Development and validation of a clinical scoring system for the prediction of methotrexate treatment response in rheumatoid arthritis patients: a multicenter real-world study.

Abstract

Background: Methotrexate (MTX) remains the anchor drug in rheumatoid arthritis (RA) management. However, treatment response varies widely.

Objectives: This study aimed to develop and validate a simple scoring model for predicting MTX monotherapy treatment success at 6 months.

Methods: We conducted a multicenter, retrospective cohort study of newly diagnosed, disease-modifying antirheumatic drug-naïve RA patients treated with MTX monotherapy. Treatment success was defined as achieving remission or low disease activity at 6 months based on Disease Activity Score-28 with erythrocyte sedimentation rate (DAS28-ESR). Multivariable logistic regression was used to identify baseline predictors and construct a scoring model. Internal validation was performed using stratified 10-fold cross-validation and bootstrap resampling, while external validation was conducted in an independent cohort from three hospitals. Model performance was evaluated by discrimination and calibration.

Results: The development cohort included 840 patients. Four independent predictors of MTX success were identified: age >60 years, normal ESR, lower baseline disease activity (DAS28-ESR ⩽5.1), and lower glucocorticoid dose. The model demonstrated an area under the ROC curve (AUC) of 0.66. Internal validation yielded a mean AUC of 0.65, with a calibration intercept of -0.001 and slope of 0.943. Bootstrap analysis showed minimal optimism (0.003). External validation in 265 patients showed moderate discrimination (AUC 0.63; 95% confidence interval 0.56-0.70) with acceptable calibration (intercept 0.16; slope 0.890) and a stable Brier score of 0.23. No significant difference in AUC was observed between cohorts (p = 0.31). Decision curve analysis demonstrated that the scoring system provided a higher net benefit compared to default strategies across a threshold probability range of 35%-80%.

Conclusion: The MTX response scoring system demonstrated stable performance and may serve as an adjunctive tool to aid in early risk stratification for MTX treatment success, particularly in resource-limited settings.