Associations between the physical environment and obesity

The physical environment is considered a contributing factor to elevated body mass index (BMI) and odds of obesity in adulthood. However, associations within current literature are inconsistent in scale and direction. This thesis used individual-level data from The Yorkshire Health Study including self-reported BMI (wave one: 2010-12, n=27,806 and wave two: 2013-15, n=11,164). The physical environment was characterised using measures of the food (fast-food, large supermarkets, convenience and other food outlets) and physical activity (PA) environment (PA facilities, parks, green, and blue space) corresponding with the baseline individual-level data. Home addresses were geocoded to postcode zone centroids. A 2km radial buffer defined neighbourhood. Analyses used multi-level models, latent class analysis (LCA) and subgroup analyses by socioeconomic status. In most cross sectional (n=22,889, age 18-86 years), and especially longitudinal findings (n=8,864), the physical environment, BMI, and obesity were inconsistently related. While PA facilities were associated with reduced BMI in longitudinal and cross-sectional findings, effects were very small. An original measure of neighbourhood typologies, was associated with BMI and obesity in cross-sectional findings, yet was unrelated in longitudinal evidence. Acknowledging the risk of residual confounding, this thesis advances evidence by suggesting that the physical environment may be relevant for BMI and obesity, but only among certain population sub-groups, for instance, low socioeconomic status individuals. Compared to current evidence, this research provides an original and rigorous longitudinal perspective that utilises advanced analytical approaches in a large sample. An extensive measurement of the physical environment from several different data sources, allowed a unique synthesis of evidence. However, even given these strengths, associations between the physical environment, BMI, and obesity are convincingly shown to be small in scale, and inconsistent in direction across this thesis. This provides preliminary evidence for an opportunity to reconsider the current trajectory of both research and policy in this field.