Few studies have examined individual differences in BMI across the entire lifespan. The purpose of this study was to create BMI trajectory groups using latent class mixture models, and in a subset analysis, examine how these BMI trajectory groups are related to cardiometabolic risk (CMR).


The study sample consisted of 2,396 white participants from the Fels Longitudinal Study (1,148 males, 1,248 females), who had from 1 to 45 BMI visits between ages 2 to 75 years. Participants were classified into 4 BMI groups using Proc Traj in SAS. Groups initially started with quintic polynomial distributions (5, 5, 5, 5) across age, and then iteratively estimated with lower order combinations until all groups were linear (1, 1, 1, 1). Covariates of birth year, sex, and exact age differences at each age were included in all models. The best model with the highest BIC and a smallest group membership of ≥5% was chosen. Linear regressions were used to examine the influence of each BMI grouping on CMR at the last visit, adjusting for age and sex.


The 4 trajectory groups consisted of 39%, 39%, 17%, and 5% of the individuals at orders 5,3,5,5 respectively. Average posterior probabilities (≥0.82) showed good fit for all groups. Group 2 individuals were never overweight at any age. Group 1 individuals were never obese, but became overweight at age ~30 years. Group 3 individuals became obese at ~30 years of age. Group 4 individuals became obese in adolescence. Subset analysis indicates that CMRs are significantly worse in order for groups 2-1-3-4 for high density lipoprotein, insulin, and systolic blood pressure (BP), while groups 3 and 4 are generally significantly worse than groups 1 and 2 for triglycerides, glucose and diastolic BP.


Four patterns of BMI change were distinguished by their different ages at onset of overweight and obesity. These BMI trajectories are significant predictors of increased CMR, where earlier onset were associated with worse cardiometabolic health.