**4. Discussion**

Consistent with the literature e.g., [2,9], the present findings provide clear evidence that overweight adolescents represent a specific risk group for mental health problems. Elevated psychopathological symptoms were observed primarily in the internalizing domain while a similar study in overweight/obese adolescents, which also used the YSR instrument, reported increased psychopathology in both internalizing and externalizing behavioral domains [15]. Most obvious, the eating disorder risk in overweight adolescent was about twice as high as in the normal weight population, which supports previously published literature emphasizing the high prevalence of binge eating and compensatory and unhealthy weight control behavior in overweight and obese adolescent populations [8,49,50]. In this regard, the present results mirror other results in the literature pointing to the shared risk factors for obesity and eating disorders [19,21]. Furthermore, the reduced quality of life scores regarding satisfaction with one's own body and appearance, relationship to peers and parents and satisfaction with the school environment reported by overweight adolescents and which has been also found in other studies [10,11], points to the urgent need for interventions promoting well-being and mental health in this group.

The question whether underweight adolescents also have an increased risk for mental health problems is much less easy to answer based on the present results. Elevated psychopathological symptoms were found for internalizing problems only (with more socially withdrawn behavior), while quality of life scores were comparable to those of normal weight adolescents and even lower levels of externalizing problems were reported compared to the reference sample. Indeed, whether or not underweight is associated with increased mental health concerns has been controversially discussed in the literature [10,14,51]. As in our study, Drosopoulou et al. [15] reported increased socially withdrawn behavior in underweight adolescents, while they found no differences in other psychopathological symptoms compared to normal weight youth. While underweight is a core characteristic of anorexia nervosa, we found that the eating disorder risk was minimally but significantly lower in underweight compared to normal weight adolescents. This may be surprising; however, a similar result was also found in another large population study where a different instrument to assess eating disorder risk was used [49]. This indicates that underweight per se is not a sign of increased eating disorder risk and that underweight adolescents (provided there are no additional risk factors) would not need specific (preventive) interventions targeting eating disorder symptoms.

The core aim of this study, which also represents the novelty of this research, was the use of psychological network analysis to identify central factors among mental health and well-being variables which inform about potential key targets for interventions for overweight and underweight adolescents. As psychopathological symptoms were most prevalent in overweight adolescents, the following discussion primarily focuses on what can be done for this risk group. The most central variables in the network were anxious/depressed mood and attention problems, while variables associated with eating disorder risk and body dissatisfaction were rather peripheral nodes. At first glance, this seems surprising, but this finding is consistent with another network analysis based on adult individuals with obesity showing that variables directly related to eating disorders were rather placed in the periphery of a network including different psychological characteristics [30]. This may—to some extent—reflect that particularly anxiety problems, but also symptoms of attention-deficit-hyperactivity disorder constitute the most prevalent

mental health problems among children and adolescents in general [35,52]. However, it must be emphasized that variables that turn out as most central in psychological network analyses do not necessarily correspond to the most prevalent symptoms of mental health disorders. The main finding from this study (depressive/anxious mood, attention problems as central symptoms in a psychological network, thus representing promising key targets for intervention) contradicts the current practice of many psychological (preventive) interventions among overweight/obese adolescents which primarily aim to reduce disordered eating behavior and weight/shape concerns [53]. Rather, our results indicate that broader intervention approaches, not solely focusing on eating disorder symptoms but (also) incorporating contents to positively impact mood and reduce feelings of depression and anxiety might be most promising. This seems all the more appropriate considering the role emotions and emotion regulation play in individuals with overweight and obesity. Negative affect and stress, for example triggered by weight-related teasing and negative body image, may challenge existing emotion regulation strategies which in turn may result in maladaptive coping such as emotional eating that is often reported in overweight individuals [54,55]. Strengthening skills towards awareness, understanding and acceptance of emotions, self-support and self-compassion may improve resilience, self-efficacy, selfesteem and assertiveness among overweight and obese adolescents [54]. This is in line with a systematic review pointing to the causal link between negative emotions (depression, anxiety, stress) and the development of obesity concluding that adolescents' anxiety and depression are therefore important targets for preventive interventions of obesity [56].

Apart from anxious/depressed mood, the 'attention problems' subscale of the YSR was also a central variable in the network analysis. On the one hand, this may be linked to the association between attention-deficit-hyperactivity disorder and obesity (e.g., higher levels of impulsivity which may reinforce disregulated eating behaviors) often reported in the literature [57]. Apart from impulsivity and problems with concentration, this scale also assesses inner restlessness and tension. This indicates that intervention components tackling these problems, like the use of relaxation techniques, may be promising concerning promoting well-being in overweight or obese adolescents. This is in line with previous randomized-controlled trials that have shown a beneficial effect of stress management and relaxation intervention (progressive muscle relaxation, guided imagery, diaphragmatic breathing) to reduce general psychopathology, anxiety and depression symptoms in children and adolescents with obesity compared to interventions solely focusing on the change of dietary and physical activity habits [58,59].

Of note, the factors contributing to the development and maintenance of overweight and obesity in childhood and adolescents are manifold as, for example, shown in the 'Foresight Obesity System Map' where different biological, medical, psychological, developmental, social and economic factors as well as factors related to diet, physical activity, media and infrastructure have been put together and correlated [60]. The present study provides a contribution to the question of the relative importance of psychological factors in overweight and obesity.

Finally, we found that the network structure of the overweight, underweight and normal weight groups was quite similar. While social problems, socially withdrawn behavior and satisfaction with the school environment tend to play a slightly more central role in the networks of underweight and normal weight compared to overweight adolescents, anxious/depressed mood and attention problems were by far the most important factors within the networks across all groups. Interestingly, in a usability study and survey assessing the adolescents' and stakeholders' perspectives on Internet-based prevention for mental health problems in general and for eating disorders and obesity specifically, coping with stress and negative mood were mentioned as the most important topics to address while topics directly related to eating disorders (e.g., healthy nutrition, physical activity) were perceived as less relevant [61,62], which supports the findings of the present study. This has implications for the conceptualization of prevention initiatives in general. Rather than having to focus on different psychological targets for different weight groups, focusing

on mood, depression, anxiety and inner restlessness might be promising targets for mental health promotion and preventive interventions across the whole weight spectrum. This is especially important for large-scale interventions to prevent obesity and eating disorders in school settings [63,64], where individualized interventions dependent on the individuals' weight status are difficult to implement.

The findings from this study must be interpreted in line with the following limitations: First, as in every network analysis, the findings strongly depend on the (variety of) variables that are considered. In this study, we focused on general psychopathological symptoms and well-being variables. Eating disorder symptoms were obtained with a brief screening questionnaire only and we did not obtain detailed information on restraint eating, binge eating or weight/shape concerns using more specific instruments. Due to time constraints, we used the YSR questionnaire assessing different behavioral and emotion problems rather than different instruments assessing different psychopathological symptoms (e.g., depression, anxiety, conduct problems, attention problems) separately (and probably more specifically). Moreover, other (non-psychological) variables like physical activity or dietary habits, which were not addressed in this study, might have provided additional information regarding the interplay between mental health and lifestyle behaviors among overweight and underweight adolescents. Second, weight and height information to calculate the BMI and classify the individual into the overweight, underweight and normal weight groups was obtained via adolescent self-report; thus, these data might lack accuracy to some extent. However, a study based on a general sample of adolescents demonstrated that the difference between self-reported and objectively measured height and weight is marginal [65]. Hence, self-report information should be sufficiently accurate for the purpose of the present study where adolescents were classified into broad weight categories and these data are not used for the diagnosis of anorexia nervosa or obesity. Third, it may be argued that the size of the subsamples of overweight and underweight adolescents may be small given the large number of variables and associations to be estimated. However, to tackle this potential limitation, we used the LASSO estimation which is particularly suitable for smaller samples as it returns a sparse network model where the number of parameters that need to be estimated is reduced [43]. Furthermore, the network stability and edge accuracy measures indicate that the achieved centrality indices and edge weights can be reliably interpreted with the obtained sample size. Finally, it should be noted that we have drawn a community sample of adolescents. Thus, the network analysis primarily informs about promising targets for indicated preventive interventions implemented in community (e.g., school) settings. Future studies may also focus on treatment seeking samples of adolescents with severe obesity or severe underweight which may better inform about important targets for clinical interventions for more severely ill adolescents.
