Our study, including 1 643 individuals with longitudinal data, showed that a GPS, based on nine established obesity-susceptibility loci, was significantly associated with BMI and WC in adolescence as well as in young adulthood. No significant associations between the GPS and change in BMI or WC between baseline and follow-up could be verified. Four of nine loci reached nominal significant associations when tested individually, whereas most of the associations of the five remaining loci showed directionally consistent associations. None of the interactions between physical activity and the obesity-susceptibility GPS could be statistically significant derived from our data, neither on the level nor the change in body fat estimates from adolescence into young adulthood.
Comparing our findings to those of others is difficult due to differences in age span, and selection of obesity-susceptibility variants. In spite of these differences, our data are in concordance with previous studies 
, also concluding that obesity-susceptibility loci found to be associated in adulthood, already affect anthropometric traits in adolescence. The GPS in our study showed an effect size in the same order of magnitude in relation to BMI in adolescence (B: 0.046) as den Hoed et al. 
reported from the EYHS study (B: 0.044). In agreement with Zhao et al. 
, we found the locus near MC4R
to be associated with BMI in adolescents.
We did not find an association between the GPS and change in body fat estimates from adolescence to adulthood. This might be due to the fact that our study had only one follow-up or did not capture enough time in the life course of the adolescents. Another reason might be that in concordance with Hardy et al. our data suggests that the genetic effects are slightly larger in adolescence and seem to decrease afterwards 
Li el al. 
, on the contrary, found an association between a GPS and change in BMI, but only in physically inactive participants. Haworth 
discussed that the genetic influence on BMI becomes progressively stronger over childhood, because of stronger expression of genes (FTO) or due to the fact that children increasingly select environments which fit to their genetic propensities. Other studies 
have suggested that genetic variants, affecting BMI level, may be distinct from those affecting BMI change. This could maybe be supported by the finding that the single variant in MTCH2
showed an association with change in both BMI and WC, not found at the cross sectional level. However, this finding was no longer statistically significant when controlling for multiple testing. Studies have reported different results for the attenuating effect of physical activity on obesity susceptibility markers 
. Examining a 12 BMI-loci GPS in a large study population, a previous study found an attenuating effect of physical activity on obesity susceptibility genes in adults 
. A recent meta-analysis 
reported that physical activity attenuated the influence of FTO
variants on obesity risk in adults, but not in adolescents and children. Our data did not indicate any attenuating effect of physical activity on obesity susceptibility in adolescence or in young adulthood. This is consistent with the recent meta-analysis in regard to adolescents 
. However, it might be noteworthy that our data all along the line indicated a larger effect size in the group of adolescents who exercised less in their leisure time. However, the differences did not reach statistical significance.
Due to the repeated cross sectional design of a total population used in HUNT and a low participation-rate for the age-group 23–29 year old in HUNT3, relatively few of the total number of participants in Young-HUNT 1 were followed-up. Comparisons of participants with non-participants at baseline (Young-HUNT 1) displayed no essential differences in the general characteristics (BMI, WC, gender), suggesting no major selection bias of the follow-up sample.
The main limitation to our study is the rather low number of participants and thus not finding statistical significance might be due to lack of power. Nevertheless, the associations for the four loci that that reach nominal significance are directionally consistent with those reported for theses loci in the original papers 
(i.e. the same allele increases obesity-susceptibility). Furthermore, most of associations for the loci that did not reach significance are also directionally consistent with those reported in the original paper. Therefore, we believe that our observations are not a mere reflection of chance. Larger sample sizes will be needed to firmly replicate (or refute) our observations.
This may be especially prominent when testing interaction effects. Besides, others have pointed out that replication studies may, besides providing insurance from the errors and biases that may be unavoidably afflict any study, also amplify confidence that any associations uncovered reflect processes that are biological interesting, rather than methodological inadequacies 
In conclusion, we replicated partly the association between previously published obesity-susceptibility loci and adiposity-related traits in both adolescence and young adulthood. Loci known to affect obesity-susceptibility in adulthood already seem to do so in adolescence. Understanding the different mechanisms behind weight increase from childhood and adolescence to adulthood is important both in a clinical and prevention perspective. Our study might possibly show some indications of effects, which could not all be verified with statistically significance. More large population studies with focus on associations between genes and weight gain over time and possible attenuating effects as physical activity are needed.