MedElevator
Войти в исследования Rheumatology (Oxford, England) · 2026
Predicting mycophenolic acid exposure in paediatric lupus nephritis: a real-world study across three clinical scenarios.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Liu Baojing, Chen Lizhi, Wang Yuqi, Lin Jingxuan, Guo Yutong, Tang Kejing, Jiang Xiaoyun, Chen Pan
Аннотация
OBJECTIVES: Childhood-onset systemic lupus erythematosus (cSLE) with lupus nephritis (LN) presents therapeutic challenges due to complex mycophenolic acid (MPA) pharmacokinetics. This study developed machine learning (ML)-based models to predict MPA area under the concentration-time curve (MPA-AUC) across three scenarios: determining therapeutic window attainment, precise AUC estimation and forecasting post-dose-adjustment exposure.
METHODS: Data were collected from cLN on MPA with two cohorts: routine treatment (group 1) and dose adjustment (group 2). Comprehensive patient data, including demographics, medication details and laboratory test results, were collected. Models were constructed using algorithms such as XGBoost, LightGBM and Random Forest (RF). Model performance was assessed using multiple metrics and the Shapley Additive Explanations (SHAP) method was employed to interpret the models.
RESULTS: Data from 154 patients (1376 follow-ups) were included. In Scenario 1, XGBoost and GBM models demonstrated superior performance, with the LightGBM model excelling in predicting therapeutic window attainment (AUC = 0.67, Precision = 0.64, Recall = 0.66, F1 = 0.65). In Scenario 2, the RF model incorporating clinical indicators and C0.5, C1.5, C4 concentrations achieved the most accurate predictions (R2 = 0.84, mean absolute error (MAE) = 3.97 and root mean square error (RMSE) = 5.97). In Scenario 3, which included data from 246 follow-up records of 99 patients, the RF model showed the best performance (R2 = 0.51 and MAE = 20.57). SHAP analysis highlighted dose, weight, renal function and concentration timepoints as key predictors. A web-based tool was developed to support clinical decisions.
CONCLUSION: This study establishes the first ML-driven framework for personalized MPA dosing in paediatric LN, addressing TDM limitations and enhancing precision medicine.
METHODS: Data were collected from cLN on MPA with two cohorts: routine treatment (group 1) and dose adjustment (group 2). Comprehensive patient data, including demographics, medication details and laboratory test results, were collected. Models were constructed using algorithms such as XGBoost, LightGBM and Random Forest (RF). Model performance was assessed using multiple metrics and the Shapley Additive Explanations (SHAP) method was employed to interpret the models.
RESULTS: Data from 154 patients (1376 follow-ups) were included. In Scenario 1, XGBoost and GBM models demonstrated superior performance, with the LightGBM model excelling in predicting therapeutic window attainment (AUC = 0.67, Precision = 0.64, Recall = 0.66, F1 = 0.65). In Scenario 2, the RF model incorporating clinical indicators and C0.5, C1.5, C4 concentrations achieved the most accurate predictions (R2 = 0.84, mean absolute error (MAE) = 3.97 and root mean square error (RMSE) = 5.97). In Scenario 3, which included data from 246 follow-up records of 99 patients, the RF model showed the best performance (R2 = 0.51 and MAE = 20.57). SHAP analysis highlighted dose, weight, renal function and concentration timepoints as key predictors. A web-based tool was developed to support clinical decisions.
CONCLUSION: This study establishes the first ML-driven framework for personalized MPA dosing in paediatric LN, addressing TDM limitations and enhancing precision medicine.