肝移植受者免疫抑制剂西罗莫司全血谷浓度脱靶风险预测模型的构建

The Development of a Prediction Model for Off-target Risk of Sirolimus Trough Concentration in liver transplant recipients

关键词:  肝移植  西罗莫司  谷浓度  列线图模型  血红蛋白

Key words:  liver transplant  sirolimus  trough concentrations  nomogram model  hemoglobin

摘要:

目的 探讨肝移植受者免疫抑制剂西罗莫司全血谷浓度脱离靶范围（4 ~ 10 ng.mL-1）的危险因素，并构建列线图风险预测模型。方法 回顾性收集2023年12月至2025年8月于北京清华长庚医院进行西罗莫司治疗药物监测（therapeutic drug monitoring, TDM）的肝移植患者临床资料，包括人口学特征、用药方案、谷浓度、实验室检查及不良反应等。采用高效液相色谱-串联质谱法（high-performance liquid chromatography-tandem mass spectrometry, HPLC-MS/MS）测定西罗莫司全血谷浓度，采用Mann-Whitney U检验、χ2检验及Lasso回归分析谷浓度脱靶的危险因素，并建立列线图模型。结果 共纳入68例患者，西罗莫司TDM监测643例次。西罗莫司日剂量中位数2 mg，谷浓度中位数4.81 ng.mL-1（1.05 ng.mL-1 ~ 19.86 ng.mL-1），处于靶目标范围内(4 ~ 10 ng.mL-1)的监测为349例次（54.28%）。Lasso回归分析结果显示，药物剂量、体重、血红蛋白、血浆总蛋白及估算肾小球滤过率是西罗莫司是否脱靶的独立预测因素。基于上述变量构建列线图模型，内、外部验证显示受试者工作特征 (receiver operating characteristic, ROC)曲线下面积（an area under the ROC curves, AUC）分别为0.66、0.75，具有较好的区分度。结论 肝移植受者西罗莫司谷浓度达标率较低，建议常规进行西罗莫司TDM监测。基于Lasso-logistic回归构建的列线图模型具有较好的预测效能，有助于临床早期识别西罗莫司谷浓度脱靶风险。

Abstract:

ABSTRACT: OBJECTIVE To analyze the clinical factors influencing sirolimus trough concentrations (target range: 4 ~ 10 ng.mL-1) in liver transplant recipients and develop a model for off-target risk prediction. METHODS Demographics, clinical factors, sirolimus trough concentrations, laboratory test results and adverse effects of liver transplant recipients underwent sirolimus therapeutic drug monitoring From December 2023 to August 2025 in Beijing Tsinghua Changgung Hospital were collected. High-performance liquid chromatography-tandem mass spectrometry (HPLC-MS/MS) method was used to detect the sirolimus trough concentrations. The correlation between the clinical variables and sirolimus trough concentration was assessed by Mann-Whitney test, chi-square test and Lasso regression, and a model for off-target risk prediction was established. RESULTS In this study, 643 cases of sirolimus steady-state trough concentrations from 68 patients were enrolled. The median daily dose of sirolimus was 2 mg and the median whole blood concentration in these cases was 4.81 ng.mL-1, ranging from 1.05 ng.mL-1 to 19.86 ng.mL-1. 349 cases (54.28%) got concentrations within the therapeutic range (4 ng.mL-1 to 10 ng.mL-1). The results of Lasso regression analysis showed that dose, hemoglobin, weight, total plasma protein and estimated glomerular filtration rate were risk factors for off-target sirolimus trough concentrations. The nomogram model was constructed with area under the receiver operating characteristic curves (AUC) of 0.66, 0.75 for internal and external validation. CONCLUSION The proportion of sirolimus trough concentrations achieving target in liver transplant recipients is relatively low, suggesting a need for routine therapeutic drug monitoring. The nomogram model based on Lasso-logistic regression could help to identify the off-target risk of sirolimus trough concentrations in liver transplant recipients.