[Construction of a risk prediction model for diabetes after kidney transplantation based on genome-wide association study].

Objective: To explore the clinical risk factors and susceptibility genes of diabetes after kidney transplantation (PTDM) and construct a risk prediction model for PTDM. Methods: The data of kidney transplant recipients who underwent follow-up in the Affiliated Lihuili Hospital, Ningbo University and Sir Run Run Shaw Hospital, Zhejiang University School of Medicine from January 2001 to December 2022 were retrospectively analyzed. The recipients were divided into PTDM group and Non-PTDM group according to whether they were complicated with PTDM. The differences in clinical indicators between the two groups were compared, the risk factors affecting the incidence of PTDM were determined, and susceptibility genes of PTDM were screened by genome-wide association study (GWAS). PTDM risk prediction models based only on clinical indicators (Model 1) and clinical indicators combined with susceptibility genes (Model 2) were established respectively, and the predictive performance of the two prediction models was compared. Finally, the Nomogram of the optimal model was drawn, and the discrimination, calibration and clinical applicability of the model were evaluated. Results: A total of 113 kidney transplant recipients (70 males and 43 females) were included, with an average age of (46.2±10.8) years. There were 51 cases in PTDM group and 62 cases in Non-PTDM group. The related factors screened by GWAS and logistic regression analysis included family history of diabetes ( OR =88.912, 95% CI : 5.827-1 356.601, P =0.001), preoperative triglyceride (TG) ( OR =1.888, 95 % CI : 1.150-3.098, P =0.012), uric acid (UA) ( OR =1.011, 95% CI : 1.000-1.022, P =0.045) and rs802707 ( OR =10.046, 95% CI : 1.462-69.042, P =0.019). The area under the curve (AUC) of the receiver operating characteristics analysis (ROC) predicted by Model 1 for PTDM was 0.891 (95% CI : 0.811-0.972), with the sensitivity of 0.889 and the specificity of 0.742. The AUC of ROC curve predicted by Model 2 for PTDM was 0.930 (95% CI : 0.864-0.995), with the sensitivity of 0.885 and the specificity of 0.900. Conclusions: Family history of diabetes, preoperative TG and UA, and rs802707 are significantly associated with the occurrence of PTDM. In addition, the combination of susceptibility genes could improve the predictive ability of clinical indicators for the risk of PTDM.