Background The association of social relationships and health may be mediated by adiposity. However, the association of social relationships with body mass index (BMI) and waist-to-hip ratio (WHR) is poorly examined. Close social relationships influence BMI and WHR levels at different points in the life course, yet have not been studied in relation to BMI and WHR trajectories over time. Here using the Whitehall II cohort of middle-aged British adults (n = 5,773) we examine whether marital status and perceptions of positive and negative aspects of support from the closest person are associated with BMI and WHR trajectories and modified by gender.

Methods This analysis examines baseline social support and covariate data from phase 2 of the Whitehall II study (1989–1990, ages 37–60) and trajectories of BMI and WHR over phases 3 (1991–1994), 5 (1997–1999), 7 (2002–2004), 9 (2007–2009) and 11 (2012–2013, ages 59–83). Person-level trajectories of BMI and WHR are estimated using a multilevel growth model with measurement occasion nested within participant. Basic models include: sex + baseline age + time + time2 + baseline support + support × time + support × time2 (in BMI models only); these models are tested for an interaction with gender and adjusted for ethnicity, socioeconomic status, health behaviours and General Health Questionnaire.

Results Using data from over 20 years of follow up, we demonstrate that high confiding support compared to low, low negative aspects compared to high and being married compared to being single are associated with less steep linear increase in BMI and WHR, independent of health behaviours and other covariates (e.g. linear BMI slope coefficient: married 0.219, SE 0.005 vs single 0.236, SE 0.006, p < 0.002). Gender modifies the associations, with greater magnitudes in women (e.g. linear BMI slope coefficient high confiding support vs. low: men 0.192, SE 0.009 vs. 0.221, SE 0.008, p < 0.05 and women 0.211, SE 0.019 vs. 0.277, SE 0.017, p < 0.05).

Conclusion This study extends previous studies by demonstrating that BMI and WHR trajectories by social support and marital status are different in men and women during mid-life independently of covariates. The Whitehall II study is not representative of the general population, yet if replicated in other cohorts, these results could inform designing intervention studies aiming to facilitate maintaining healthy BMI and WHR levels by improving support networks of middle-aged people.