Mai Xiaoyou, Li Mingli, Jin Xihui, Huang Shengzhu, Xu Mingjie, Yan Boteng, Wei Yushuang, Long Xinyang, Wu Yongxian, Mo Zengnan
School of Public Health, Guangxi Medical University, Nanning 530021, China.
Center for Genomic and Personalized Medicine, Guangxi Key Laboratory for Genomic and Personalized Medicine, Guangxi Collaborative Innovation Center for Genomic and Personalized Medicine, Guangxi Medical University, Nanning 530021, China.
Healthcare (Basel). 2025 Apr 23;13(9):969. doi: 10.3390/healthcare13090969.
Our study aims to develop a personalized nomogram model for predicting the risk of nonalcoholic fatty liver disease (NAFLD) in hypertension (HTN) patients and further validate its effectiveness. A total of 1250 hypertensive (HTN) patients from Guangxi, China, were divided into a training group (875 patients, 70%) and a validation set (375 patients, 30%). LASSO regression, in combination with univariate and multivariate logistic regression analyses, was used to identify predictive factors associated with nonalcoholic fatty liver disease (NAFLD) in HTN patients within the training set. Subsequently, the performance of an NAFLD nomogram prediction model was evaluated in the separate validation group, including assessments of differentiation ability, calibration performance, and clinical applicability. This was carried out using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA). The risk-prediction model for the HTN patients concomitant with NAFLD included oral antidiabetic drugs (OADs) (OR = 2.553, 95% CI: 1.368-4.763), antihypertensives (AHs) (OR = 7.303, 95% CI: 4.168-12.794), body mass index (BMI) (OR = 1.145, 95% CI: 1.084-1.209), blood urea nitrogen (BUN) (OR = 0.924, 95% CI: 0.860-0.992), triglycerides (TGs) (OR = 1.474, 95% CI: 1.201-1.809), aspartate aminotransferase (AST) (OR = 1.061, 95% CI: 1.018-1.105), and AST/ALT ratio (AAR) (OR = 0.249, 95% CI: 0.121-0.514) as significant predictors. The AUC of the NAFLD risk-prediction model in the training set and the validation set were 0.816 (95% CI: 0.785-0.847) and 0.794 (95% CI: 0.746-0.842), respectively. The Hosmer-Lemeshow test showed that the model has a good goodness-of-fit (-values were 0.612 and 0.221). DCA suggested the net benefit of using a nomogram to predict the risk of HTN patients concomitant with NAFLD is higher. These results suggested that the model showed moderate predictive ability and good calibration. BMI, OADs, AHs, BUN, TGs, AST, and AAR were independent influencing factors of HTN combined with NAFLD, and the risk prediction model constructed based on this could help to identify the high-risk group of HTN combined with NAFLD at an early stage and guide the development of interventions. Larger cohorts with multiethnic populations are essential to verify our findings.
我们的研究旨在开发一种个性化列线图模型，用于预测高血压（HTN）患者非酒精性脂肪性肝病（NAFLD）的风险，并进一步验证其有效性。总共1250名来自中国广西的高血压（HTN）患者被分为训练组（875例患者，70%）和验证组（375例患者，30%）。采用LASSO回归结合单因素和多因素逻辑回归分析，在训练集中识别与HTN患者非酒精性脂肪性肝病（NAFLD）相关的预测因素。随后，在单独的验证组中评估NAFLD列线图预测模型的性能，包括区分能力、校准性能和临床适用性评估。这通过受试者操作特征（ROC）曲线、校准曲线和决策曲线分析（DCA）来进行。HTN合并NAFLD患者的风险预测模型包括口服降糖药（OADs）（OR = 2.553，95%CI：1.368 - 4.763）、抗高血压药（AHs）（OR = 7.303，95%CI：4.168 - 12.794）、体重指数（BMI）（OR = 1.145，95%CI：1.084 - 1.209）、血尿素氮（BUN）（OR = 0.924，95%CI：0.860 - 0.992）、甘油三酯（TGs）（OR = 1.474，95%CI：1.201 - 1.809）、天冬氨酸转氨酶（AST）（OR = 1.061，95%CI：1.018 - 1.105）和AST/ALT比值（AAR）（OR = 0.249，95%CI：0.121 - 0.514）作为显著预测因素。训练集和验证集的NAFLD风险预测模型的AUC分别为0.816（95%CI：0.785 - 0.847）和0.794（95%CI：0.746 - 0.842）。Hosmer-Lemeshow检验表明该模型具有良好的拟合优度（-值分别为0.612和0.221）。DCA表明使用列线图预测HTN合并NAFLD患者风险的净效益更高。这些结果表明该模型具有中等预测能力和良好的校准。BMI、OADs、AHs、BUN、TGs、AST和AAR是HTN合并NAFLD的独立影响因素，基于此构建的风险预测模型有助于早期识别HTN合并NAFLD的高危人群并指导干预措施的制定。需要更大规模的多民族人群队列来验证我们的研究结果。
