Development and validation of cardiovascular risk scores for haemodialysis patients

Abstract Background A simple clinical tool to predict cardiovascular disease risk does not exist for haemodialysis patients. The long-term coronary risk Framingham Heart Study Risk score (FRS), although used in this population, may be inadequate. Therefore, we developed separate risk-scores for cardiovascular mortality (CVM) and cardiovascular morbidity & mortality (CVMM) in a Fresenius Medical Care-based haemodialysis patient cohort (AROii). Methods Applying a modified FRS approach, we derived and internally validated two-year risk-scores in incident European adult patients randomly assigned to a development (N = 4831) or a validation dataset (N = 4796). External validation was conducted in the third Dialysis Outcomes and Practice Patterns Study (DOPPS III) cohort. Additional discrimination comparing to the Framingham score was performed. Results The overall two-year CVM and CVMM event rates were 5.0 and 22.6 per 100 person-years respectively. Common risk predictors included increasing age, cardiovascular disease history, primary diabetic nephropathy, low blood pressure, and inflammation. The CVM score was more predictive in AROii (c-statistic 0.72) and in DOPPS III (c-statistic 0.73–0.74) than the CVMM (c-statistic 0.66–0.67 & 0.63 respectively). The FRS was not predictive of either CVM (c-statistic 0.54) or CVMM in AROii (c-statistic 0.56). Conclusions We describe novel, easy-to-apply and interpret CV risk-scores for haemodialysis patients. Our improved cardiovascular prediction performance over traditional (FRS) scores reflected its tailored development and validation in haemodialysis populations, and the integration of non-classical cardiovascular risk factors. The lower expected versus observed CVM and CVMM risk suggests the existence of novel cardiovascular risk factors in this patient population not measured in this study.