Introduction: Cumulative exposure to cardiovascular disease (CVD) risk factors during young adulthood is associated with later life CVD risk. Few prospective cohort studies measured exposures in young adulthood. We sought to develop and validate a method to impute trajectories of CVD risk factors across the life course.
Methods: 36,546 participants (55% women, 25% black, average exams 5.1/participant) from 6 studies (ARIC, CARDIA, CHS, Framingham Offspring, Health ABC, and MESA) were included. Demographics and CVD risk factors (BMI, smoking, BP, lipids, glucose, medications for BP, lipids and glucose) were collected at each exam and harmonized across cohorts. We multiply imputed complete risk factor trajectories from age 18 to 99 years for each participant using an extension of linear mixed modeling (for continuous variables) and interval-censored survival modeling (for categorical variables), taking into account the multilevel structure of data. For validation, we randomly selected 25% of all participants and deleted their observed data for exam age 20-35, 50-65, or 80-95 years. We then imputed risk factor values for deleted age periods and compared imputed values with directly observed values.
Results: Imputed values were relatively consistent with observed values for BMI, SBP, LDL, and glucose, particularly in young and middle ages (Figure). The mean (standard deviation) of the difference between imputed vs. observed values for BMI, SBP, LDL, and glucose were 0.1 (2.7) kg/m2, 0.9 (16.3) mm Hg, -1.1 (30.2) mg/dL, and -0.6 (23.0) mg/dL. The prevalence of imputed smoking, diabetes, and medications were also consistent with observed data.
Conclusions: We demonstrated a validated method for estimating CVD risk factor trajectories across the life course. This approach may advance understanding of potential impact of cumulative early risk factor exposures on later life CVD risk, and inform primary prevention strategies over the life course.
Figure. Mean and prevalence of observed vs. imputedrisk factorsby age periods