Correlation analysis between plasma concentration of nilotinib and clinical efficacy and safety in patients with chronic myeloid leukemia: a single–center retrospective cohort study

Objective This study aimed to investigate the correlation between nilotinib plasma concentrations and clinical efficacy and safety in patients with chronic myeloid leukemia (CML), thereby supporting personalized therapeutic optimization. Methods We conducted a retrospective cohort study of 121 CML patients receiving nilotinib with therapeutic drug monitoring (TDM) at Nanfang Hospital of Southern Medical University between March 2021 and February 2024. Major molecular response (MMR) and adverse events (CTCAE v5.0) were analyzed against concentrations. Receiver operating characteristic (ROC) curves defined thresholds; dose-normalized exposure (C norm = C min /(Dose/600 mg)) addressed TDM bias. Results The effective group showed higher nilotinib concentrations than those in the ineffective group (1,036.40 ± 463.67 vs. 737.14 ± 518.97 ng mL -1 ; P < 0.001), confirmed post-normalization (1,045.10 ± 468.08 vs. 858.34 ± 723.66 ng mL -1 ; P < 0.05). The ROC-derived efficacy threshold was 636.99 ng mL -1 (AUC = 0.693, 95% CI: 0.596–0.791), which was identical after normalization (AUC = 0.655, 95% CI: 0.554–0.755). Although adverse events were common (76.9% of patients), they showed no overall concentration dependence (P = 0.288). However, hyperbilirubinemia risk was significantly elevated in patients with concentrations >1,273.98 ng mL -1 cohort (50.0% vs. 20.0%–22.6%; P = 0.030), with a toxicity threshold identified at 1,290.34 ng mL -1 (AUC = 0.656, 95% CI: 0.540–0.771). Longer treatment duration was also associated with higher drug exposure (P = 0.029). Conclusion Nilotinib concentrations predict MMR attainment independent of TDM-driven dose adjustments (validated via C norm ). We recommend targeting early-phase concentrations >636.99 ng mL -1 , monitoring bilirubin above 1,290.34 ng mL -1 , and integrating dynamic TDM with pharmacogenetic profiling.