Abstract
The purpose of the present report was to examine the association of recent maltreatment experiences with cortisol reactivity in young adolescents. The ethnically diverse sample consisted of boys and girls 9 to 12 years of age. The maltreatment group (N = 303) all had recent, substantiated reports to protective services for neglect, physical abuse, sexual abuse, and/or emotional maltreatment. The comparison group (N = 151) was recruited from the same neighborhoods and was demographically similar to the maltreatment group. Cortisol reactivity was assessed by a laboratory stressor, a modified version of the Trier Social Stress Test for Children. Statistical analyses indicated that the maltreated young adolescents showed a blunted or attenuated response to the stressor as compared with those in the comparison group. This attenuated response was especially pronounced for those whose maltreatment included physical and/or sexual abuse. A main effect for gender was also found with boys having higher cortisol than girls. Implications for treatment of mental and physical health problems associated with child maltreatment and for prevention of developmental problems across the life span are discussed.
Child abuse and neglect are experienced by millions of infants, children, and adolescents in America (Sedlak et al., 2010). Several decades of research indicate that experiencing such maltreatment is associated with a diverse array of problems, including mental health problems (especially depression and anxiety); aggression, delinquency, and other antisocial behavior, such as substance use and abuse; cognitive delays and school achievement problems; and problems with interpersonal relationships (Margolin & Gordis, 2000; Trickett & Negriff, 2011; Trickett & Schellenbach, 1998). Some recent studies also indicate associations between child maltreatment and physical health problems (e.g., Arias, 2004; Dube, Felitti, Dong, Giles, & Anda, 2003; Noll, Zeller, Trickett, & Putnam, 2007).
Child Maltreatment and the Hypothalamic–Pituitary–Adrenal Axis
The traumatic nature of child maltreatment is thought to influence the functioning of physiological stress response systems and, in particular, the hypothalamic–pituitary–adrenal (HPA) axis. Typical HPA axis functioning is assessed by measuring basal, diurnal, or reactive cortisol. Basal cortisol levels refer to mean levels, whereas the diurnal cycle is characterized by cortisol levels peaking within 30 to 45 min of awakening and then declining throughout the day. Reactive cortisol levels are assessed prior to and after introduction of a known stressor. Against this background, acute episodes of stress cause increases in circulating free cortisol within approximately 10–15 min followed by a return to baseline (Susman, Dorn, Inoff-Germain, Nottelmann, & Chrousos, 1997). Specifically, stress stimulates the paraventricular nucleus of the hypothalamus to release corticotropin-releasing hormone (CRH). CRH stimulates the anterior pituitary to release adrenocorticotropic hormone, leading to the release of cortisol from the adrenals. Cortisol circulates back to the pituitary, inhibiting CRH release via negative feedback loops. Thus, cortisol is initially elevated in response to stress but then returns to a baseline, homeostatic state as a result of the HPA negative feedback loop (Chrousos & Gold, 1992).
In the short term, cortisol increases in response to a stressor are essential for adaptation, maintenance of homeostasis, and survival (McEwen, 2000). However, sustained periods of chronic stress, such as living in an abusive home, have no natural end to bring about the return to baseline of cortisol. Chronically high levels of circulating cortisol exert damage on organ systems, including the hippocampus, a brain area rich in corticosterone receptors in nonhumans (Sapolsky, Uno, Rebert, & Finch, 1990) and cortisol receptors in humans (Pruessner, Dedovic, Pruessner, Lord, Buss, Collins, et al., 2010). According to the attenuation theory, because chronic high levels of circulating cortisol are damaging, the HPA axis downregulates with lower cortisol as an indicator of the attenuation (Susman et al., 1997). Thus, initially maltreatment may lead to increased cortisol levels and reactivity but over time the chronic stress that occurs in the lives of maltreated youth may lead to blunted HPA axis responses, lower basal levels, and blunted diurnal cortisol rhythms due to allostatic load (Goldstein & McEwen, 2002; Susman, 2006). Because the HPA axis is involved in regulating several key physiological systems, including the central nervous system, the metabolic system, and the immune system (Miller, Chen, & Zhou, 2007), disrupted HPA axis functioning can have numerous behavioral, emotional, and physical consequences.
Cortisol Level and Maltreatment
Experimental studies on nonhuman species suggest causal links between early stress during developmentally sensitive windows and long-term alterations in HPA axis functioning (e.g., Parfitta, Waltona, Corriveaub, & Helmreich, 2007; Renard, Rivarola, & Suárez, 2007).
Recently, a number of studies with children and adolescents have been conducted to examine the association between many forms of child maltreatment and cortisol level, especially basal or morning level or diurnal cortisol. Trickett, Negriff, Ji, and Peckins (2011) reviewed these studies and reported multiple inconsistencies in the findings. Although many studies have shown elevated levels of cortisol in maltreated children and adolescents as compared with their nonmaltreated peers (e.g., Bruce, Fisher, Pears, & Levine, 2009; Cicchetti, Rogosch, & Sturge-Apple, 2007), other studies report lower levels of cortisol in the maltreatment group (King, Mandansky, King, Fletcher, & Brewer, 2001), and some show no differences between groups (Cicchetti, Rogosch, Gunnar, & Toth, 2010; Murray-Close, Han, Cicchetti, Crick, & Rogosch, 2008). The reasons for these inconsistencies are not clear but may well include sample differences relating to the nature (e.g., physical or sexual abuse), severity, or chronicity of the maltreatment, age or developmental stage, the diurnal variability of cortisol, or covariates or mediators (such as comorbid posttraumatic stress disorder or externalizing behavior problems) that were included in some models. One replicated finding is that early abuse leads to a flattening of diurnal variation in cortisol (Cicchetti et al., 2010). Recent findings suggest that maltreatment interacts with variation in the CRH-1 receptor genotype in predicting the flattening of the diurnal variability in cortisol (Cicchetti, Rogosch & Oshri, 2011). Thus, maltreatment may lead to attenuated diurnal cortisol slopes at least among certain youth who are genetically susceptible. One longitudinal study showed that sexually abused girls had higher cortisol levels as compared with nonabused girls at younger ages and closer in time to the sexual abuse but lower levels at older ages further in time from the abuse (Trickett, Noll, Susman, Shenk, & Putnam, 2010). These findings were interpreted as supportive of the attenuation hypothesis (e.g., Susman, 2006). That is, initially cortisol was elevated in the sexually abused females but over time the chronic stress associated with this abuse resulted in attenuated cortisol levels.
Cortisol Reactivity and Maltreatment
As contrasted to these studies examining basal cortisol level, fewer studies have examined cortisol reactivity, which is often indexed by a change from basal cortisol levels in response to an acute stressor administered in a laboratory setting. Studies specifically focusing on child maltreatment are especially scarce. Hart, Gunnar, and Cicchetti (1995) found cortisol reactivity, defined as “ … the tendency to produce cortisol levels markedly higher than one’s typical basal levels” (p. 17), assessed daily over 31 days to be blunted in maltreated preschoolers. A few studies have examined cortisol reactivity in adults reporting retrospectively on their maltreatment history (e.g., Carpenter et al., 2007; Heim et al., 2000). The findings of these two studies were also inconsistent with each other. Carpenter and colleagues (2007) found an attenuated response whereas Heim et al. (2000) found a heightened response to a stressor. Only two studies were found that examined cortisol reactivity in maltreated adolescents. Harkness, Stewart, and Wynn-Edwards (2011) examined cortisol reactivity in a sample of depressed (N = 41) and nondepressed (N = 30) male and female Canadian adolescents (ages 12 to 21; 89% Caucasian). A history of child maltreatment was obtained by self-report. Results indicated that the adolescents who reported a history of maltreatment (37%) exhibited heightened reactivity than did those who did not but that this was only true for those with mild/moderate levels of depression. The second study, also with a sample of Canadian adolescents, compared cortisol reactivity in maltreated (N = 67) and nonmaltreated (N = 25) females aged 12 to 16 years (MacMillan et al., 2009). The maltreatment group, recruited from child protection agencies, exhibited a significantly attenuated cortisol response even after controlling for posttraumatic stress disorder and major depressive disorder. In sum, the extant evidence about cortisol reactivity among maltreated adolescents is scant, limited to small ethnically nondiverse samples. Like studies examining cortisol level, findings from reactivity studies are inconsistent. That is, one study shows increased reactivity among a subgroup of the maltreated male and female adolescents and the other shows attenuated reactivity among maltreated female adolescents.
Links between maltreatment and cortisol reactivity are particularly relevant because of the potential role of the HPA axis in the development of emotional and behavioral problems. For example, although findings are not completely consistent, low cortisol levels are linked with aggressive and disruptive behaviors in children and adolescents (McBurnett, Lahey, Rathouz, & Loeber, 2000; Shoal, Giancola, & Kirillova, 2003; Snoek, Van Goozen, Matthys, Buitelaar, & Van Engeland, 2004; van Goozen et al., 1998). Findings are particularly inconsistent regarding the relation between cortisol reactivity and aggressive behavior among children and adolescents (Alink et al., 2008). Cortisol reactivity is also found to be related to internalizing problems such as anxiety, behavioral inhibition, and depression (Nachmias, Gunnar, Mangelsdorf, Parritz, & Buss, 1996).
The Present Study
This report addressed some of the gaps in the literature by examining acute cortisol reactivity in response to a laboratory stressor, the Trier Social Stress Test for Children (TSST-C), among maltreated and nonmaltreated multi-ethnic young adolescent boys and girls from a large urban area. The maltreated sample consisted of substantiated cases recruited from a child protection agency and the nonmaltreated (comparison) group was recruited from the same neighborhoods. Because maltreated youth tend to have repeated and chronic exposures to stress and maltreatment, we expected the maltreated youth to exhibit an attenuated cortisol response to the laboratory stressor as compared with the nonmaltreated group. Because physical and sexual abuse may be more consistently traumatic than neglect (Margolin & Gordis, 2000; Trickett, Negriff, et al., 2011), we expected that maltreated youth who experienced that form of maltreatment (whether or not in company with neglect) would show a stronger pattern of attenuation than maltreated youth who did not (that is, were neglected and/or emotionally maltreated but not physically or sexually abused).
Our study also aimed to shed light on the association between child maltreatment and HPA axis functioning in male versus female youth. Susman et al. (2010) report a more consistent cortisol response to the TSST among boys than among girls. However, among youth exposed to violence, boys were more likely to show an attenuated response in cortisol, whereas girls tended to show an elevated response (Peckins, Dockray, Eckenrode, Heaton, & Susman, 2012). Thus, we anticipated that among comparison youth, boys would have a larger response to the TSST than would girls, whereas among maltreated youth, boys would show a more attenuated response than would girls. The rationale for focusing on young adolescents was based on psychosocial theories of development suggesting that the transition from childhood to adolescence is a period characterized by high risk of emotional and behavioral problems. The multiple neurobiological changes that accompany puberty during this transition add to the risk for problems (Susman & Dorn, 2012). Furthermore, there is evidence that younger (prepubertal) children may be protected from psychobiological effects of stress that diminishes as they transition into adolescence (Gunnar & Quevedo, 2007).
Method
Participants
The present study used data from the initial assessment of a longitudinal study examining the effects of maltreatment on adolescent development. The participants who comprised the maltreatment group (N = 303) were recruited from active cases in a county Department of Children and Family Services (DCFS) of a large western U.S. city. The inclusion criteria were (1) a new substantiated referral to DCFS in the preceding month for any type of maltreatment (i.e., neglect, physical abuse, sexual abuse, emotional maltreatment); (2) child age of 9–12 years; (3) child identified as Latino, African American or Caucasian (non-Latino); (4) child residing in 1 of 10 urban zip codes within the county at the time of referral to DCFS. With the approval of DCFS and the University’s Institutional Review Board, the parent or guardian of potential participants was contacted by post card about willingness to be contacted about participating in a research study. Contact via mail was followed up by a phone call: 77% of those contacted participated in the study.
The nature of each maltreated adolescent’s experience of abuse and neglect was obtained by examining the DCFS case record for each individual and entering this information into the Maltreatment Case Record Abstraction Instrument (MCRAI), which was a modification of the instrument developed by Barnett, Manly, and Cicchetti (1993; see Mennen, Kim, Sang, & Trickett, 2010, for further information). For each adolescent, information about the current DCFS referral was abstracted as was information for all other referrals to DCFS in the past 4 years (this was the extent of the case record information available from the agency). The record abstraction was performed by social work graduate students. Training (by one of the authors and one advanced doctoral student) consisted of an initial 2-hr extensive orientation and close supervision of the first 4–5 case abstractions until the abstractor achieved at least 90% interrater agreement with the supervisors. In cases of inconsistency, original DCFS records were rechecked and, if necessary, reconciled by group decisions. Twenty cases were chosen at random to test interrater agreement on the four major types of maltreatment for each referral for each youth. This process indicated good reliability: κ of .82, .82, .79, and .75 for physical abuse, sexual abuse, emotional maltreatment and neglect, respectively. Analysis of the MCRAI data indicated that 77% of the maltreated adolescents in this sample had experienced some form of neglect; 52% had experienced physical abuse; 52% emotional maltreatment; and 20% sexual abuse. The average number of reports to DCFS for each child in the prior 4 years was 3.9 (standard deviation [SD] = 2.7) with a range from 1 to 16 referrals. The average number of maltreatment types was 2.0 (SD = 1.0), and multiple types of maltreatment were experienced by 67% of the adolescents. Of the maltreated youth, 60.5% had been physically and/or sexually abused and 39.5% had not. Of those physically and/or sexually abused, 78% had also been neglected and 61% had been emotionally maltreated. On average, youth in this subgroup had 4.6 (SD = 2.8) reports to DCFS in the prior 4 years. For the maltreated youth who were not physically or sexually abused, 77% had been neglected and 38% had been emotionally maltreated. This subgroup, on average, had 2.8 (SD = 1.0) reports to DCFS in the prior 4 years. Thus, overall the maltreated group can be characterized as multiply maltreated, and those whose maltreatment included physical or sexual abuse were more severely maltreated, at least as indicated by number of reports to protective services, as compared with maltreated youth who did not experience physical or sexual abuse.
The comparison group (N = 151) was recruited via mail using names from school lists of children aged 9–12 years residing in the same 10 zip codes as the maltreatment group. These children were screened for DCFS referrals for abuse or neglect. The maltreatment and comparison groups were not significantly different on age (maltreated: M = 10.84 years, SD = 1.16); comparison: M = 11.11 years, SD = 1.15); gender (50% male maltreated, 60% male comparison); or ethnicity (African American maltreated 45%, comparison, 32%; Latino maltreated 35%, comparison 47%; Caucasian maltreated 12%, comparison 10%; biracial maltreated 13%, comparison 11%). Because about half of the maltreated group were living in foster care (and we were unable to obtain any basic demographics on their biological parents), we could not compare the groups on the usual indices of socioeconomic status (e.g., parent education level or household income). Instead we compared the census track characteristics of the neighborhoods in which the adolescents resided. (For the maltreatment group, this was the address where they were living at the time of referral to DCFS.) Specifically, we compared the census blocks of the residential addresses on nine census categories characterizing the social, educational, economic, and demographic nature of the neighborhoods and deemed important for child development (Duncan & Aber, 1997). These comparisons indicated that the two groups were residing in very similar neighborhoods (Trickett, Mennen, Negriff, & Horn, 2011).
Measures
TSST-C
This modification (Buske-Kirschbaum et al., 1997) of a common method used to elicit an HPA stress response in a laboratory context (Kirschbaum, Pirke, & Hellhammer, 1993) includes a social challenge (composing and performing a story in front of a panel of judges) and a cognitive challenge (oral serial subtraction). The TSST-C has proved to be a reliable stressor across different childhood and adolescent age groups and populations (e.g., Buske-Kirschbaum et al., 1997; Dorn et al., 2003; Gilissen, Bakermans-Kranenburg, van IJzendoorn, & Linting, 2008; Kudielka, Hellhammer, & Kirschbaum, 2007; Susman et al., 2010).
Saliva samples for cortisol assays
We collected saliva samples via passive drool through a short straw into a vial. Saliva samples were immediately frozen and then transported on ice to Salimetrics, Inc., and stored frozen at −80°C until assayed for cortisol. On the day of testing, all samples were centrifuged at 3000 rpm for 15 min to remove mucins. Samples were assayed for salivary cortisol using a highly sensitive enzyme immunoassay U.S. Food and Drug Administration (510k) cleared for use as an in vitro diagnostic measure of adrenal function (Salimetrics, State College, PA). The test used 25 μl of saliva and had a lower limit of sensitivity of 0.007 μg/dl, with a range of sensitivity from 0.007 to 3.0 μg/dl. Average intra- and interassay coefficients of variation are less than 5% and 10%, respectively. The mean of the two duplicates was used in the statistical analyses.
Covariates
Pubertal stage was measured using self-report Tanner staging (Marshall & Tanner, 1969, 1970), the gold standard for assessing pubertal onset and development. The youth were given detailed training on the stages of pubertal development and how to assess the salient characteristics of each stage. Using sets of serial line drawings that depict the five stages of pubertal development of two different secondary sex characteristics (i.e., breast and genital growth for girls and boys, respectively, and pubic hair growth for both girls and boys) from prepubertal (stage 1) to postpubertal (stage 5). Youth were asked to rate which stage they were closest to in their development. Previous research has shown that youth can provide reliable self-reports after training (Dorn, Susman, Nottelmann, & Chrousos, 1987). Ratings for the two characteristics were averaged. Caregiver status indicated whether the youth was residing with a biological, step-, or adoptive parent versus being in foster care. Time of day of the baseline saliva sample was recorded at the time it was obtained.
Procedures
The university Institutional Review Board approved all procedures that were carried out with the understanding and written assent of the adolescent participants and consent of their parent or guardian. All participants were paid for their participation using the guidelines of the National Institutes of Health Healthy Volunteer Program. Each adolescent and a parent or a guardian attended a lab session lasting approximately 4 hr, which included the administration of a number of interviews, questionnaires, and standardized tests. During part of the assessment, adolescents engaged in the TSST-C. For the administration of this measure, the interviewer and a two-judge panel entered the room. The interviewer told the participant that he or she would read the beginning of a story. The participant would have 5 min to compose the end of the story, after which the participant would present it aloud to the panel of judges for 4 min. The interviewer and judges then left the room for 5 min. When they re-entered, the interviewer asked the adolescent to present the story. If the youth did not fill the 4 min, the interviewer used a standard set of prompts to encourage further narration. After the 4 min, the interviewer asked the youth to perform a serial subtraction task in the presence of the judges. This task was designed to be challenging for the particular age of the adolescent (e.g., subtracting 7 serially beginning with 758). If a mistake was made, the youth was told to stop and start again at the last correct number. Immediately following the serial subtraction task, the adolescents were asked to rate how nervous these tasks had made them on a scale from 1 (not at all nervous) to 10 (very nervous).
We collected six saliva samples, two before and four after the TSST-C. The rationale for the number of samples and the timing of intervals was based on the importance of accurately assessing both baseline levels of cortisol and reactivity to the stressor. Since this procedure took place in a research laboratory setting new to the participants, we were concerned that the novelty of this experience might be stressful enough that a true baseline would be hard to determine. To produce a more accurate baseline, a relaxation exercise was done with the adolescents after the initial sample and prior to the TSST-C. Thus, the first saliva sample occurred immediately after the informed consent procedure and 30 min before the stressor (i.e., the TSST-C). The second sample was collected 10 min before the stressor, immediately after a 5-min relaxation protocol involving listening to soft music while viewing a still slide of a beach scene. The TSST-C procedure lasted 14 min. The third sample was collected immediately after the stressor, and the fourth, fifth, and sixth samples were collected 10, 20, and 30 min after the end of the stressor, respectively. All questionnaires completed during the period prior to the TSST-C and during the period following the stressor when we were taking saliva samples involved measures selected for content unrelated to trauma and stress. Data collection occurred primarily in the afternoon, with 96% of participants providing the baseline saliva sample between 1200 and 1700 hr. The earliest baseline sample was collected at 1223 hr, and the latest was collected at 1735 hr. We control for time of day in all analyses.
Data Analysis
Raw cortisol data were substantially positively skewed, ranging from skew values of 4.77 to 10.54 (standard error [SE] = .12). We log transformed cortisol values and then winsorized outliers (> 3 SD from mean, all positive) to the next highest value within 3 SD from the mean. We also excluded 12 cases due to the use of steroidal medications that might influence cortisol levels. Skew of the transformed and winsorized variables ranged from .34 to .50 (SE = .12).
We examined the effect of maltreatment status on the response curve of cortisol using several strategies. First, we conducted a 6 (sampling time) × 2 (maltreatment status) × 2 (sex) repeated measures analysis in a mixed-modeling framework. The models were estimated using maximum likelihood (method = ML) procedures to account for missing data, with an unstructured covariance structure (type = UN; Singer, 1998). In these analyses, we controlled for child’s pubertal stage, caregiver status, and time of day. A significant group by sampling time interaction would suggest that the response curves varied by maltreatment status. Finally, we conducted parallel analyses comparing three groups: the comparison group, the subset of youth who had experienced physical and/or sexual abuse (with or without neglect or emotional maltreatment), and the subset who had experienced neglect and/or emotional maltreatment but not physical or sexual abuse. Post hoc analyses were performed on the significant group by sampling time and maltreatment type by sample time interactions by slicing the interaction in Proc Mixed. Slicing the interaction (e.g., group by sample time) tests whether the cortisol response at each sampling time differs by group.
Results
Figure 1 shows the means for the six salivary cortisol samples for the total sample and separately for the maltreated and comparison groups (see also Table 1 for means and SDs for all groups). On average, the highest salivary cortisol levels were measured at 10 min poststressor. Of the whole sample, 56% exhibited at least a 10% increase in cortisol from postrelaxation to peak. Of the maltreated youth, 55% exhibited a 10% increase from postrelaxation to peak value and 59% of the comparison youth exhibited such an increase; a chi-square test showed no significant difference between these groups.
Mean salivary cortisol levels at six time points for the whole sample, the maltreatment group, and the comparison group.
Means and Standard Deviations for Cortisol at Six Sampling Times for Abused Group (With or Without Neglect), Neglected Group (No Abuse), Comparison Group, and Full Sample.
Note. Mean (standard deviation). Cortisol measured in µg/dL.
The 6 (sampling time) × 2 (maltreatment status) × 2 (sex) repeated measures analysis, controlling for pubertal stage, time of day, and caregiver status, revealed a significant within subjects’ effect for sampling time, F(5, 426) = 2.26, p = .048, a marginally significant between-group main effect for maltreatment group, F(1, 426) = 3.16, p = .076, and a main effect for sex, F(1, 426) = 14.36, p = .002. These main effects indicate that, on average, the release of cortisol is blunted in the maltreated adolescents as compared with those in the comparison group and in girls compared to boys. A significant maltreatment group × sampling time interaction effect, F(5, 426) = 5.78, p < .001, was found, suggesting the cortisol response curve differed between maltreated and nonmaltreated adolescents (see Figure 2).
Mean salivary cortisol levels at six time points for maltreated and comparison boys and girls.
The significant Maltreatment Group × Sampling Time interaction effect was probed using post hoc analysis to test whether maltreated and nonmaltreated youth differ in the cortisol response curve by slicing the interaction by sampling time. The results suggest that maltreated youth exhibit a blunted cortisol response at +10 min, F(1, 426) = 7.11, p = .008, and +20 min, F(1, 426) = 4.26, p = .040, poststressor. The difference in cortisol at +30 min poststressor, F(1, 426) = 2.73, p = .099, between maltreated and nonmaltreated youth was marginally significant (see Table 2).
Results of a Sampling Time × Group (Maltreatment vs. Comparison) × Sex (Male vs. Female) Repeated Measures ANCOVA in a Mixed-Modeling Framework.
Note. ANCOVA = analysis of covariance; AIC = Akaike information criterion; BIC = Bayesian information criterion. Mean (standard deviation). Cortisol measured in µg/dL. Indices of fit: −2 Log Likelihood = 1,176.1, AIC = 1,302.1, BIC = 1,558.7; degrees of freedom = 1,426. Covariates not presented: pubertal stage, time of day, and caregiver status.
†p < .10; * p < .05; ** p < .01.
Next, a repeated measures analysis, with covariates, tested for differences in the cortisol response between nonmaltreated youth, maltreated youth with a history of physical and/or sexual abuse, and maltreated youth who only experienced neglect and/or emotional maltreatment (see Table 3 and Figure 3). The analysis revealed a significant within-subjects’ effect for sampling time, F(5, 424) = 2.28, p = .046, a significant between-group main effect for maltreatment type, F(2,424) = 3.89, p = .021, a significant between-group main effect for sex, F(2, 424) = 13.58, p < .001, and a significant maltreatment type × sampling time interaction effect, F(10, 424) = 3.56, p < .001, after controlling for pubertal stage, time of day, and caregiver status. The interaction was further explored and revealed that the cortisol response differed by maltreatment type at prerelaxation, F(2, 424) = 3.25, p = .040, postrelaxation/prestressor, F(2, 424) = 3.46, p = .032, +10 min poststressor, F(2, 424) = 4.67, p = .010, and +20 min poststressor, F(2, 424) = 3.58, p = .029. Physically/sexually abused youth exhibit a blunted cortisol response compared to nonmaltreated youth at prerelaxation, t(424) = 2.20, p = .028, +10 min poststressor, t(424) = 3.04, p = .003, and +20 min poststressor, t(424) = 2.57, p = .011. Similarly, physically/sexually abused maltreated youth exhibit a blunted cortisol response compared to the maltreated youth who did not experience physical or sexual abuse at prerelaxation, t(424) = 2.04, p = .042, and postrelaxation/prestressor, t(424) = .252, p = .012. The cortisol response of the maltreated group without physical or sexual abuse did not differ significantly from the comparison group.
Mean salivary cortisol levels at six time points for physically and/or sexually abused youth (with or without neglect), neglected but not abused youth, and comparison youth.
Results of a Sampling Time × Sex × Three-Group (Abused With or Without Neglect vs. Neglect Only vs. Comparison) Repeated Measure ANCOVA in a Mixed-Modeling Framework.
Note. ANCOVA = analysis of covariance. Mean (standard deviation). Cortisol measured in µg/dL. Indices of fit: −2 Log Likelihood = 1159.8, AIC = 1305.8, BIC = 1611.2; degrees of freedom: 2,424. Main effects for sex not shown. Covariates not presented: pubertal stage, time of day, and caregiver status.
†p < .10; *p < .05; **p < .01.
In an attempt to disentangle the type of maltreatment from the number of referrals to child protective services, which were associated in this sample, additional analyses were run without the nonmaltreated comparison group. First, the same maltreatment-type model was run including only two groups (maltreated youth with a history of physical and/or sexual abuse and maltreated youth who only experienced neglect and/or emotional maltreatment). There was a main effect of type but no main effect of sampling time and no Sampling Time × Maltreatment Type interaction. Second, the same model was run, with number of referrals added as an additional covariate. In this analysis, the main effect of maltreatment type was no longer significant.
On the self-ratings of subjective stress or nervousness induced by the TSST-C task, a two-way analysis of variance (maltreatment vs. comparison and boys vs. girls) indicated no significant differences in the ratings of the maltreated adolescents (M = 6.9, SE = 3.0, range = 1–10) and the comparison group (M = 6.7, SD = 2.7, range = 1–10). A main effect for sex was found, with girls reporting a higher level of nervousness/stress (M = 7.2, SD = 2.8, range = 1–10) than boys (M = 6.5, SD = 2.9, range = 1–10; F = 7.46, p = .007). The interaction term was not significant (F = .33, p = .57). The same results held when the maltreatment group was divided into those with physical/sexual abuse and those without. That is, there was a significant main effect for sex but not for maltreatment group and no group × sex interaction. These subjective stress or nervousness scores were not correlated with cortisol levels.
Discussion
The current report is one of the first studies to examine stress reactivity, in contrast to cortisol basal levels or diurnal patterns, in maltreated children or young adolescents. It is also the first to examine cortisol reactivity in youth who experienced different types of maltreatment. Findings indicated that for these young adolescents, maltreatment was associated with attenuated cortisol reactivity overall and that this result was pronounced in those youth who experienced more severe maltreatment as indicated by maltreatment that included physical and/or sexual abuse as well as, frequently, other forms of maltreatment, and as noted in the description of the sample, by more reports to DCFS. Particular strengths of this study include the relatively large multiethnic sample that includes both boys and girls and a careful and specific definition of the types of maltreatment experienced by the participants.
Although the design of this study precluded specifically testing Susman’s (2006) attenuation hypothesis because no data were available on the initial cortisol response of the youth (in the period immediately following trauma), our results provide support for the notion that early adverse and traumatic experiences can result in lower cortisol reactivity over time. Differences in cortisol reactivity between maltreated and comparison youth were evident on a number of dimensions. First, the cortisol response curve for the TSST-C varied for the maltreatment group as a whole compared to the comparison youth, indicating that differences were sustained from 10 to 20 min poststressor. The maltreatment group as a whole showed less increase in cortisol poststressor at 10 and 20 min poststressor (and marginally at 30 min poststressor) than did the comparison group. Results from previous studies had conflicting results for cortisol reactivity in individuals maltreated as children. In concordance with our findings, Carpenter et al. (2007) reported hypocortisol reactivity in adults who had been exposed to maltreatment as children, as did MacMillan et al. (2009) in a sample of maltreated female adolescents. In contrast, Heim and colleagues (Heim et al., 2009) found that hypercortisol reactivity was typical of maltreated children assessed as adults, and Harkness et al. (2011) similarly found heightened reactivity in maltreated youth but only among those rated as moderately depressed. Thus, even in studies of adolescents such as this one, there is still inconsistency in findings that cannot as yet be explained.
The role that depression may play as a mediator to explain these contrasting findings is not clear. According to Heim and colleagues (2009), neuroendocrine alterations secondary to maltreatment may be a discernible risk of psychopathology in adults. Specifically, altered HPA axis functioning is implicated in disorders like depression, with depression being characterized by hypo- or hyperactivation in atypical or typical depression, respectively (Chrousos & Gold, 1992). Future research should examine whether levels of depression and cortisol hypo- or hyperreactivity interact in maltreated children and adolescents.
Another factor that may well be involved in the conflicting findings of heightened or attenuated reactivity is the nature of the maltreatment experienced by individuals in the sample. Studies differ markedly in whether maltreatment status comes from self-report (e.g. Harkness, Stewart, & Wynn-Edwards, 2011) or from protective service records (e.g., MacMillan et al., 2009) and in whether status is reported concurrently or retrospectively (e.g., Heim et al., 2009). This study carefully abstracted complex DCFS case records in order to categorize the nature of the maltreatment and the findings indicated that the type of maltreatment experienced was relevant here. Thus, in this sample it was the youth who experienced physical and/or sexual abuse, with or without neglect and/or emotional maltreatment, who showed the attenuation of the stress response, not those individuals who experienced neglect and/or emotional maltreatment but neither physical nor sexual abuse. That is, most of the significant differences found in the three-group repeated measures analyses were between the comparison group and the physically/sexually abused maltreatment group, with the maltreatment group without sexual or physical abuse showing few significant differences from either of the other two groups. Although neglect and emotional maltreatment certainly can be very stressful, physical abuse and sexual abuse are forms of maltreatment that are uniformly highly stressful. As noted earlier, in this study, those youth who experienced physical and/or sexual abuse can be said to have been more severely maltreated than those who had not in that they were also likely to have experienced neglect or emotional maltreatment as well and to have a larger number of prior reports to protective services.
Limitations
A limitation of this study was that because of the complex and overlapping nature of the maltreatment experienced by the young adolescents in this sample, it is not possible to say to what degree the experience of physical and sexual abuse, per se, rather than other indices of severity of maltreatment was responsible for the findings. Our analyses indicated that covarying the number of referrals to protective services when comparing the stress reactivity of the two maltreatment subgroups negated the few significant differences between these two groups. However, this analysis cannot really answer this complex question. Another related limitation was that, because of the overlap of physical and sexual abuse and sample size, it was not possible to examine cortisol levels of those youth who experienced physical abuse only in contrast to those who experienced sexual abuse only.
In this study, novel sex differences were found. Although girls reported being more nervous than boys after the TSST-C task, in the repeated measures analyses a main effect for sex was found, indicating lower levels of cortisol for girls as compared with boys. The sex × sampling time and sex × group interactions were not significant however. Sex differences in cortisol are far from consistent, and the majority of studies assess sex differences in basal versus reactivity cortisol. For instance, higher cortisol and increased distress were reported for boys in a study of adjustment and cortisol in post-Katrina youth (Vigil, Geary, Granger, & Flinn, 2010). Another study showed no differences in cortisol reactivity between young adolescent boys and girls (Susman et al., 2010). Our sample size was not large enough to allow examination of sex differences in more detail. One particular issue that needs addressing is illustrated by the fact that in this study, as in many others, female youth were more likely to be identified as sexually abused than males; therefore, sex and type of maltreatment may be confounded. Future research must attempt to tease apart these factors as moderators of stress reactivity.
There are also important developmental issues that need to be addressed, which are beyond the scope of this study. For example, cortisol reactivity can interact with characteristics such as timing of puberty (Natsuaki et al., 2009), and the age or developmental level at the time of assessment of cortisol reactivity may affect secretory levels. A recently published long-term follow-up of sexually abused girls showed that basal cortisol concentrations of the abused females became lower than the cortisol concentrations in nonabused girls across development from childhood into adolescence (Trickett et al., 2010). Similar studies are not available for cortisol reactivity, but it may well be that the developmental stage at the time of assessment of cortisol reactivity as well as the basal levels across development may be associated with cortisol secretion. The finding that cortisol reactivity was lower in youth who experienced certain forms of maltreatment indicates that the severity of maltreatment may differentially be related to the HPA axis.
The notion framing the attenuation hypothesis is that under conditions of enduring or chronic stress, cortisol secretion decreases as an adaptive process. Continuing secretion of cortisol under enduring stress can lead to allostatic load, whereby organ systems may become damaged (Goldstein & McEwen, 2002). The exact mechanism whereby chronic stress attenuates a cortisol response remains unknown, but speculation is that trauma may affect brain functioning at the hypothalamus, pituitary, or neuronal receptor levels. Attenuated reactivity may reflect an effort to cope with activating stimuli by withdrawing from these stimuli. Overall, low HPA activity may be a protective factor against emotional perturbations for maltreated youth. However, low HPA activity may have adverse health consequences as well. Although most attention to date has been on the health consequences of heightened cortisol levels, some recent research indicates that attenuation of basal cortisol levels has physical health as well as mental health implications. For instance, low cortisol disrupts metabolism, resulting in low blood sugar and a lack of glycogen required for energy. In addition, low cortisol is related to atypical depression and autoimmune disorders (Sternberg & Gold, 2002), Although there is scant research examining the relation between basal cortisol levels in contrast to cortisol reactivity to a social stressor, it is possible that both indices are different ways of measuring the same phenomena and thus have similar implications for physical and mental health problems across the life span. This hypothesis should be examined in future studies.
Despite its limitations, this study has important implications for future treatment research. Heim et al. (2009) showed that a history of childhood trauma was related to treatment outcomes, potentially due to neurobiological changes associated with childhood trauma. Our findings regarding differences in cortisol reactivity in the different maltreatment subgroups indicate the possible need for different and/or more intense interventions. Lower cortisol reactivity at a later point in time may be a diagnostic indicator of the need for more intensive treatment, particularly because it is linked to disruptive and antisocial behavior as well as other emotional problems that have important implications for an adolescent’s development. Research on the effects of treatment with at-risk youth suggest that effects of psychosocial interventions on cortisol reactivity may even mediate the protective effect of treatment on behavioral problems later (Brotman et al., 2007; O’Neal et al., 2010). Understanding maltreatment as a risk factor for HPA axis functioning is an important step in identifying and intervening effectively with youth who may be particularly at risk based on abnormal stress reactivity.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported, in part, by NIH grant R01 HD 39129, Penelope K. Trickett, principal investigator, and a grant from the Alcohol Beverage Medical Foundation (ABMRF), Elana Gordis, principal investigator.