Abstract / Summary
Background and Objectives: Methotrexate (MTX) is a key drug for treating haematologic malignancies. Due to its narrow therapeutic window, Population pharmacokinetic (PopPK) models help predict MTX plasma concentrations (Cp) and optimize folinic acid rescue. This study evaluated and compared the predictive performance and clinical utility of two PopPK models in adults with lymphoma or acute lymphoblastic leukaemia (ALL). Methods: A retrospective PK study was conducted in adults with lymphoma/ALL treated with MTX. Cp at 24 h (lymphoma) and 42 h (ALL) were estimated by Bayesian forecasting using the Gallais et al. (2021) and Kawakatsu et al. (2019) PopPK models with Levenberg–Marquardt least squares (M2) and Strict Adaptive Metropolis (SAA), using frequency-based (FB) or concentration-based (CB) weighting. Predictive performance was assessed using median percentage error (MDiPE), median absolute percentage error (MDiAPE), and the percentage of individual prediction errors within ±20% (iF20) and ±30% (iF30) of observed concentrations. Clinical utility was evaluated by concordance with folinic acid rescue. Results: Thirty-seven patients (72 MTX cycles) were included: 17 with lymphoma (37 cycles; 52% men; median age of 66 (34–73) years) and 20 with ALL (35 cycles; 55% men; median age of 46 (17–68) years), with a median of two cycles per patient. In lymphoma, the Kawakatsu model showed the most favourable predictive performance with SAA-FB (MDiPE of 8.5%, MDiAPE of 30.3%, iF20 of 37.83%, and iF30 of 48.65%), whereas the Gallais model showed lower predictive accuracy across the evaluated approaches. Concordance with folinic acid rescue ranged from 75.68% to 86.49%, with the highest value observed with M2-CB. In ALL, both models showed lower and more variable predictive performance, with MDiAPE ranging from 43.0% to 73.8%, iF30 from 11.43% to 42.86%, and concordance with folinic acid rescue from 40.00% to 54.29%. Conclusions: PopPK models for MTX showed variable performance in external cohorts, with numerically more favourable results for the model by Kawakatsu et al. External validation is essential prior to clinical application, particularly given the relationship between MTX exposure, delayed clearance and toxicity. Validated models could contribute to safer monitoring and more personalised treatment, although larger studies are needed to confirm their clinical utility.