Residential Density and Adiposity: Findings from the UK Biobank

Manuscript Number: THELANCET-D-17-04013R1 Background Obesity has emerged as a global pandemic, however the evidence for identifying the optimum residential density in relation to obesity has been far from compelling. High residential density may be hypothesized to constitute leptogenic multi-functional environments promoting active living. We examine the association between adiposity and housing unit density. Methods This cross-sectional study involved 450,433 adults from the UK Biobank aged 38-73 years with full data. Residential unit density was objectively assessed within one-kilometer street catchment of participants' residence. Other activity-influencing built environment included density of retail, public transport and street movement density modelled from network analyses of through-movement of street links within the defined catchment. Adiposity is expressed in-terms of measured body mass index (BMI; Kg/m²), waist circumference (WC; cm), whole body fat (WBF; Kg), and obesity as defined by WHO. We fitted linear and non-linear (restrictedcubic-spline) models after adjusting for activity-influencing built environment, neighbourhood deprivation, socio-demographics, lifestyle and co-morbidities and investigated effect modification by gender, age, and physical activity. Findings Restricted-cubic-spline model with three knots best fitted the data identifying two inflexion points at residential densities of 1600 and 3400 units/Km². Below a density of 1600 units/Km², increment of 1000 units/Km² was significantly associated with higher BMI (βBMI=0.24, 95% CI: 0.19 to 0.30), WC (βWC=0.55, 0.40 to 0.69), WBF (βWBF=0.57, 0.46 to 0.68) and odds of obesity (ORObesity=1.13, 1.09 to 1.13). Between 1600-3400 units/Km², it was associated with lower BMI (βBMI=-0.13, -0.18 to -0.08), WC (βWC=-0.19, -0.32 to -0.07), WBF (βWBF=-0.20, -0.30 to -0.10) and obesity (ORObesity=0.96, 0.94 to 0.99). Above 3400 units/Km², each increment of 1000 units/Km2 was leptogenic, being associated with lower BMI (βBMI=-0.15, -0.19 to -0.11), WC (βWC=-0.50, -0.60 to -0.40), WBF (βWBF=-0.26, -0.34 to -0.18) and obesity (ORObesity= 0.93, 0.91 to 0.95). Stronger leptogenic effects of housing density were observed among younger, female and participants doing higher physical activity. Interpretation High residential density is associated with lower adiposity in a large and diverse population sample. The evidence point to the value of housing-level policy related to densification as an upstream-level candidate for public health intervention against adiposity. Further longitudinal evidence are needed to establish causality