Abstract:

High-dose methotrexate (HD-MTX) remains a cornerstone in the treatment of central nervous system lymphoma (CNSL). However, delayed MTX clearance markedly increases the risk of toxic adverse events, thereby restricting its clinical use. To address this challenge, we developed a risk prediction model for delayed MTX clearance in patients with CNSL, aiming to enhance medication safety. This study analyzed data from 34 CNSL patients who underwent 104 HD-MTX treatment cycles between July 2020 and July 2024. Variables with P < 0.05 in univariate analysis were incorporated into a multivariate logistic regression model, and stepwise selection was performed based on the Akaike Information Criterion (AIC). Based on the principle of the minimum AIC value and further considering the statistical significance of the variables, two independent risk factors were identified: serum albumin level (OR = 0.834) and age > 60 years (OR = 4.166), which were used to construct the model. Model validation demonstrated robust performance, with an AUC of 0.717 (95% CI: 0.605–0.830). The Hosmer-Lemeshow goodness-of-fit test (χ² = 6.229, P = 0.622) and calibration curve analysis confirmed good calibration. Decision curve analysis (DCA) further supported the model’s clinical utility, while bootstrap validation yielded consistent accuracy (AUC = 0.717, 95% CI: 0.602–0.832). Additionally, delayed MTX clearance was significantly associated with acute kidney injury (P < 0.001). In summary, the risk prediction model developed in this study provided reliable estimates of delayed MTX clearance in CNSL patients undergoing HD-MTX therapy, offering a valuable reference to guide clinical decision-making and improve patient safety.

Key words: Nomogram, High-dose methotrexate, Central nervous system lymphoma, Acute kidney injury