Establishing a model for predicting type 2 diabetes mellitus with cognitive impairment

Objective To screen and evaluate early clinical predictors of type 2 diabetes mellitus with mild cognitive impairment and dementia in Hunan, China. Method This is a cross-sectional study. A total of 164 patients in Hunan Province with T2DM who were treated in the Department of Geriatric Endocrinology, Xiangya Hospital, Central South University from May 2023 to October 2023 were enrolled. The average age was ( 59.18 ± 10.67 years ), 113 males and 51 females. The duration of T2DM was ( 10.28 ± 7.48 ) years. According to the Montreal Cognitive Scale ( MoCA ), the patients were divided into T2DM group ( MoCA ≧ 26 ), T2DM & MCI group ( 19 ≤ MoCA < 26 ), T2DM & De group ( MoCA < 19 ).The height, weight, waist circumference, hip circumference, living habits, food habits and diabetes history of the patients were collected, and the serum levels of PDGFRβ, HDL-3, Nesfatin-1, GDF-15, FPN and 1,5-AG were detected by ELISA kit. Analysis of variance or KW test was used to analyze the differences of each index between groups with or without MCI or De as the outcome. Multivariate COX proportional hazard regression analysis was used to screen out the independent risk factors of T2DM complicated with MCI or De, and the risk Nomogram model of T2DM complicated with MCI or T2DM complicated with De was constructed. The training set and validation set were constructed according to the 7 / 3 splitting principle.The efficiency of each model was evaluated by Random forest algorithm. Results The differences in years of education ( p < 0.01 ), weight ( p = 0.014 ), serum HDL-3 ( p = 0.017 ), FPN ( p < 0.01 ) and 1,5-AG ( p = 0.013 ) levels were statistically significant. Correlation analysis indicated that MoCA score was significantly correlated with years of education ( r = 0.44, p < 0.01 ), weight ( r = -0.16, p < 0.05 ), HDL-3 ( r = -0.2, p < 0.05 ), FPN ( r = -0.17, p < 0.05 ), 1,5-AG ( r = -0.23, p < 0.01 ). Multiple linear regression demonstrated that the patient 's years of education ( p < 0.01 ) and serum 1,5-AG ( p < 0.01 ) levels affected the patient 's MoCA score to a certain extent ( 31 % of the total impact ). Logistic regression analysis proved that the years of education ( p < 0.01 ) and serum 1,5-AG ( p = 0.013 ) level were important factors for the progression of T2DM patients to T2DM & MCI, and the weight ( p = 0.032 ), years of education ( p = 0.016 ) and serum 1,5-AG ( p = 0.019 ) level were important factors for the progression of T2DM & MCI patients to T2DM & De. Nomogram risk assessment models were established based on the above indicators, and the prediction accuracy was about 75 %. The decision curve analysis of the training set and the validation set of the two models showed that the clinical benefit of the model was higher than that of all patients with clinical intervention, and the calibration curves of each model were symmetrically distributed near the diagonal.Conclusion Weight, years of education, serum HDL-3, FPN and 1,5-AG levels are important influencing factors for MCI and De in T2 DM patients. The Nomogram evaluation model based on the above indicators can accurately predict the risk of MCI and dementia in T2 DM patients, and provide a theoretical basis for clinical decision-making.