Associations of diabetes subgroups and lifestyle with incident dementia: insights from unsupervised phenotype clustering analysis.

BACKGROUND: To investigate heterogeneity in dementia risk across diabetes subgroups and examine whether lifestyle mitigates this risk. METHODS: We did an unsupervised clustering analysis (self-organizing maps) in patients with diabetes from the Health and Retirement Study (HRS). Clustering variables included age at diabetes onset, body mass index (BMI), hemoglobin A1c (HbA1c), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-c), systolic blood pressure (SBP), estimated glomerular filtration rate (eGFR), and C-reactive protein (CRP). Healthy lifestyle factors included never smoking, low to moderate alcohol consumption, and top tertile of physical activity. Incident dementia was defined as a self-reported diagnosis or Telephone Interview for Cognitive Status (TICS) scores ≤6. RESULTS: We included 10,705 participants (40.2% men and mean age 67.6 years) in this analysis. Three subgroups were identified among diabetes: subgroup 1 (high BMI and elevated CRP), subgroup 2 (late-onset diabetes, high blood pressure, and low eGFR), and subgroup 3 (early-onset diabetes, high HbA1c, and high cholesterol levels). Compared to normoglycemia, subgroup 3 of diabetes showed a significantly higher risk of dementia (hazard ratio [HR] 1.43, 95% confidence interval [CI] 1.15-1.77), while the other subgroups did not. However, adopting a favorable lifestyle was associated with a reduced risk of dementia in participants with normoglycemia (HR 0.70, 95% CI 0.57-0.87), prediabetes (HR 0.67, 95% CI 0.50-0.90), and subgroup 1 of diabetes (HR 0.33, 95% CI 0.14-0.79). CONCLUSIONS: Diabetes-related dementia risk is heterogeneous, and benefits of lifestyle may vary by diabetes subgroups.