Multi-method confirmatory factor analyses of puberty in early adolescent girls

Abstract

(full):Correctly measuring and operationalizing pubertal processes during development is important for capturing variation in both normative development and trajectories associated with adverse outcomes. The aim of the current study was to understand how this variance may be captured by different pubertal processes (adrenarche and gonadarche) and how multiple methods may contribute to variance (and how they are correlated). The study explored multi-method cross-sectional pubertal data, including self-reported physical characteristics from the Pubertal Development Scale (PDS) and Morris & Udry line drawings, and levels of hormones dehydroepiandrosterone (DHEA), testosterone (T), and estradiol (E2) in saliva and hair, from a cohort of 174 early adolescent girls aged 10-13 years. Authors developed a theory-driven structural equation modelling framework of puberty for girls that includes multiple measures of puberty and considers both processes of adrenarche and gonadarche. Self-reported puberty and hormone levels from saliva and hair can estimate an overall pubertal process as well as each process of adrenarche and gonadarche. The self-reported PDS item on height growth had the lowest standardized loading of the self-report items. In full models with all measures, the more parsimonious one-factor model was not a significantly worse fit than the two-factor model. The shared variance between the hormonal latent puberty factor and the self-reported latent puberty factor was 32%. This suggests fairly good reliability between hormones and self-report puberty data but also suggests that measuring puberty via hormones and via self-report questionnaires is not entirely redundant, and models of puberty require collecting both. Exploratory analyses showed 34%, 48%, and 52% shared variance between chronological age and the adrenarcheal, gonadarcheal, and overall pubertal factors, respectively, suggesting that there is additional variance in these pubertal processes that is not explained by age alone. When controlling for age, the adrenarcheal and gonadarcheal factors were still strongly associated (90% shared variance), suggesting that these factors are correlated when calculated as pubertal timing (i.e., stage compared to same-age peers), as well. Finally, correlations with age were higher when using the latent factors, whether for self-report questionnaire data alone, or for hormones, and correlations were highest when using all available self-report and biological measures. Abstract (short):This structural equation modelling study explored how variance in pubertal development is captured by different processes (adrenarche and gonadarche) and methods, using cross-sectional pubertal data from a cohort of 174 early adolescent girls aged 10-13 years. Self-reported puberty and hormone levels from saliva and hair can estimate an overall pubertal process as well as each process of adrenarche and gonadarche. The self-reported PDS item on height growth had the lowest standardized loading of the self-report items. Hormonal and self-reported latent puberty factors had 32% shared variance, suggesting that research should collect both. There was 34%, 48%, and 52% shared variance between chronological age and the adrenarcheal, gonadarcheal, and pubertal factors, respectively, suggesting these processes are not explained by age alone.

Associating puberty, metabolism and cognition

“…The pubertal latent factor found here, which included 5 out of the 6 explored markers of sexual maturation chosen because they are considered not to be entirely redundant (Van Hulle et al, 2015;Worthman et al, 2019; see also Byrne et al, 2019), confirmed findings from Shirtcliff et al (2009) regarding the associations of Tanner ratings, PDS scores and DHEA, but not the significant relations of most of these variables with testosterone. Shirtcliff et al (2009), however, separated PDS scores into gonadal and adrenal indicators, so their results cannot be directly contrasted with ours.…”

“…To the best of our knowledge, a latent factor using various subjective and objective pubertal indicators has only been built once, in a recent non-peer reviewed preprint (Byrne et al, 2019) that sampled 10-to 13-year-old girls and used different pubertal indicators, which is not directly comparable to our own work. Still, the theory driven inclusion of variables in the model and its internal consistency (i.e.…”

Intergenerational transmission of depression risk: Mothers’ neural response to reward and history of depression are associated with daughters’ neural response to reward across adolescence.

“…Relatedly, another potential limitation of the present study was the measure of pubertal development. Recent evidence indicates that self-report (e.g., PDS) and hormonal measures both meaningfully contribute to assessment of pubertal development (Byrne et al, 2019), and so future work will benefit from employing multiple sensitive self-report and physiological measures of pubertal development (Ladouceur et al, 2019). It is also notable that many of the daughters reported being in relatively advanced puberty, suggesting that puberty may not be an ideal measure of developmental progress in a sample aged 10 to 19; however, supplemental analyses conducted with age instead of pubertal development (see Tables A and B in the online supplemental material) obtained comparable results to those derived using the PDS, thereby providing further support for the findings presented here.…”

The effect of pubertal status on self-regulation of behavior – a systematic review

“…The recognition of these inter-individual changes in terms of age of onset and end of puberty, which do not follow a linear trajectory, has contributed to the growing interest in exploring pubertal effects on cognition and behavior in the last few decades (Marceau et al, 2019). During this period, many recommendations have been proposed regarding ways to incorporate pubertal measures in developmental studies (Byrne et al, 2019;Cheng et al, 2021;Dorn, 2006Dorn, , 2015Dorn and Biro, 2011;Dorn et al, 2006;Huang et al, 2009;Mendle et al, 2019). However, conceptual difficulties remain.…”

“…Latent pubertal measures could even be improved over time by adding more measures as new methods and biomarkers are developed. The incorporation of latent constructs in studies involving puberty has been called for before (Dorn and Biro, 2011) and we have seen recent steps in this direction (Byrne et al, 2019;Herting et al, 2021), although there is still a long way to go before this strategy becomes widespread. We discuss this issue further in França et al (2022).…”