Childhood Poly-victimization Is Associated With Elevated Body Mass Index and Blunted Cortisol Stress Response in College Women

Abstract

Background

Childhood interpersonal violence is linked to obesity and central adiposity in adulthood. Victimization experiences are likely to co-occur within individuals, but few studies have examined poly-victimization in the context of obesity and central adiposity.

Purpose

The aim of this study is to evaluate the relationship between poly-victimization and body mass index (BMI) and waist circumference (WC) and to explore whether dysregulation of the hypothalamic-pituitary-adrenal axis, as measured by the cortisol stress response, mediates the relationship.

Methods

Healthy undergraduate women were recruited for a laboratory study from an online survey that assessed six different childhood victimization experiences: physical abuse, sexual abuse, peer violence, intimate partner violence, community violence, and witnessing violence. Forty-four women were categorized as poly-victims (2–5 types of violence exposures) and 48 were controls (0 types of violence exposures). Salivary cortisol was measured before and after the Trier Social Stress Test. Cortisol stress response was analyzed by the area under the curve with respect to ground (AUC_G).

Results

Compared with controls, poly-victims had higher BMI (B = 2.03, p = .04) and lower cortisol AUC_G (B = −6.70, p < .01), independent of covariates. Poly-victims showed blunted AUC_G in response to the stress task compared with controls, independent of covariates (B = –6.70, p < .01). Bootstrapping tests of mediation showed that cortisol response was not a significant mediator of the relationship between poly-victimization and BMI. Secondary analyses among poly-victims showed that the more frequent the violence exposure the higher the BMI.

Conclusions

Childhood poly-victimization is linked to higher BMI and blunted cortisol responses in young adult women. Assessment of multiple forms of childhood victimization is recommended.

Young adult women with histories of poly-victimization, compared to those without, have higher body mass index and blunted cortisol response to stress.

Introduction

Interpersonal violence is defined as behavior that threatens, attempts, or causes physical harm against another person. The literature on childhood interpersonal violence consistently reports positive relationships between exposures to violence (physical abuse, sexual abuse, peer bullying, community violence) and adult obesity or elevated body mass index (BMI) [1–3].

Childhood physical abuse and sexual abuse are also linked to central adiposity [4–7], as measured by waist circumference (WC), suggesting that childhood interpersonal violence may predict abdominal deposition of body fat in adulthood.

Research in the field of childhood interpersonal violence suggests that children who are exposed to one form of violence are two to three times more likely to be exposed to other types of violence [8]. Many previous studies examine only one type of violence exposure without considering how other types may influence BMI. The co-occurrence of different types of violence, termed poly-victimization, may partly explain the similar elevated BMI outcomes across different types of violence. Furthermore, some evidence has suggested a dose–response relationship between the number of exposures to adverse experiences and BMI [9, 10]. These findings suggest that it would be useful to examine the associations between exposure to multiple forms of interpersonal violence and BMI.

Of primary importance is the consideration of how poly-victimization may “get under the skin” and lead to later risk for obesity and central adiposity. Poly-victimization may be linked to increasing BMI through the hypothalamic-pituitary-adrenal (HPA) axis. Greater cortisol response to laboratory stressors is associated with higher BMI in children 8–13 years old [11]. Individuals with Cushing syndrome, sometimes called hypercortisolism, have marked abdominal obesity and an adrenalectomy reverses obesity. Cortisol promotes differentiation of adipocyte precursors into adipocytes and stimulates lipogenesis in the presence of insulin [12]. Visceral fat is particularly sensitive to circulating cortisol because visceral adipose tissue has greater blood flow and up to four times the number of glucocorticoid receptors compared with peripheral fat [13]. A recent qualitative review suggested a consistent pattern of cortisol hyperreactivity and abdominal adiposity, even independent of BMI [14].

Nonetheless, the empirical evidence has been inconsistent when examining the relationship between childhood interpersonal violence or poly-victimization and HPA axis activity. Several studies suggest that physical or sexual abuse predict hyperactivity in response to mild laboratory stressors [15, 16], but other studies of maltreatment showed a null relationship [17] or cortisol hypoactivity [18–22].

The current study examines the relationship between childhood poly-victimization and BMI, and the potential mediating role of cortisol responsivity, while addressing limitations of previous literature. These include measuring underexamined victimization experiences (e.g., intimate partner violence, witnessing violence), measuring height and weight rather than using self-report, and controlling for confounding variables (e.g., socioeconomic status [SES] and family history of obesity). This is also the first study that examines the relationship between poly-victimization and central adiposity, measured by WC, which predicts morbidity and mortality independent of BMI [23]. This study tests the following hypotheses: Women reporting childhood poly-victimization would have higher BMI and WC compared with women who report no poly-victimization (controls). We did not have directional hypotheses about cortisol responses to a laboratory stressor but did expect that the cortisol response would statistically mediate the relationship between poly-victimization and BMI/WC.

Methods and Materials

Participants

The laboratory study included 92 participants; 44 were identified as poly-victims (two to five types of violence exposures; see definition below) and 48 were controls (zero types of violence exposures). Eligible participants were recruited from the psychology subject pool at the University of Pittsburgh and were asked to complete a survey on childhood experiences. Inclusion criteria for responding to the survey were being female and being at least 18 years of age. A total of 876 participants provided complete data for the survey, and 513 (58.6%) agreed to be contacted for a follow-up laboratory study.

Inclusion criterion for the laboratory study was endorsing zero or two or more exposures to violence. Within the subset of 513 women, ~52% reported zero exposure to violence and 21% reported two or more exposures to violence. Participants who reported one exposure to violence (27%) were excluded from laboratory participation. Women were also excluded from laboratory participation if they (i) used medications associated with weight gain/loss (e.g., Haldol, Lithium, Depakote, Glucocorticioids, Orlistat), (ii) had been diagnosed or treated for Post-Traumatic Stress Disorder, (iii) had Cushing’s disease, high blood pressure, type II diabetes, or immune disorders, or (iv) used recreational/illegal drugs more than two times per week. These specific criteria have been shown to be associated with altered cortisol levels [24] or BMI.

Once eligible participants were identified, names and contact information were copied into a document without subsequent labeling of group membership. Thus, at the point of the recruitment phone calls for the laboratory session, investigators were blinded to group membership, and recruitment ceased once cells were filled. The goal was to recruit 90–100 participants (45–50 each in control and poly-victim groups) so that analyses could detect a medium effect size (f² = .09–.12). Power analyses were calculated assuming 1–4 independent variables, degrees of freedom for the error variance, significance level of .05, and power of .80. A total of 228 eligible participants were contacted and asked to be a part of a laboratory study to obtain the 92 participants who completed the laboratory study protocol. Many participants who chose not to participate reported that they had already completed subject pool requirements. Cortisol analyses included 87 participants: three were taking steroid medication for acute infections, one had unmeasurable cortisol, and one was sick on the day of the laboratory session. Each survey and laboratory participant signed separate informed consent forms approved by the University of Pittsburgh Institutional Review Board and received course credit for participation.

Recruitment Survey

A short questionnaire addressed demographic characteristics and background information, including participants’ age, race/ethnicity, current medications, health conditions, recreational/illegal drug use, physical activity, and parental obesity. Alcohol use was addressed by the question: “How many servings of alcohol do you drink per week, on average? A serving is equal to 12 oz of beer, 4 oz of wine, or 1.25 oz of hard liquor.” Physical activity was assessed by the following item: “Structured physical activity is activity that lasts at least 10 minutes in a row without breaks, with intensity equal to or greater than a brisk walk. On average, how many minutes (or hours) per week do you engage in structured physical activity?” Parental obesity was identified by a positive to the following question: “Is your mother or father obese?” This item has been suggested as a reliable marker of family history of obesity [25].

All interpersonal violence questionnaires inquired about experiences before the age of 18. Participants completed the physical and sexual abuse scales, five items each, from the Childhood Trauma Questionnaire (CTQ) [26], with responses on a five-point scale ranging from never true to very often true. Pre-established clinical cutoff scores were used to categorize women as positive for physical or sexual abuse. Peer violence and intimate partner violence were assessed by three items each [27], with responses of no, once or twice, and more often. Women who responded with more often to any item were categorized as scoring positive for peer or intimate partner violence. Community violence (occurring to participant) and witnessing violence (occurring to others in the community) were measured with 11 items and 14 items, respectively, by the Community Experiences Questionnaire [28], with items rated on a 4-point scale from never to lots of times. Participants were classified as positive for community violence if they reported (at a minimum) any item of direct victimization occurring a few times or three separate items occurring once. Due to high reported rates of witnessing violence, the threshold for this scale was increased to identify individuals who were victimized by witnessing violence. Participants were classified as positive for witnessing violence if they reported (at a minimum) three items that occurred with a frequency of a few times or lots of times. See Supplementary Materials for additional information regarding violence questionnaires. To permit secondary analyses of frequency of victimization and to equate across the response formats, the continuous scores for each scale were standardized within the poly-victim group and averaged.

Study Protocol

The laboratory session followed a protocol (with minor modifications) suggested by the MacArthur Foundation for conducting the Trier Social Stress Test (TSST) and measuring salivary cortisol response [29, 30]. Laboratory sessions began between 2:30 and 3:30 pm and lasted for 80 min. Participants were instructed to refrain from eating or drinking (excluding water) for 3 hr before the session. They were also asked to refrain from smoking, drinking alcohol, or exercising on the day of the laboratory session. Upon arrival, participants began a 20-min pre-task rest period. During this time, they answered a questionnaire measuring health behaviors and current mood and relaxed with magazines. After the pre-task rest period, the first saliva sample was obtained.

At the beginning of stress protocol, the experimenter informed the participant that she would have to deliver a speech in front of an “expert committee” with the goal of convincing the committee that she should be hired for her ideal job, which would be followed by a math task of subtracting serial sevens from 100. The participant was then given 5 min each for speech preparation, speech delivery, and math subtraction. The expert committee was trained to respond to the participant’s performance in a nonaccepting manner (e.g., not smiling, stony-faced, and occasionally writing comments on a clipboard).

Following completion of the stressor, the participant completed questionnaires evaluating the task and labeling current mood. The participant then began a 30-min post-stress recovery period during which she had ad libitum access to snack food while watching a travel video. Approximately 10 min within the recovery period (~30 min after initiation of the stressor), the second saliva sample was obtained. At the end of the recovery period (~50 min after initiation of the stressor), the participant provided the third and final saliva sample. At the conclusion of the video, participants were weighed, measured, and debriefed.

Laboratory Measures

The following health behaviors occurring on the day of and the day previous to the laboratory session were measured due to their associations with cortisol stress response [24]: exercise, eating, tobacco use, alcohol use, medication use, previous night sleep duration, menstrual cycle (days since the beginning of last menstruation), and perceived health status. Childhood SES was measured using the Subjective Social Status Ladder [31]. The state version of the Positive and Negative Affect Schedule (PANAS) [32] was administered to participants during the pre-task rest period, post-task recovery period, and post-video period. Responses within the positive affect and negative affect categories were summed, with higher scores indicating higher levels of respective affect. After completing the stress protocol, participants responded to a 10-item task-rating questionnaire used in previous studies [33]. Responses were on a seven-point Likert scale. Scores were averaged with higher scores indicating strong task performance/positive interpersonal evaluation/low task difficulty and lower scores indicating poor task performance/negative interpersonal evaluation/high task difficulty. See Supplementary Materials for additional information on laboratory questionnaires.

Weight was measured without shoes and in light indoor clothing using a balance beam scale. Height was measured without shoes using a fixed stadiometer. BMI was calculated as weight over height (kg/m²). WC was measured in centimeters at the level of the natural waist, defined as the narrowest part of the torso as seen from the anterior aspect.

Three cortisol samples were obtained: (i) just prior to the stressor, (ii) ~30 min after initiation of the stressor, and (iii) ~50 min after initiation of the stressor. Salivary free cortisol, unbound cortisol that is biologically active, reflects the amount of free cortisol in plasma [34]. Saliva samples were collected with salivettes, plastic vials with cotton dental rolls inside, and frozen until laboratory analysis by Dr. Clemens Kirschbaum’s laboratory at the University of Dresden. Salivary cortisol concentrations were analyzed using a commercially available chemiluminescence assay (CLIA; IBL-International, Hamburg, Germany). The intra- and inter-assay coefficients of variance were both below 8%, suggesting accurate and reliable lab measurement of salivary cortisol. The lower limit of sensitivity was 0.4 nmol/L, and the upper limit of sensitivity was over 110 nmol/L, suggesting adequate lab sensitivity to detect salivary cortisol. Cortisol output was measured by area under the curve with respect to ground (AUC_G) or zero. This is the total area under the curve of all measurements and reflects total cortisol output. AUC_G takes into account both the overall intensity of the response (distance of the measures from the ground) and sensitivity of response (difference between the measurements). AUC_G was calculated using the logged cortisol values to reduce skew.

Statistical Analyses

The poly-victimization variable was dichotomized; control participants were given a score of 0 and poly-victims (summary score of two or more exposures to violence) were given a score of 1. BMI and WC were highly correlated in the full sample (r = .94, p < .001); therefore, BMI and WC were examined as dependent variables in separate models. Study hypotheses were tested using linear regression, with poly-victimization (categorical: controls vs. poly-victims) as the independent variable and BMI or WC as the dependent variable. Secondary analyses examined a continuous measure of victimization based on frequency of exposure within the poly-victim group.

Mediation tests examined whether an intermediate variable (called a mediator) helps to account for the relationship between an independent variable and dependent variable. See Fig. 1 for a visual representation of the proposed mediation paths in this study. Mediation was tested using the bootstrap method [35, 36]. To compute path a, the independent variable was poly-victimization and the dependent variable was cortisol response (AUC_G). To compute path b, the independent variable was cortisol response and the dependent variable was BMI or WC, adjusting for poly-victimization. From the bootstrap sampling distribution, one can obtain the mediated effect (a * b) and 95% confidence interval estimate [37]. Data were analyzed with SPSS software version 19.

Fig. 1.

Proposed model of mediation. BMI body mass index; WC waist circumference.

Covariates

The following variables were examined for significant correlations with poly-victimization, BMI, or cortisol: age, race/ethnicity, childhood SES, family history of obesity, birth control pill (BCP) use, alcohol use, physical activity, laboratory start time, health over the past day, health over the past week, previous night sleep duration, and days since last menstrual cycle. Correlations of p ≤ .10 were included as covariates in regression analyses. Childhood SES was associated with poly-victimization; family history of obesity and alcohol use were associated with BMI; BCP use was associated with cortisol. Thus, analyses predicting BMI or WC controlled for childhood SES, family history of obesity, and alcohol use. Analyses predicting cortisol response controlled for childhood SES and BCP use. See Supplementary Materials for additional information on methods and materials.

Results

Table 1 shows participant characteristics for the full sample and by group. The sample was mostly White and ~18 years of age. About a third of the sample reported having lower childhood SES and about a quarter of the sample had at least one parent with obesity (BMI ≥ 30). Half the participants were taking BCPs. On average, participants consumed less than three servings of alcohol per week and exercised more than 4 hr per week. Sixteen percent of the sample had BMI between ≥25 and <30 (overweight) and 11% of the sample had a BMI ≥30. Six women reported taking selective serotonin reuptake inhibitors or serotonin–norepinephrine reuptake inhibitors (four controls and two poly-victims). Poly-victims and controls showed no differences on age, BCP use, alcohol use, physical activity per week, or family history of obesity.

Table 1.

Sample characteristics

	Full sample (n = 92)	Controls (n = 48)	Poly-victims (n = 44)
Age, M (SD)	18.6 (1.5)	18.4 (0.64)	18.9 (2.1)
Race/ethnicity, % (n)
White	80 (74)	85 (41)	75 (33)
Asian	9 (8)	6 (3)	11 (5)
Black	4 (4)	2 (1)	7 (3)
Mixed	5 (5)	6 (3)	5 (2)
Hispanic	1 (1)		1 (1)
% (n) Lower childhood SES (2–5 on the SES ladder)	33 (30)	25.0 (12)	40.9 (18)
% (n) Taking BCPs	45.7 (42)	39.6 (19)	52.3 (23)
Alcohol servings/week, M (SD)	2.7 (4.0)	2.4 (3.6)	3.0 (4.5)
Physical activity hours/week, M (SD)	4.2 (3.1)	4.7 (3.3)	3.7 (2.7)
% (n) Family history of obesity	23.9 (22)	18.8 (9)	29.5 (13)
BMI, kg/m2	23.9 (4.8)	22.7 (3.7)	25.3 (5.4)
Underweight, BMI < 18.5, % (n)	6.5 (6)	10.4 (5)	2.3 (1)
Normal weight, BMI < 25.0, % (n)	66.3 (61)	66.7 (32)	65.9 (29)
Overweight, BMI 25.0–29.9, % (n)	16.3 (15)	16.7 (8)	15.9 (7)
Obese, BMI ≥ 30.0, % (n)	10.9 (10)	6.3 (3)	15.9 (7)
WC, cm	73.7 (9.6)	71.5 (7.6)	76.1 (11.0)
Average total cortisol level, nmol/L	9.4 (4.8)	10.6 (5.5)	8.1 (3.6)

SES socioeconomic status; BCPs birth control pills; WC waist circumference; BMI body mass index.

The distribution of violence exposure for the 44 poly-victims is presented in Table 2. Community violence and witnessing violence were the most frequent types of experiences, while sexual abuse was the most infrequent experience. Table 3 shows correlations between violence subtypes. Peer violence was correlated with most other subtypes of violence exposure, and intimate partner violence was correlated with the least number of other subtypes of violence exposure.

Table 2.

Characteristics of interpersonal violence within poly-victims (n = 44)

	% (n)
Number of exposures
2	63.6 (28)
3	29.5 (13)
4	4.5 (2)
5	2.3 (1)
Type of exposures
Physical abuse	36.4 (16)
Sexual abuse	9.1 (4)
Peer victimization	18.2 (8)
Intimate partner victimization	20.5 (9)
Community violence	79.5 (35)
Witnessing violence	81.8 (36)

Table 3.

Correlations between violence exposures

	Physical abuse	Sexual abuse	Peer violence	Intimate partner violence	Community violence	Witnessing violence
Physical abuse	–	.18	.27**	–.15	.29**	.46**
Sexual abuse	.05	–	.31**	–.07	.05	.27**
Peer violence	.13	.17	–	.29**	.24*	–.01
Intimate partner violence	–.08	–.03	.66**	–	.19	.04
Community violence	.34**	.05	.33**	.13	–	.66**
Witnessing violence	.37**	.23*	.27**	.08	.58**	–

Numbers above the diagonal are Spearman’s correlations between measures using threshold scores. Numbers below the diagonal are Pearson’s correlations between measures using continuous scores.

*p ≤ .05, **p ≤ .01.

There were no group differences on the task-rating questionnaire, t(90) = –0.52, p = .60. In addition, controls and poly-victims reported similar decreases in PANAS positive mood, t(90) = –1.06, p = .29, and increases in PANAS negative mood, t(90) = –0.23, p = .82, across the stress task.

Univariate linear regressions showed that poly-victimization was associated with higher BMI (B = 2.61, p < .01) and higher WC (B = 4.65, p = .02). The relationship between poly-victimization and BMI remained significant after controlling for childhood SES, family history of obesity, and alcohol use (B = 2.03, p = .04). The relationship between poly-victimization and WC became a trend after adjusting for childhood SES, family history of obesity, and alcohol use (B = 3.63, p = .07). Within poly-victims, secondary analyses showed that poly-victimization measured continuously was significantly associated with BMI in univariate (B = 4.19, p = .03) and adjusted models (B = 4.27, p = .04). Poly-victimization measured continuously was significant when predicting WC in univariate models (B = 8.09, p = .04) but became nonsignificant in adjusted models (B = 6.90, p = .11).

Cortisol (logged values) at baseline did not significantly differ between controls and poly-victims, t(85) = 1.22, p = .22. Univariate linear regressions showed that poly-victims had significantly lower AUC_G (B = –5.78, p = .02), and this relationship persisted after adjusting for childhood SES and BCP use (B = –6.70, p < .01). In sum, poly-victims had lower total cortisol output compared with controls. See Fig. 2 for a visual representation of the cortisol stress response. Within poly-victims, secondary analyses showed that poly-victimization measured continuously was not significantly associated with AUC_G in univariate (B = .77, p = .85) or adjusted models (B = .42, p = .92). AUC_G did not significantly predict BMI (B = 0.01, p = .83) or WC (B = 0.02, p = .81). As such, bootstrapping methods to test AUC_G as a mediator between poly-victimization and BMI/WC were nonsignificant.

Fig. 2.

Cortisol output in response to the Trier Social Stress Test. Mean and standard error of cortisol concentrations before and after stress induction in controls (n = 45) and poly-victims (n = 42).

Discussion

Rather than examining the influence of individual types of interpersonal violence, this study examined the construct of poly-victimization in relationship to BMI and WC in young adult women. Our definition of poly-victimization examined a combined index of community violence, witnessing violence, and intimate partner violence as well as peer and intimate partner violence and physical and sexual abuse. Frequency rates from the recruitment survey showed that among the participants reporting at least one violence exposure, 45% of the participants endorsed two or more exposures. A study using similar thresholds with cutoff scores in the CTQ found that 38% of victims were poly-victims in a sample of 967 men and women [38]. Poly-victimization is not rare; in fact, a substantial portion of victims experience multiple types of violence from various sources.

As predicted, results showed that women reporting poly-victimization had greater BMI than women reporting no victimization. On average, poly-victims had BMIs that were 2.6 units higher. For example, a woman who is of average height (5 feet, 4 inches) would weigh 15 pounds more if she were a poly-victim than if she were a control. Furthermore, within poly-victims, there appears to be a dose–response relationship. Increasing frequency of exposure to victimization is related to increasing BMI. The finding that poly-victimization is positively associated with BMI is consistent with previous literature linking physical abuse, sexual abuse, or peer bullying to BMI [1–3]. Many studies do not control for other variables that may confound the relationship between poly-victimization and BMI. When our analyses adjusted for childhood SES, family history of obesity, and alcohol use, poly-victimization continued to be associated with BMI.

Our findings are similar to the results from Lovallo et al. [21], which reported a trend for significantly different BMI means in their study of lifetime adversity in young men and women. Female poly-victims in the Lovallo et al. study had a mean BMI of 25.0, and the no victimization group had a mean BMI of 23.0. Lovallo et al. assessed poly-victimization using one item each to assess mugging, robbery, sexual assault by relative, sexual assault by nonrelative, and not living with biological mother/father for 6 months. In contrast, the present study used measures ranging from 3 to 14 items per violence subtype, increasing validity and reliability of violence measurement. In addition, we utilized cutoff scores that generally reflected at least moderate frequency of victimization experiences.

Similar results were found for the outcome of WC; women reporting poly-victimization had 4.62 cm (1.8 inches) higher WC than women reporting no victimization. Mean WCs for both groups were in the healthy range (i.e., WC < 88 cm). When covariates were included in the model, the relationship between poly-victimization and WC only trended at p < .07. This is consistent with the pattern in the literature of a somewhat weaker relationship between violence exposure and WC, compared with the outcome of BMI [6, 7]. Results approached significance in the full model, suggesting value in examining the relationship between poly-victimization and WC in future studies.

We hypothesized that poly-victimization would be related to cortisol response to stress. The findings showed that women reporting poly-victimization had lower AUC_G, which reflects the intensity and sensitivity of the cortisol response. Based on the suggested normative cortisol response to stress [39], it appears that poly-victims in the present sample show a blunted cortisol response. The effect of lower overall cortisol response does not seem to be driven by lower baseline levels, as they did not differ between controls and poly-victims. Furthermore, the finding is not explained by diminished psychological responses to the stressor, as there were no group differences regarding impressions of the TSST or changes in negative or positive affect across the stress task. Interestingly, previous literature on interpersonal violence exposures tends to primarily focus on childhood physical and sexual abuse and underemphasize community and witnessing violence [3], perhaps missing an important and relatively large subgroup of individuals at increased physiological risk.

Several studies show a similar pattern of blunted cortisol reactivity in individuals with a history of poly-victimization. Lovallo and colleagues [21] found a dose-dependent cortisol response to the TSST, such that cortisol response in men and women was progressively lower with increasing lifetime adverse events (most occurring before the age of 18). Another study of adults reported that the total CTQ score, which included measures of emotional and physical neglect, was related to lower overall cortisol response to the TSST [18]. Finally, MacMillan et al. [22] completed a study of female adolescents about 14 years of age. Although there was no direct test of poly-victimization and cortisol response, a majority of participants in the childhood maltreatment group had more than one exposure to victimization (including physical and emotional neglect). The authors reported that maltreated children had flattened cortisol reactivity and recovery slopes after the TSST compared with controls.

There may be several reasons why the present study, as well as others, found that poly-victims showed hypoactivity rather than hyperactivity. The relationship between poly-victimization and cortisol hypoactivity may be more common in healthy subjects who were not recruited for studies of depression [16, 40]. Duration or timing of victimization may be another factor related to differential HPA axis responses to stress. For example, the number of early life traumatic events occurring before puberty showed an inverse relationship with peak plasma cortisol response to the TSST, but the number of traumas after puberty showed no significant relationship with cortisol [41]. The women in the present study may have experienced victimization earlier in the lifespan.

Previous literature has suggested ways in which victimization leads to blunted cortisol response to stress. Poly-victimization, possibly indicating a history of chronic victimization, may cause persistently elevated HPA axis activity which then develops into dampened cortisol responses to stress [42–45]. The dampened response to stress may be a result of down-regulation of the HPA axis. Some studies have attempted to identify the point in the lifespan in which HPA axis dampening may occur. For example, one research group followed a sample of sexually abused and nonabused girls from age 6 through 32. They reported that the abused participants showed basal hypercortisol in childhood and then hypocortisol in adulthood, in addition to blunted cortisol response to stress in late adolescence [44, 45]. The present sample may provide additional evidence that an attenuated cortisol response is measurable by late adolescence.

The blunted cortisol response to stress did not mediate the positive association between poly-victimization and BMI or WC. Although many previous studies speculate that HPA axis dysregulation is the mechanism linking poly-victimization to obesity/central adiposity, this study empirically tested the hypothesis. Cortisol response to stress was unrelated to BMI or WC. Another study also failed to show an association between cortisol reactivity and BMI [46]. However, the results are inconsistent with two previous reports that cortisol reactivity is positively related to overweight/central adiposity in children and women [11, 47]. Overall, results from studies suggest that HPA axis activity may be at best a weak physiological pathway contributing to overweight and central adiposity among those exposed to victimization. Other physiological systems may be better targets as possible mediators between poly-victimization and BMI/WC, for instance, inflammatory processes or sympathetic nervous system activity.

The study has several limitations. First, the study was cross-sectional and unable to make claims regarding causality of the observed relationships. The data are consistent with longitudinal studies that link victimization to steeper trajectories of weight gain by young adulthood [48]. However, it is also possible that heavier children are more likely to be victimized, whether because it makes them targets for peer bullying [49] or because higher rates of obesity are found in neighborhoods with increased likelihood of community violence [50]. As only undergraduate women participated in this study, the results may not be generalized to men or to participants with varying age, SES, and differing proportions of race/ethnicity groups. Despite being less consistent and generally weaker, several studies that included men have shown that victimization is linked to BMI and cortisol response to stress [21, 51]. The women were able to attend college, which may reflect a level of resilience or economic advantage that could be protective in the relationship between poly-victimization and health outcomes; one might expect stronger relationships in samples of low-SES participants. A measure of dietary intake was not included and thus could not be examined in models with the outcome of BMI. Finally, measures assessing peer, intimate partner, community, and witnessing violence did not have validated cutoff scores based on the psychometric literature; thus, cutoffs were based on a priori categories by authors taking into account sample distributions and face validity.

Strengths of the current study include measurements of underexamined violence exposures such as community and witnessing violence, as well as identifying violence exposures of at least moderate frequency. Validated measures of violence exposure were included, such as the CTQ and the Community Experiences Questionnaire. Because type of violence exposure does not occur in isolation, this study had a more ecologically valid measure of victimization. Models were also adjusted for confounding factors that influence poly-victimization, BMI, and cortisol. Finally, this study furthered our exploration into how childhood adversity affects physiological pathways and important health outcomes.

Continuing to work on the mechanisms connecting poly-victimization to BMI and WC will improve understanding of how poly-victimization affects morbidity and mortality. Chronic and stress-induced inflammation has been linked to victimization and BMI and warrants further study as a plausible mechanism [52, 53]. Interview-based assessments may improve poly-victimization construct validity and reliability and allow for questions on timing, duration, perpetrator, and severity of victimization experiences. Finally, a longitudinal assessment of children and adolescents may clarify the theory of changing HPA dysregulation during childhood and adolescent development. Future research would benefit from adding various health-related measures to the longitudinal assessments, such as BMI, visceral adiposity, immune factors, and insulin sensitivity. Understanding the trajectory of mental and physical health after poly-victimization would enable interventions to be tailored to development stage and critical periods of health risk.

In conclusion, the results from this study showed that poly-victimization is associated with BMI and that the more frequent the violence exposure the higher the BMI. This relationship does not appear to be accounted for by cortisol activity in response to stress. Considering multiple sources of interpersonal victimization in childhood is important when examining relationships with physical health.

Compliance With Ethical Standards

Authors’ Statement of Conflict of Interest and Adherence to Ethical Standards The authors declare that they have no conflict of interest.

Ethical Approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed Consent Informed consent was obtained from all individual participants included in this study.