**Background**
Alzheimer's disease (AD), type 2 diabetes mellitus (T2DM), and depression are three common non-communicable diseases with substantial comorbidity. T2DM carries approximately a 1.5-fold higher risk of non-vascular dementia, and T2DM more than doubles the odds of comorbid depression. Mid-life is thought to be the critical period when modifiable risk factors such as depression and poor glucose control exert their largest effects on later dementia risk. Up to 35% of dementia cases are attributable to preventable risk factors. Despite this, the inter-relationships among insulin resistance, depression, and cognition have not been directly characterised in middle-aged adults prior to dementia onset.
**Methods**
This study used baseline cross-sectional data from 665 cognitively healthy participants aged 40-59 years (mean age 51.20, SD=5.44; 61.65% female; mean education 16.67 years, SD=3.39) recruited from four UK centres (West London, Edinburgh, Cambridge, Oxford) in the PREVENT cohort (2014-2019). Cognition was assessed using the COGNITO battery, including phonemic fluency (mean 11.32, SD=4.12), semantic fluency (mean 16.40, SD=4.15), processing speed (mean 5756.29 ms, SD=1516.21), recall (mean 6.90, SD=1.47), the 4 Mountains Test (mean 10.41, SD=2.34), and intraindividual SD of reaction time (mean 0.14, SD=0.07). Insulin resistance was measured using HOMA-IR from fasting plasma samples (mean 60.02, SD=40.85). Depression was measured using the CES-D (mean 9.26, SD=8.46). Body mass index averaged 27.75 (SD=5.60). A structural equation model (SEM) estimated via maximum likelihood with missing values was constructed with executive function as a latent construct from phonemic and semantic fluency tasks, and direct paths from insulin resistance to cognition and depression, and from depression to cognition, adjusting for age, sex, and education. The model was repeated by age decade (40-49 vs 50-59).
**Key Results**
The full model demonstrated good fit: RMSEA=0.02, CFI=0.983, TLI=0.925, χ²=15.80, p=0.148. In the full sample, higher insulin resistance significantly predicted lower executive function (b=-0.12, p<0.01) and higher depressive scores (b=0.15, p<0.001). Higher depressive scores predicted poorer 4 Mountains Test performance (b=0.14, p<0.01). In pairwise correlations, insulin resistance correlated with semantic verbal fluency (r=-0.18, p<0.01), sex (r=0.16, p<0.01), and depression (r=0.15, p<0.01). When stratified by age, the insulin resistance-executive function relationship was significant for the older group (b=-0.15, p<0.01) but non-significant for the younger group (b=-0.09, p=0.126). Insulin resistance predicted depression in both age groups (p<0.01). Depression predicted poorer 4 Mountains Test performance in the older group only (p<0.05).
**Clinical Implications**
The findings support a multi-domain framework for dementia prevention focused on mid-life, especially in adults aged 50-59 years. Insulin resistance may contribute to executive dysfunction via cerebrovascular rather than AD-specific neurodegenerative pathways, consistent with literature linking T2DM to vascular dementia risk. The bidirectional metabolic-affective pathway, whereby insulin resistance predicts depressive symptoms across mid-life, suggests that combined interventions targeting glycaemic control and mood may be more effective than addressing either risk factor alone. Limitations include the cross-sectional design, use of HOMA-IR rather than the hyperinsulinaemic euglycaemic clamp, inability to assess central insulin resistance, and lack of dementia-specific biomarkers. Future prospective research using the PREVENT cohort can clarify longitudinal causality and inform targeted multi-domain interventions for at-risk middle-aged adults.