior time-dependent discrimination compared with Sokal and remained the baseline predictor of initial treatment failure. Importantly, once "Warning" occurred, baseline ELTS no longer predicted deterioration. Instead, attainment of "Warning" within 6 months after TKI initiation independently predicted transition from "Warning" to "Unfavorable" (HR 7.31, 95% CI 2.94-18.2; p<0.001). These findings distinguish baseline risk from post-milestone disease evolution and support a two-stage risk framework integrating baseline ELTS stratification with "Warning" timing to guide dynamic risk-adapted management in CP-CML.