Abstract P017: Development And Validation Of A Body Composition-based Prediction Model For Cardiovascular Outcomes Among Older Adults
Chenkai Wu,Lihui Chen,Xinran Wang
DOI: https://doi.org/10.1161/circ.145.suppl_1.p017
IF: 37.8
2022-03-01
Circulation
Abstract:Background: Cardiovascular disease (CVD) has long been a leading cause of death in the world and a major public health concern. Precise stratification of cardiovascular risk is critically important to deliver person-centered treatment decisions. Several risk prediction models of CVD have been developed, mainly focusing on demographics, healthy lifestyles, and blood-based biomarkers (mostly lipids). Recent evidence suggests that components of body composition are associated with the risk of CVD. However, it remains unknown whether the use of body composition data could improve the performance of a well-established CVD prediction model. Objectives: We aimed to derive and validate a model for predicting the risk of incident CVD based on body composition assessed by the Dual-energy X-ray absorptiometry (DXA) among older black and white adults. Methods: Participants were 1,994 adults initially free of CVD from the Health, Aging, and Body Composition Study. We developed and validated a model for risk of myocardial infarction, stroke, heart failure, and cardiovascular death over 10 years based on body composition data. We used the LASSO and stepwise backward elimination for variable selection of body composition data in the Cox model. Weibull accelerated failure time model was used to obtain the risk score. We compared the model performance (goodness-of-fit, discrimination, and calibration) between the body composition model and the Framingham model, in which sex, age, total cholesterol, high-density lipoprotein cholesterol, systolic blood pressure, diabetes, and smoking were included. We also examined whether the inclusion of body composition data in the Framingham model could improve model performance. Results: The average age was 73.0 years (SD=2.8 years); 1,118 (56.1%) were females and 802 (41.3%) were black. A total of 8 body composition variables were retained after variable selection (8-DXA model). The C statistics were 0.608 and 0.616 for the 8-DXA model and the Framingham model, respectively. The 8-DXA plus Framingham model significantly improved the C-statistic by 0.026 (95% CI: 0.011, 0.042). When compared with the Framingham model, the 8-DXA plus Framingham model had an integrated discrimination index of 0.016 (95% CI: 0.009, 0.023) indicating an absolute increase of 12% in average risk for participants with events compared with those without. the 8-DXA plus Framingham model had a net reclassification index (NRI) of 0.167 (95% CI: 0.008, 0.259), with event-specific NRI of 0.018 (95% CI: -0.063, 0.094) and no event-specific NRI of 0.149 (95% CI: 0.098, 0.205). Conclusion: Among older adults free of CVD, the inclusion of body composition data assessed by DXA could moderately improve the prediction performance of a well-established CVD prediction model. Future research is needed to validate the usefulness of body composition data in predicting the risk of CVD in other cohorts.
cardiac & cardiovascular systems,peripheral vascular disease