Abstract
Aims/hypothesis
This study aimed to explore the added value of subgroups which categorize individuals with type 2 diabetes by K-means clustering for two primary care registries (Netherlands and Scotland), inspired by Ahlquist’s novel diabetes subgroups and previously analyzed by Slieker et al.
Methods
We used two Dutch and Scottish diabetes cohorts (n = 3,054 and 6,145; median follow-up = 11.2 and 12.3 years) and defined five subgroups by K-means clustering with age, BMI, HbA1c, HDL, and C-peptide. Cluster consistency over follow-up was assessed. We investigated differences between subgroups by trajectories of risk factor values (random intercept models), medication patterns (multinomial logistic models), and time to diabetes-related complications (log-rank tests and Cox models). We also compared directly using the clustering indicators as predictors of progression versus the K-means discrete subgroups.
Results
Clusters were consistent over eight years with an accuracy ranging from 59% to 72%. Subgroups’ risk factors were significantly different, and these differences remained generally consistent over follow-up. Among all subgroups, individuals with severe insulin-resistance faced a significantly higher risk of myocardial infarction both before (HR 1.65; 95% CI 1.40 - 1.94) and after adjusting for age effect (HR 1.72; 95% CI 1.46–2.02) compared to mild diabetes with high HDL. Individuals with severe insulin-deficient diabetes were most intensively treated with more than 25% prescribed insulin at 10 years of diagnosis. For severe insulin-deficient diabetes relative to mild diabetes, the relative risks for using insulin relative to no treatment would be expected to increase by a factor of 3.07 (95% CI 2.73-3.44), holding other factors constant. Clustering indicators were better predictors of progression variation relative to subgroups but prediction accuracy improved after combining both.
Conclusions/interpretation
Data-driven subgroups’ allocations were consistent over follow-up and captured significant differences in risk factor trajectories, medication patterns, and complication risks. Subgroups serve best as a supplement rather than the compression for continuous clustering indicators.
This study aimed to explore the added value of subgroups which categorize individuals with type 2 diabetes by K-means clustering for two primary care registries (Netherlands and Scotland), inspired by Ahlquist’s novel diabetes subgroups and previously analyzed by Slieker et al.
Methods
We used two Dutch and Scottish diabetes cohorts (n = 3,054 and 6,145; median follow-up = 11.2 and 12.3 years) and defined five subgroups by K-means clustering with age, BMI, HbA1c, HDL, and C-peptide. Cluster consistency over follow-up was assessed. We investigated differences between subgroups by trajectories of risk factor values (random intercept models), medication patterns (multinomial logistic models), and time to diabetes-related complications (log-rank tests and Cox models). We also compared directly using the clustering indicators as predictors of progression versus the K-means discrete subgroups.
Results
Clusters were consistent over eight years with an accuracy ranging from 59% to 72%. Subgroups’ risk factors were significantly different, and these differences remained generally consistent over follow-up. Among all subgroups, individuals with severe insulin-resistance faced a significantly higher risk of myocardial infarction both before (HR 1.65; 95% CI 1.40 - 1.94) and after adjusting for age effect (HR 1.72; 95% CI 1.46–2.02) compared to mild diabetes with high HDL. Individuals with severe insulin-deficient diabetes were most intensively treated with more than 25% prescribed insulin at 10 years of diagnosis. For severe insulin-deficient diabetes relative to mild diabetes, the relative risks for using insulin relative to no treatment would be expected to increase by a factor of 3.07 (95% CI 2.73-3.44), holding other factors constant. Clustering indicators were better predictors of progression variation relative to subgroups but prediction accuracy improved after combining both.
Conclusions/interpretation
Data-driven subgroups’ allocations were consistent over follow-up and captured significant differences in risk factor trajectories, medication patterns, and complication risks. Subgroups serve best as a supplement rather than the compression for continuous clustering indicators.
| Original language | English |
|---|---|
| Publication status | Unpublished - 3-Apr-2022 |
| Event | The 56th annual meeting of the European Diabetes Epidemiology Group (EDEG) - Hersonissos, Greece Duration: 2-Apr-2022 → 5-Apr-2022 |
Conference
| Conference | The 56th annual meeting of the European Diabetes Epidemiology Group (EDEG) |
|---|---|
| Country/Territory | Greece |
| City | Hersonissos |
| Period | 02/04/2022 → 05/04/2022 |