Yu Lisheng, Cao Shunshun, Song Botian, Hu Yangyang
Neurosurgery, The Second Affiliated Hospital of Wenzhou Medical University, Wenzhou, Zhejiang, China.
Wenzhou Municipal Key Laboratory of Neurodevelopmental Pathology and Physiology, Wenzhou Medical University, Wenzhou, Zhejiang, China.
BMC Public Health. 2025 Aug 27;25(1):2934. doi: 10.1186/s12889-025-24212-y.
Testosterone deficiency (TD) is a clinically significant condition strongly associated with aging and metabolic syndrome. While previous studies have established links between muscle mass and TD, evidence regarding the relationship between muscle quality index (MQI) and TD remains limited. This study aimed to investigate the association between MQI and TD in adult males in the United States and to develop an interpretable machine learning (ML) model based on SHapley Additive exPlanation (SHAP) for predicting TD risk.
We conducted a cross-sectional study using weighted multivariate logistic regression and subgroup analysis to assess the association between MQI and TD. Six ML models incorporating MQI were developed to predict TD risk. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), confusion matrix, F1 score, Brier score, and precision-recall curve. The optimal model was selected based on these metrics and further interpreted using SHAP to elucidate feature importance and decision-making processes.
The study included 2,628 eligible male participants, with a TD prevalence of 25.76%. After adjusting for confounders, each unit increase in MQI was associated with a 52% reduction in TD risk (OR = 0.480, 95% CI: 0.362-0.636, P < 0.001), demonstrating a dose-response relationship. Among the six ML models, the Light Gradient Boosting Machine (LGBM) exhibited the best predictive performance, achieving an AUC of 0.746 (95% CI: 0.707-0.790). SHAP analysis revealed that body mass index (BMI) was the most influential feature in the LGBM model, followed by high-density lipoprotein and MQI. Notably, lower MQI values were consistently associated with a higher risk of TD.
Our findings indicate that MQI is an independent and reliable predictor of TD in males. The interpretable LGBM model provides a cost-effective and clinically applicable tool for early TD risk assessment. These results underscore the importance of muscle quality in testosterone regulation and may inform preventive strategies to mitigate TD risk in adult males.
睾酮缺乏（TD）是一种临床上具有重要意义的病症，与衰老和代谢综合征密切相关。虽然先前的研究已经确立了肌肉量与TD之间的联系，但关于肌肉质量指数（MQI）与TD之间关系的证据仍然有限。本研究旨在调查美国成年男性中MQI与TD之间的关联，并基于夏普利加法解释（SHAP）开发一种可解释的机器学习（ML）模型，用于预测TD风险。
我们进行了一项横断面研究，使用加权多元逻辑回归和亚组分析来评估MQI与TD之间的关联。开发了六个纳入MQI的ML模型来预测TD风险。使用受试者工作特征曲线下面积（AUC）、混淆矩阵、F1分数、布里尔分数和精确召回曲线评估模型性能。根据这些指标选择最佳模型，并使用SHAP进一步解释以阐明特征重要性和决策过程。
该研究纳入了2628名符合条件的男性参与者，TD患病率为25.76%。在调整混杂因素后，MQI每增加一个单位，TD风险降低52%（OR = 0.480，95% CI：0.362 - 0.636，P < 0.001），显示出剂量反应关系。在六个ML模型中，轻梯度提升机（LGBM）表现出最佳的预测性能，AUC为0.746（95% CI：0.707 - 0.790）。SHAP分析表明，体重指数（BMI）是LGBM模型中最具影响力的特征，其次是高密度脂蛋白和MQI。值得注意的是，较低的MQI值始终与较高的TD风险相关。
我们的研究结果表明，MQI是男性TD的独立且可靠的预测指标。可解释的LGBM模型为早期TD风险评估提供了一种经济高效且临床适用的工具。这些结果强调了肌肉质量在睾酮调节中的重要性，并可能为减轻成年男性TD风险的预防策略提供依据。
