Examining the bidirectional associations between adiposity and cognitive function using population-level data

Abstract

Background: The association between adiposity and cognitive function has been extensively explored in previous literature, and numerous cross-sectional and longitudinal analyses suggest a reliable association. However, most previous studies on this topic were predominantly executed with a narrow, unidirectional assumption that baseline adiposity predicts future cognitive function (i.e., the “brain-as-outcome” perspective). Literature within neuropsychology, the cognitive neurosciences and cognitive epidemiology suggests that baseline cognitive function may also predict the development of adiposity (i.e., the “brain-as-predictor” perspective), although this reverse directionality has not been extensively explored to date using population-level datasets. Instead, relatively small-scale experimental studies have shown that temporary attenuation of some facets of cognitive function, particularly the executive control domain, could result in disinhibited eating. Therefore, it is plausible that impaired cognitive function affects the implementation of behaviors that confer downstream risk for adiposity. Taken together, these findings suggest that the association between adiposity and cognitive function could be reciprocal, but bidirectional effects have not been explored systematically in previous literature. This dissertation aimed to examine the hypothesized bidirectional associations between adiposity and cognitive function and their possible mediation paths using population-level datasets in three age groups: older adults, middle-aged adults, and adolescents.

Methods: Studies 1 and 2 were conducted using the Canadian Longitudinal Study on Aging (CLSA) datasets. Study 1 was a cross-sectional analysis of the baseline CLSA comprehensive cohort (N = 30,097), whereas Study 2 was a prospective analysis of the baseline and first follow-up datasets. The bidirectionality hypotheses were examined using three indicators of cognitive function (animal fluency, Stroop interference, and mean reaction time) and four indicators of adiposity (body mass index [BMI], total fat mass, waist circumference [WC] and waist-hip ratio [WHR]). Hierarchical multivariable regression, multivariate multivariable regression and cross-lagged panel model with latent variable modeling (CLPM-L) were employed to test the study hypotheses. Mediation analyses were conducted for lifestyle (e.g., diet, physical activity) and physical health status (e.g., hypertension, blood pressure and diabetes) variables.

Study 3 was a prospective analysis of the Adolescent Brain Cognitive Development (ABCD) dataset (N = 11,878). The above-mentioned bidirectional hypotheses were examined using two indicators of adiposity (e.g., BMI z score [zBMI] and WC) and five indicators of cognitive function included in the NIH Toolbox Cognitive Battery (e.g., Flanker, pattern recognition, picture sequence, picture vocabulary and oral reading tasks). Multivariate multivariable regression and CLPM-L were employed to test the study hypotheses. Mediation analyses were conducted for lifestyle (e.g., diet, physical activity) variables, physical health status (e.g., blood pressure) variables, and lateral prefrontal cortex (PFC) morphology features (volume and thickness).

Results: Study 1 showed that measures of cognitive functions were significantly associated with adiposity after controlling for confounders in cross-sectional analysis of the CLSA baseline datasets. In general, superior performance on animal fluency, Stroop, and reaction time tasks was associated with lower adiposity by most metrics. These associations were more substantial for moderate- and high-income sub-populations and mediated through lifestyle behavior (e.g., diet and physical activity) and physical health conditions (e.g., diabetes and diet).

Study 2 suggested that higher baseline adiposity was associated with higher Stroop interference at follow-up for both middle-aged and older adults. Similarly, higher baseline Stroop interference was associated with higher follow-up adiposity, but only in middle-aged adults. Effects involving semantic fluency and processing speed were less consistent. The above effects persisted following covariate adjustments and when used latent variable modeling of the adiposity variable. Significant mediation effects were observed for blood pressure, diabetes, and diet.

Study 3 revealed that higher baseline zBMI and WC were associated with worse follow-up picture sequence and better picture vocabulary task performance, respectively. Likewise, superior baseline performance on Flanker and picture sequence tasks was associated with better follow-up adiposity status. A bidirectional association was observed between episodic memory and zBMI. Latent adiposity modeling showed a bidirectional association with executive function (measured by Flanker task) but not with other cognitive domains. Significant mediation effects were observed for blood pressure, physical activity, and lateral PFC volume/thickness.

Conclusion: This dissertation examined the possibility of bidirectional associations between adiposity and cognitive function among older adults, middle-aged adults, and adolescents. Findings suggested that bidirectional associations between adiposity and cognitive function exist among adolescents and middle-aged individuals. In contrast, findings involving older adult population supported primarily a “brain-as-outcome” perspective on the association between adiposity and cognitive function.