Back

Cardiovascular Risk Models in Chronic Kidney Disease: A Systematic Review and External Validation of Models

Major, R. W.; Grant, R.; Shepherd, D.; Medcalf, J. F.; Jesus-Silva, J.; Gray, L. J.; Brunskill, N. J.

2022-01-23 cardiovascular medicine
10.1101/2022.01.22.22268958 medRxiv
Show abstract

Background and ObjectivesRisk factors for cardiovascular disease in chronic kidney disease differ to the general population due to the increased role of inflammation, calcification and arteriosclerosis. Over 350 cardiovascular risk models ("models") exist in the general population but most have not undergone testing ("external validation") in other populations, such as cohorts with chronic kidney disease. We aimed to update a previous systematic review of models in chronic kidney disease and then perform external validation of these and general population models in a chronic kidney disease cohort. Design, Setting, Participants and MeasurementsWe searched Medline up to 26th August 2020 for models in chronic kidney disease. We performed external validation of models using a primary care chronic kidney disease cohort of 17,248 individuals with 2,072 (12.0%) cardiovascular events, 5,108 (29.6%) deaths and a median follow-up of 5.0 years. Model discrimination and calibration was assessed and where appropriate models were re-calibrated. Multiple imputation was used to account for missing data. ResultsSeven chronic kidney disease specific models were identified. These models and three general population models underwent external validation. All models had worse discrimination of events than in their original cohorts, particularly in the chronic kidney disease specific models. General population models were miscalibrated and overpredicted risk. The major contributor to this was the high competing risk of death from non-cardiovascular causes. ConclusionsExisting chronic kidney disease cardiovascular risk models performed poorly in external validation. General population cardiovascular disease risk models should be interpreted with caution in individuals with chronic kidney disease as they may overestimate risk due to the competing risk of death. Further development of models to include simultaneous risk prediction for cardiovascular and non-cardiovascular disease risk is required.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.