Wu Ya, Dong Danmeng, Liu Yang, Xie Xiaoyun
Department of Endocrinology and Metabolism, Shanghai Tenth People's Hospital, Tongji University School of Medicine, Shanghai, 200072, China.
School of Medicine, Anhui University of Science and Technology, Huainan, 232001, China.
Nutr Metab (Lond). 2025 Mar 25;22(1):26. doi: 10.1186/s12986-025-00917-0.
The prognostic nutritional index (PNI), an indicator of nutritional status, has been linked to various diabetic complications. However, its relationship with diabetic peripheral neuropathy (DPN) remains unclear. This study aimed to explore the association between PNI and DPN using machine learning (ML) approaches.
A total of 625 patients with type 2 diabetes (T2D) were enrolled, with 282 diagnosed with DPN. PNI was calculated based on serum albumin and lymphocyte count. Random forest (RF) and eXtreme Gradient Boosting (XGBoost) models were developed to predict DPN using clinical and biochemical data. SHapley Additive exPlanations (SHAP) were applied to determine feature importance. Multivariate logistic regression was used to evaluate the relationship between PNI quartile and DPN risks.
Both RF and XGBoost models exhibited strong performance. The RF model achieved a recall of 78.4%, specificity of 87.8%, and accuracy of 84.0%, while the XGBoost model showed a recall of 77.4%, specificity of 92.1%, and accuracy of 84.8%. SHAP analysis identified lower PNI as a key factor for DPN. Multivariate logistic regression revealed that patients in the lowest PNI quartile had a significantly higher DPN risk compared to those in the highest quartile (OR: 3.271, 95% CI: 1.782-6.006, P < 0.001). Additionally, lower PNI levels were associated with impaired peripheral nerve function, including reduced motor and sensory nerve conduction velocity and action potential amplitudes.
Lower PNI levels were associated with increased DPN risk and poorer nerve function, highlighting the importance of nutritional status in DPN management. Further longitudinal studies are needed to confirm these findings.
预后营养指数（PNI）作为营养状况的一项指标，已与多种糖尿病并发症相关联。然而，其与糖尿病周围神经病变（DPN）的关系仍不明确。本研究旨在采用机器学习（ML）方法探究PNI与DPN之间的关联。
共纳入625例2型糖尿病（T2D）患者，其中282例被诊断为DPN。基于血清白蛋白和淋巴细胞计数计算PNI。利用临床和生化数据建立随机森林（RF）和极端梯度提升（XGBoost）模型来预测DPN。应用SHapley值相加解释法（SHAP）确定特征重要性。采用多因素逻辑回归评估PNI四分位数与DPN风险之间的关系。
RF和XGBoost模型均表现出强大性能。RF模型的召回率为78.4%，特异性为87.8%，准确率为84.0%，而XGBoost模型的召回率为77.4%，特异性为92.1%，准确率为84.8%。SHAP分析确定较低的PNI是DPN的关键因素。多因素逻辑回归显示，PNI四分位数最低的患者与最高四分位数的患者相比，DPN风险显著更高（比值比：3.271，95%置信区间：1.782 - 6.006，P < 0.001）。此外，较低的PNI水平与周围神经功能受损有关，包括运动和感觉神经传导速度及动作电位幅度降低。
较低的PNI水平与DPN风险增加及神经功能较差相关，凸显了营养状况在DPN管理中的重要性。需要进一步的纵向研究来证实这些发现。
