Bidirectional Associations Between Adiposity and Cognitive Function and Mediation by Brain Morphology in the ABCD Study

This cohort study assesses bidirectional associations between adiposity and cognitive function in youth in the United States and tests mediational pathways through the lateral prefrontal cortext.

provided as two different measures. A sum of these measures was taken to create the MVPA variable.
Sleep duration. The sleep duration was also taken from the wrist-worn accelerometer device. Sleep periods were defined using the average of all minutes classified as being any kind of sleep (i.e., light + deep + REM) for all included days, based on (low) movement counts (14).
Diet. Dietary behaviors were assessed in the "ABCD Child Nutrition Assessment" module (14). Intake of whole grains, green leafy vegetables, other vegetables, berries, beans and nuts were coded as "healthy" food choices. To query the intake of healthy foods, parents of the participants were asked: "In a typical week, does your child eat (i) Whole grains 3 or more times per day, (ii) Green leafy vegetables 6 or more times per week, (iii) Other vegetables 1 or more time per day, (iv) Berries 2 or more times per week", (v) Beans 4 or more times per week, and (vi) Nuts 5 or more times per week? Responses for each of these probes were coded as 1 = Yes or 0 = No.
Intake of fast food and pastries were coded as "unhealthy" food choices. To query intake patterns of unhealthy foods, parents were asked: "In a typical week, does your child eat (i) fast food or fried food less than 1 time per week, (ii) Pastries or sweets less than 5 times per week? These variables were also coded as 1 = Yes and 0 = No (14).
Blood pressure (BP). Systolic and diastolic BP was measured three times at wave 3 data collection (14). The average of systolic BP and the average of diastolic BP were provided separately in the dataset as continuous variables.

eAppendix 4. Description of limitations because of the use of default atlas for MRI
Morphological features of the cortex were estimated using Freesurfer v5.3.0 (17). Freesurfer utilizes an automated, atlas-based, volumetric segmentation procedure for cortical surface reconstruction and subcortical segmentation. Images obtained from reconstruction were visually inspected, and only the images of sufficient quality were included in the study. Finally, morphological features were processed for the Deskian-Killiany atlas as part of the standard FreeSurfer pipeline (18). The use of the default atlas could have some influence on the mediation analyses; therefore, a stronger or weaker mediational effect could be observed using other atlases.

eAppendix 5. Description of limitations in cross-lagged panel model and of alternative models
With only two time points of data available during which cognition and adiposity were measured, the only type of dynamic, bidirectional model that can be identified is the cross-lagged panel model (CLPM) employed in the current study. With additional waves of data collection, more sophisticated models can be identified, including the random intercept-CLPM (19) and the dual change score model (20). In short, these models are capable of distinguishing between-from within-person effects, as well as modeling group-level change over time in the constructs of interest, both of which are not possible when applying the CLPM to two waves of data. For a more comprehensive discussion and comparison of these models, see Mund and Nestler (2019) (21). All that said, the CLPM is an important tool for the behavioural researcher that allows one to address questions such as: "Does a child with a high level of adiposity (relative to others) experience a subsequent lower development in cognition?" Furthermore, more sophisticated models often suffer from poor model convergence and/or only examine prospective withinperson effects (22        Path c: the association between baseline cognition and follow-up adiposity. Four decimal places were retained because of the smaller values of the indirect effects and to clarify the direction of coefficient and CI. Total indicates the sum of individual indirect effects for the respective mediators. All estimates are unstandardized coefficients. Significant indirect effects indicate that the association between adiposity and cognition is, in part, mediated through that respective variable.