Baseline cross-sectional data (2006–10) from UK Biobank were used (502 664 participants aged 40–69 years, of whom 264 341 commuted). Self-reported commuting mode was operationalised across seven mode categories, ordered to indicate typical physical activity. Outcomes were objectively measured body-mass index (BMI) and percentage body fat (BF%). Hypothesised confounders were identified: commute frequency, income, area deprivation, urban or rural commute, education, alcohol use, smoking, leisure physical activity, occupational physical activity, general health, and limiting illness or disability. Interaction effects were found for sex, so sex-stratified, nested, multivariate linear regression models were estimated. Final complete case sample sizes were 71 353 men, 82 365 women (for BMI analysis); and 70 627 men, 81 504 women (BF%).
Active commuting was significantly and independently predictive of lower BMI and BF% for both men and women, with a dose-response pattern apparent across the seven mode categories. In the fully adjusted models, mixed active and public transport commuters had significantly lower BMI than did car-driving commuters (men −1·0 kg/m2, 95% CI −1·13 to −0·84; women −0·7, −0·85 to −0·48). Cycling predicted lower BMI (men −1·72 kg/m2, 95% CI −1·87 to −1·57; women −1·66, −1·91 to −1·42). A similar pattern and magnitude of effect was observed for BF%. Mixed active and public transport commuters had significantly lower BF% than did car users (men −1·3%, 95% CI −1·50 to −1·10; women −1·09, −1·38 to −0·80). Walking predicted lower BF% (men −1·2%, 95% CI −1·54 to −0·91; women −1·2, −1·34 to −0·98), and cyclists had significantly lower BF% (men −2·3%, 95% CI −3·03 to −2·47; women −3·3, −3·81 to −2·73). Sensitivity analyses to investigate the effects of item missingness, interactions, distance and frequency, and effects of energy intake did not affect the results.
This study is the first to our knowledge to use UK Biobank data to address the topic of active commuting and obesity; we found robust, independent associations between active commuting and healthier body weight and composition. This research supports the promotion of active commuting as a population-level policy response. Further research should make use of longitudinal data to provide stronger causal inference.
Medical Research Council Strategic Skills Postdoctoral Fellowship in Population Health (awarded to EF)