Systematic review and meta-analysis of cardiovascular event risk prediction models in maintenance hemodialysis patients.

This research pursues a systematic review and meta-analysis concerning the cardiovascular event risk prediction models for maintenance hemodialysis patients. Through systematic literature searching, the titles and abstracts of 23,707 related papers were initially screened, ultimately including 16 papers covering 17 prediction models. The results reveal that among these models, a total of 16 predictive variables were chosen at least twice, with age, diabetes history, and history of cardiovascular disease being the primary predictors. Regarding model validation, 14 models underwent internal validation, 3 models underwent external validation, while 3 models were not subjected to any form of validation. Additionally, calibration testing was performed on 14 models. Risk of bias assessment showed that only 1 model was rated as low risk bias, while the other models were rated as high risk bias due to issues with study cohort characteristics and methodology. Meta-analysis results showed that the combined C-statistic for 13 prediction models was 0.80 (95%CI = 0.74, 0.86), and no significant publication bias was detected. Thus, future construction and validation of prediction models should strictly follow reliable methodological standards and enhance external validation to provide more reliable evidence-based guidance for predicting cardiovascular event risk in maintenance hemodialysis patients.