why choose us

300×250 Ad Slot

Research Article: Prognostic value of the atherogenic index of plasma for early-stage diabetic kidney disease in type 2 diabetes: a retrospective cohort study using supervised machine learning

Date Published: 2026-06-30

Abstract:
Diabetic kidney disease (DKD) is the leading cause of end-stage renal disease (ESRD). Its insidious onset means that once it progresses to the stage of heavy proteinuria or significant decline in renal function, treatment difficulty and burden increase dramatically. Therefore, identifying high-risk individuals in the early-stage DKD (ES-DKD, stages 1–2) and intervening promptly is crucial for delaying disease progression and improving prognosis. The atherogenic index of plasma (AIP) is a novel composite indicator reflecting dyslipidemia and insulin resistance. However, its prognostic value in ES-DKD among patients with type 2 diabetes mellitus (T2DM) remains unclear. This study aims to investigate the association between AIP and the risk of developing ES-DKD in T2DM patients, and to develop prognostic prediction models using supervised machine learning (SML) algorithms. Clinical data of 1,006 patients with T2DM were extracted from the hospital information system. The study population was divided into three groups according to the tertiles of baseline AIP values. The study endpoint was the occurrence of ES-DKD during follow-up. Cox proportional hazards regression models and subgroup analyses were used to assess the association between AIP and the risk of ES-DKD. Restricted cubic spline (RCS) analysis was applied to examine the linear trend of the dose–response relationship. Differences between groups were compared using Kaplan–Meier survival curves and the log-rank test. The Boruta algorithm was employed to evaluate the importance of AIP among the predictor variables. Feature selection was further performed using least absolute shrinkage and selection operator (LASSO) regression and backward stepwise selection. Prognostic prediction models were constructed based on six SML algorithms, including random survival forest (RSF), LASSO-Cox, CoxBoost, extreme gradient boosting (XGBoost), supervised principal components (superpc), and partial least squares regression for the Cox model (plsRcox). Model performance was evaluated using time-dependent area under the curve (AUC), calibration curves, and decision curve analysis (DCA). Among the 1,006 T2DM patients, the median follow-up time was 49?months, and 404 ES-DKD events occurred. Multivariable Cox regression analysis showed that for each one-unit increase in AIP, the risk of ES-DKD increased by 74% (HR?=?1.74, 95% CI: 1.19–2.54, p =?0.004). Compared with the lowest AIP tertile group, patients in the highest tertile group had a 33% increased risk of ES-DKD (HR?=?1.33, 95% CI: 1.01–1.74, p =?0.039). Subgroup analyses yielded robust results, and RCS analysis indicated a linear positive correlation between AIP and the risk of ES-DKD. Kaplan–Meier curves showed that the long-term event-free survival rate was significantly lower in the high AIP group. Boruta feature selection results demonstrated that AIP had a high importance score. Among the six SML models, the prognostic prediction model based on the XGBoost algorithm exhibited the best performance, achieving the highest time-dependent AUC. In the training and validation cohorts, the AUCs for predicting the 1-, 3-, and 5-year risk of ES-DKD ranged from 0.779 to 0.809 and 0.752 to 0.784, respectively. Calibration curves showed good agreement between predicted probabilities and observed probabilities, and DCA indicated that the model provided high clinical net benefit. To facilitate the application of ES-DKD risk prediction in clinical practice, a web-based tool was also developed in this study. AIP is an independent prognostic factor for the occurrence of ES-DKD in patients with T2DM. Integrating this low-cost, easily accessible biomarker into SML frameworks may provide effective tools for early risk stratification and individualized management.

Introduction:
Diabetic kidney disease (DKD) is the leading cause of end-stage renal disease (ESRD). Its insidious onset means that once it progresses to the stage of heavy proteinuria or significant decline in renal function, treatment difficulty and burden increase dramatically. Therefore, identifying high-risk individuals in the early-stage DKD (ES-DKD, stages 1–2) and intervening promptly is crucial for delaying disease progression and improving prognosis. The atherogenic index of plasma (AIP) is a novel composite indicator…

Read more

300×250 Ad Slot