Keywords

Family socioeconomic status (SES), Hypothalamic-pituitary-adrenal (HPA) axis, Cortisol, Social support, Stress-sensitization, Longitudinal studies

Introduction

Evidence shows that children raised in lower socioeconomic households (e.g., lower income) are at higher risks of experiencing physical and mental health problems later in life (Bradley and Corwyn, 2002; Kim et al., 2018). These children are also more likely to face chronic stressors within and outside of their family environment, which may further tax their capacity to adapt to stress (Kim et al., 2018). Accordingly, several researchers have proposed that early stressors, such as lower family socioeconomic status (SES), jeopardize later functioning in part through the dysregulation of the neurophysiological stress systems (Shonkoff, 2010).

The hypothalamic-pituitary-adrenal (HPA) axis is one of the main physiological systems responsible for maintaining adaptation to stress through the secretion of the glucocorticoid hormone cortisol (Koss and Gunnar, 2018). Cortisol secretion follows a diurnal cycle characterized by higher levels after awakening reaching a peak approximately 30 to 40 minutes followed by progressive declining levels during the day. Alternatively, hair cortisol concentration (HCC) is thought to capture prolonged cortisol production over months, typically three (Koss and Gunnar, 2018). Previous research has shown that socioeconomic disadvantage often co-occurs with dysregulation in HPA axis activity (Koss and Gunnar, 2018). Yet, important discrepancies exist regarding the magnitude and direction of these associations. Indeed, prior studies have reported both lower and higher diurnal and hair cortisol secretion among youth living in impoverished households (Chen et al., 2010; Gray et al., 2018; Lupien et al., 2001; Tarullo et al., 2020) and nonsignificant findings (Malanchini et al., 2020; McLachlan et al., 2016; Young et al., 2019). This mixed pattern of findings may be due, among other things, to past studies’ failure to explore the possibility that these associations may vary according to changes in SES (Cantave et al., 2021; Desantis et al., 2015). Accounting for stability and changes in SES across childhood and adolescence may help clarify whether these experiences have cumulative (i.e., additive) or synergic (i.e., interaction) effects on cortisol secretion. Based on the stress sensitization model (Daskalakis et al., 2013), family socioeconomic disadvantage experienced during early childhood may prime a stronger response to similar contexts in adolescence, taking the form of a stronger association between adolescence SES and cortisol secretion for youth who grew up in impoverished households during early childhood (Young et al., 2020). This hypothesis was partially tested in a previous study showing that neither early childhood or adult SES, taken independently or jointly, covaried with adult diurnal cortisol (Young et al., 2019). However, it remains unknown whether these findings also apply to adolescents. As the HPA axis’ is highly responsive to social cues during adolescence (Gunnar et al., 2019), family socioeconomic disadvantage might more readily affect cortisol secretion during that period, especially for those who also experienced lower-SES conditions in early childhood.

The magnitude of the association between SES and cortisol may also depend on the perceived availability of social support within youth’s social network. Growing evidence suggests that social support can buffer or offset the association between stressful environments and HPA axis activity (Hostinar and Gunnar, 2015; Shonkoff, 2010). However, evidence corroborating the buffering role of social support in the SES-cortisol association is scarce. To the best of our knowledge, only one study has examined this hypothesis, targeting young adults (Hooker et al., 2018). This study revealed that lower-SES was associated with higher cortisol secretion during stress recovery, but only for participants with low social support (Hooker et al., 2018). However, another study, which assessed the moderating role of SES in the association between social support and diurnal cortisol secretion, reported nonsignificant results (Hooker et al., 2020). Further investigation of whether social support buffers the association between family socioeconomic disadvantage and (salivary) diurnal cortisol or HCC is warranted to further understand how social support may promote youth resilience.

Building from previous work on the same cohort of youth that examined associations between early childhood (0–5 years old), adolescence (age 14) family SES, and either diurnal (Cantave et al., 2021) or hair cortisol secretion (age 19) (Cantave et al., 2022), this study tested for the first time the hypothesis that the association between family SES in mid-adolescence and cortisol secretion depends on the socioeconomic hardship experienced in early childhood. We also examined whether social support buffered these associations.

Methodology

Participants

Participants were part of the Quebec Newborn Twin Study (QNTS), recruited between 1995 and 1998 in the greater Montreal area. Of the 989 families with twins initially contacted, 662 agreed to participate (68%). Participants were first seen at 5 months of age and then prospectively assessed for a variety of children and family characteristics every one or two years on average (for more information, see (Boiivn et al., 2019). At the time of participants’ birth, 17% of mothers and 14% of fathers had not graduated from high school, one in 10 families received social welfare or unemployment insurance and approximately a third of the families (30%) had an annual income of less than $30,000. During the twins' preschool years and adolescence, 16-28% of families reported income below the low-income cut-offs (LICOs) set by Statistics Canada (2016). LICOs represent an income-to-need threshold where families must allocate a higher proportion of their income towards necessities compared to the average Canadian family (for more information, see (Cantave et al. 2021). This study focused on a subsample of participants with valid cortisol data collected from saliva at age 14 years [Mean(Standard Deviation or SD)=14.0(.3)] or from hair at age 19 years [Mean(SD)=19.1(.3)]. Information on saliva samples was collected in 2009-2010 for 569 participants [52% girls] from whom most (74%) had collected samples across all four collection days. Hair samples were collected in 2014-2015 in 704 participants [61% girls; see section 2.4 for analyses examining attrition].

Procedure

At age 14, letters detailing the objectives of the study were sent to the families, followed by a home visit. After informed consent (parents) and assent (youths) were obtained, the research assistants explained the saliva collection protocol, which consisted in sampling saliva at four-time points during the day (awakening, 30 min later, late in the afternoon and bedtime) on four collection days (Tuesdays and Thursdays on two consecutive weeks) and the completion of an interview-based questionnaire with the youths and their parents. During the home visit, the research assistants ensured that participants and their parents were familiar with the collection material. Families were visited a second time to gather the saliva tubes. At age 19, youths were invited to our laboratory, although approximately 11% chose a home visit (with research assistants) instead and about 4% opted to collect hair samples themselves and to mail them back to the laboratory. After youth provided their informed consent, hair sample of at least 3 cm long and 1 cm wide were collected from the posterior vertex area of the participants’ scalp by trained research assistants or according to the illustrated instructions for those who collected the samples themselves. No mean differences in HCC were detected between samples measured from hair collected at home and laboratory (Ouellet-Morin et al., 2016). All instruments and study procedures were approved by the Ethics Committee of the Sainte-Justine Hospital Research Center.

Measures

Early childhood and mid-adolescence SES were derived from the parents’ highest educational level and family income (see Figure S1 in the Supporting information). Parents’ highest educational level was categorized as high school diploma or less (scored as 0) or postsecondary diploma (scored as 1). Family income was reported in early childhood (i.e., 5 and 18 months, and again at 2.5, 4 years; see Figure S1 in the Supporting information) in seven categories, ranging from 0 to $\geq$ $80,000, with a $10 000 increment for each category. The categories at age 14 were comparable (i.e., 1=“$0 to $40,000”, 2=“$> \$$40,000 to $60,000”, 3=“$> \$$60,000 to $80,000”, 4=“$>$$80,000 to $100,000”, 5=“$>$$100,000”). The scores were averaged to create a mean family income in early childhood [Mean(SD)=4.46(1.92), corresponding to a $40,000-to-$49,999 average] and mid-adolescence [Mean(SD)=3.06(1.49), corresponding to a $60,000-to-$80,000 average] and then partitioned into five groups due to the scores’ asymmetric distributions. Information about the highest parental educational level and family income were included in a confirmatory factor analysis (CFA) to derive robust and cohesive SES latent indicators in early childhood (age 0–5 years) and mid-adolescence (age 14 years). Models confirmed that a unique factor fit well the data at each developmental period [(see Cantave et al. (2021), for more information].

Social support was self-reported using the widely-used Network of Relationships Inventory (Furman and Buhrmester, 1985) at age 14 and once again at age 19. In this study, three items assessing perceived emotional and instrumental support (i.e., When you’re feeling down or upset, how often do you depend on this person to cheer you up? How often do you depend on this person for help, advice, or sympathy? How often do you turn to this person for support with personal problems?) were used to rate the participants’ support from their mother, father, close friend, co-twin, and teacher at age 14, as well as from their romantic partner at age 19. We did not consider perceived support from a romantic partner at age 14 because most participants did not report any romantic partner at this age (i.e., proportion at age 14: 89% and at age 19: 44%). Social support from teachers was not available at age 19. Items were rated on a 5-point Likert-type scale ranging from 1 (little or none) to 5 (the most) and averaged to create a total score at age 14 (mid-adolescence; 15 items; Mean=2.74; SD=.84; $\propto$=.90) and at age 19 (late adolescence; 15 items; Mean=2.93; SD=.81; $\propto$=.86), respectively. Given the moderate correlation between mid- and late-adolescence social support scales (r=.38, p<.001), both scores were averaged into a single adolescence social support index (Mean=2.84; SD=.73) to ensure a more reliable assessment of social support during adolescence. The mid-adolescence (age 14) perceived social support was used in the analyses of cortisol measured at age 14 and the aggregated index of perceived social support (i.e., averaged ages 14 and 19) was used for analyses of cortisol at 19 years so that associations could be examined according to a clear temporal sequence (i.e., the moderator precedes the outcome).

Diurnal cortisol. At age 14, participants were provided saliva tubes, diaries to report collection times and instructions for collection. Saliva samples were placed in the participants’ refrigerator during data collection days and then stored in -20^o C freezers once returned to the laboratory until cortisol determination using a high sensitivity enzyme immune assay kit (Salimetrics® State College, PA, Catalog No. 1–3102). Frozen samples were brought to room temperature to be centrifuged at 15,000 $\times$ g (3000 rpm) for 15 min and analyzed on 96-well plates. The range of detection for this assay was between 0.007–3 µg/dl (.19–82.76 nmol/L) and the intra- and inter-assay coefficients of variation were 4.8% and 8.2%, respectively. We identified 1% of cortisol samples with a value greater than 3 times the SD above, which were winsorized. Participants were considered “compliant” if their awakening and +30 min saliva samples were separated from at least 20 min and less than 40 min and that their awakening saliva collection was completed within the first 15 min following awakening and not distinct from the co-twin ($\leq$8 min). A total of 8.61% of the samples were discarded due to noncompliance, as based on the participants’ written records of collection time. Cortisol values were converted from µg/dl to nmol/L (i.e., multiplied by 27,588) and natural log-transformed prior to analyses.

Using CFA, we estimated three stable (i.e., trait-like), robust and cohesive estimates of diurnal cortisol: the cortisol awakening response (CAR), the mean level of cortisol at awakening (intercept) and the diurnal change levels (slope). Specifically, the CAR was calculated by subtracting the awakening cortisol level from the level collected 30 minutes later for each collection day. Growth curve analyses using mixed modeling were estimated to capture the cortisol diurnal rhythm by estimating the mean awakening cortisol level (intercept) and the change that took place from awakening to bedtime (slope). An unspecified curve model was chosen to accommodate slightly varying assessment times between individuals without assuming a specific shape of change across individuals (Duncan et al., 1997). The model contained fixed and random estimates, corresponding to the parameters’ mean and variance between individuals (see Ouellet-Morin et al. (2016)). Significant confounders (awakening time, hours of sleep, sleeping problems, exercises and alcohol or drug consumption) were accounted for in these analyses.

Hair cortisol. Hair samples collected at age 19 were washed and steroids were extracted according to a previously validated protocol (Kirschbaum et al., 2009). The range of detection for this assay was between 005-4 μg/dL and the intra-and inter-assay coefficients of variation were 5.54 and 18.74, respectively. All samples were assayed in duplicates and averaged. Nine participants were discarded because of unusually high scores (the highest outlier greater than 14 SD). The remaining (1,6%, n=13) were winsorized to 3 SD and the cortisol variable was natural logarithmically transformed. Several confounders (body mass index, hair wash frequency, anxiolytic medication use, as well as cocaine and ecstasy consumption in the last three months) were uniquely related to HCC. Standardized residuals were computed to account for these potential confounders.

Statistical analyses

Preliminary analyses indicated that diurnal cortisol data were not missing completely at random [$\chi$²(7)=22.20, p=.002] and that DZ twins and those reporting more severe depressive symptoms at age 13 years were less likely to have participated in saliva collection (awakening cortisol [$\chi$²(2)=45.65, p=.001], CAR [$\chi$²(2)=36.42, p=.001], diurnal slope [$\chi$²(2)=46.69, p=.001]). HCC data were found to be missing completely at random [χ²=20.60, df=19, p=.36]. Moreover, significant mean differences in social support between boys and girls were detected at mid-adolescence [t(798)=-7.70, p<.001; boys: Mean=2.51; girls: Mean=2.95] and for the adolescence aggregated index [t(1011)=-11.36, p<.001; boys: Mean=2.59; girls: Mean=3.08]. Sex effects were controlled for in subsequent analyses.

To account for the hierarchical structure of the data (i.e., twins clustered within families), we conducted multilevel regression analyses using a full maximum-likelihood fit function in Mplus, version 8.1.7. The main analyses were conducted in two steps. First, we specified a null model in which variation in cortisol was partitioned into within (individual-level) and between (family-level) components. This model contained one fixed-effect (intercept) and one random-effect (variation in intercepts across families) and allowed us to estimate the intra-class correlation (ICC). The ICC represents the proportion of variance of cortisol that is explained by the grouping variable (i.e., family; Heck et al. (2013)). Higher ICCs indicate greater variability between families, thereby suggesting that the adoption of a multilevel approach is warranted. Second, the unique effects of the predictors as well as their joint (interaction) effects were introduced sequentially in the models as fixed effects, while still accounting for between-family variation. Of note, the mid-adolescence social support scale was included as an independent variable in the diurnal cortisol models (all variables measured at age 14), whereas the aggregated index of social support (averaged 14 and 19 years) was included as a predictor of HCC models (age 19). For ease of interpretation, all variables were Z-standardized before the analyses.

Results

Prospective and concurrent associations between the main study variables

As reported elsewhere using the same cohort (Cantave et al., 2021), a moderate-to-strong association was observed between early childhood and mid-adolescence SES, indicating stability but also changes in youth’s family SES across these developmental periods. As shown in Table 1, childhood SES did not predict social support, nor diurnal cortisol or HCC. In contrast, mid-adolescence SES significantly covaried with mid-adolescence social support and the aggregated index of adolescence social support. This suggests that youth from wealthier households reported receiving more emotional and instrumental support from their social network. As reported previously within the same cohort (Cantave et al., 2021), adolescents living in higher-SES families exhibited higher awakening cortisol levels and a steeper decreasing diurnal slope. No other associations were detected between SES and cortisol indicators. Moreover, youth reporting higher social support in mid-adolescence had higher awakening cortisol levels. While the diurnal cortisol indicators all covaried, none were associated with HCC.

[ INSERT TABLE 1 ]

Independent and joint contributions of early childhood and mid-adolescence SES to cortisol

Model 1 (Table 2) presents a significant variation in the CAR accounted for by both between-and within-family differences. The ICC indicated that around 25% of the total variation in the CAR is accounted for by differences between families, while the remaining 75% is explained by differences within families (i.e., within twin pairs). Similar findings were observed for the other cortisol indicators, thereby indicating that the adoption of a multilevel approach is warranted.

Next, we tested the additive contribution of early childhood and mid-adolescence SES to cortisol by including these variables as fixed effects (i.e., within-family differences, Model 2). After adjusting for sex, between-family effects (and awakening cortisol for diurnal slope models), early childhood and mid-adolescence SES were not uniquely associated with the diurnal cortisol indicators. In contrast, lower mid-adolescence SES predicted higher chronic (i.e., hair) cortisol secretion at age 19.

The interaction term between both SES indicators included in model 3 showed that the strength of the association linking mid-adolescence SES to HCC significantly varied as a function of early childhood SES levels. Simple slope analyses revealed that lower mid-adolescence SES predicted higher HCC for twins who were raised in lower (-1 SD) [b=-.25(SE=.07), p=.001] and moderate (average level) [b=-.12(SE=.05), p=.02] SES households during early childhood, whereas this association was not significant for those from higher-SES families (+1 SD) during early childhood [b=.01(SE=.07), p=.87]. To further illustrate this moderation, the conditional effect of mid-adolescence SES on HCC was plotted at lower (-1 SD), moderate (mean) and higher (+1 SD) levels of early childhood SES (see Figure 1, Panel A). Among youth who grew up in lower-SES households during early childhood, those who continued to live in disadvantage families at age 14 showed higher HCC at age 19 in comparison to those for whom family SES increased (to moderate or higher) SES during mid-adolescence (i.e., upward social mobility). Among those raised in the sample’s average SES in early childhood, youth who experienced downward social mobility (i.e., lower mid-adolescence SES) also had higher HCC than those for whom family SES did not change or increased. Notably, no differences in HCC were detected among youths living in lower-, moderate-, or higher-SES families during mid-adolescence for those who were raised in higher-SES households during early childhood. This result suggests that growing up in favourable socioeconomic circumstances during early childhood shield children from the adverse impact of downward socioeconomic mobility during mid-adolescence on later HCC. Alternatively, this interaction may signal that lower-SES conditions in early childhood may sensitized HCC levels in the face of increasing socioeconomic adversity.

We observed a similar interaction between early childhood and mid-adolescence SES for cortisol diurnal change measured at age 14 (see Table 2, Model 3). Simple slope analyses and Figure 1 (Panel B) revealed a pattern of findings that was partially consistent with the one found for HCC. Specifically, a flatter diurnal slope was noted in youth growing up in lower-SES families in mid-adolescence, but only among those who experienced socioeconomic disadvantage in early childhood (b=-.15(SE=.07), p=.03).

[ INSERT TABLE 1 ]

[INSERT FIGURE 1]

Adolescence social support as moderator of the associations between SES and cortisol

Table 3 shows that the bivariate association previously observed between mid-adolescence social support and awakening cortisol remained significant after adjusting for SES, sex and between-family effects (see Model 2). Moreover, an interaction emerged between mid-adolescence SES and social support (Table 3, Model 3). As depicted in Figure 2 (Panel A), youth who were living in lower-SES households in mid-adolescence had lower awakening cortisol levels, but only if they reported higher levels (+1 SD) of social support [b=.19(SE=.08), p=.01]. In contrast, no significant association was detected between mid-adolescence SES and awakening levels for youth who reported lower (-1 SD) [b=.01(SE=.08), p=.98] or moderate (i.e., average) levels of support [b=.10(SE=.06), p=.12].

A significant interaction was also noted between early childhood SES and mid-adolescence social support predicting the CAR (Table 3, Model 3). Simple slope analyses and Figure 2 (Panel B) revealed that youth raised in lower SES households during early childhood evidenced a lower CAR, but this association was only significant for those reporting higher social support in mid-adolescence (+1 SD [b=.19(SE=.09), p=.04]). Social support did not moderate the associations between SES and other cortisol indicators. Of note, models including only age 19 social support (measured concurrently with HCC) as the moderator led to similar findings as those performed with the aggregated index of social support (see Supporting information). Also, considering that SES is measured at the family-level, we further explored the associations between SES and cortisol or social support at the family-level to complement the analyses presented at the individual-level. The findings remain unchanged (see Table S2 and S3 in Supporting information).

[ INSERT TABLE 3 ]

[ INSERT FIGURE 2 ]

Discussion

This prospective study assessed the sensitization role of early childhood SES on the association linking mid-adolescence SES to multiple cortisol indicators and evaluated whether social support mitigated these associations. Our results revealed that youth living in lower-SES households at age 14 had higher HCC at age 19. This is consistent with past studies reporting higher HCC in low-SES children, but that did not distinguish the contribution of earlier vs. later family socioeconomic disadvantage in that association (Bryson et al., 2021; Kao et al., 2019; Merz et al., 2019; Tarullo et al., 2020). The repeated assessments of family SES in the current study enabled to formally test the impact of stability and change in SES in predicting cortisol secretion. Youth raised in more impoverished households during early childhood and who were still exposed to lower family SES in mid-adolescence (i.e., chronic exposure) had higher HCC at age 19 years and a trend for flatter diurnal slope at age 14 years. In contrast, mid-adolescence SES did not relate to these cortisol indicators among youth who grew up in wealthier families in early childhood. Altogether, these findings are in line with the stress-sensitization hypothesis, which proposes that early adversity may enhance HPA axis sensitivity to stress experience later in time (Daskalakis et al., 2013; Young et al., 2019). Furthermore, we found that youth raised in lower-SES families during childhood but who experienced upward social mobility during mid-adolescence had lower HCC and a more dynamic diurnal slope than those who experienced prolonged socioeconomic disadvantage. This finding echoes another feature of the stress-sensitization hypothesis; future stress ought to be present to trigger the embedded diathesis brought about by early socioeconomic disadvantage. Contrastingly, youth who were exposed to downward social mobility (i.e., from the sample’s average early childhood SES to lower mid-adolescence SES) also showed higher HCC. This is consistent with previous findings showing that women who experienced a deterioration of income during a 4-year period had higher HCC in comparison to those who experienced no change or an improvement in income (Serwinski et al., 2016). Our study extends this prior evidence by showing that downward social mobility is not predictive of age 19 HCC among children who grew up in wealthier families in early childhood. This finding points to the long-term protective effect associated with growing up in a higher-SES context in early life and strengthens calls for fortifying the socioeconomic welfare of young families.

It is noteworthy that the synergistic effect of early childhood and mid-adolescence SES was only detected for HCC, and to some extent, the diurnal slope. These indicators both reflect patterns of cortisol secretion during an extended period in comparison to the awakening cortisol and the CAR (Koss and Gunnar, 2018; Stalder et al., 2017). Similarly to others (Malanchini et al., 2020), we speculate that the diurnal slope and HCC may better capture exposure to chronic adversity and prolonged HPA axis adaptation to it. More generally, these results underline the importance of investigating complementary indicators of HPA axis activity to provide much-needed insights into the pathways by which early adversity may similarly or differentially disrupts HPA axis activity.

We found that youth raised in higher-SES families in early childhood or concurrently living in higher-SES contexts in mid-adolescence had a higher CAR and higher awakening cortisol levels when they perceived higher levels of social support. Contrastingly, the cortisol levels of youth from lower-SES backgrounds did not appear to be buffered by social support, which call in question the stress-buffering hypothesis. The beneficial effect of perceived social support on adjustment has been shown to be reduced in highly stressful environments such as lower-SES households, as social support also tends to co-occur with stress contagion, negative interactions and reciprocal obligations, which may lead to higher stress (Belle, 1983; Cattell, 2001; Moskowitz et al., 2013; Stringhini et al., 2012; Tigges et al., 1998). Nonetheless, in line with the biological sensitivity to context theory (Boyce and Ellis, 2005) and the adaptive calibration model (Del Giudice et al., 2011), the higher CAR and awakening cortisol level detected in youth from wealthier families who reported higher social support may also reflect a heightened sensitivity to environmental stimuli. This may enable youths to benefit more from these social resources. In contrast, the lower CAR and awakening cortisol level observed among lower-SES youth –regardless of their levels of social support– might be indicative of a reduced sensitivity to expected stress likely to occur in socioeconomic less advantageous contexts (Boyce and Ellis, 2005; Del Giudice et al., 2011; Evans, 2004). More research is needed to examine whether this constellation of higher awakening cortisol and CAR, higher-SES, and social support promote vulnerable or resilient socioemotional and behavioral functioning.

Strengths and Limitations

This study had several strengths, including its large sample size, the assessment of multiple indicators of cortisol, as well as the access to repeated measurement of family SES at key developmental periods (early years and adolescence). It has also some limitations. First, given that its sample is largely composed of Whites from middle-to-higher SES families, with only about a quarter of families reporting an income less than CAN$30K (≈US$24K), our findings may not readily generalize to more diverse or socioeconomically deprived populations. Second, perceived social support and cortisol secretion were only measured during adolescence, limiting our ability to establish the temporal sequence of family SES-cortisol associations. Future studies with repeated measures of SES, cortisol, and social support from preschool years to adolescence will provide additional insights into how these associations unfold over time. Third, noncompliance to the saliva collection protocol was verified through written records provided by the participants instead of relying on information drawn from electronic devices. Consequently, our CAR measurement does not fully adhere to current guidelines (Stalder et al., 2022). Nevertheless, based on the participants’ reported sampling times, most complied with the protocol. The remaining small differences were accounted for in our analyses to minimize potential bias.

Conclusion

Using a prospective study design, this study found evidence that early childhood family SES exacerbated the association between mid-adolescence SES and HCC and, to a lesser extent, its association with the diurnal slope, suggesting a sensitization effect of early socioeconomical adversity to subsequent one. Furthermore, higher perceived social support magnified the association between mid-adolescence family SES and awakening cortisol and the CAR in youth growing up in wealthier families. Collectively, these findings highlight the relevance of examining patterns of stability and change in youths’ adverse experiences to better understand the complex and protracted temporal dynamics that may disrupt HPA axis activity.

Acknowledgments

Funding for this study was provided by the Social Sciences and Humanities Research Council of Canada and the Canadian Institutes of Health Research (grant number: MOP 142350). Mara Brendgen is supported by the Fonds de Recherche du Québec-Santé (FRQS), Sonia J Lupien is a Canada Research Chair in Human Stress, Michel Boivin is a Canada Research Chair in Child Development, and Isabelle Ouellet-Morin is a Canada Research Chair in the Developmental Origins of Vulnerability and Resilience. We thank Alain Girard for his help in the analyses and Marie-Elyse Bertrand for coordinating the data collection and Hélène Paradis for data management and preparation. We also thank the twins and their families as well as their classmates for participating in this study. The first author confirm that she had full access to all the data in the present study and takes responsibility for the integrity of data and accuracy of the data analysis. The authors have declared no competing or potential conflicts of interest. Statements concerning informed consent and ethical approval are available at pages 6-7.

Corresponding Author

Isabelle Ouellet-Morin, Ph.D., School of Criminology, University of Montreal, Research Center of the Montreal Mental Health University Institute & the Research Group on Child Maladjustment | C.P. 6128, succursale Centre-ville, Montréal QC, H3C 3J7, Canada | Tel : (001) 514-343-6111 ext. 6191| Fax: 514-343-5650 | [email protected]

Key points

• Youth raised in chronically deprived households had a flatter diurnal slope at age 14 and higher HCC at age 19.

• Lower early childhood SES exacerbated the association between adolescence SES and cortisol indicators, whereas higher early childhood SES had a protective effect.

• Higher-SES youth who perceived higher levels of social support had higher awakening cortisol and CAR levels.

• Depicting stability and changes in SES will help to refine the complex association between socioeconomic adversity and HPA axis activity.

References

Boivin, M., Brendgen, M., Dionne, G., Ouellet-Morin, I., Dubois, L., Pérusse, D., Robaey, P., Tremblay, R. E., & Vitaro, F. (2019). The Quebec Newborn Twin Study at 21. Twin Research and Human Genetics, 22(6), 475–481. https://doi.org/10.1017/thg.2019.74

Boyce, W. T., & Ellis, B. J. (2005). Biological sensitivity to context: I. An evolutionary–developmental theory of the origins and functions of stress reactivity. Development and Psychopathology, 17(2), 271–301. https://doi.org/10.1017/S0954579405050145

Bradley, R. H., & Corwyn, R. F. (2002). Socioeconomic status and child development. Annual Review of Psychology, 53(1), 371–399. https://doi.org/10.1146/annurev.psych.53.100901.135233

Bryson, H. E., Goldfeld, S., Price, A. M. H., & Mensah, F. (2019). Hair cortisol as a measure of the stress response to social adversity in young children. Developmental Psychobiology, 61(4), 525–542. https://doi.org/10.1002/dev.21840

Bryson, H. E., Price, A. M., Goldfeld, S., & Mensah, F. (2021). Associations between social adversity and young children’s hair cortisol: A systematic review. Psychoneuroendocrinology, 127, 105176. https://doi.org/10.1016/j.psyneuen.2021.105176

Cantave, C. Y., Brendgen, M., Lupien, S., Dionne, G., Vitaro, F., Boivin, M., & Ouellet-Morin, I. (2022). Association between the timing of family socioeconomic deprivation and adolescence hair cortisol among adolescent twins: A study of the genetic and environmental processes involved. Psychoneuroendocrinology, 144, 105889. https://doi.org/10.1016/j.psyneuen.2022.105889

Cantave, C. Y., Brendgen, M., Paquin, S., Lupien, S., Dionne, G., Vitaro, F., Boivin, M., & Ouellet-Morin, I. (2021). The phenotypic associations and gene–environment underpinnings of socioeconomic status and diurnal cortisol secretion in adolescence. Development and Psychopathology, 1–11. https://doi.org/10.1017/S0954579421001048

Chen, E., Cohen, S., & Miller, G. E. (2010). How Low Socioeconomic Status Affects 2-Year Hormonal Trajectories in Children. Psychological Science, 21(1), 31–37. https://doi.org/10.1177/0956797609355566

Cohen, S., & Wills, T. A. (1986). Stress, social support, and the buffering hypothesis. Psychological Bulletin, 98(2), 310. https://doi.org/10.1037/0033-2909.98.2.310

Daskalakis, N. P., Bagot, R. C., Parker, K. J., Vinkers, C. H., & de Kloet, E. R. (2013). The three-hit concept of vulnerability and resilience: Toward understanding adaptation to early-life adversity outcome. Psychoneuroendocrinology, 38(9), 1858–1873.

Del Giudice, M., Ellis, B. J., & Shirtcliff, E. A. (2011). The Adaptive Calibration Model of stress responsivity. Neuroscience & Biobehavioral Reviews, 35(7), 1562–1592. https://doi.org/10.1016/j.neubiorev.2010.11.007

Desantis, A. S., Kuzawa, C. W., & Adam, E. K. (2015). Developmental origins of flatter cortisol rhythms: Socioeconomic status and adult cortisol activity. American Journal of Human Biology, 27(4), 458–467. https://doi.org/10.1002/ajhb.22668

Duncan, T. E., Duncan, S. C., Alpert, A., Hops, H., Stoolmiller, M., & Muthen, B. (1997). Latent Variable Modeling of Longitudinal and Multilevel Substance Use Data. Multivariate Behavioral Research, 32(3), 275–318. https://doi.org/10.1207/s15327906mbr3203_3

Evans, G. W. (2004). The Environment of Childhood Poverty. American Psychologist, 59(2), 77. https://doi.org/10.1037/0003-066X.59.2.77

Evans, G. W., & English, K. (2002). The Environment of Poverty: Multiple Stressor Exposure, Psychophysiological Stress, and Socioemotional Adjustment. Child Development, 73(4), 1238–1248. https://doi.org/10.1111/1467-8624.00469

Furman, W., & Buhrmester, D. (1985). Children’s perceptions of the personal relationships in their social networks. Developmental Psychology, 1016–1024.

Gray, N. A., Dhana, A., Van Der Vyver, L., Van Wyk, J., Khumalo, N. P., & Stein, D. J. (2018). Determinants of hair cortisol concentration in children: A systematic review. Psychoneuroendocrinology, 87, 204–214. https://doi.org/10.1016/j.psyneuen.2017.10.022

Gunnar, M. R., DePasquale, C. E., Reid, B. M., Donzella, B., & Miller, B. S. (2019). Pubertal stress recalibration reverses the effects of early life stress in postinstitutionalized children. Proceedings of the National Academy of Sciences, 116(48), 23984–23988. https://doi.org/10.1073/pnas.1909699116

Heck, R. H., Thomas, S. L., & Tabata, L. N. (2013). Multilevel and Longitudinal Modeling with IBM SPSS. Routledge. https://doi.org/10.4324/9780203701249

Hooker, E. D., Campos, B., Hoffman, L., Zoccola, P., & Dickerson, S. S. (2020). Is Receiving Social Support Costly for Those Higher in Subjective Socioeconomic Status? International Journal of Behavioral Medicine, 27(3), 325–336. https://doi.org/10.1007/s12529-019-09836-w

Hooker, E. D., Campos, B., Zoccola, P. M., & Dickerson, S. S. (2018). Subjective Socioeconomic Status Matters Less When Perceived Social Support Is High: A Study of Cortisol Responses to Stress. Social Psychological and Personality Science, 9(8), 981–989. https://doi.org/10.1177/1948550617732387

Hostinar, C. E., & Gunnar, M. R. (2015). Social Support Can Buffer Against Stress and Shape Brain Activity. AJOB Neuroscience, 6(3), 34–42. https://doi.org/10.1080/21507740.2015.1047054

Kao, K., Tuladhar, C. T., Meyer, J. S., & Tarullo, A. R. (2019). Emotion regulation moderates the association between parent and child hair cortisol concentrations. Developmental Psychobiology, 61(7), 1064–1078. https://doi.org/10.1002/dev.21850

Kim, P., Evans, G. W., Chen, E., Miller, G., & Seeman, T. (2018). How Socioeconomic Disadvantages Get Under the Skin and into the Brain to Influence Health Development Across the Lifespan. In Handbook of Life Course Health Development (pp. 463–497). Springer, Cham. https://doi.org/10.1007/978-3-319-47143-3_19

Kirschbaum, C., Tietze, A., Skoluda, N., & Dettenborn, L. (2009). Hair as a retrospective calendar of cortisol production—Increased cortisol incorporation into hair in the third trimester of pregnancy. Psychoneuroendocrinology, 34(1), 32–37. https://doi.org/10.1016/j.psyneuen.2008.08.024

Koss, K. J., & Gunnar, M. R. (2018). Annual Research Review: Early adversity, the hypothalamic–pituitary–adrenocortical axis, and child psychopathology. Journal of Child Psychology and Psychiatry, 59(4), 327–346. https://doi.org/10.1111/jcpp.12784

Lupien, S. J., King, S., Meaney, M. J., & Mcewen, B. S. (2001). Can poverty get under your skin? Basal cortisol levels and cognitive function in children from low and high socioeconomic status. Development and Psychopathology, 13(3), 653–676. https://doi.org/10.1017/S0954579401003133

Malanchini, M., Engelhardt, L. E., Raffington, L. A., Sabhlok, A., Grotzinger, A. D., Briley, D. A., Madole, J. W., Freis, S. M., Patterson, M. W., Harden, K. P., & Tucker-Drob, E. M. (2020). Weak and uneven associations of home, neighborhood, and school environments with stress hormone output across multiple timescales. Molecular Psychiatry. https://doi.org/10.1038/s41380-020-0747-z

McEwen, B. S., & Seeman, T. (1999). Protective and Damaging Effects of Mediators of Stress: Elaborating and Testing the Concepts of Allostasis and Allostatic Load. Annals of the New York Academy of Sciences, 896(1), 30–47. https://doi.org/10.1111/j.1749-6632.1999.tb08103.x

McLachlan, K., Rasmussen, C., Oberlander, T. F., Loock, C., Pei, J., Andrew, G., Reynolds, J., & Weinberg, J. (2016). Dysregulation of the cortisol diurnal rhythm following prenatal alcohol exposure and early life adversity. Alcohol (Fayetteville, N.Y.), 53, 9–18. https://doi.org/10.1016/j.alcohol.2016.03.003

Merz, E. C., Desai, P. M., Maskus, E. A., Melvin, S. A., Rehman, R., Torres, S. D., Meyer, J., He, X., & Noble, K. G. (2019). Socioeconomic Disparities in Chronic Physiologic Stress Are Associated With Brain Structure in Children. Biological Psychiatry, 86(12), 921–929. https://doi.org/10.1016/j.biopsych.2019.05.024

Ouellet-Morin, I., Brendgen, M., Girard, A., Lupien, S. J., Dionne, G., Vitaro, F., & Boivin, M. (2016). Evidence of a unique and common genetic etiology between the CAR and the remaining part of the diurnal cycle: A study of 14 year-old twins. Psychoneuroendocrinology, 66(Supplement C), 91–100. https://doi.org/10.1016/j.psyneuen.2015.12.022

Ouellet-Morin, I., Laurin, M., Robitaille, M.-P., Brendgen, M., Lupien, S. J., Boivin, M., & Vitaro, F. (2016). Validation of an adapted procedure to collect hair for cortisol determination in adolescents. Psychoneuroendocrinology, 70, 58–62. https://doi.org/10.1016/j.psyneuen.2016.05.002

Serwinski, B., Salavecz, G., Kirschbaum, C., & Steptoe, A. (2016). Associations between hair cortisol concentration, income, income dynamics and status incongruity in healthy middle-aged women. Psychoneuroendocrinology, 67, 182–188. https://doi.org/10.1016/j.psyneuen.2016.02.008

Shonkoff, J. P. (2010). Building a new biodevelopmental framework to guide the future of early childhood policy. Child Development, 81(1), 357–367. https://doi.org/10.1111/j.1467-8624.2009.01399.x

Stalder, T., Lupien, S. J., Kudielka, B. M., Adam, E. K., Pruessner, J. C., Wüst, S., Dockray, S., Smyth, N., Evans, P., Kirschbaum, C., Miller, R., Wetherell, M. A., Finke, J. B., Klucken, T., & Clow, A. (2022). Evaluation and update of the expert consensus guidelines for the assessment of the cortisol awakening response (CAR). Psychoneuroendocrinology, 146, 105946. https://doi.org/10.1016/j.psyneuen.2022.105946

Stalder, T., Steudte-Schmiedgen, S., Alexander, N., Klucken, T., Vater, A., Wichmann, S., Kirschbaum, C., & Miller, R. (2017). Stress-related and basic determinants of hair cortisol in humans: A meta-analysis. Psychoneuroendocrinology, 77, 261–274. https://doi.org/10.1016/j.psyneuen.2016.12.017

Tarullo, A. R., Tuladhar, C. T., Kao, K., Drury, E. B., & Meyer, J. (2020). Cortisol and socioeconomic status in early childhood: A multidimensional assessment. Development and Psychopathology, 32(5), 1876–1887. https://doi.org/10.1017/S0954579420001315

Young, E. S., Doom, J. R., Farrell, A. K., Carlson, E. A., Englund, M. M., Miller, G. E., Gunnar, M. R., Roisman, G. I., & Simpson, J. A. (2020). Life stress and cortisol reactivity: An exploratory analysis of the effects of stress exposure across life on HPA-axis functioning. Development and Psychopathology, 1–12. https://doi.org/10.1017/S0954579419001779

Young, E. S., Farrell, A. K., Carlson, E. A., Englund, M. M., Miller, G. E., Gunnar, M. R., Roisman, G. I., & Simpson, J. A. (2019). The Dual Impact of Early and Concurrent Life Stress on Adults’ Diurnal Cortisol Patterns: A Prospective Study. Psychological Science, 30(5), 739–747. https://doi.org/10.1177/0956797619833664

Tables

Table 1. Linear correlation estimates between SES, social support and salivary and hair cortisol indicators

Note. ***p$\leq$.001, **= p$\leq$.01, *= p$\leq$.05. CAR = Cortisol awakening response; HCC = Hair cortisol concentration.

Table 2. Multilevel regression models linking early childhood and mid-adolescence SES to the salivary and hair cortisol indicators

Note. SES=socioeconomic status, CAR=Cortisol awakening response; HCC=Hair cortisol concentration, $\beta$=Standardized regression coefficient, SE=Standard error, Df=Degree of freedom, ICC=Intra-class correlation. Significant parameters are indicated in boldface. Due to space limitations, the between-family variation parameters were not included in Models 2 and 3 notations.

Table 3. Multilevel regression models linking early childhood, mid-adolescence SES and social support to the cortisol indicators

Note. SES=socioeconomic status, CAR=Cortisol awakening response; HCC=Hair cortisol concentration, $\beta$=Standardized regression coefficient, SE=Standard error, Df=Degree of freedom, ICC=Intra-class correlation. The mid-adolescence social support scale was included in diurnal cortisol models, while the aggregated index of social support (14- and 19-years average) was included in HCC models to account for the measures’ timeline. Significant estimates are indicated in boldface. Due to space limitations, the between-family variation parameters were not included to Models 2 and 3 notations.

Figures

Figure 1. Associations between mid-adolescence SES and HCC (Panel A) and (salivary) diurnal cortisol (Panel B) according to early childhood SES.

Note. **= p$\leq$.01, *= p$\leq$.05. The graphs indicate at which levels of early childhood SES (Lower = -1 SD; Moderate =mean; Higher =+1 SD) the associations linking mid-adolescence SES to diurnal and hair cortisol were significant.

Figure 2. Association of early childhood SES with the CAR (Panel A) and of mid-adolescence SES with awakening cortisol levels (Panel B) according to mid-adolescence social support (age 14).

Note. ***= p$\leq$.001,**= p$\leq$.01,*= p$\leq$.05. Panel A indicates that the association between mid-adolescence SES and awakening cortisol was only significant at higher levels of mid-adolescence social support (+1 SD). The graph also revealed that perceived social support is associated with awakening cortisol, but only among youth from moderate(mean)-to-higher (+1 SD) SES families. Panel B indicates at which levels of mid-adolescence social support (Lower = -1 SD; Moderate =mean; Higher =+1 SD) the association linking early childhood SES to the CAR was significant.

Supplementary Material

Figure S1. Overview of the timeline of the main study variables

Note. Childhood SES was computed using the parents’ highest educational level (3 assessments at 5, 30 and 60 months, respectively) and family income (4 assessments at 5, 18, 30 and 48 months) during the preschool years. Diurnal salivary cortisol was derived from saliva samples collected at four time points during the day (awakening, 30 min later, late afternoon and bedtime) on four collection days (Tuesdays and Thursdays over two consecutive weeks). The remaining measures were assessed once, as indicated.

Table S1. Multilevel regression models linking early childhood, mid-adolescence SES and age 19 social support to hair cortisol

Note. SES=socioeconomic status, CAR=Cortisol awakening response; HCC=Hair cortisol concentration, $\beta$=Standardized regression coefficient, SE=Standard error, Df=Degree of freedom, ICC=Intra-class correlation. Significant estimates are indicated in boldface. Due to space limitations, the between-family variation parameters were not included to Models 2 and 3 notations.

Table S2. Multilevel regression models linking early childhood and mid-adolescence SES to the salivary and hair cortisol indicators

Note. SES=socioeconomic status, CAR=Cortisol awakening response; HCC=Hair cortisol concentration, $\beta$=Standardized regression coefficient, SE=Standard error, Df=Degree of freedom, ICC=Intra-class correlation. Significant parameters are indicated in boldface. Due to space limitations, the between-family variation parameters were not included in Models 2 and 3 notations. Grand mean-centered family SES variables were included at the between-level. a = simple slopes analyses indicated that family-level variation in mid-adolescence SES was significantly related to family-level variation in the diurnal slope, but only for those who reported lower SES during early childhood (i.e., -1 SD [b=-.35(SE=.15), p=.02]). This association was thus not significant for families who reported average (i.e., mean [b=-.08(SE=.07), p=.27]) to higher (i.e., +1 SD [b=.19(SE=.16), p=.23]) SES in early childhood. b = simple slopes analyses revealed that the association between family-level variation in mid-adolescence SES and family-level variation in HCC was significant within families who reported lower (i.e., -1 SD [b=-.50(SE=.16), p=.001]) and average (i.e., mean [b=-.17(SE=.07), p=.02]) SES during early childhood. This was not the case for those who reported higher SES (i.e., +1 SD [b=.17(SE=.15), p=.28]) during this developmental period.

Table S3. Multilevel regression models linking early childhood and mid-adolescence SES to social support

Note. SES=socioeconomic status, CAR=Cortisol awakening response; HCC=Hair cortisol concentration, $\beta$=Standardized regression coefficient, SE=Standard error, Df=Degree of freedom, ICC=Intra-class correlation. Significant parameters are indicated in boldface. Due to space limitations, the between-family variation parameters were not included in Models 2 and 3 notations. Grand mean-centered family SES variables were included at the between-level.

Supplementary analyses: associations of mid-adolescence SES with awakening cortisol according to perceived social support from specific providers

Additional analyses were carried out to test whether perceived social support from distinct providers modulated the association linking mid-adolescence SES to awakening cortisol. A significant interaction was observed between mid-adolescence SES and perceived social support from the cotwin [$\beta$=.11(SE=.05), p=.03], while it reached a trend level for perceived social support from the mother [$\beta$=.10(SE=.05), p=.052] and the father [$\beta$=.11(SE=.06), p=.06] respectively. Specifically, youth from higher SES families showed higher awakening cortisol only if they reported higher social support from either their mother [b=.20(SE=.07), p=.01], father [b=.23(SE=.09), p=.01] or cotwin [b=.22(SE=.08), p=.01]. This association was, however, not significant for those who reported moderate (Mother: [b=.09(SE=.06), p=.14]; Father: [b=.12(SE=.06), p=.06], cotwin: [b=.10(SE=.06), p=.10]) to lower support (mother: [b=-.01(SE=.09), p=.91]; father: [b=.01(SE=.09), p=.87], cotwin: [b=-.01(SE=.08), p=.92]) from these providers. While higher social support from youth’s friend [$\beta$=.11(SE=.05), p=.03] and teacher [$\beta$=.15(SE=.05), p=.01] were significantly related to higher awakening cortisol, they were not found to moderate the mid-adolescence SES-awakening cortisol association (friend: [$\beta$=.04(SE=.05), p=.50]; teacher: [$\beta$=.03(SE=.07), p=.67].

Similar analyses were performed for the CAR, which revealed a significant interaction between early childhood SES and mid-adolescence perceived social support from the mother [$\beta$=.14(SE=.06), p=.01]. Youth raised in higher SES families during early childhood had a higher CAR in mid-adolescence only if they reported higher social support from their mother [b=.22(SE=.08), p=.01]. The association was not otherwise conditional on youth’s perceived social support from their father [$\beta$=.11(SE=.06), p=.08], friend [$\beta$=.01(SE=.05), p=.79], teacher [$\beta$=.01(SE=.10), p=.98] and cotwin [$\beta$=.09(SE=.08), p=.25].