Using a comprehensive model, simultaneously incorporating child and family variables in a prospective design, we examined associations between a number of risk factors experienced in childhood/adolescence and BMI in early adulthood, adjusting for respondents’ age, sex, education, income and health. This study provides several important findings at the family level: 1) socioeconomic adversity (measured by receipt of social assistance) was related to increased BMI, whereas parental education was associated with lower BMI in early adulthood; 2) parental immigrant status was associated with lower BMI; 3) family functioning was negatively related to BMI (higher family function was associated with lower BMIs) and 4) parental mental health problems were associated with increased BMI. At the child level, presence of child mental disorders and poor school performance were both related to higher BMI, even after controlling for current education and mental health status. Altogether, the predictor variables explained about 10% of respondent variability in BMI at young adulthood. In the fully adjusted model, the effects of variables such as immigrant status and social assistance converted to standard deviation units exhibited what would be considered small to medium effects (d = 0.25 and 0.40) based on Cohen’s criteria .
In this study, we found that 39.05% of the variation in BMI was associated with between-family differences. The familial aggregation of BMI reflects a multifaceted interplay between genetic susceptibility to weight gain and shared environmental influences within families. Some of these shared influences are linked to measured family risk factors, such as socioeconomic disadvantage, parental education and family functioning. Recent studies have found that genetics play an increasingly important role in explaining variation in BMI over time, with its greatest influence in late adolescence; whereas environmental influences decrease over time, exerting their strongest effects early in childhood and adolescence [54–56].
Our finding that family status factors (parental education and receipt of social assistance) are associated with BMI is consistent with previous research [5–8, 10]. In children and adolescents, lower SES, regardless of how it is measured (parental education, occupation, or income), is associated with increased BMI and obesity [57–59]. Economic disadvantage in families may be directly related to a number of factors that pose a risk for higher BMI, such as engaging in less physical activity, poorer nutrition and eating habits, and lack of participation in organized sports [60, 61]. In addition, early in childhood, healthy lifestyle trajectories may be set via modeling by caregivers . We also found that parental immigrant status was negatively related to BMI in adulthood. This finding is in agreement with a recent report indicating that first generation immigrants have lower BMI compared to second generation or Canadian born children . Interestingly, children from that study lived in a multi-ethnic, disadvantaged inner city community, where many immigrants initially settle. These areas are typically characterized by lower levels of education and income. Despite these socioeconomic risks, immigrant status still conferred a protective influence in this sample; however, this protective health advantage may be lost over time with exposure to unhealthy lifestyle habits in host communities.
Family functioning was associated with lower BMI in early adulthood. Our measure of family functioning was comprised of questions on problem solving, affective responsiveness and involvement and behavioral control within the family. These questions tap into parenting practices. Research investigating the impact of parenting practices on BMI in children and adolescents has found that authoritative homes, characterized by a family context of warmth, high emotional support, encouragement, monitoring and bidirectional communication, is related to healthier eating habits, increased physical activity and lower BMI [63–67]. Interestingly, our distal family status factors (parental education and receipt of social assistance) remained significant even after adding family functioning to the model. This implies that family functioning does not act as a mediator between family status and BMI but exerts its own independent effects. Pathways relating family functioning to BMI are likely complex and may impact weight through its direct influence on diet and physical activity  or through more indirect mechanisms such as child self-regulation capabilities .
Parental mental health problems predicted elevated BMI, although this association became non-significant when childhood variables were added to the model. Consistent with previous findings [15, 19, 27], we found that childhood psychiatric disorder and school difficulties were associated with greater BMI in early adulthood, even after controlling for current mental health status and years of education. Although there is some debate regarding the magnitude of the effect, starting at an early age  mental health problems are established predictors of elevated BMI through the lifespan [16–18]. Given the established nature of this association, there is likely an underlying mechanism shared by both mental health and obesity. Potential candidates include genetic, behavioral, and/or psychological factors which are common to both phenotypes [27, 69]. Contrary to previous research [21, 22], we did not find an association between history of childhood physical or sexual abuse and adult BMI. This may be due to the fact that we have such a comprehensive array of risk factors measured. Also contrary to previous research [12, 13], we did not find an association between low birth weight and adult BMI. This may be due to reporting inaccuracies: birth weights were based on self-reports by mothers several years after the birth of their child. The prevalence of low birth weight in this sample (2.5%) is lower than prevalence rates reported for Ontario (4.8%), suggesting that low birth weight may be under represented in our sample.
We found that gender moderated the effect of two risk factors on BMI: receipt of social assistance and presence of a medical condition in childhood. In females, but not in males, the presence of these risk factors was associated with higher BMI in early adulthood. Our results support other findings that childhood socioeconomic disadvantage is associated with later obesity in women [70–72]. The association between BMI and later chronic disease in adulthood is fairly well characterized; however research linking medical problems in childhood to later BMI is relatively scarce. One recent study found that childhood leukemia was associated with increased BMI in adulthood but only in females . Presence of medical conditions or chronic illnesses in childhood may place greater limitations on physical activity, leading to a more sedentary lifestyle. In addition, it is noteworthy that within the OCHS, the prevalence of a chronic illness or medical condition in 1983 was not evenly distributed across socioeconomic groups. Children from low income households, defined as below the Statistics Canada poverty line, had higher rates of medical conditions compared to children of families above the poverty line . It is possible that these gender differences are linked to variations in underlying biological mechanisms, such as HPA axis function given that there are sex differences in stress system structure and function  and that the HPA is linked to BMI [75, 76], various medical conditions  and disparities in SES . Theoretically, it is argued that females have a heightened predisposition to stress-related disease with exaggerated sensitivity to stress which pushes females over the “disease threshold” . If this is true, medical conditions or socioeconomic disadvantage experienced in childhood may lead to biological sensitivities to stress, especially in females, putting them at greater risk of elevated BMI later in life.
Despite strengths of the OCHS in assessing childhood risks associated with BMI in early adulthood, this study has limitations. First, approximately 30% of 1983 participants were lost over the 18-year follow-up. Because this loss was selective to socioeconomic disadvantage, Boyle and colleagues  devised attrition weights that integrated original sample selection probabilities from 1983. We believe that any potential systematic bias is likely to be small and would be more related to underestimation of the influence of risk factors on outcomes. Second, there are a few measurement limitations. All risk factors were measured at one point in time, and we are not able disentangle their temporal associations or to assess intervening variables between 1983 and 2001. Our assessment of risk factors was collected prospectively over twenty years ago. Despite this lapse of time, we believe our findings are currently applicable to young adults who are likely to experience the same risk factors. We are unaware of any changes that would mitigate the association between childhood risk factors and BMI, as found in this study. Two, we do not have measures of child or parent BMI in 1983. It is well established in studies tracking weight status in children that those with higher BMIs early on tend to maintain these trajectories throughout adolescence and into adulthood, indicating some stability for most individuals [11, 80, 81]. Our measure of BMI was also based on self-report. Self-reported BMI yields lower rates of obesity and overweight [35–37]. Nevertheless, self-reported BMI remains an important tool for health surveillance ; commonly used because it is a simple, economical, and non-invasive method of collecting data from large samples . Moreover, self-reported BMIs are related to morbidity and mortality [83–85]. Applying a BMI correction factor did not alter our findings. Three, our abuse risk indicators were measured retrospectively. In general population studies, such assessments are very difficult to obtain from children prospectively and there is no reason to believe that current BMI would influence the recall of these experiences. Fourth, in exploring interactions between gender and childhood risk factors, we are vulnerable to obtaining a significant effect because of multiple testing. We chose to explore these interactions because of the lack of consistent research evidence on this question and the recognition that differential weight and body fat composition are integral to gender differences.