Latent Class Trajectory Analysis of Risk Factors Uncovers Progression to Type 2 Diabetes

Abstract

We identified trajectories of diabetes risk factors in the Longitudinal Epidemiologic Assessment of Diabetes Risk (LEADR) cohort analyzing 8 years of electronic health records on 1.4 million patients, and investigated associations between trajectories and progression to new onset Type 2 diabetes. Design and Methods: Analyzing LEADR data (2010-2016), we applied Latent Class Trajectory Analysis (LCTA) to classify patterns of risk factor change. There were 824,043 patients with BMIs; 955,128 patients with systolic blood pressures; 957,491 patients with diastolic blood pressures; 300,137 patients with HDLs; 267,553 patients with non-HDL cholesterols; and 297,026 patients with triglycerides. Patients had to have data for all risk factors being assessed. Association between trajectories and incidence of type 2 diabetes for 94,551 patients was assessed using negative binomial regression analysis. Results: Compared to a static BMI trajectory, those with a sustained weight increase (25%+ from starting BMI) were at higher risk of type 2 diabetes over 4.8 years of follow-up (range 2.0 to 8.0 years) (adjusted rate ratios ranged 1.53-1.62, p-value<0.05). Patients with a BMI decrease trajectory (of ~10%), were at reduced risk of diabetes (adjusted rate ratios ranged 0.54-0.74, p-value<.05). BP and lipid trajectories had significant associations with diabetes onset. Conclusions: Regardless of the starting BMI, those who increased their BMI by 25% within two years and maintained the higher weight were significantly at increased risk of type 2 diabetes. Monitoring BMI change and other known risk factor trajectories, BP and lipids, are additional tools for identifying patients at risk for type 2 diabetes.