Construction and validation of a predictive model for the risk of kidney injury in human immunodeficiency virus/acquired immunodeficiency syndrome patients
10.3760/cma.j.cn311365-20250126-00033
- VernacularTitle:HIV感染/AIDS患者肾损伤风险预测模型的构建与验证
- Author:
Xiaoyun QIN
1
;
Guoxian LI
;
Simei LUO
;
Jiaguang HU
;
Kai FU
;
Peng ZHANG
;
Xu LI
;
Zhongsheng JIANG
Author Information
1. 柳州市人民医院感染性疾病科，柳州 545006
- Publication Type:Journal Article
- Keywords:
HIV infections;
Acquired immunodeficiency syndrome;
Nomograms;
Anti-retroviral therapy;
Kidney injury;
Predictive model
- From:
Chinese Journal of Infectious Diseases
2025;43(2):90-97
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To investigate the risk factors for kidney injury during anti-retroviral therapy (ART) with zidovudine (AZT) or tenofovir disoproxil fumarate (TDF) in human immunodeficiency virus/acquired immunodeficiency syndrome (HIV/AIDS) patients, and to construct and validate a prediction model for the risk of kidney injury in HIV/AIDS patients based on a nomogram.Methods:A total of 923 HIV/AIDS patients admitted to Liuzhou People′s Hospital between January 1st, 2004 and December 31st, 2020 were included in this study. The modeling set (647 cases) and the validation set (276 cases) were divided in a 7∶3 ratio. Risk factors were screened using the least absolute shrinkage and selection operator (LASSO) regression analysis, and a nomogram prediction model for renal impairment risk in HIV/AIDS patients was constructed based on the selected variables. The model′s predictive performance was assessed by calculating the area under the curve (AUC) using the receiver operating characteristics curve (ROC curve). The performance of this model was evaluated using calibration curves. The clinical utility of the model was assessed using decision curve analysis (DCA).Results:Among 923 HIV/AIDS patients, there were 91 cases with kidney injury, including 67 in the modeling set and 24 in the validation set. AZT was used in 29 cases, and TDF was used in 62 cases. LASSO regression analysis was employed to screen seven non-zero variables, including age, ART regimen, baseline estimated glomerular filtration rate (eGFR), baseline CD4 + T lymphocyte count, baseline human immunodeficiency virus (HIV) RNA, baseline hemoglobin, and baseline aspartate aminotransferase (AST), their LASSO regression coefficient were 1.296, 0.250, 1.443, 0.240, 0.120, 0.395, and 0.002, respectively. Based on these variables, a visual nomogram model was constructed and subsequently validated. Through ROC curve analysis, the AUC for the modeling set was 0.826 (95% confidence interval ( CI) 0.767 to 0.884), with a sensitivity of 0.731 and a specificity of 0.809. For the validation set, the AUC was 0.872 (95% CI 0.807 to 0.956), with a sensitivity of 0.875 and a specificity of 0.778. The calibration curve results for the modeling set showed a mean absolute error (MAE) of 0.012 and a consistency index of 0.826, while the validation set had an MAE of 0.021 and a consistency index of 0.872. These results indicated that the model had a high goodness-of-fit, excellent calibration performance, and was reliable and stable. When the risk threshold for the modeling set ranged from 2% to 73%, the model demonstrated favorable net benefits, indicating its excellent clinical utility. Conclusion:The nomogram-based risk prediction model for kidney injury in HIV/AIDS patients is constructed using seven variables including age, ART regimen, baseline eGFR, baseline CD4 + T lymphocyte count, baseline HIV RNA, baseline hemoglobin, and baseline AST, which provides a valuable tool for early identification of individuals at risk of kidney injury and supports timely clinical interventions.
