Mental health in children and adolescents with overweight or obesity

The data for the study were collected between 2011 and 2020 within the LIFE Child study conducted in Leipzig. The LIFE Child study is a cohort study examining the healthy development of children and adolescents and the development of civilization diseases. Participants are examined every year from the prenatal period to young adulthood. From the age of 6 years, LIFE Child recruits children with obesity in a specific cohort (obesity cohort), leading to a higher proportion of obesity from this age on. The children and adolescents complete an elaborate study program including the collection of bio samples, body function tests, the assessment of personal characteristics, and many other examinations. For this purpose, a broad-based team of physicians, psychologists, nutritionists, and sports scientists is involved [28, 29]. The study was designed according to the criteria of the Declaration of Helsinki and approved by the Ethics Committee of the Medical Faculty of the University of Leipzig (Reg. No. 264/10-ek). All families participate voluntarily, and parents provide written informed consent before the inclusion of their children. Additionally, adolescents aged 12 years and older provide their own written informed consent.

In the present project, children and adolescents aged 4 to 18 years were assessed. The original dataset comprised 11631 data points. Data of children and adolescents for whom information on BMI (body mass index), sex, age, or socioeconomic status (SES) was missing were excluded (n=785). In the case of multiple visits of the same participant, we excluded all but the last study visits (n=7566). Furthermore, we restricted our analysis to the youngest child in the family, i.e., we excluded older siblings (n=930). This strategy prevented biological dependencies and led to a homogeneous age distribution within the sample. The final sample comprised 2350 children and adolescents (1150 girls and 1200 boys, mean age 10.87 years, age range 4.00 - <19). Based on child/adolescent age and the person who completed the questionnaires, we divided the total sample into two subsamples, a child sample (aged 4 to 10 years, questionnaires completed by parents, n=1213 (591 girls), mean age 7.32 years (SD=2.06)) and an adolescent sample (aged 11 to 18 years, questionnaires completed by adolescents themselves, n=1137 (559 girls), mean age 14.66 (SD=2.17)). However, the specific analyses were only performed in those participants who had completed the respective questionnaires.

The total sample included 2350 participants (1150 (48.9%) girls, mean age=10.87 years, SD=4.24) (Table 1). On average, participants had a BMI-SDS of 0.34 (SD=1.25). Most children and adolescents were categorized as "normal weight" (n=1660, 70.6%), whereas 352 (15.0%) were categorized as “obese”, 169 (7.2%) as “overweight”, and 169 (7.2%) as “underweight”. Regarding SES, most children and adolescents (n=1335, 56.8%) had middle SES, while 730 (31.1%) had high SES, and 285 (12.1%) had low SES. Table 2 shows the average scores of the single scales of the questionnaires stratified by weight status. With a few exceptions (quality of life regarding school environment, somatoform complaints, prosocial behavior, and hyperactivity/inattention (only in adolescent sample)), all differences between children/adolescents with normal or underweight and children/adolescents with overweight or obesity were statistically significant (p < .05).

In the total sample, children and adolescents with low SES had a significantly higher BMI-SDS than those with high SES (b=0.95, p<0.001), or those with middle SES (b=0.65, p<0.001). In children with a high SES, the average BMI-SDS was estimated 0.05 (SD=1.04), compared to 0.35 (SD=1.25) in those with middle SES, and 1.00 (SD=1.48) in those with a low SES.

We observed no significant association between somatoform complaints in GBB-KJ and BMI-SDS, neither in the child sample (beta=0.20, p=0.14, see table 4), nor in the adolescent sample (beta=0.19, p=0.35, see table 3). However, in the child sample, a significant interaction with age (p=0.002) indicated that the association changed from negative at kindergarten age (beta=-0.41, p=0.09) into positive at primary school age (beta=0.47, p=0.003). We did not find significant interactions with sex or SES.

Table 3 (adolescent sample) and Table 4 (child sample) show the coefficients for the associations between the BMI-SDS and the different SDQ scales.

In the child sample (4–10 years), there was a significant positive association between BMI-SDS and parent-reported conduct problems (beta=0.15, p<0.001). Moreover, we found a significant positive association between BMI-SDS and parent-reported peer relationship problems (beta=0.24, p<0.001). Here, a significant interaction with age (p<0.001, see Figure 1) indicated that the positive relation existed only at primary school age (beta=0.33, p<0.001), not at kindergarten age (beta=-0.01, p=0.92). The scores on the SDQ scales “prosocial behavior”, “hyperactivity/inattention”, and “emotional symptoms” were not significantly associated with BMI-SDS in the child sample. However, regarding the association between BMI-SDS and prosocial behavior (ps) and emotional symptoms (es), respectively, significant interactions with age (p_ps=0.009 and p_es=0.006) revealed a change from negative associations at kindergarten age (beta_ps=-0.16 (p=0.049) and beta_es=-0.14 (0.08) into positive associations at primary school age (beta_ps=0.09 (p=0.08) and beta_es =0.15 (p=0.007). Furthermore, the associations between BMI-SDS and hyperactivity/inattention (hi) and emotional symptoms, respectively, revealed significant interactions with SES (p_hi=0.02 and p_es=0.02), showing changes in the associations from negative at low SES (beta_hi=-0.24 (p=0.10) and beta_es=-0.14 (p=0.21) into positive at middle SES (beta_hi=0.14 (p=0.08) and beta_es=0.16, (p=0.01)). The analyses revealed no interactions with sex.

In the adolescent sample, we found a significant positive association between BMI-SDS and scores on the SDQ scale emotional symptoms (beta=0.10, p=0.04). A significant interaction with age (p=0.006) revealed that the positive relation at early pubertal age (beta=0.24, p<0.001) was no longer present at late pubertal age (beta=-0.01, p=0.85). As in the child sample, we observed a significant positive association between BMI-SDS and the score on the conduct problems scale (beta=0.14, p<0.001). A significant interaction with age (p=0.007, see Figure 1) showed that this association was weaker at late (beta=0.07, p=0.11) than at early puberty age (beta=0.25, p<0.001). Comparable with the child sample, there was a significant positive association between BMI-SDS and scores on the peer relationship problems scale (beta=0.27, p<0.001). Sex and SES were not found to moderate the associations between BMI-SDS and behavioral difficulties in the adolescent sample.

In the present sample, 7.2% of children were overweight (but not obese), and 15.0% were obese. The large proportion of children and adolescents with obesity can be explained by the fact that there is an obesity cohort in the LIFE Child study for which children and adolescents with obesity are explicitly recruited [28].

As hypothesized, the analyses revealed a negative association between BMI-SDS and SES, consistent with results of previous studies conducted in high-income countries [6, 9]. A study by Gibbs et al. identified unhealthier feeding practices as the most important explanation for the association between SES and early childhood obesity [37]. Healthy foods are often more expensive and therefore less likely to be purchased by families with low SES/low income [38].

Our analyses showed lower HRQoL in adolescents with higher BMI-SDS. However, only the associations between BMI-SDS and physical and psychological well-being were significant. Regarding the other domains of quality of life, the associations with BMI-SDS pointed in the expected direction but did not reach statistical significance.

The finding that physical well-being was lower in adolescents with higher BMI-SDS is in line with previous findings [12, 13, 15]. An explanation might be reduced physical activity and fitness in adolescents with higher BMI [39]. A study by Mozzillo et al. was able to show that lower HRQoL was associated with less physical activity among adolescents with overweight or obesity [40].

Regarding psychological well-being, previous studies showed lower levels in children and adolescents with obesity compared to children and adolescents with normal weight [10, 15] but no differences between adolescents with overweight (including obesity) and normal weight [13]. The present finding, however, suggests a linear relationship between BMI-SDS and psychological well-being. One explanation is that adolescents with higher BMI suffer more frequently from social withdrawal, social isolation or bullying [41‐43]. Furthermore, adolescents with higher BMI might be exposed to stigmatization by parents and teachers [44, 45]. These factors are likely to have a negative effect on the psychological well-being of the affected person.

In line with another study that found greater HRQoL reductions of higher weight status relative to normal weight in younger than in older adolescents [14], associations between BMI-SDS and physical as well as psychological well-being were stronger at early than in later puberty age. Regarding autonomy and parents, peers and social support, and school environment, we observed the same trend. These findings indicate that associations between BMI-SDS and HRQoL are particularly strong in young adolescents, i.e., during early puberty. In that specific age group, adolescents might be particularly sensitive to appearances and teasing, e.g., due to limited social functioning. In line, Riazi et al. showed lower social functioning in pre-pubescents with obesity than in pubescents and post-pubescents with obesity [46].

In contrast to age, sex or SES did not moderate associations between BMI-SDS and HRQoL. Some previous studies suggest that the overweight/obesity-related HRQoL impairment is higher in girls than in boys [13, 14]. The fact that we did not find a difference could be due to the fact that our sample was smaller compared to the two studies mentioned above.

While previous studies pointed to a significant association between BMI and somatoform complaints [17, 18, 20‐22, 47], here, associations pointed in the expected direction but did not reach significance. One reason could be the smaller sample size (compared to the samples on associations with HRQoL and behavioral strengths and difficulties). Furthermore, our sample included healthy children and adolescents that rarely reported somatoform complaints. Interestingly, in the child sample, a significant interaction revealed that the association between BMI-SDS and somatoform complaints becomes apparent only from the primary school age onwards. Similar to the association between BMI-SDS and HRQoL, associations between BMI-SDS and somatoform complaints might be more relevant in late pre-puberty/early puberty.

As expected, children and adolescents with higher BMI-SDS showed more behavioral difficulties than those with lower BMI-SDS, especially more internalizing problems (emotional symptoms (the latter only in the adolescent sample) and peer relationship problems) and more conduct problems.

As already highlighted above, the associations between higher BMI-SDS and internalizing problems might be explained by more social withdrawal, social isolation, bullying, and exposed stigmatization in children and adolescents with higher BMI-SDS [41‐45].

The finding of significantly more conduct problems in children and adolescents with higher BMI-SDS is more surprising, as previous studies did not find such association [23, 24]. Various studies have documented an altered hormone balance in children and adolescents with obesity [48, 49]. Since various hormones can also influence behavior, this could be a reason for more frequent conduct problems in children and adolescents with higher BMI-SDS. As already shown for HRQoL and somatoform complaints, significant interactions with age indicate that associations between BMI-SDS and internalizing and conduct problems are more serious in late prepuberty or early puberty than in earlier childhood or later adolescence.

Regarding hyperactivity and inattention, we did not find a significant association with BMI-SDS. This is in line with previous studies using the same instrument (SDQ) [23, 24]. In contrast, several studies assessing clinically relevant amounts of hyperactivity/inattention (e.g., diagnosis of ADHS) showed positive associations with higher BMI [24, 25, 27]. Therefore, it is possible that associations between BMI and symptoms of hyperactivity/inattention are only observable in a range that is not captured by the SDQ.

In the child sample, we found a significant interaction with SES for the associations between BMI-SDS and hyperactivity/inattention and emotional problems. The associations for low SES pointed in a negative direction and for middle SES in a positive direction, whereas the trend for high SES was in between. We have no explanation for this inconclusive result, since we expected a falling or rising trend from low to middle to high SES.

One strength of this study was the examination of many different psychosocial factors in a large German sample of children and adolescents. A special characteristic was the consideration of interactions with sex, age, and SES.

As low SES was underrepresented in our sample, the findings can be transferred to the general population only to a limited extent [32]. Two further limitations concern the collection of data via questionnaires. First, responses might be biased, e.g., by social desirability. Second, responses provided by parents and children themselves might differ and, therefore, results of both the child sample and the adolescent sample are not necessarily comparable.