8-OR: Longitudinal Patterns of Heterogeneity in Metabolic Decompensation among Diabetes Prevention Program Outcomes Study (DPPOS) Participants over 20 Years

Introduction & Objective: Type 2 Diabetes (T2D) complications cause morbidity and mortality, but occur heterogeneously among those at risk. While some individuals maintain stable health over time, others decompensate. Using data from the DPPOS, this study aimed to identify unbiased, data driven subgroups of adults at risk for T2D with heterogeneous metabolic trajectories and clinical outcomes over time.

Methods: Data from 1736 DPP participants over 39 semi-annual visits with repeated clinical parameters (glycemia, insulin, lipids, body mass index (BMI), waist circumference, blood pressure (BP), microalbuminuria (MA) and eGFR) were analyzed. A 3-dimensional tensor of individuals x clinical parameters x visits was constructed. Nonnegative tensor decomposition was performed and the resulting factors were clustered using K-means. Cohesion and separation were computed to select an optimal number of clusters. Cumulative incidence of T2D and kidney dysfunction (eGFR < 60 mL/min) were tabulated for each cluster.

Results: Nine longitudinal clusters of varying sizes were identified. One cluster represented a healthy group with stable clinical parameters (n=556). Kidney dysfunction was found in two clusters with differing patterns and trajectories: 1) slowly declining eGFR without MA or T2D (n=198), 2) MA, elevated BP and delayed eGFR decline (n=138). Other clusters segregated by patterns of glycemia and insulin resistance (IR) ranging from stable IR, glycemia and elevated BMI (n=253), to rapidly accelerating IR, dysglycemia and T2D (n=34).

Conclusion: Temporal trajectories of clinical parameters revealed marked heterogeneity in metabolic decompensation over time for DPPOS participants. A minority had accelerated metabolic and renal decompensation, important groups to target for early intervention. These groups also form the basis for novel biomarker discovery efforts to improve prediction of future complications.