Bibian van der Voorn

150 CHAPTER 10 methodological accuracy 13 . Fixed-effect meta-analysis has the advantage of increasing the impact of large studies on the effect estimate. For comparison, results of random- effects meta-analyses, which put more weight on studies with small sample sizes, were also included. ( Appendix 3 ). A limitation of this study is that only a subset of studies (16%) considered gender differences as the primary outcome. In addition, in 22 studies (28%) samples were not collected specifically during mornings. Both could have led to a selection or performance bias, which we accounted for in our sensitivity analysis. Furthermore, 21 articles with data on 3,985 subjects could not be included in the meta-analysis due to lack of gender-stratified quantitative data, while most of these articles reported no significant gender differences. However, funnel plots of the articles included in the meta-analysis were not suggestive of publication bias. Instead, the plots seem to indicate that most articles reported on the nonexistence of gender differences, which might be a result of the common idea that gender differences are nonexistent at this early age. Nonetheless, ourmeta-analysis shows that significant gender differences are already present early in life. Another limitation is that almost all studies that reported on salivary or serum cortisol used immunoassays. Due to its superior specificity, liquid chromatography-tandemmass spectrometry is the method of choice for steroid hormone analysis 34 . Furthermore, we stratified studies based on the mean age or age range of the study group. Since study samples differed in age range, we have probably included some subjects < 8 yr of age in the 8-18 yr groups, and vice versa. An overview of the age ranges of studies included in the meta-analysis is presented in Appendix 5 . Moreover, only a minority of the included studies assessed Tanner pubertal staging. Therefore, we were unable to address the question at which maturational stage the direction of the gender-specific dimorphism in cortisol changes. CONCLUSIONS In conclusion, gender differences in HPA axis activity are present early in life, with higher salivary cortisol concentrations in boys. A gender-specific evolution of cortisol metabolism is suggested to be induced by puberty, resulting in lower bioavailability of cortisol in boys. Although results from random-effects analyses were inconclusive for serum cortisol, the gender difference in cortisol production seems to be consistent between genders with age. Future research should take gender differences in HPA axis activity into account, regardless of age. Whether gender differences in stress- induced cortisol levels also exist is unknown and remains to be explored.