Clinical phenotypes of type 2 diabetes and their association with microvascular complications in primary care: a cluster analysis

Background Type 2 diabetes mellitus is a heterogeneous condition associated with a substantial burden of microvascular complications. Conventional assessment based on isolated clinical variables may not fully capture this heterogeneity. Cluster analysis offers a potential strategy to identify clinically relevant diabetes phenotypes using routinely collected primary care variables. Objective To identify clinical phenotypes of type 2 diabetes in a primary care population using cluster analysis and to examine their association with prevalent diabetic retinopathy and nephropathy. Methods A cross-sectional study was conducted using routinely collected data from primary care centres in the Cantabrian Health Service. Centres were randomly selected using probability proportional to size sampling. People with type 2 diabetes mellitus were identified from clinical records, and those with missing key variables were excluded. K-means clustering was applied using routinely collected clinical variables, including body mass index, systolic and diastolic blood pressure, HbA1c, age, and years since diagnosis. Associations between clinical phenotypes and prevalent microvascular complications were assessed using logistic regression models adjusted for age, sex, and years since diagnosis. Model discrimination and calibration were evaluated using the area under the receiver operating characteristic curve (AUC), Hosmer–Lemeshow test, Nagelkerke R², and Brier score. Results Of 742 initially identified individuals, 680 were included in the final analytical sample. Phenotype assignment was possible for 674 individuals, who were classified into three clinically interpretable diabetes phenotypes: controlled ( n = 492), metabolic ( n = 146), and hypertensive ( n = 36). The prevalence of diabetic retinopathy was 20.5%, 58.2%, and 27.8%, respectively, while the prevalence of diabetic nephropathy was 25.9%, 51.1%, and 27.8%. After adjustment, the metabolic phenotype was associated with higher odds of prevalent retinopathy (OR 4.69, 95% CI 2.99–7.36; p < 0.001) and prevalent nephropathy (OR 2.52, 95% CI 1.67–3.79; p < 0.001) compared with the controlled phenotype. The hypertensive phenotype was associated with prevalent retinopathy (OR 2.43, 95% CI 1.03–5.69; p = 0.042), but not with prevalent nephropathy (OR 1.23, 95% CI 0.56–2.68; p = 0.611). The retinopathy model showed good discrimination (AUC 0.82, 95% CI 0.78–0.86), whereas the nephropathy model showed lower discrimination (AUC 0.67, 95% CI 0.63–0.72). Sensitivity analyses yielded consistent results. Conclusions Three clinically interpretable diabetes phenotypes were identified in a real-world primary care population and were associated with different patterns of prevalent microvascular complications. The metabolic phenotype showed the highest prevalence and higher adjusted odds of both retinopathy and nephropathy, while the hypertensive phenotype showed a more selective association with retinopathy. These findings suggest that routinely collected primary care data can be used to describe clinically meaningful patterns of co-occurrence between type 2 diabetes phenotypes and microvascular complications. Given the cross-sectional design, these findings should not be interpreted as predictive or causal, and longitudinal studies and external validation are required before these phenotypes can be considered for clinical implementation.