The moderating role of eating behaviour traits in the association between exposure to hot food takeaway outlets and body fatness

The study addresses how individual eating behaviour traits may modify the relationship between exposure to hot food takeaway outlets in residential environments and dietary/adiposity outcomes. In an obesogenic environment where access to energy-dense foods is easy and takeaway outlets have proliferated, prior evidence on associations between outlet exposure and obesity-related outcomes is mixed. Psychological traits such as cognitive restraint, uncontrolled eating, and emotional eating may influence susceptibility to food cues and help explain heterogeneity in responses to the food environment. The study aims to test whether these eating behaviour traits moderate the associations between neighbourhood takeaway outlet exposure, takeaway food consumption, and adiposity.

The obesogenic environment shapes diet and physical activity, contributing to obesity prevalence. Takeaway outlet density has increased over time, with disparities by area deprivation. Evidence linking residential exposure to takeaway/fast-food outlets with diet and adiposity is mixed, though some UK studies report positive associations. Prior work in Cambridgeshire associated greater exposure with more takeaway consumption and higher BMI/obesity odds. Eating behaviour traits measured by TFEQ (cognitive restraint, uncontrolled eating, emotional eating) influence diet and weight: higher uncontrolled and emotional eating associate with poorer diet quality and higher weight, while findings for cognitive restraint are mixed. Limited studies have examined psychological moderation of food environment effects; one found greater fast-food consumption with higher external eating in children, and another reported stronger associations among adults with higher reward sensitivity.

Design and sample: Cross-sectional analysis of the Fenland Study (Cambridgeshire, UK), recruiting adults born 1950–1975 between 2005–2015. Of 46,024 invited, 12,435 participated; a sub-sample with eating behaviour data (n=4,791; TFEQ-R18 collected 2011–2015) was analyzed. Ethics approval and informed consent obtained. Exposure assessment: Number of hot food takeaway outlets within a 1-mile Euclidean radius of participants’ residential postcode, using local council data (December 2011) mapped in ArcGIS. Takeaway outlets defined as selling hot food primarily for off-premises consumption, ordered/paid at counter, minimal/no table service; included chains and independents. Exposure was operationalized in quartiles due to non-linear associations and to mitigate temporal misclassification (Q1=0 outlets; Q2=1–2; Q3=3–12; Q4=13–51). Outcomes: Primary outcomes were (1) intake of takeaway-like foods (g/day) from FFQ—sum of pizza, burgers, fried fish, and French fries (marker of takeaway consumption); and (2) adiposity by DEXA-derived total body fat percentage (measured at clinical visit after overnight fast; standard protocols). Secondary adiposity measures included Fat Mass Index (FMI=kg fat/m²) and BMI (kg/m²). Moderators: Eating behaviour traits from TFEQ-R18: cognitive restraint (6 items; α=0.75), uncontrolled eating (9 items; α=0.85), emotional eating (3 items; α=0.87). Raw scores transformed to 0–100 scale; higher scores indicate greater respective trait. Covariates: Age, sex, household income (<£20k; £20–39,999; ≥£40k), age at completion of full-time education, occupational social class (professional/intermediate/working class), and count of supermarkets within 1-mile of home. Data handling and statistics: Outliers were winsorized to mean ±3 SD. Participants lacking DEXA (n=68) were excluded. Missingness ranged up to 2.2% for income; multiple imputation by chained equations (20 datasets; seed 1234) was used for covariates and body fat percentage, assuming data missing at random. Associations between takeaway outlet exposure quartiles and outcomes, and between eating traits and outcomes, were estimated using multiple linear regression adjusted for covariates. Effect modification tested by including interaction terms (trait × exposure quartile); traits were mean-centered. Interaction significance assessed with F-tests pooled across imputations using the Median P Rule. Interaction alpha set at 0.10 due to lower power; other tests at 0.05. Conditional effects probed at trait levels of mean ±1 SD. Analyses performed in STATA v16.

Sample characteristics: n=4,791; mean age 51.0 (SD 7.2) years; 53.9% female. Median takeaway consumption 30.0 g/day (IQR 17.5–47.5); mean body fat percentage 33.3% (SD 9.2). Adjusted associations (per Table 2; reference Q1=0 outlets): Takeaway consumption (g/day): Q2 β=0.6 (95% CI -1.4, 2.6); Q3 β=2.6 (0.5, 4.8); Q4 β=4.0 (0.3, 7.6). Body fat percentage: Q2 β=0.3 (-0.3, 0.9); Q3 β=1.8 (1.2, 2.5); Q4 β=1.8 (0.7, 2.9). Eating behaviour traits (per 10-point increase): Cognitive restraint: takeaway consumption β=-2.2 (-2.6, -1.8); body fat % β=0.2 (0.1, 0.3). Emotional eating: takeaway consumption β=0.6 (0.4, 0.9); body fat % β=0.9 (0.8, 1.0). Uncontrolled eating: takeaway consumption β=1.3 (0.9, 1.7); body fat % β=1.1 (0.9, 1.2). Interaction (moderation): Uncontrolled eating did not moderate associations with either outcome. Cognitive restraint moderated the association between exposure and takeaway consumption (F-test p≈0.03): the positive exposure–consumption association was stronger among those with higher cognitive restraint; differences by restraint attenuated at highest exposure quartile. Emotional eating moderated the exposure–body fat percentage association (F-test p≈0.05): the positive association was stronger among low emotional eaters; e.g., among low emotional eaters, Q4 vs Q1 body fat difference was 2.8% (95% CI 1.6, 4.0), versus 1.5% (0.3, 2.7) among high emotional eaters. Effect sizes for interactions were small. Secondary outcomes (FMI, BMI) showed similar patterns to body fat percentage; complete-case analyses were consistent.

The study demonstrates that both environmental exposure to takeaway outlets and individual eating behaviour traits are associated with higher takeaway-like food consumption and greater adiposity. This supports the notion that obesogenic environments contribute to unhealthy dietary patterns, while psychological predispositions further influence outcomes. Moderation analyses provided limited evidence that individual traits alter susceptibility to the food environment: cognitive restraint slightly strengthened the association between outlet exposure and takeaway consumption, potentially reflecting that individuals attempting to restrict intake face greater difficulty in high-exposure areas. Emotional eating slightly modified the exposure–adiposity association, with stronger environmental effects among low emotional eaters, suggesting that high emotional eaters may seek takeaway foods irrespective of local density. These findings align with prior work linking emotional and uncontrolled eating to poorer diet quality and higher weight, and they suggest bi-directionality for cognitive restraint (higher adiposity may increase restraint efforts). Overall, moderation effects were small and inconsistent across outcomes, indicating that food outlet exposure explains only a modest fraction of adiposity variance. The results underscore the need for combined strategies targeting both individual behaviours and environmental factors to reduce unhealthy diet and obesity risk.

Eating behaviour traits (emotional and uncontrolled eating, cognitive restraint) and residential exposure to takeaway outlets were each associated with greater takeaway consumption and higher adiposity. Evidence that certain traits confer greater susceptibility to takeaway outlet exposure was weak: no moderation by uncontrolled eating; slightly stronger exposure–consumption association among those with higher cognitive restraint; and slightly stronger exposure–adiposity association among low emotional eaters. Integrated interventions addressing both individual-level behaviours and the food environment are warranted. Future research should include longitudinal designs in diverse populations, employ improved or complementary measures of eating traits (e.g., TFEQ-R21, ecological momentary assessment), examine additional psychological moderators (e.g., external eating, food responsiveness), and consider broader dietary outcomes.

Cross-sectional design limits causal inference. Temporal mismatch between exposure (food outlet data from 2011) and participant measurements (2005–2015) may cause exposure misclassification and unknown exposure duration. Generalizability may be limited: sample from Cambridgeshire, relatively older adults, likely less ethnically diverse. Measurement considerations include potential floor/ceiling effects for emotional eating (3-item subscale), possible under-capture of some takeaway foods (e.g., fried chicken, some Asian dishes) and potential misclassification of supermarket-prepared foods as takeaway-like items. Despite these, associations were consistent across outcomes. Some participants lacked DEXA at one site (excluded n=68). Although multiple imputation addressed missing data (up to 2.2% for income; 0.8% for body fat), assumptions of missing at random may not fully hold. Interaction tests had lower power; observed moderating effects were small and may be due to chance given multiple tests.