This study showed that growing up in families with parent’s having a low level of education in early or late childhood increased the risk of overweight and obesity at age 18 and 21 in both genders, where especially father’s low level of education appeared to be a quite strong risk factor in both genders, despite somewhat wide confidence intervals. We also found that girls, who reported poor “family functioning” in early or late childhood had increased risk of overweight and obesity at age 18 and 21, which was not seen among boys. Among boys, results showed that growing up in families with parent’s having low LMP during early childhood increased their risk of overweight and obesity at age 18 and 21.
To our knowledge, this is the first study to examine how childhood SEP relates to overweight and obesity, using a longitudinal study-design with 14 years of register-based exposure information. Furthermore, this was supplemented with the social factor “family functioning” to facilitate the subjective perception of childhood social conditions on the family level, which is not captured by the objective SEP measures.
Our results were in line with findings from the study by Kestila et al., who examined the association between childhood social circumstances and overweight and obesity in early adulthood in a cross-sectional design. The authors found a strong inverse association between parental educational level and obesity in both genders . These results are also supported by Mathiessen et al., who found that educational level of the parents was inversely associated with their off-spring being overweight . Morgen et al. found that 14–16-year-old girls of lower parental SEP had more than four times the risk of developing overweight/obesity at age 21, compared to girls of higher parental SEP . In our study, we found that parental lower educational level in early or late childhood may increase the risk of overweight and obesity at age 18 and 21 up between 1.8 and 3-fold among both genders.
Al-Emranie et al. examined the association between five-year weight gain among adults and SEP in childhood and adulthood. They found a significant association between childhood SEP and obesity among males aged 29–39, thereby suggesting that the socioeconomic gradient is even more prominent in relation to obesity . Results from our study showed that parental low LMP in early childhood was associated with increased risk of overweight and obesity in primarily boys, with a more than 2-fold increased risk of obesity at the age of 18 and 21.
Bann et al. examined how childhood and adult SEP relates to BMI across adulthood in three national British birth cohorts. They found that father’s occupational class at age 10/11 was associated with higher adult BMI in both genders . These findings are partly supported by results from our study concerning girls showing that low parental LMP in late childhood was associated with increased risk of overweight and obesity at age 18, although the findings were inaccurate. Among boys our results indicated that parental LMP in late childhood may be less important for boys’ risk of later overweight and obesity.
Overall, our findings indicate that childhood low SEP at the family level is associated with increased risk of overweight and obesity in adolescence and early adulthood. As mentioned in the background, a recent theory suggests that obesity may be a healthy active response to a future lack of energy caused by the sense of e.g. social insecurity in the family . This could be a plausible explanation for a possible pathway between low childhood SEP and the development of obesity in a well-fare society with easy accessibility to rich calories-dense food. We did not find strong associations between low household income in childhood and later overweight/obesity, which may be due to this population living in a well-fare society, where a family may have a reasonable living despite a rather low income. However, we saw a tendency towards an increased risk of obesity at age 18 and 21 among girls, who grew up in low income families in late childhood, when the associations were adjusted for the early childhood income indicating that the timing of this exposure may be relevant among girls, but not boys.
Parental lower educational level(s) during early and late childhood were quite consistent risk factors for overweight and obesity in both genders in this youth cohort. Parental low LMP in early childhood was primarily a risk factor for boys, and for girls there was a tendency in late childhood to influence girls’ future risk of overweight and obesity. Parental lower educational level and parental low LMP may negatively affect the psychosocial security experienced in families due to e.g. job insecurity, living in poorer residential area and perhaps also an unhealthy life style, which may affect the children. Due to role modeling, children reflect themselves in their parents, so when boys experience their father having low LMP during early childhood, this may increase boys’ feeling of perceived social insecurity in daily life, which may be translated into psychological processes with possible future biological consequences . Lower educated parent’s and parent’s with low LMP are perhaps also more likely to pass on poorer eating habits to the children , which combined with increasing sedentary behavior and risk behavior may tract into adolescence and adulthood and thereby also contribute to an enhanced risk of overweight and obesity.
A recent review conducted primarily on cross sectional studies concluded, that poor “family functioning” was associated with increased risk of overweight and obesity among children and adolescent aged 3–17 . We observed gender differences in our study, where reporting poor “family functioning” at age 15 was a risk factor for overweight and obesity in adolescence and early adulthood in girls, but not among boys. Perhaps weight-gaining in boys during adolescence and early adulthood are less affected by how the nearest family function, compared to girls due to e.g. different coping strategies or life styles .
This cohort study had several strengths. The initial study response rate was 78%, which somewhat declined at the latter rounds. The study covered up to 21 years of follow up and used register-based information to define most of the exposure variables, which resulted in few missing values. The exposure variable “family functioning” was applied to uncover the child’s experiences of the social conditions in the family during childhood. By adding the subjective perspective in terms of this social factor, we emphasize the importance of this influence on the physical health of young people.
The prospective design is suitable to observe potential changes over time and we consider this to be an appropriate way of studying this kind of associations. Also, applying register-based exposure variables diminishes the risk of differential information bias on these variables.
Participating in surveys may be prone to selection bias; that is if non-participation is associated with both exposures and outcomes. In this cohort, we found non-participants to be significantly different from participants with respect to the exposure variables, however, we do not have any information on height and weight from the non-participants, so it is not possible to disentangle whether any selection was differential. Non-participation and drop-outs in the same cohort was examined in a previous study by Winding et al. and results showed that neither non-participation nor drop-outs influenced significantly on the size of the measured associations .
The main limitation of the study was that the outcome was based on self-reported height and weight and consequently prone to misclassification. Participants, who are overweight, are probably more likely to underestimate their weight , which may be most pronounced in girls . This increases the risk of underestimating the associations between the exposures and the outcome and hence bias towards the null-hypothesis. We believe that due to the study design the risk of differential misclassification of the outcome was small. We applied the self-reported variable “family functioning” along with the outcome reported from the baseline questionnaire at age 15 and we are aware that these findings are cross-sectional and cannot tell us anything about causality.
We decided to imply the additional cut-offs for obesity from the Global database on BMI due to a relatively low prevalence of obese participants in this cohort according to conventional World Health Organization-guidelines. We believe that applying the additional cut-off seems reasonable in this young healthy population.
All the associations in the study were mutually adjusted for the other exposure variables, but these adjustments did not alter the results much. We did not find strong correlations between e.g. household income and highest educational level in this study. This may be explained by the fact that household income in Denmark not necessarily reflects a person’s level of education. An unskilled worker in a factory often earns a rather high salary compared to e.g. health care workers with a short or medium long education. For the early childhood adjustment’s we applied split home 1989, however we repeated the analyses with split home 1991 instead, because the first couple of years after the birth of a child may be a difficult time for the parents’ relationship and one could suspect that more families may split up during these years. Applying split home 1991 did not change the estimates.
A previous examination of the study setting concluded that the participants of this youth cohort are comparable to young people in other parts of Denmark . Therefore, the results of this study may be transferred to young people with similar environmental and social conditions to this Danish cohort, when taking the above-mentioned limitations into account.