Abstract 17868: Longitudinal Trajectories of Clinical Risk Factors Are Associated With Incident Cardiovascular Disease in the Community
Introduction: Cardiovascular disease (CVD) risk prediction is traditionally based on values from a single time point. The relative associations of longitudinal trajectories of clinical risk factors (i.e., cumulative exposure, rate of change, and variability) with incident CVD are largely unknown.
Hypothesis: Antecedent risk factor trajectories are associated with incident CVD after adjustment for baseline (single examination) values.
Methods: Framingham Offspring study participants totaling 12,080 observations (mean age 60±7 years, 54% women) from 6 examination cycles were followed up to 6 years. At each examination, individuals >50 years without prevalent CVD had risk factors measured. We used stepwise Cox regression models to relate the baseline CVD risk factors, and their corresponding longitudinal trajectory variables, with incident CVD. The trajectory variables included the mean, standard deviation (SD) and slope for continuous variables, and the sum and mean for categorical variables.
Results: We observed 634 incident CVD events (226 in women) during 3.5±0.8 years of follow-up. Traditional risk factors (ascertained at a single examination) including age, male sex, systolic BP (blood pressure), diabetes, lower high-density lipoprotein cholesterol and hypertension treatment were directly related to incident CVD (all p ≤0.01). In a multivariable model adjusted for these factors, mean body mass index, mean total cholesterol and mean smoking, and SD of diastolic BP were significantly associated with incident CVD (all p ≤0.03; Table) and attenuated the associations of baseline risk factors. Prediction of future CVD was marginally improved by incorporating trajectory variables (c-statistic 0.74, difference versus baseline values only = 0.009, 95% CI 0.003-0.015).
Conclusion: Longitudinal trajectories, particularly cumulative exposure, of CVD risk factors predict incident CVD and suggest a role in improving current risk prediction methods.
Author Disclosures: M. Nayor: None. A. Lyass: None. M.J. Pencina: None. R.S. Vasan: None. C.W. Tsao: None.
- © 2016 by American Heart Association, Inc.