Childhood Trauma and Hair Cortisol Response over the Year following Onset of a Chronic Life Event Stressor

Abstract

Objective:

Childhood trauma may contribute to poor lifelong health in part through programming of the HPA-axis response to future life stressors. To date, empirical evidence shows an association of childhood trauma with dysregulation of the HPA-axis and blunted cortisol reactivity to acute stressors. Here, we conduct an initial examination of childhood trauma as a moderator of changes over time in perceived stress levels and HPA-axis response to a major chronic stressor in adulthood.

Methods:

Participants were 83 maternal caregivers of children newly diagnosed with cancer who completed the Childhood Trauma Questionnaire (CTQ), and who, over the year following their child’s cancer diagnosis, had hair samples collected up to 7 times for the assessment of cortisol and completed monthly measures of perceived stress.

Results:

CTQ scores were in the expected range for a community sample and associated with changes in perceived stress and cortisol concentration over time (γ = .003, p = .002; γ = −.0004, p = .008, respectively) independently of age, education, treatment intensity and randomization to stress management intervention. Maternal caregivers who endorsed lower childhood trauma showed a steeper decline in perceived stress and a larger increase in cortisol levels across the year than caregivers who recalled more childhood trauma.

Conclusions:

Findings extend animal models and studies that examine cortisol reactivity to acute stressors and suggest that childhood trauma may program a phenotype that is more psychologically reactive but shows a blunted HPA-axis response to chronic stress. While adaptive in the short-term, this early life programming may incur long-term costs for health. Further work is warranted to examine this possibility.

1.1. Introduction

Evidence links childhood adversity to increased lifelong risk for mental and physical health problems and earlier mortality (Miller, Chen, & Parker, 2011; Norman et al., 2012; Shonkoff, Boyce, & McEwen, 2009). Pathways linking early life experience to future health remain unclear; however, theory and research suggest that the hypothalamic-pituitary- adrenal (HPA) axis (Kuhlman, Chiang, Horn, & Bower, 2017; McEwen, 1998; Sapolsky, Romero, & Munck, 2000) and exposure to psychological stress (Miller et al., 2011) play a role. In this paper, we examine whether recalled childhood adversities are related to the magnitude of HPA-axis response and perceptions of psychological stress following the major life stressor of having a child newly diagnosed with cancer. Specifically, we examine whether childhood adversity predicts changes in caregiver hair cortisol levels and trajectories of subjective stress across the year following the child’s initial cancer diagnosis.

It remains unclear how early life exposures contribute to lifelong health. Of the many pathways that likely contribute, recent theoretical and research interest has focused on the biological embedding of childhood adversity through programming of the HPA-axis (Miller, Chen, & Zhou, 2007). The HPA-axis is a primary pathway linking the central nervous system to the systemic release of hormones that regulate biological, psychological and behavioral responses to environmental demands (Sapolsky et al., 2000). Coordinated activation of these processes is thought to promote an adaptive response to real or perceived life threats (stressors). In response to acute stressors, activation of the HPA-axis cascade results in the peripheral release of cortisol in humans (Sapolsky et al., 2000). However, if the HPA-axis is activated repeatedly or over prolonged periods, as in the context of chronic or repeated stress exposure, or during sensitive periods of development, this pathway can become dysregulated, which may result in reduced cortisol production (Lupien, McEwen, Gunnar, & Heim, 2009; Miller et al., 2007; Sapolsky et al., 2000). Thus, chronic stress can associate with both hyper- and hypo-cortisolism, with each having consequences for negative health outcomes (Chrousos, 2009).

The stress sensitization model (Daskalakis, Lehrner, & Yehuda, 2013) proposes that early life exposures during a critical period when the central nervous and neuroendocrine systems are developing and show heightened plasticity may program lifelong psychological and HPA axis sensitivity to stressors (Bunea, Szentagotai-Tatar, & Miu, 2017; Lupien et al., 2009; McLaughlin et al., 2015; Miller et al., 2011). Adverse environmental exposures during this critical developmental period can result in structural changes to the amygdala, hippocampus, and prefrontal cortex, which are regions of the brain that respond to threat and exert top-down control over activation of the HPA-axis (Lupien et al., 2009; McEwen & Gianaros, 2011; Shonkoff et al., 2012). Thus, it has been proposed that exposure to adversity in childhood results in a phenotype that is threat sensitive and initially shows hyperactivity of the HPA-axis in association with life challenges (Miller et al., 2007; Shonkoff et al., 2012). However, over time, this hyperactivity can result in compensatory down-regulation of glucocorticoid receptor sensitivity, and ultimately to suppressed levels of cortisol and blunted physiological reactivity to stress in later life (Dempster, O’Leary, MacNeil, Hodges, & Wade, 2021; Koss & Gunnar, 2018; McEwen, 1998; Miller et al., 2007). According to the stress sensitization model, the effects of early adversity on cortisol levels in adulthood should only be detectable in the context of high life event stress.

Empirical evidence provides support for both hyper- and hypo-activity of the HPA-axis in association with life stress in adulthood (Miller et al., 2007) and for both heightened and blunted cortisol reactivity to acute stressors (Miller et al., 2007; Stroud, Chen, Doane, & Granger, 2016). In support of the stress sensitization model, exposure to early life adversity appears to contribute to the direction of these effects, being associated with heightened psychological reactivity to life stress (McLaughlin, Conron, Koenen, & Gilman, 2010) and dysregulation of the HPA-axis, typically observed as hypoactivity or null relationships at rest (Bernard, Frost, Bennett, & Lindhiem, 2017) and blunted cortisol reactivity to acute challenge (Bunea et al., 2017; Khoury, Bosquet Enlow, Plamondon, & Lyons-Ruth, 2019; Lupien et al., 2009; Miller et al., 2007; Voellmin et al., 2015; Young et al., 2019). However, to date, no studies have examined whether childhood stress moderates HPA-axis response over time in response to the onset of major life event stress in adulthood.

Most research examining the HPA-axis has assessed cortisol levels in blood, urine, or saliva, providing a method of assessing acute HPA-axis responses to stressors and diurnal rhythms of cortisol release. These measures do not reliably characterize cortisol levels over longer periods, and vary as a function of circadian rhythm and other contemporaneous factors (Meyer & Novak, 2012). Hair cortisol concentration (HCC) offers a more temporally stable measure of HPA-axis activation and is better suited to examining responses to chronic stressors (Kirschbaum, Tietze, Skoluda, & Dettenborn, 2009; Stalder & Kirschbaum, 2012). Evidence supports the validity and reliability of HCC as an index of chronic HPA-axis activation, and as a biological marker of chronic stress (D’Anna-Hernandez, Ross, Natvig, & Laudenslager, 2011; Sauve, Koren, Walsh, Tokmakejian, & Van Uum, 2007; Staufenbiel, Penninx, Spijker, Elzinga, & van Rossum, 2013; Xie et al., 2011). Meta-analytic evidence shows a positive association between HCC and stress exposure in adulthood (Khoury et al., 2019; Stalder et al., 2017); however, the direction of effects varies considerably across individual studies with further evidence for stress-related hypo- and hyper-activation of the HPA-axis. Whether childhood adversity moderates hair cortisol response over time in response to major life stress in adulthood as proposed in the stress sensitization model remains to be established.

1.1.1. The Current Study

The primary goal of the current study was to conduct an initial examination of the stress sensitization model in predicting hair cortisol response across the 12 months following onset of a major life stressor. In accordance with the model and existing evidence, it was hypothesized that childhood adversity would associate with blunted cortisol responses and heightened perceptions of psychological stress in response to having a child diagnosed with cancer. Because data are drawn from a randomized controlled trial assessing the efficacy of a stress management intervention (Marsland et al., 2020), secondary analyses were conducted to explore if changes in perceived stress and hair cortisol over the year were moderated by intervention group.

1.2. Material and Methods

1.2.1. Participants

Data for this secondary analysis were drawn from a randomized controlled trial examining the efficacy of a supportive stress management intervention (ClinicalTrials.gov Identifier: NCT02022449 (Marsland et al., 2020)). Participants were 120 female primary caregivers of a child (0–17 years) newly diagnosed with cancer (mean caregiver age = 36 years; 14% racial/ethnic minority; See Table 1 for sample characteristics) recruited from the Division of Hematology and Oncology, Children’s Hospital of Pittsburgh between 2010 and 2014. Eligibility criteria evaluated by maternal report included (a) no history of psychotic or bipolar illness, neurological disorder, or chronic disease known to impact immune function; (b) not taking medications that might alter responses to questionnaires or indices of immune function (including major sedatives or glucocorticoid, anti-inflammatory, anti-retroviral, or immunosuppressant medication); (c) everyday use of English for >10 years; (d) at least 18 years of age; (e) not working nightshifts exclusively; (f) not pregnant at baseline. The current analyses are based on a subsample of 83 caregivers who completed a measure of childhood adversity that was added after the study started. This subsample was slightly older than the full sample but did not differ on other key variables (see Table 1 for subset and full sample characteristics). Informed consent was obtained in accordance with the University of Pittsburgh Institutional Review Board.

Table 1.

Sample descriptive statistics

	Full sample (n = 120)	Subsample with CTQ Data (n = 83)	Significant group difference
	n (%) or M(SD)	n (%) or M(SD)
Randomized to CTC group	60 (50.0)	42 (50.6)
Randomized to control group	60 (50.0)	41 (49.4)
Age	35.94 (7.95)	36.99 (7.89)	t(118) = 2.21, p = .029
Racial/ethnic minority	17 (14.2)	11 (13.3)
Years of education	13.99 (2.12)	14.19 (2.13)
Child’s cancer treatment intensity*
Least intensive	0 (0)	0 (0)
Moderately intensive	34 (28.6)	26 (31.3)
Very intensive	67 (56.3)	47 (56.6)
Most intensive	18 (15.1)	10 (12.0)
Childhood Trauma (CTQ sum)	-	39.95 (17.28)
Number of hair cortisol samples (max = 10)	4.97 (2.91)	5.25 (2.68)
Number of perceived stress scores (max = 13)	8.84 (3.16)	9.82 (2.30)	t(48.7) = 4.80, p < .001
Mean hair cortisol (pg/mg)	12.84 (10.00)	13.39 (11.02)
Mean hair cortisol (log)	.97 (.30)	.99 (.30)
Mean perceived stress	7.09 (2.50)	6.93 (2.63)

CTQ = Childhood Trauma Questionnaire; M = mean; SD = standard deviation; CTC = Connections to Coping* one participant in the full sample was missing information on their child𠆓s treatment intensity. Hair cortisol range = 1.73 to 82.50 pg/mg.

1.2.2. Procedure

Female caregivers were followed closely over the first year following their child’s diagnosis. Baseline assessment was conducted an average of 1.84 (SD = .82) months after the diagnosis and included completion of questionnaires at home and collection of hair samples in the outpatient clinic or hospital. Following home and clinic assessments, maternal caregivers were randomized to a stress management intervention or a no-treatment control condition. The intervention included six in-person sessions and six telephone check-ins over a ~six-month period. Sessions used a cognitive behavioral framework to provide social support and teach coping skills. The primary study outcomes were 1) intervention feasibility and acceptability, and 2) maternal psychological function and physical health (See (Marsland et al., 2020) for details of the parent project). In-person follow-up assessments were conducted approximately 3-, 6-, 9-, and 12-months post-diagnosis and 2 weeks and 6 months post-intervention. To reduce caregiver burden, these assessments were scheduled around the child’s treatment visits and the absolute timing varied considerably by person. In addition to the in-person assessment visits, caregivers completed monthly questionnaires via email across the 12-month follow-up period. Seventeen caregivers dropped out of the study after the baseline assessment; n = 11 dropped out before the end of the intervention period and an additional n = 6 before the 12-month follow-up visit (dropout rate = 14%).

1.2.3. Measures

1.2.3.1 The Childhood Trauma Questionnaire (CTQ) was administered at the 6-month assessment (mean = 6.81 [SD=1.94] months post diagnosis). It is a retrospective, self-report measure comprised of 28 items (25 clinical and 3 validity) assessing experiences of abuse and neglect prior to age 18 years, including emotional, physical, and sexual abuse and emotional and physical neglect (D. Bernstein & Fink, 1998; D. P. Bernstein et al., 1994). Participants recorded the frequency with which these experiences occurred on a 5-point scale (1=never true to 5=very often true). Responses to the 25 clinical items were summed to create a total score ranging from 25 to 125, with higher scores indicating more childhood trauma. Cronbach’s α in current sample = 0.83.

1.2.3.2. The Perceived Stress Scale (PSS):

The abbreviated PSS was included as part of the monthly assessment that participants were asked to complete via email. This 4-item measure assesses how unpredictable, uncontrollable, and overloaded respondents perceive their lives to have been over the past month (Cohen, Kamarck, & Mermelstein, 1983). Participants recorded how often they have felt or thought a certain way on a five-point scale (0 = never to 4 = very often). Responses were scored to create a total score ranging from 0 to 16, with higher scores reflecting more perceived stress. Mean Cronbach’s α across the 12 assessments = 0.82.

1.2.3.3. Hair Cortisol concentration (HCC).

Hair samples for the assessment of cortisol levels were collected at in-person study visits on up to seven occasions across the 12-month follow-up (baseline, and ~ 3-, 6-, 9-, and 12-months post diagnosis and 2 weeks and 6 months post intervention). Hair samples were collected from a posterior vertex position by cutting close to the scalp. Samples were wrapped in aluminum foil and stored in individual plastic bags in the dark. At the time of analysis, two segments of hair were cut from each sample: 0–2 cm and 2–4 cm from the scalp end. Samples were assayed following the methods described by (Laudenslager, Jorgensen, Grzywa, & Fairbanks, 2011). First, the hair was washed twice in isopropyl alcohol and allowed to dry before being ground in a ball mill (Retsch MM200). At least 50mg of powdered hair was extracted for 24 hours in 1.5mL 100% HPLC grade methanol on a plate shaker. Samples were spun down and 1mL of the extraction medium was dried under a nitrogen stream. Samples were reconstituted in 0.4mL of PBS and cortisol was quantified in pg/mg using an enzyme immunoassay run according to manufacturer’s directions (High sensitivity salivary cortisol EIA # 3002, Salimetrics, LLC PA). All samples were run in duplicate with an average inter-assay coefficient of variation (CV) of 4.74% and intra-assay CV < 10%. In a cross-lab validation, we found high correlations for HCC assays from the same individuals when compared with results from Dr. Laudenslager’s laboratory in Colorado (r = 0.96, p< .001).

Timing of hair collection varied considerably across participants. For this reason, we first organized the date of hair sample collection chronologically within each subject across the 12 months. Hair grows at a rate of ~1 cm/month and we analyzed two segments of hair 0–2 cm and 2–4 cm from the scalp from each sample. To estimate the month of each of these cortisol assessments, one and three months were subtracted from the date of hair collection to provide a measure of HCC across the prior 0–2 and 2–4 months, respectively. This resulted in up to 14 assessments of HCC for each participant. We excluded the two baseline measures because they partially represented levels before diagnosis and organized the remaining up to 12 assessments by month since diagnosis across the 12-month follow-up. This permitted the consideration of time as a continuous variable in our statistical models. HCC was log transformed prior to analysis. We did not observe any biologically implausible HCC values.

1.2.3.4. Time:

Time was measured continuously and reflected the number of months since the child’s cancer diagnosis. HCC and PSS were first ordered chronologically for each participant based on the number of days since the child’s cancer diagnosis. Days were then converted to months, with the first time point reflecting HCC and PSS measured within 1 month of the cancer diagnosis and the last time point reflecting values measured 12 or more months (range 12–15 months) following the cancer diagnosis.

1.2.3.4. Covariates:

Control variables included caregiver age, years of education, intervention condition, and the child’s level of treatment intensity. Treatment intensity was coded from 0 [least] to 4 [most] intensive by a pediatric oncologist using information abstracted from the child’s medical record and the Intensity of Treatment Rating Scale (ITR-3; (Kazak et al., 2012)), which considers diagnosis, stage or risk level, and treatment modality, and has high reliability (r = .90) and content validity (r = .88). We assessed caregiver BMI, but a priori did not include it as a covariate due to its inconsistent association in the literature with systemic levels of cortisol; there were no significant associations of BMI with HCC at any timepoint (p’s>.10).

1.2.4. Statistical analyses

To account for nesting of repeated measures within participants, multilevel models were fit using the lmer package in R (4.2.2) to examine interactions between time (i.e., months since diagnosis) and childhood trauma on perceived stress and HCC over the study period. Prior to fitting multilevel models, independent sample t-tests and chi-squares were used to compare participants with and without childhood trauma data. Next, unconditional means models were fit to calculate intraclass correlations (ICC) to determine the amount of variance in perceived stress and hair cortisol attributed to within- and between-person differences. A model building approach was then used to establish the functional form across time for each outcome variable and tested 1) whether the addition of a random slope improved model fit when using time to predict perceived stress and HCC, comparing the fit of the random intercept only model to the random intercept and slope model using ANOVA; and 2) whether the associations between time and changes in perceived stress and hair cortisol were non-linear by adding the effect of quadratic time (time²) to the models.

Primary analyses examined whether childhood trauma interacted with time to predict perceived stress and HCC and to what extent this pattern held after adjusting for covariates. Significant interactions were probed at one standard deviation above and below the sample mean for childhood trauma. Gamma weights (γ) with their corresponding standard error (SE) and p-values are provided. Gamma weights can be interpreted analogous to unstandardized beta weights in regression.

Secondary analyses tested whether changes in perceived stress and HCC over time were moderated by intervention group.

1.3. Results

1.3.1. Available and Missing Data

Table 1 depicts a summary of sample characteristics. A subsample of 83 participants had available data on childhood trauma (mean = 39.95; range = 25–95). Compared to the full sample, this subsample was slightly older and completed more perceived stress assessments. They did not differ significantly with respect to race/ethnicity, years of education, the child’s cancer treatment intensity, number of HCC samples, mean perceived stress, or mean HCC (p’s >.05). The childhood trauma scores in this sample are in the range typically observed in normative samples in the United States (means typically 35–40; (D. P. Bernstein et al., 2003; Scher, Stein, Asmundson, McCreary, & Forde, 2001). Using validated cutoffs for clinically elevated maltreatment exposure (D. Bernstein & Fink, 1998; Walker et al., 1999), childhood abuse and neglect histories were reported by 55% of the sample (sexual abuse >8= 8%; physical abuse >8 = 19%; emotional abuse > 10 = 37%; emotional neglect > 15 = 24%; physical neglect > 8 = 34%).

Over the study period, participants with childhood trauma data completed an average of 9.81 (range = 3–12) assessments of their perceived stress and had 5.25 (range = 0–10) HCC samples collected. Participants who were missing a larger percentage of HCC samples had higher mean HCC across the study period (r = .411, p < .001). The percentage of missing perceived stress assessments and HCC samples were not correlated with mean perceived stress, childhood trauma, years of education, age, or the intensity of their child’s cancer treatment (p’s > .05).

1.3.2. Bivariate Correlations among study variables

Given the longitudinal nature of the study with repeated measures of perceived stress and HCC nested within individuals, we examined both between- and within-person correlations. Between-person correlations depict participants’ average levels of perceived stress and HCC over the study period in relation to time-invariant factors including childhood trauma, caregiver age, caregiver education, and children’s cancer treatment intensity. Within-person correlations depict correlations based on participants’ deviations from their own mean perceived stress levels and HCC over the study period (e.g., when participants reported higher perceived stress levels than what was typical for them, did they also have higher HCC than usual for them). Based on between-person correlations (Supplemental Table 1), participants who reported greater childhood trauma reported higher perceived stress levels, but not HCC, on average. Specifically, higher average perceived stress over the study period was positively correlated with childhood trauma (r = .35, p < .001) and negatively correlated with years of education (r = −.23, p < .001), but not significantly correlated with average HCC over the study period, caregiver age, or the intensity of the child’s cancer treatment (r’s = |.03-.10|; p’s > .05). Higher average HCC over the study period was positively correlated with caregiver age (r = .22, p < .001),, but not average perceived stress over the study period, childhood trauma, years of education, or the intensity of the child’s cancer treatment(r’s = |.02-.20|; p’s > 05). When considering within-person correlations (Supplemental Table 2), participants’ deviations from their own average perceived stress over the study period were weakly correlated with their deviations from their own average HCC over the study period (r = .01) and with childhood trauma, age, education, and the intensity of their child’s cancer treatment (r’s < .001). Because within-person correlations are based on data nested within individuals and thus, are not independent observations, we do not indicate their statistical significance.

1.3.3. ICC and Functional Form

Unconditional means/ICC models indicated that 58% of variance in perceived stress was attributed to between-person differences and 42% to within-person differences. HCC was more stable, with 73% of the variance attributed to between-person differences and 27% to within-person differences. We conducted nested model comparisons to establish an optimal fitting functional form. The addition of a random slope improved model fit for the effect of time on perceived stress (χ² = 7.68, p = .005) but not on HCC (χ² = .010, p = .92). Thus, a random intercept and random slope were used to model changes in perceived stress across the study period while a random intercept and fixed slope were used to model changes in HCC. On average, perceived stress levels decreased across the study period such that each additional month since child diagnosis was associated with .151 unit decrease in PSS (γ = −.151, SE = .027, p < .001). HCC did not change significantly across time (γ = .002, .SE = 002, p =.40), but there was adequate within-person variability in HCC trajectories over time to test our moderation hypotheses. We also probed but did not find evidence of non-linear effects of time on perceived stress (p = .31) or HCC (p = .93).

1.3.4. Childhood trauma × Time on Perceived Stress

Changes in perceived stress over the study period were moderated by childhood trauma (γ = .003, SE = .001, p = .002), which remained statistically significant after adjusting for intervention group assignment, caregiver age, years of education, and the child’s cancer treatment intensity (γ = .003, SE = .001, p = .002; see Table 2). As displayed in Figure 1, although perceived stress levels decreased on average across the study period, participants with lower levels of childhood trauma (−1 SD) had a steeper decline in perceived stress levels across the study period (simple slope estimate = −.208, SE = .030, p <.001) compared to participants who reported greater levels of childhood trauma (+1 SD; simple slope estimate = −.091, SE = .031, p = .005). Including the minimization/denial subscale of the CTQ as an additional covariate in the model did not change these results (γ =.003, SE = .001, p = .002).

Table 2.

Results of multilevel models examining time by childhood trauma interactions on perceived stress and hair cortisol levels across the study period

	Perceived Stress			Hair Cortisol (log)
	γ	SE	p	γ	SE	p
Intercept	8.265	0.620	<.001	0.868	0.069	<.001
Time	−0.149	0.024	<.001*	0.002	0.003	0.567
Intervention group	−0.775	0.553	0.166	0.119	0.062	0.061
Age	0.023	0.040	0.565	0.011	0.004	0.016*
Education	−0.173	0.138	0.216	0.037	0.016	0.024*
Very intensive cancer treatment	0.056	0.670	0.934	0.035	0.076	0.648
Most intensive cancer treatment	−0.112	0.949	0.906	0.055	0.107	0.604
Childhood trauma	0.030	0.019	0.120	0.002	0.002	0.361
Childhood trauma × Time	0.003	0.001	0.002*	0.000	0.000	0.008*

γ = fixed effect coefficient; SE = standard error; Time = months since diagnosis; Moderately intensive cancer treatment is the reference level.

^*p = .05

Figure 1.

Scores on the Perceived Stress Scale over the 12 months following the onset of a major chronic stressor (being the maternal caregiver of a child newly diagnosed with cancer) by recalled level of childhood trauma. Plots represent 1 SD below and 1 SD above the mean level of childhood trauma.

1.3.5. Childhood Trauma × Time on HCC

Changes in HCC over the study period were moderated by childhood trauma (γ = −.0004, SE = .0001, p = .008), and this remained statistically significant when accounting for intervention group assignment, caregiver age, years of education, and the child’s cancer treatment intensity (γ = −.0004, SE = .0001, p = .008; see Table 2). As displayed in Figure 2, among participants with lower levels of childhood trauma, HCC significantly increased across the study period (simple slope estimate = .008, SE= .004, p = .030) whereas there was a non-significant decrease in HCC over time among participants with higher levels of childhood trauma (simple slope estimate = −.005, SE= .003, p = .13). Inclusion of the minimization/denial subscale of the CTQ did not change these results (γ = <.001, SE < .0001, p = .008).

Figure 2.

Levels of hair cortisol over the 12 months following the onset of a major chronic stressor (being the maternal caregiver of a child newly diagnosed with cancer) by recalled level of childhood trauma. Plots represent 1 SD below and 1 SD above the mean level of childhood trauma.

1.3.6. Secondary Analyses: Main and moderating effects of Intervention group

In secondary analyses using all participants (N=120), we tested whether changes in perceived stress and hair cortisol over time were moderated by intervention group. There was no evidence of an interaction between intervention group and time on HCC (p = .84; see Table 3). However, there was a statistically non-significant trend of an interaction between intervention and time on perceived stress (γ = −.061, SE = .034, p = .08; see Supplemental Figure 2), such that participants in the intervention group had a marginally steeper decrease in perceived stress levels over the 12-month follow-up (simple slope estimate = −.176, SE = .035, p = .001) relative to those in the control condition (simple slope estimate = −.119, SE = .028, p = .001).

Table 3.

Results of multilevel models examining time by intervention group interactions on perceived stress and hair cortisol levels across the study period

	Perceived Stress			Hair Cortisol (log)
	γ	SE	p	γ	SE	p
Intercept	7.780	0.541	<.001	0.889	0.063	<.001
Time	−0.116	0.032	0.001*	0.003	0.003	0.445
Intervention group	−0.221	0.502	0.660	0.058	0.063	0.356
Age	−0.019	0.030	0.539	0.006	0.004	0.081
Education	−0.268	0.113	0.019*	0.017	0.014	0.226
Very intensive cancer treatment	0.537	0.536	0.319	0.053	0.066	0.427
Most intensive cancer treatment	0.059	0.724	0.935	0.054	0.091	0.554
Intervention × Time	−0.062	0.035	0.078	−0.001	0.005	0.843

γ = fixed effect coefficient; SE = standard error; Time = months since diagnosis; Moderately intensive cancer treatment is the reference level.

^*p = .05

1.4. Discussion

The stress sensitization model proposes that exposure to adversity during critical developmental periods in childhood programs lifelong sensitivity of the HPA-axis to stressors (Daskalakis et al., 2013; Lupien et al., 2009; Shonkoff et al., 2009). In support of this model, the current study provides novel longitudinal evidence that exposure to childhood trauma moderates cortisol responses over the 12 months following the onset of a chronic stressor in adulthood. Specifically, among female caregivers of a child newly diagnosed with cancer, those with lower levels of childhood trauma showed the expected increase in hair cortisol concentration across the year following diagnosis. However, this response was blunted among mothers who endorsed higher levels of childhood trauma. These findings were independent of caregiver age, years of education, the child’s treatment intensity, and intervention group. In contrast, maternal caregivers with a history of more childhood trauma showed persistent psychological stress compared to caregivers with lower childhood trauma. Thus, in accordance with the model, the current findings suggest that childhood trauma prolongs psychological reactivity to chronic stressors in adulthood while blunting the HPA response to stress.

The current results are consistent with one of the only other longitudinal studies to examine whether childhood stress moderates HPA-axis dysregulation in adults. In this prospective study individuals who experienced high levels of life event stress in early life (e.g., family financial troubles, family relationship problems, death/illness of a family member) had flatter diurnal cortisol patterns at age 37 years, but only in the context of high life event stress in adulthood (Young et al., 2019). Similar results are observed in studies that examine cortisol reactivity to brief laboratory stressors, with adults adopted from orphanages/institutions or who endorse a history of childhood adversity being hyporeactive when compared with those exposed to less childhood adversity (Bunea et al., 2017; Carpenter et al., 2007; McLaughlin et al., 2015; Ouellet-Morin et al., 2011). The current findings extend this work to show a similar pattern of blunted HPA-axis reactivity in response to chronic stress among adults who recall early life trauma.

Cross-sectional evidence supports an association of chronic life stress in adulthood with dysregulation of the HPA-axis; however, the direction of observed effects is mixed. Multiple aspects of the stress exposure have been proposed to moderate the HPA-axis response to chronic stress, including the type, severity, chronicity, and controllability of the stress exposure, whether the stressor is ongoing at the time of cortisol assessment, and time since the onset of the stressor (Khoury et al., 2019; Koss & Gunnar, 2018; Miller et al., 2007). In addition, the methodological challenges of reliably assessing cortisol levels in saliva, blood, or urine likely contribute to inconsistent findings (Khoury et al., 2019; Stalder et al., 2017). To address these challenges, the current study assessed HCC, which provides a more stable measure of cortisol levels over time and examined response following the onset of a single chronic, uncontrollable, ongoing, and severe life stressor. Yet, findings still showed considerable variability across individuals. Among maternal caregivers who endorsed lower levels of childhood trauma, the current results show a progressive increase in hair cortisol levels across the year following diagnosis. This is consistent with meta-analytic evidence showing an overall positive association between HCC and chronic stress exposure in adulthood (Khoury et al., 2019; Stalder et al., 2017) and with current understanding of the chronic stress response (Agorastos & Chrousos, 2022). Under conditions of chronic stress, extended periods of elevated cortisol levels are believed to provide metabolic support for physiological, psychological, and behavioral responses to effectively address the demands of the situation (Agorastos & Chrousos, 2022). This response is considered adaptive and termed allostasis (McEwen, 2019). However, prolonged elevations of cortisol or inability to turn off the response when the level of threat decreases can result in allostatic load, with deleterious long-term consequences for physical and mental health (McEwen, 2019). It remains to be determined whether cortisol levels would return to baseline among the current sample of maternal caregivers as the threat of the initial cancer diagnosis and treatment declines.

In contrast to maternal caregivers who recalled lower levels of childhood trauma, participants with higher levels of trauma did not show the expected increase in HCC over time. This pattern of results is consistent with the stress sensitization model and the possibility that adverse experiences in early life shape HPA responses to stress in later life. Interestingly, cortisol levels assessed over the first month after the child’s diagnosis, when distress was uniformly high across participants (Marsland et al., 2020), did not relate to childhood trauma. Rather, trauma moderated the HPA-axis response across the subsequent year. Thus, the full impact of childhood adversity on HPA-axis function may only be evident in the context of longer-term adjustment to chronic life stressors.

Having a child diagnosed with cancer is a life-altering stressor that includes possible threat to life and handling the many uncontrollable life circumstances that accompany repeated hospitalizations and clinic visits, intensive treatment regimens, and alterations of roles and responsibilities (Long & Marsland, 2011). Survival rates are high; however, initial treatments for most childhood cancers are grueling and last longer than 12 months. On average, female caregivers adjust well to this unexpected life challenge, with only transient increases in distress (Sharp et al., 2022). Our findings are consistent with this pattern, with a mean decrease in distress as measured 1-, 6-, and 12-months following diagnosis (Marsland et al., 2020; Walsh et al., 2018) and, in the current analysis, a mean decrease in perceived stress across the year. Although most caregivers adapt well, convergent evidence identifies a subset who experience more persistent adjustment difficulties, including symptoms of anxiety, depression and cancer-related post-traumatic stress (Sharp et al., 2022). This variability in psychological response is also observed in response to other potentially traumatic events, such as war, death of a significant other, or natural disasters (Galatzer-Levy, Huang, & Bonanno, 2018). The current results suggest that experiences in childhood may contribute to this variability, with childhood trauma predicting a more sustained elevation of perceived stress. These results are consistent with epidemiological evidence linking childhood adversity to increased risk of PTSD in response to traumatic stressors later in life, possibly as a result of heightened emotional responses and amygdala reactivity to threat (El-Khoury, Rieckmann, Bengtsson, Melchior, & Rod, 2021; McLaughlin et al., 2017).

The stress sensitization model proposes that childhood adversity programs both psychological and physiological sensitivity to stressful life events in adulthood, resulting in heightened emotional responses (El-Khoury et al., 2021; McLaughlin et al., 2017) and hypocortisolism and blunted HPA-axis reactivity to acute stress (Meewisse, Reitsma, de Vries, Gersons, & Olff, 2007; Zorn et al., 2017). Thus, we anticipated an inverse relationship between perceived stress and hair cortisol levels across the study period among those exposed to childhood trauma. Contrary to expectations, individual differences in the two responses were unrelated over time despite childhood adversity moderating patterns of change in both perceived stress and hair cortisol concentrations. Reasons that hair cortisol and perceived stress responses were unrelated are unclear; however, others have similarly observed no consistent association of hair cortisol concentrations with self-reports of perceived stress (Stalder et al., 2017). Thus, evidence suggests that biological markers of stress sensitivity may operate independently of or on different timelines from perceptions of stress.

The current results are largely consistent with our earlier findings (Marsland et al., 2020) and show a marginally steeper decline in perceived stress across the 12 months for participants who received the supportive stress management intervention when compared with the usual care group, albeit not statistically significant. Interestingly, the benefits of the intervention for psychological adjustment observed in our prior work did not extend to biological markers of stress, including hair cortisol levels (current analysis) and markers of inflammation (Marsland et al., 2020).

The mechanism(s) through which childhood adversity can affect HPA-axis reactivity to life stressors remain unclear. The stress sensitivity model proposes that exposure to adversity in early life impacts the development of brain regions that control both psychological and HPA-axis responses to meet environmental demands, resulting in a phenotype that is more reactive to life threats. Recent evidence suggests that prenatal and early postnatal years are particularly sensitive periods for this environmental programming of stress-related physiology (Khoury et al., 2019; Koss & Gunnar, 2018; Raymond et al., 2021). While adaptive in the short term, heightened activation of the HPA-axis over prolonged periods or in response to repeated stress in early life is proposed to result in compensatory blunting of HPA-reactivity in adulthood (Dempster et al., 2021; McEwen, 2007). This down-regulation may result from enhanced negative feedback sensitivity of the HPA-axis to cortisol (Fries, Hesse, Hellhammer, & Hellhammer, 2005) and be an adaptive response to counteract the physiological consequences of chronic or repeated HPA-axis activation for physical health, including cardiometabolic function (McEwen, 1998, 2007). In support of this pathway, a meta-analysis of 47 studies showed enhanced negative feedback sensitivity of the HPA-axis among individuals exposed to traumatic life events (Morris, Compas, & Garber, 2012). Interestingly, prior findings, including those from the current study, show an association of exposure to chronic stress with decreased sensitivity of glucocorticoid receptors in immune cells (Miller, Cohen, & Ritchey, 2002; Walsh et al., 2018). However, it remains to be determined whether exposure to stress in childhood moderates this response. It is also unknown whether glucocorticoid sensitivity of peripheral immune cells reflects responsiveness of central glucocorticoid receptors that modulate HPA-axis activity.

Programming of the sensitivity of the HPA-axis by early life experiences may also reflect epigenetic changes to genes that encode the glucocorticoid receptor and regulate the cortisol response and its impact on tissues (Bustamante et al., 2016; Romens, McDonald, Svaren, & Pollak, 2015). Early life environmental exposures have also been shown to relate to the structural development of brain regions and circuits involved in fear learning that exert top-down control over the HPA-axis (Callaghan & Tottenham, 2016; Lupien et al., 2009). For example, animal models show that stressful experiences during critical periods of early development can permanently modify corticolimbic brain systems that control stress responsivity (Balouek et al., 2023; Gutman & Nemeroff, 2003; Ladd et al., 2000). Initial human evidence shows that adults exposed to adversity between the ages of 3 and 7 years, a period when the amygdala, hippocampus and prefrontal cortex are developing, mount a blunted HPA-axis response to acute stress (Raymond et al., 2021). Further work is warranted to examine critical periods of sensitivity of developing systems that control stress reactivity.

In much of the literature, the impact of exposure to childhood adversity on biological programming of stress regulatory systems is viewed from a perspective of increased risk for morbidity and mortality (McEwen, 2007). While there is considerable evidence that childhood adversity relates to lifelong health risk (Grummitt et al., 2021), recent theoretical models have begun to consider the benefits of a phenotype that is adapted to handle the demands of an unpredictable and uncontrollable environment (Ellis et al., 2022; Grummitt et al., 2021). From a life-history framework, adverse conditions in childhood may program a response that has evolved to maximize reproductive fitness in the context of uncertain environmental conditions (Ellis & Del Giudice, 2019). In adverse environments, reproductive success may result from allocating metabolic energy to early reproduction rather than somatic maintenance, resulting in accelerated life-history strategies at the cost of long-term health and longevity. According to the Adaptive Calibration Model (Ellis & Del Giudice, 2019), programming of the HPA-axis and other stress response systems that contribute to the control of metabolism, energy allocation, and reproduction may play a key role in guiding life history strategies over the life course. In sum, early sensitization of stress pathways may program faster life history strategies that are adaptive for short-term health, reproductive success, and survival, but incur increased risk for accelerated aging, health conditions in later life, and premature mortality that accompany childhood adversity.

There are several noteworthy strengths and limitations of the results presented here. The parent project was a trial of a psychosocial intervention for mothers of children newly diagnosed with cancer. This provided an opportunity to conduct the first examination of changes in cortisol levels across 12 months following the onset of the same extreme life stressor for all participants and addressed many limitations of the extant literature, which is largely cross-sectional, focuses on a wide range of differing life experiences, and varies considerably in the time elapsed since the stressor occurred and whether it is ongoing at the time of cortisol assessment. However, the current study is limited by lack of a comparison group of mothers who were experiencing more normative levels of stress. Furthermore, although intervention-related changes in cortisol levels were not observed and analyses controlled for intervention condition, the intervention may have complicated the interpretation of longitudinal changes. Indeed, it is possible that participation in the intervention may have influenced the relationship between recollection of childhood trauma and cortisol response over time. The current study was not sufficiently powered to be able to examine this possibility. Another limitation of the current work is the retrospective assessment of childhood trauma from birth to age 18 years. This assessment lacks developmental specificity and may be subject to recall bias (Hardt & Rutter, 2004; Reuben et al., 2016) . Evidence suggests that retrospective and prospective measures of childhood maltreatment are at best only moderately correlated within person (Baldwin, Reuben, Newbury, & Danese, 2019), with emerging evidence that risk for adverse mental health outcomes more closely relates to subjective reports of childhood adversity than to objective measures (Francis, Tsaligopoulou, Stock, Pingault, & Baldwin, 2023). In this regard, it is possible that the subjective memory of adverse childhood experiences impacts the appraisal of threat in response to major life stressors. It is also possible that negative emotional responses to life stressors bias the recall of past experiences. Interestingly, measures of perceived stress and recalled childhood adversity were unrelated (r = .07) when assessed approximately 6 months after the onset of the stressor in the current study. Although challenging to accomplish, prospective studies following individuals across critical developmental periods of early life with repeated assessment of childhood adversity are needed to inform current understanding of sensitivity to stressors in adulthood.

The current study is strengthened by the repeated assessment of cortisol levels in hair, which addresses many of the measurement issues that accompany momentary assessment of cortisol in blood or saliva and permits a more temporally stable examination of levels across the weeks/months prior to sample collection. It was hoped that collecting the first hair sample close to the time of the child’s diagnosis would permit assessment of pre-stressor cortisol levels. Unfortunately, however, variability in the timing of the child’s first medical appointment relative to the caregiver’s knowledge of the diagnosis prevented a clean examination of baseline levels. To address this limitation, future work would benefit from standardizing the timing of hair collection close to the onset of the chronic life stressor.

The external validity of the current study is also limited by the inclusion of only female caregivers and results cannot be generalized to other demographics. In this regard, existing evidence shows that men have higher hair cortisol levels on average than women and may mount different cortisol stress reactivity (Stalder et al., 2017). This suggests that biological sex warrants careful consideration in examining patterns of stress sensitivity. Finally, the current sample is relatively small and limited by missing data, which largely reflects the challenge of accessing caregivers at the time of the child’s medical appointments. As a result, the presented findings should be considered preliminary and in need of replication before firm conclusions are drawn.

Despite these limitations, the current study makes a novel contribution to the literature. To our knowledge it is the first longitudinal study to examine cortisol response across the year following onset of an overwhelming and ongoing life stressor in adulthood. Findings are consistent with growing evidence from animal models and with studies examining reactivity to acute stress and suggest that early life stress may program an adult phenotype that is more psychologically reactive but shows blunted HPA-axis response to chronic stress. These results suggest that individuals who experience early adversity may be biologically adapted to survive in the short-term. However, there may be a cost of these short-term adaptations to stress for long-term mental and physical health. Future work is warranted to determine how best to promote long-term health among individuals who report adverse childhood experiences.

Highlights.

1. Retrospective recall of childhood adversity relates to prolonged increases in psychological stress across the year following the major life stressor of having a child diagnosed with cancer.

2. Retrospective recall of childhood adversity relates to blunted hair cortisol levels across the year following the major life stressor of having a child diagnosed with cancer.

3. Childhood trauma may program a phenotype that is more psychologically reactive but shows blunted HPA-axis response to chronic stressors in adulthood.