Figure 1

A latent linear spline SEM to test modification of GRS associations with BMI trajectories by physical activity (PA). GRS was computed for each individual as the summation of risk alleles across 11 obesity variants and PA was ascertained at 36 and 53 years of age during interviews with a research nurse. Leisure time PA assessment was based on the Minnesota leisure time PA questionnaire at the age of 36 years and on more basic questions at the age of 53 years. At each age, participants who reported no leisure time PA were classified as ‘inactive’ and those who reported any relevant activity (in the previous month at the age of 36 years and in the previous 4 weeks at the age of 53 years) were classified as ‘active’. Thin arrows are used for the latent linear spline that summarises the serial BMI data for each individual as an intercept (i.e., kg m⁻² at the age of 36 years) and two slope terms (i.e., kg m⁻² per year between 36 and 53 and between 53 and 63 years of age). Solid thick arrows are used for main associations of GRS and PA with the intercept and slope terms. Dashed thick arrows are used for GRS-by-PA interaction associations with the intercept and slope terms. Each thick arrow is labelled with a letter that corresponds to an estimated parameter in Table 2.