Effect modification of the association between BMI and cognition in India
Emma Nichols, Jennifer A Deal, Alden L. Gross, Jinkook Lee- Psychiatry and Mental health
- Cellular and Molecular Neuroscience
- Geriatrics and Gerontology
- Neurology (clinical)
- Developmental Neuroscience
- Health Policy
- Epidemiology
Abstract
Background
High body‐mass index in midlife is an established risk factor for poor cognition and dementia. While most evidence supporting the association between high BMI and dementia comes from high‐income countries, 60% of global dementia cases are expected to be in low‐ and middle‐income countries by 2050. Existing evidence from India suggests higher BMI is associated with better cognition and lower odds of dementia. We sought to replicate these findings and investigate effect modification by age, rurality, and education to better contextualize results.
Method
We used cross‐sectional data from the Longitudinal Aging Study in India (N = 57,164). We used weighted linear regression models to assess associations between latent cognitive functioning, estimated from confirmatory factor analysis models, and BMI category. We compared unadjusted models to models controlling for age, sex, rurality, state, literacy, education, per capita consumption quintile, parental education, smoking, and depressive symptoms. We investigated effect modification by age, rurality, and education.
Result
Compared to unadjusted models, estimated associations from multivariable models were substantially attenuated but remained statistically significant. Underweight BMI was associated with lower cognitive scores and overweight and obesity were associated with higher cognitive scores (underweight = ‐0.16 SD units [‐0.18‐ ‐0.14]; overweight = 0.10 [0.09‐0.12]; obese = 0.13 [0.11‐0.16]). For participants in urban settings or with higher educational attainment, associations between cognition and overweight and obesity were smaller and not statistically significant for those with tertiary education. A global ANOVA test for effect modification was statistically significant for rurality (p<0.01), trending towards significance for education (p = 0.09), but not significant for age category (p = 0.31).
Conclusion
Counter to findings in high‐income settings, but consistent with prior findings in India, high BMI was associated with higher cognitive performance. Lack of evidence for effect modification by age and the association between low BMI (underweight) and lower cognition, indicate reverse causation due to weight loss in early disease stages is unlikely to explain results. Instead, strong attenuation of effects with confounder control and effect modification by rurality and education indicate that associations could be due to the clustering of access to fat‐enriched foods with other factors that are beneficial for health, such as higher education and access to healthcare.