The relationships between chronological age, pubertal maturation and PA were analysed for the first time in a large, representative sample of German adolescents. Overall, the results point to chronological age instead of absolute or relative pubertal maturation as dominant PA predictor. Puberty effects were only found in boys but were less marked than age effects. BDS also contributed to the prediction in both sexes. It did not mediate puberty effects in girls, while in boys the positive effect of pubertal maturation on PA seems to be partially mediated by a lower BDS.
Chronological age versus absolute pubertal status
A consistent independent contribution to the prediction of PA was only accomplished by chronological age of both, girls and boys. Absolute pubertal status did not add to the prediction beyond chronological age concerning girls. For boys, after accounting for age and the other predictors, the effect of pubertal status changed direction: more mature boys were more likely to be physically active on a daily basis than prepubescent boys. However, this was attributable mainly to a much lower risk of inactivity in more mature migrant boys. Since the type of bodily changes does not differ with ethnic background, it has to be assumed that these account for higher PA levels rather indirectly (i.e. via culturally shaped social expectations or psychological reactions).
In girls, only irregular menses was associated with lower PA. Irregular periods are normal during the first 1-2 post-menarcheal years and, as such, are an indicator of ongoing pubertal development. However, there are also other reasons for irregularities than puberty that can be associated with pain and strong bleeding and may lead to lower PA in the affected girls . An independent association of absolute pubertal development with infrequent PA in girls, therefore, could not be established.
A similar pattern of results as we observed for boys with exercise levels increasing with puberty, was found for the baseline data of the UK HABITS study . No difference in this association was reported by ethnicity. Others [18, 19] found a decline in PA with pubertal maturation in Canadian boys and girls without considering ethnicity. In a study by Bradley et al.  more mature US American girls reported less PA and more sedentary activities. There was, however, no such association found in boys. Activities differed by grade and ethnicity in both sexes.
Other cross-sectional [21, 23] and longitudinal  studies from the UK could not find any associations of pubertal status with PA for any gender.
One explanation for the inconsistency of results may be a different methodology. Studies that analysed puberty effects differ in that they did or did not control for age, body fat and/or relative pubertal timing. Furthermore, different measures of puberty are in use, each reflecting different aspects of physical maturation (i.e. growth spurt and/or various secondary sex characteristics) . Moreover, measures of PA behaviour are also diverse and cover different facets of this complex behaviour. These methods all have certain advantages and deficiencies, and the concordance of results has been shown to be modest at best .
Not all studies that found statistically significant associations between maturity and PA among other things adjusted the analyses for age [e.g. [18, 19, 40]]. Consistent with the results of these studies we found pubertal status and PA clearly associated in the univariate analyses. However, our multivariate models show that age markedly confounded this relation. There are also unadjusted analyses that did not find any associations [21, 22]. Yet, these were restricted to a narrow age range.
However, in case of the HABITS study, cross-sectional analyses that used similar self-report measures found boys' PA levels decline with maturity after accounting for age effects , whereas we observed PA levels increasing with pubertal maturation. One explanation might be that the HABIT study measured PA by taking into account total days of PA per week (including school sports) versus exercising on weekends (yes/no) while our index of PA frequency was restricted to PA beyond physical education classes. Together, these results indicate that associations with pubertal maturation are existent only for boys' leisure time PA.
To what extent associations with pubertal status differ by ethnicity may also depend on the culture of origin as well as the culture of the study country. The situation of migrants in Germany, who set up a very heterogeneous group, may not be directly comparable to the situation of black people in the US or UK. Anyhow, our results indicate that ethnicity or culture to some extent influences the relation between puberty and PA. Attitudes towards maturation culturally differ [17, 42] and may, therefore, also shape effects of puberty on PA.
Relative pubertal timing
We found an independent effect of pubertal timing for boys. While early maturers had a more than doubled risk of being inactive than average maturing boys, this risk was nearly halved in late maturers. This result is in line with the hypothesis that early maturing adolescents are at a higher risk to behave in unhealthy ways [6, 43] but discredits the "maturational deviance hypothesis" which states that both, early as well as late maturers, tend to behave in more health-risky ways. One reason for early maturers to behave more like older adolescents is, that because of a more mature appearance, they tend to be treated like adults by peers, parents, and other adults, including expectations to behave in more mature ways. More mature looking adolescents also tend to have older friends whose interests they may adopt - including sedentary leisure activities [10, 43].
A study by Kemper and colleagues  showed higher activity levels in late maturing versus early maturing adolescents for both sexes. By contrast, other studies did not find associations between pubertal timing and PA after controlling for age [24, 45].
In longitudinal analyses of the HABITS study , contrary to our results, early maturing boys had statistically significant higher rates of PA compared to average or late-maturing boys after controlling for age and other sociodemographic variables. Early maturing girls had significantly higher rates of PA compared to average-developing girls in early adolescence but this effect attenuated over the time. Longitudinal data from another study indicated that early timing of puberty at age 11 in girls predicted lower levels of objectively measured PA at age 13 .
In summary, the little evidence available for timing effects also turns out highly inconsistent. Again, this may be due to the diverse methodology of the studies, where timing and PA are measured in different ways and different covariates are controlled . Furthermore, it is worth mentioning that the classification in pubertal timing groups in our and most other studies was based on within-sample comparisons. These differ between samples and are not equivalent to accelerated or retarded puberty in clinical terms.
All in all, our interpretation is that physical changes during puberty in males facilitate PA. The association of inactivity with early pubertal timing may be explained by more mature boys adopting interests of their older friends as we have pointed out before.
Body dissatisfaction (BDS)
We found an independent association of BDS with lower PA for both sexes. While in the univariate models boys and girls who judged themselves as "much too fat" most notably were affected by a higher risk of less frequent PA, after accounting for body fat percentage this risk was far more pronounced in adolescents with a body image "(much) too thin". This result is explained by the fact that many girls and boys, who rated their bodies "too fat", indeed, had higher body fat percentages (data not shown). Nevertheless, BDS had an independent association, and it may be in part artificial to separate this effect from actual body composition. On the other hand, the lack of controlling for objective weight in studies on body image and health behaviour has been criticised .
Associations of BDS or a negative body image with low PA levels have been observed in a number of previous studies of adolescents. For example, Niven et al.  found significant associations of all subscales of the Children and Youth's Physical Self-Perception Profile with self-reported PA in early adolescent girls. However, longitudinal analyses  for the same sample revealed no significant association between changes in perceived body attractiveness and PA changes over 12 months. Neumark-Sztainer and colleagues  found from Project EAT that youths' BDS predicted lower PA levels after five years. After adjusting for BMI, this longitudinal association remained significant only in girls.
We found no indications of a mediating function of BDS that would suggest psychological reactions to physical changes as cause of inactivity in girls. Although relative pubertal timing predicted BDS, and BDS in turn predicted PA, the associations were for different forms of dissatisfaction (i.e. early maturing girls feeling too fat while girls feeling too thin were more likely to be inactive).
This result is in line with the study by Simon et al. , who also found no hints that effects of puberty on PA behaviour are mediated by psychological factors (stress and psychological difficulties). It is contrary to results of other studies that documented such a psychological mediation of puberty effects in girls [25, 49]. In one study, perceived body attractiveness clearly mediated maturity effects on PA . However, in the study by Davison et al  body esteem, which was similar to our measure of BDS, was not among the variables that qualified as a mediator.
Results for boys point to a partially mediated effect: puberty was associated with a decrease in feeling "too thin" and feeling too thin was associated with an increase in the odds of infrequent PA. This finding is in accordance with the presumption that pubertal development, encompassing muscular development, for boys fosters PA by lowering the risk of BDS due to feeling "too thin" (i.e. insufficiently muscular).
In fact, even if girls in general are less satisfied with their bodies, BDS seems to be highly relevant for adolescent boys, too . However, while girls most often feel being too fat, boys may be dissatisfied when feeling too fat as well as too thin or falling short of the muscular male ideal, respectively [13, 51, 52].
Nevertheless, it is not clear if lower PA levels determine negative physical self-perceptions or if, vice versa, a negative body image leads to lower PA levels. Both direction of effects seem plausible and may be, in fact, contributing to the associations found . In addition, BDS may also act as a motivator for exercising and, as such, the direction of the relationship may be opposite in subgroups.
Either way, body image seems to be an important factor during the adolescent years, which is associated with pubertal development and timing, PA, as well as with body composition.
Future research may untangle how adolescents with a negative body image (and/or with a high percentage of body fat) may be best motivated for PA and how changes in body image affect PA. Overweight as well as a negative body image may both act as PA impediments and motives at once [48, 51].
Strengths and limitations
The study is characterised by several strengths. First, data from a large representative sample were used. The KiGGS study will further be continued as a longitudinal cohort study, which implies that results from the reported survey may be verified longitudinally. Second, the employed questionnaires were extensively pretested and validated . Third, the anthropometric measures were taken by trained medical professionals, and instead of BMI, which often is used as a proxy for adiposity, body fat percentage was used as a more direct measure. Last but not least, most of the existing studies, that linked aspects of body image to PA behaviour, were confined to female samples. Since our sample consisted of both sexes, it was possible to compare the results of boys and girls.
There are, however, also some limitations that have to be taken into account. The main variables studied, namely PA and pubertal maturation, were self-reported by the adolescents. It is well recognised, that the reliability of PA self-reports is limited . The use of accurate objective measures (e.g. accelerometers) is preferable, but often not feasible in large epidemiological studies. Even so, a screening measure similar to the one we used performed best compared to other short measures practicable in large epidemiological surveys .
The possibility of assessing pubertal maturation by clinical examination of secondary sex characteristics was discarded in the KiGGS study because of a low participation rate in the pilot test, especially for girls with migrant background . However, as a review by Coleman and Coleman  showed, self-assessment of the development of secondary sex characteristics with the aid of drawings is a relative good proxy.
We only used two aspects of pubertal status in our analyses. For a comprehensive picture of developmental status other measures like age at peak height velocity, hormone levels or ratings of further secondary sex characteristics would have to be considered and might come up with a different picture. Furthermore, it is to be considered that due to sex-specific differences in timing and order of the appearance of secondary sex characteristics, boys and girls within the same pubic hair stage need not to be of the same maturity status .
Only cross sectional data were analysed. It is, therefore, not possible to infer the causal direction of the found associations. However, in the case of chronological age a reverse causal influence can be ruled out.
BDS was the only psychological variable that was examined as possible mediator of puberty effects on PA. It may be that in girls other constructs would still explain associations found, as they did in other studies [7, 25, 49]. More studies are needed that compare sex-specific BDS effects. Since the concept of mediation implies a causal chain, the aspect of causality in relation to the role of BDS should also be clarified further.