Establishment of metabolite prediction model for the risk of diabetic retinopathy in Chinese type 2 diabetic population

Abstract

Objective The aim of this study is to examine the relationship between multiple plasma metabolites and the risk of diabetic retinopathy, and to develop a risk prediction model for diabetic retinopathy in Chinese type 2 diabetes patients. Methods This study conducted a hospital-based cross-sectional study. According to medical records, 743 patients from tertiary hospital were retrieved, and were divided into diabetic retinopathy group (DR group) and non-diabetic retinopathy group (non-DR group). The levels of metabolites such as amino acids and acylcarnitine in plasma were measured by Liquid Chromatography-Mass Spectrometry (LC-MS). Partial least squares regression was used to rank the importance of these metabolites. Receiver operating characteristic curve (ROC) curves were made for the variables at the top of the importance list. Area under curve (AUC) was used to evaluate the performance of the prediction model. Results We retrieved 743 hospitalized patients with type 2 diabetes, 94 of which had diabetic retinopathy. Compared to non-DR group, there were more diabetic nephropathy patients (P = 0.019) and stroke patients (P < 0.001) in DR group. A DR risk prediction model consisting of histidine, citrulline, phenylalanine, methionine, tyrosine, C3, C24 was obtained with an AUC of 0.770(95%CI: 0.701–0.839). Conclusions A DR risk prediction model consisting of 7 plasma metabolites is a simple and effective tool to detect the risk of DR in Chinese diabetic patients.