Clustering of HbA1c Longitudinal Trajectories in Type 2 Diabetes Mellitus (T2DM)

Background: Attaining glycaemic control in type 2 diabetes mellitus (T2DM) is essential to reduce the risk of morbidity and mortality. Glycated haemoglobin (HbA1c) is often used as a clinical target in T2DM management. Identifying distinct clusters of HbA1c trajectories may help to develop more personalized strategies for T2DM management.

Objectives: To identify and characterize common patterns of HbA1c progression among T2DM patients who initiate a non-insulin antidiabetic drug in the UK.

Methods: The IQVIA Medical Research Data incorporating data from THIN, a Cegedim database of anonymized electronic health records, was used to identify all patients with ≥4 HbA1c measurements after a first-ever prescription for a non-insulin antidiabetic drug (NIAD; index date) between 2006 and 2019. Patient trajectory clusters were identified using an Expectation-Maximization algorithm by iteratively fitting k thin-plate splines and reassigning each patient to the nearest cluster. Clusters were characterised based on evidence of initial response (adequate/inadequate) and long-term control. A maintained decline in HbA1c for six months after index defined adequate response. HbA1c values at 10-years defined long-term response in 4 categories: optimally controlled ( < 7.0%), controlled (7.0-7.5%), sub optimally controlled (7.6—9.0%), and poorly controlled (>9.0%). Cox proportional hazards models estimated the risk of retinopathy between clusters.

Results: We identified 116,251 new users of NIADs with ≥4 HbA1c measurements. We found five distinct clusters of HbA1c progression: adequate—controlled (A-C), adequate—optimally controlled (A-OC), adequate—sub optimally controlled (A-SOC), inadequate —sub optimally controlled (I-SOC), and inadequate—poorly controlled (I-PC). Compared to the A-OC cluster with optimal HbA1C ( < 7.0%), the I-PC cluster was younger (53.8 vs 63.8 years), more obese (mean body mass index 34.1 vs. 32.5 kg/m2), and had shorter time to first insulin use (6.2 vs. 8.4 years). The I-PC and A-C clusters had similar index HbA1C (>10%), but different trajectories — the A-C cluster achieved and maintained control (HbA1C < 7.5%), while the I-PC HbA1c remained >9.0%. A significant risk of retinopathy was observed in I-PC and A-C clusters (HR: 1.3, 95% confidence interval [CI] 1.2-1.4, and HR: 1.2, 95% CI 1.1-1.3, respectively), but not the A-SOC or I-SOC, when compared to the A-OC cluster.

Conclusions: This study identified five distinct HbA1c trajectories with different patient and risk profiles. Notably, young patients with high obesity (I-PC cluster) had the worst HbA1c progression. However, despite achieving diabetic control, patients with high index HbA1C (A-C cluster) had a similar increased risk of retinopathy to those with poor control (I-PC).