Predicting sugar-sweetened beverage intake from the brain and known risk factors in adolescents.

BACKGROUND

Low socio-economic status, male sex, and body mass index (BMI) are known risk factors for high sugar sweetened beverage (SSB) consumption in adolescents. The present analysis aimed to predict SSB intake based on known risk factors and resting-state functional magnetic resonance (rsfMRI) connectivity from the Adolescent Brain Cognitive Development study.

METHODS

Using the year-2 follow up visit Block Kids Food Screener data, participants were categorized as low SSB consumers (<8 floz/day) or high SSB consumers (>16 floz/day). The high and low groups were matched on baseline age, BMI percentile (BMI%) and combined household income (CHI; n/SSBgroup = 841). We used a grid search linear support vector classifier (SVC) to select baseline features from BMI%, sex, age, CHI, and resting state functional connectivity associated with SSB intake. With the selected features we used a binary logistic model to predict high SSB consumption at year-2.

RESULTS

The SVC identified sex and 58 functional connections as relevant features and predicted SSB intake at year-2 with 57 % accuracy. Logistic regression revealed visual-right caudate connectivity (predicted probability [PP]: 0.71, pFDR = 0.02) and sex (PPmale: 0.56, PPfemale: 0.72; pFDR < 0.0001) as significant predictors. Post hoc analysis revealed head motion quality control outcomes were independently associated with low CHI (OR = 1.102, p < 0.0001), higher BMI% (OR = 0.93, p < 0.0001), and high SSB intake (OR = 0.842, p = 0.008).

CONCLUSION

The average correlation between visual cortex and right caudate is significantly related to high SSB consumption in adolescents. Participants from low CHI families are at higher risk of exclusion due to excessive motion.