Background: Whether diabetes and adipokine-driven inflammation explain the association of obesity to cognitive impairment is unknown.

Methods: Structural equation models estimated the total effects of waist circumference on cognitive outcomes among African American participants cross-sectionally (index exam) and longitudinally. Total effects were deconstructed into direct pathways of waist circumference to cognitive impairment and indirect mediation pathways through leptin, soluble tumor necrosis factor receptor 2 (sTNFR2), and diabetes. Waist circumference, leptin, and sTNFR2 were standardized. Cognitive impairment was defined as MMSE <21 or a z-score < -1.5 standard deviation (SD). Incident cognitive impairment was defined among those without cognitive impairment at the index exam as follow-up MMSE<21, z- score < -1.5, MMSE decline >1 point/year, or z-score decline of >0.1 SD/year.

Results: Among 1008 participants (70% women, mean age 62.9 years, 14.5% with obesity, 26% with diabetes), 132 (13%) had baseline cognitive impairment. Each SD higher waist circumference was associated with higher odds of cognitive impairment, odds ratio (OR) = 1.63; (95% confidence interval: 1.17, 2.24), with mediating pathways explaining 65% of the total effect (58% from diabetes; 7% from inflammation). At follow-up (mean 6.8 years), 106 of 535 (19.8%) had developed cognitive impairment. Each SD higher waist circumference was associated with higher odds of developing cognitive impairment (OR = 1.87; 95%CI: 1.18, 2.74); the direct effect of waist circumference explained 37% of the total effect and mediating pathways explained 63% (61% from diabetes; 2% from inflammation), although individual pathways were not statistically supported in the smaller sample.

Conclusion: Diabetes, and to a lesser degree, adiposity-driven inflammation, appear to explain a substantial proportion of abdominal adiposity relationships with cognitive impairment. The impact of preventing and treating obesity on cognitive outcomes merits study.