Supplementary Material for: Circulating Humanin Improves the Prognostic Accuracy of Cardiovascular Risk Models in Chronic Hemodialysis Patients

INTRODUCTION Cardiovascular morbidity and mortality (CVMM) remain highly prevalent among end-stage kidney disease (ESKD) patients undergoing chronic hemodialysis (HD). Nevertheless, risk prediction in this setting is limited by the complexity of factors involved and the individual patients’ characteristics. In this study, we evaluated whether circulating levels of Humanin, a micropeptide reflecting mitochondrial dysfunction, could refine CVMM prediction when included in established risk models for CVMM in ESKD-HD.
METHODS Eighty-three chronic HD patients (mean age 67±12 years; 74.7% male) were enrolled in a 24-month prospective multicenter study. CVMM occurred in 33 patients (39.7%). The performances of an internal prognostic model based on cohort-related risk predictors and four externally generated and validated risk scores (AROii-2, jDOPPS, You et al., and Zhang et al.) were compared before and after the inclusion of Humanin.
RESULTS Humanin displayed a curvilinear association with CVMM, with both low (<462 pg/mL) and high (>778 pg/mL) levels linked to an increased risk. CVMM was best predicted by the internal model (AUC 0.671; R2 0.13), followed by the Zhang (0.669; 0.12), AROii-2 (0.632; 0.05), You (0.628; 0.07), and jDOPPS (0.610; 0.03) scores. All models were well calibrated. Humanin inclusion remarkably improved discrimination (ΔAUC spanning from 0.083 to 0.109), calibration, explained variance (R2 gain from 9 to 15%), and reclassification (NRI 8 to 27.9%; IDI 4.4 to 7.3%) of all models considered.
CONCLUSIONS Integration of Humanin significantly enhances the predictive accuracy of CVMM risk models in HD patients, supporting its potential role as an additive biomarker in this high-risk population.

**引言（INTRODUCTION）**
心血管发病率与死亡率（Cardiovascular morbidity and mortality, CVMM）在接受慢性血液透析（chronic hemodialysis, HD）的终末期肾病（end-stage kidney disease, ESKD）患者中仍高发。然而，该人群的CVMM风险预测受限于复杂的影响因素与个体患者特征。本研究旨在评估：反映线粒体功能障碍的微肽人类素（Humanin），当其被纳入ESKD-HD患者的CVMM既定风险模型时，能否优化CVMM的预测效能。

**方法（METHODS）**
本研究纳入83例慢性HD患者，平均年龄为67±12岁，男性占比74.7%，并开展为期24个月的前瞻性多中心研究。最终共有33例患者（39.7%）发生CVMM事件。本研究比较了两类风险模型在纳入人类素（Humanin）前后的预测性能：一类是基于本队列相关风险预测因子构建的内部预后模型，另一类是4种已外部生成并验证的风险评分，分别为AROii-2、jDOPPS、You等人及Zhang等人评分。

**结果（RESULTS）**
人类素（Humanin）的循环水平与CVMM呈曲线关联：低水平（<462 pg/mL）与高水平（>778 pg/mL）的人类素均与CVMM风险升高相关。各模型中，内部预后模型对CVMM的预测效能最优，曲线下面积（Area Under Curve, AUC）为0.671，决定系数R²为0.13；其后依次为Zhang等人评分（0.669；0.12）、AROii-2评分（0.632；0.05）、You等人评分（0.628；0.07）以及jDOPPS评分（0.610；0.03）。所有受试模型均具有良好的校准性。纳入人类素后，所有模型的多项评价指标均得到显著改善：区分度提升（ΔAUC介于0.083至0.109之间）、校准性能优化、解释方差提高（R²提升9%至15%），且重分类能力增强，其中净重分类改进（Net Reclassification Improvement, NRI）为8%至27.9%，综合判别改进（Integrated Discrimination Index, IDI）为4.4%至7.3%。

**结论（CONCLUSIONS）**
将人类素（Humanin）纳入CVMM风险模型，可显著提升HD患者的CVMM风险预测准确性，提示其有望成为该高危人群的附加生物标志物。