Dandan Tian ¹ You Xu ^2* Ying Wang ^3* Xirui Zhu ^4* Chun Huang ⁴ Min Liu ^1* Panlong Li ^4,5* Xiangyong Li ^6*

• ¹ Henan Provincial People's Hospital, Zhengzhou, China
• ² Department of Clinical Laboratory, Third Affiliated Hospital, Southern Medical University, Guangzhou, China
• ³ School of Public Health, Sun Yat-sen University, Guangzhou, China
• ⁴ School of Electrical and Information Engineering, Zhengzhou University of Light Industry, Zhengzhou, Henan Province, China
• ⁵ Department of Medical Imaging, Henan Provincial People’s Hospital, Zhengzhou, Henan Province, China
• ⁶ Third Affiliated Hospital of Sun Yat-sen University, Guangzhou, Guangdong Province, China

Background: The risk factors of cardiovascular disease (CVD) in end-stage renal disease (ESRD) with hemodialysis remain not fully understood. In this study, we developed and validated a clinical-longitudinal model for predicting CVD in patients with hemodialysis, and employed Mendelian randomization to evaluate the causal relationships. Methods: The study included 468 hemodialysis patients, and biochemical parameters were evaluated every three months. A generalized linear mixed (GLM) predictive model was applied to longitudinal clinical data. Calibration curves and area under the receiver operating characteristic curves (AUCs) were used to evaluate the performance of the model. Kaplan-Meier curves were applied to verify the effect of selected risk factors on the probability of CVD. Genome-wide association study (GWAS) data for CVD (n=218792101866 cases), end-stage renal disease (ESRD, n=16405, 326 cases), diabetes (n=202046, 9889 cases), creatinine (n=7810), and uric acid (UA, n=109029) were obtained from the large-open GWAS project. The inverse-variance weighted MR was used as the main analysis to estimate the causal associations, and several sensitivity analyses were performed to assess pleiotropy and exclude variants with potential pleiotropic effects. Results: The AUCs of the GLM model was 0.93 (with accuracy rates of 93.9% and 93.1% for the training set and validation set, sensitivity of 0.95 and 0.94, specificity of 0.87 and 0.86). The final clinical-longitudinal model consisted of 5 risk factors, including age, diabetes, ipth, creatinine, and UA. Furthermore, the predicted CVD response also allowed for significant (p<0.05) discrimination between the Kaplan-Meier curves of each age, diabetes, ipth, and creatinine subclassification. MR analysis indicated that diabetes had a causal role in risk of CVD (β=0.088, p<0.0001) and ESRD (β=0.26, p=0.007). In turn, ESRD was found to have a causal role in risk of diabetes (β=0.027, p=0.013). Additionally, creatinine exhibited a causal role in the risk of ESRD (β=4.42, p=0.01). Conclusions: The results showed that old age, diabetes, and low level of ipth, creatinine, and UA were important risk factors for CVD in hemodialysis patients, and diabetes played an important bridging role in the link between ESRD and CVD.