Assessing the stability of psychobiological stress reactivity during adolescence: mixed-effect modelling of cortisol responses to laboratory stressors

Pubertal development

Puberty is a physiological process that enables and prompts change in the psychological and social life of the person; these biopsychosocial changes have been considered within both risk and protective frameworks (Chen & Raine, 2018; Hummel et al., 2013). The focus has often been on pubertal timing, or measurement of when children and adolescents reach specific physiological, hormonal, or other domain-dependent stages of pubertal development relative to their same age and sex peers (Mendle et al., 2010). Earlier or later onset of puberty has been linked to future and concurrent mental and physical health, for example, people, particularly girls, who commence puberty ahead of their peers are at greater risk of experiencing affect disorders (Galvao et al., 2014), behavioural problems (Dimler & Natsuaki, 2015), and health issues in later life (Day et al., 2015). Similar associations between earlier pubertal onset and adverse outcomes have also been observed in boys (Ullsperger & Nikolas, 2017), however, these negative outcomes may also be largely shaped by social contextual factors. For example, for boys with less harsh, more supportive parenting, earlier onset of puberty has been associated with enhanced social competence and positive psychosocial development, as well less aggressive behaviour (Klopack et al., 2019). In girls, timing has been shown to only predict aggressive behaviour when coupled with low maternal nurturance (Mrug et al., 2008). The interaction of parental relationships and pubertal timing has been observed in both boys and girls, with positive parenting moderating the relationship between off-time pubertal onset and adverse outcomes (Chen & Raine, 2018). These findings support the concept that health and behavioural outcomes are not solely determined by the physiological processes, or the timing or tempo of these alone, but rather situated within the context of the young person’s environment.

The impact of pubertal development on adolescent development has been examined in combination with stress responsivity, including as indexed by cortisol reactivity, and symptoms of depression have been predicted by an interaction of earlier puberty and heightened stress reactivity (Gong et al., 2019). Interestingly, one study found there to be a link between the timing of puberty, the patterns of cortisol reactivity, and the risk of depressive symptoms (Colich et al., 2015). Here, depressive symptoms were predicted by HPA hyporeactivity in those who began puberty earlier than their peers, and HPA hyperreactivity in those who had begun puberty later. What these studies may indicate, is that the timing of puberty is a complex, highly socially contextual process that is closely linked with outcomes related to stress.

Whereas pubertal timing is the onset of puberty relative to peers, pubertal tempo refers to the rate at which an individual progresses through puberty. There has been comparatively little research exploring the influence of pubertal tempo on biopsychosocial development, however research to date suggests that relatively fast or slow pubertal tempo may be associated with both depressive symptoms (Keenan et al., 2014; Mendle, 2014) and behavioural issues (Marceau et al., 2011). One recent systematic review of outcomes related to pubertal tempo found that while both boys and girls who experienced accelerated development had symptoms of depression during childhood and adolescence, there were distinct gender differences (Cheng et al., 2020). For instance, boys tended to experience more psychosocial difficulties due to accelerated development, for girls this was an opportunity to compensate for the impact of later onset puberty. These findings can be somewhat contradictory, and indicate there may be strong gender differences in the risk and protective factors associated with development (Mendle et al., 2010). In the context of cortisol reactivity, there is relatively little research examining the impact of pubertal tempo, however, accelerated development is associated with heightened stress reactivity, and this reactivity is associated with later depressive symptoms (Gong et al., 2019). It is important to note that while pubertal development and HPA maturation is associated with negative outcomes, some suggest that these should be treated as neither antecedents nor consequences of development, but rather as contributors to a system of accumulating risk, such as with allostatic load (Joos et al., 2018; Whelan et al., 2021). In other words, where pubertal timing and pubertal tempo can influence cortisol reactivity patterns, these exist within a wide system of opportunities for positive and negative growth.

Stability of cortisol reactivity during pubertal development

Patterns of cortisol secretion are reported to change during adolescence, with a normative increase in cortisol production, and differences in diurnal cortisol and cortisol reactivity partly explained by pubertal development, and differences in age and sex groups noted (Gunnar et al., 2009; Ji et al., 2016; Platje et al., 2013; Rotenberg et al., 2012). Because of this, adolescence provides a unique opportunity to study stability, as stability/instability is highly linked to this rapid development in the Hypothalamic Adrenal Pituitary axis, and the Hypothalamic Pituitary Gonadal axis, and offers a unique opportunity to study stability and change (Gunnar et al., 2019). The Pubertal Stress Recalibration Hypothesis (PSHR) describes the process wherein several bioecological and neurological factors, such as heightened plasticity in the hypothalamus, association cortex, and prefrontal cortex, and heightened sensitivity to socio-emotional environments, may allow for physiological responses to stress to be recalibrated during adolescence. There is some evidence from both cross-sectional (DePasquale et al., 2019; Zhang et al., 2021) and longitudinal (Gunnar et al., 2019) studies to indicate a recalibration of cortisol responses in the context of positive social environments, at least in children and adolescents who experienced institutionalisation. However, there has been limited exploration of how cortisol reactivity or stability may be impacted by off-timing or off-tempo pubertal development. It is suggested that cortisol reactivity is influenced by the timing of puberty, with greater associations between off-time development and moderations in both stress reactivity and recovery (Smith & Powers, 2009). The current study examines the potential role of off-time and off-tempo pubertal development on the stability of cortisol reactivity to laboratory stress challenge across a period of adolescent development.

Participants

This study uses secondary data comprised of a sample of 135 participants collected as part of the 'Physiology of Puberty and Antisocial Behavior' project (NIMH:5R01MH058393-03). Initially, a comprehensive list of all children residing in specific zip code areas was obtained from the American Student List (ASL), a commercial organization that provides names of school-aged children and adolescents. The zip codes chosen belonged to the county where the research was conducted, as well as adjacent counties that were easily accessible to the laboratory. This recruitment process was inclusive of rural and semi-rural addresses located within a reasonable distance to the laboratory. A letter was sent to 966 parents, requesting their cooperation in the study and to be contacted by the research team. Following this, the research staff personally reached out to the parents through phone calls, seeking the child / adolescents' participation. Ultimately, 85 children and adolescents were enrolled based on the responses from their parents. Among the remaining children and adolescents, 584 could not be contacted due to returned letters and lack of forwarding contact information, while 89 others did not meet the inclusion criteria. Additional participants were recruited via flyers distributed throughout the community and through telephone responses to emails sent to university staff members. All these recruitment procedures were approved by the Institutional Review Board at Pennsylvania State University.

This data set is comprised of biological, physiological, and psychological measures. Of the 135 included in the final data set, thirty-three were removed due to missing data in cortisol reactivity or pubertal measured. Participants included in the current study include 102 (49 boys, 53 girls) adolescent-aged children and adolescents. At the first wave of measurement, girls were aged 8, 10 or 12 years (Mean = 10.96, SD=1.66), and boys were aged 9, 11 or 13 years (Mean = 10.08, SD = 1.66). The difference in ages between sexes was to account for noted differences in the typical time young people enter puberty. All participants included in this study were free from chronic health problems and were not using any medication known to interfere with hormone levels. Participants were predominantly white (non-Hispanic) (94%), Hispanic (4%), African American (4%), and Asian (1).

A more detailed breakdown of this sample has been described previously (Susman et al., 2010), however, a summary of these details can be seen in the table below:

This data set is comprised of biological, physiological and psychological measures collected over the course of three years. Participants included in the current study include 102 (49 boys, 53 girls) adolescent-aged children. At the first wave of measurement, girls were aged 8, 10 or 12 years (Mean = 10.96, SD=1.66), and boys were aged 9, 11 or 13 years (Mean = 10.08, SD = 1.66). All participants included in this study were free from chronic health problems and were not using any medication known to interfere with hormone levels.

Dependent variable

Cortisol reactivity. Cortisol reactivity was measured using salivary cortisol levels in response to the Trier Social Stress Test for Children/Adolescents (TSST-C) (Allen et al., 2017). The TSST-C is a test which elicits a stress response in participants by using both social evaluative and cognitive components. Children and adolescents were asked to come up with the end to a story in front of two judges and told that their answer would be compared to those of other children and adolescents their age. Prior to the commencement of the test, two samples were gathered to estimate a baseline measure of cortisol. After the administering of the task, three further samples were collected, first immediately after the task, then in ten-minute intervals. Cortisol reactivity is operationalised as the cortisol response to stress, measured as Area Under the Curve with respect to increase (AUC_I) (Pruessner et al., 2003). This method provides a measure of change over time, as it ignores the distance from zero on all measurements, instead using baseline measures as a focal point for increase, and is thus more useful for increase response and acute measurement (Fekedulegn et al., 2007). A breakdown of demographics can be seen in Table 1. Previous research has opted to use a larger scale growth curve analysis to calculate cortisol response and recovery (Ji et al., 2016). These calculations provide a sensitive and robust model. However, this research is concerned with the overall production of cortisol response, relative to the baseline and so AUCi (Pruessner et al., 2003) was used.

Independent variables

Pubertal timing and tempo. Pubertal timing and tempo were calculated using pubertal stage. Pubertal stage was assessed by a paediatric research nurse during the research visit, and was operationalised following Tanner guidelines of genital stage and pubic hair stage for boys, and breast stage and pubic hair for girls (Marshall & Tanner, 1968), and included a self-report, parent report, and physical exam. To achieve this, the nurse first explained to the participants what each of the five stages of puberty included. The nurse then showed the parent and adolescent pictures of each of the five Tanner stages, and asked them to independently identify what stage of development the adolescent had reached. For this study, participant self-report of stage (genital for boys, breast for girls) was used in the analysis. Pubertal timing was calculated using a regression of pubertal stage on age of participants stratified by gender (Dorn et al., 2003; Dorn et al., 2006). The resulting residual value was then used as an index of pubertal timing – where a higher value indicates a later timing of puberty, and a lower value indicates an earlier timing of puberty. Pubertal tempo was calculated using linear growth curve estimation. This method was based on an adapted version of previous methods used (Marceau et al., 2011). To calculate individual tempo, each growth trajectory was modelled using linear growth curve estimation, and the total variance over time was used as an indicator of pubertal tempo.

Anxiety. Anxiety was assessed using participants Anxious/Depressed scores on the subcategory of the Child Behaviour Checklist (CBCL) (Achenbach & Edelbrock, 1991). The CBCL is a checklist which measures problem behaviour and emotional problems in children and adolescents, and has been well-validated cross-culturally (Ivanova et al., 2007). The CBCL was completed by parents / guardians of the child and the parent data was used in this analysis.

Covariates. There are several other factors that may be relevant in examining the association of pubertal development with cortisol reactivity profiles, including Socio-economic status (SES) (Reiss, 2013; Vliegenthart et al., 2016), body-mass index (BMI) (Dockray et al., 2009), and sex. SES was determined using the Hollingshead four-factor index of social status (Hollingshead, 1975). The mean score for participants was 47 (SD = 12), out of a possible 66. BMI was calculated using the standard formula of weight (km)/height² (m²) at each wave of measurement. Measurements of height and weight were obtained through nurse reports during a physical exam, with an average of three measurements taken as the operant value.

Statistical analysis. All data analyses were conducted using the IBM Statistical Package for the Social Sciences (SPSS), version 28. A linear mixed-effect model was used to examine whether there were any associations between cortisol reactivity to stress and each of the independent variables between each wave of measurement (West, 2009). The effects of SES, BMI, and sex were controlled for in the model. A summary of mean values of age, pubertal timing, pubertal tempo, cortisol reactivity, BMI, and anxiety can be seen in Table 2 below.

Stability of cortisol reactivity across adolescence

There was a significant difference in cortisol reactivity to stress between timepoints (Estimate= 3.39, t=3.67, p<.001, CI[1.56, 5.22]). We observed a general trend of increasing cortisol reactivity from wave 1 (Mean=-2.02, SD=3.89) to wave 3 (Mean = 4.23, SD =7.12) of data collection (Estimate=3.39, t=3.67, CI[1.56, 5.22]) (see Table 3).

Impact of pubertal timing and tempo on this relationship

There was no direct significant effect of pubertal timing or anxiety on cortisol reactivity profiles over time, although we did observe a significant main effect of pubertal tempo on cortisol activity (Estimate= -3.59, t=-2.13. p=.03, CI[-6.92, -0.25]). We found that more rapid pubertal tempo was associated with a reduced cortisol reactivity profile, however the width of confidence interval limits meaningful inference.