The mediating impact of PTSD symptoms on cortisol awakening response in the context of intimate partner violence

Abstract

Multi-level modeling examined the association between cortisol awakening responses (CAR) and different PTSD symptom clusters in a sample of 158 female participants presenting with intimate partner violence-related PTSD. Results revealed that arousal over the past week and month, respectively ([β = −0.124, z = −2.33, p = .028; β = −.147, z = −2.19, p = .028]) significantly moderated the trajectory of cortisol levels, and emotional numbing symptom severity (over the past week [β = −0.122, z = −2.07, p = .076]) was found to be trending toward significance. In each case higher symptom severity was associated with flatter CAR slopes compared to those with lower symptom severity.

Assessing PTSD symptom clusters in relation to cortisol may better inform future interventions compared to studies that assess PTSD globally. Our findings suggest a subtype of PTSD patients displaying higher levels of arousal may be more likely to experience alterations in HPA axis functioning.

1. Introduction

While a majority (89 %) of individuals will experience a traumatic event during their lifetimes, only a minority of trauma survivors (12 %) will develop posttraumatic stress disorder (PTSD: Kilpatrick et al., 2013). Thus, research has focused on examining risk factors associated with PTSD so that at-risk individuals can be identified for appropriate intervention (Kilpatrick et al., 2013). Given that a PTSD diagnosis requires exposure to a traumatic stressor, research into understanding biological risk factors has largely focused on examining alterations in the primary stress pathways. In general, PTSD is associated with heightened sympathetic nervous system (SNS) activity (Morris & Rao, 2013); however, studies examining alterations in hypothalamic pituitary adrenal (HPA) axis functioning have produced equivocal results (Meewisse, Reitsma, Vries, Gersons, & Olff, 2007; Pan, Wang, Wu, Wen, & Liu, 2018). Many possible explanations have been posited for these discrepant findings (i.e., differences in methodology, trauma type, and possible uncontrolled spurious variables) (Wust et al., 2000). It is also plausible, however, that the amorphous nature of a PTSD diagnosis may conceal differences in HPA axis functioning. That is, different types of PTSD symptoms may be differentially associated with alterations in cortisol levels. As individuals may meet diagnostic criteria via different symptom presentations, it is possible that a focus on cortisol levels in PTSD (incorporating all symptoms) obscures relationships between specific symptom types and cortisol. The present study aimed to address the inconsistent findings in cortisol research by exploring the relationship between cortisol awakening responses (CAR) and PTSD symptom clusters in a sample of women exposed to intimate partner violence (IPV).

1.1. PTSD

PTSD is a complex diagnosis requiring the presence of symptoms from four symptom clusters (re-experiencing, avoidance, negative alterations in cognition and mood, and arousal and reactivity). Such a diagnostic scheme allows for over 600,000 possible combinations of symptoms that can meet the diagnostic criteria for PTSD, leading researchers to criticize the amorphous nature of PTSD and the utility of a PTSD diagnosis (Galatzer-Levy & Bryant, 2013). This heterogeneity of symptom presentation also likely obscures important subtypes of PTSD patients that may respond better to different treatment approaches and may blur findings regarding biological correlates and predictors of PTSD.

1.2. Cortisol and PTSD

Results from a recent meta-analysis indicate a trend toward lower cortisol levels in those with PTSD compared to both trauma exposed and non-trauma exposed controls (Pan et al., 2018). The relationship was stronger in studies that assessed morning cortisol levels, while no significant group differences were found in cortisol levels collected in the afternoon or evening. This led the authors to hypothesize that cortisol awakening responses (CAR) are less sensitive in those with PTSD compared to controls (Pan et al., 2018). The CAR is the spike in cortisol levels observed within 30 min of awakening and is tracked by assessing cortisol typically at 4 time points (i.e., upon awakening, and 30, 45, and 60 min later) (Wust et al., 2000); thus a less sensitive CAR would be depicted by a flatter curve. Although the Pan et al. (2018) and other meta-analyses (Meewisse et al., 2007) have suggested some consistency in the relationship between PTSD and cortisol levels, individual studies continue to produce differing and often contradictory findings regarding this relationship.

Many factors may be responsible for these discrepant findings including differing relationships between each symptom cluster and cortisol. Prior research examining salivary cortisol and PTSD symptom clusters has suggested that re-experiencing (de Kloet et al., 2007; Goenjian et al., 1996), emotional numbing symptoms (Stoppelbein & Greening, 2015), and arousal symptoms (de Kloet et al., 2007; Wessa, Rohleder, Kirschbaum, & Flor, 2006) were negatively related to cortisol levels, although non-significant relationships have also been reported (de Kloet et al., 2007; Goenjian et al., 1996; Wessa et al., 2006). It is of note that the only study that did not find a significant association between cortisol and arousal symptoms (Goenjian et al., 1996) measured cortisol levels via a single basal morning cortisol assessment while the studies finding significant relationships assessed waking or diurnal curves using an area under the curve approach (de Kloet et al., 2007; Wessa et al., 2006). In contrast, a positive correlation (Wahbeh & Oken, 2013) has been reported between avoidance and salivary cortisol levels although, again, not all studies have found significant relationships (de Kloet et al., 2007; Wessa et al., 2006).

The type of traumatic event may also impact the direction or strength of the relationship between PTSD symptoms and cortisol levels. In particular, prior meta-analyses have found that PTSD symptoms stemming from interpersonal sexual and physical traumas are inversely related to cortisol levels (Meewisse et al., 2007; Morris, Hellman, Abelson, & Rao, 2016), while symptoms following other traumas were not consistently related to cortisol. The present study focused on one particularly common type of interpersonal trauma: intimate partner violence (IPV). IPV is characterized by actual or threatened sexual, physical, or psychological violence by a current or former partner (Pico-Alfonso, Garcia-Linares, Celda-Navarro, Herbert, & Martinez, 2004). Prevalence rates indicate that 15.8 % of women will experience sexual violence by an intimate partner, 22.3 % will experience severe physical violence, and 47.1 % will experience psychological aggression during their lifetimes (Breiding et al., 2014). IPV is often chronic, with revictimization rates ranging from 22 to 93 % (Cattaneo & Goodman, 2005). Additionally, approximately 64% (range = 31%–84.4%) of women who experience IPV develop PTSD (Golding, 1999), a much higher prevalence rate compared to other traumas (on average 12 %) (Kilpatrick et al., 2013). Given the higher prevalence rates of PTSD in IPV and the aforementioned relationship between interpersonal sexual/physical trauma and cortisol, women who have experienced IPV provide an ideal sample in which to examine the link between different PTSD symptoms and cortisol levels.

As prior research has suggested relatively consistent links between violence and cortisol and between PTSD and CAR, it would follow that the relationship between PTSD symptoms and CAR in IPV should be consistently strong; however, research has produced mixed findings. Although chronic IPV victimization (Johnson, Delahanty, & Pinna, 2012; Pinto, Correia-Santos, Costa-Leite, Levendosky, & Jongenelen, 2016; Seedat, Stein, Kennedy, & Hauger, 2003), severity of IPV (Kim et al., 2015; (Valladares, Peña, Ellsberg, Persson, & Högberg, 2009), and IPV-related stress (Suglia et al., 2010) have been associated with lower cortisol levels, a positive relationship has been found between CAR and meeting diagnostic criteria for PTSD (Pinna, Johnson, & Delahanty, 2014) and the severity of PTSD (Johnson, Delahanty, & Pinna, 2008). Further, others have found no relationship between CAR and IPV (Basu, Levendosky, & Lonstein, 2013; Pico-Alfonso et al., 2004). Thus, the present study attempted to explain these prior mixed findings by examining the relationship between CAR and avoidance, re-experiencing, emotional numbing, and arousal symptoms separately. The determination of reliable biomarkers associated with specific symptoms will allow for identification of possible subgroups of PTSD patients who may differentially respond to interventions. Furthermore, some evidence suggests that cortisol levels may be able to be used as a predictor for clinical response to PTSD treatment outcomes (Colvonen et al., 2017). Therefore, fully elucidating the relationship between cortisol and specific symptoms could allow for more targeted and effective treatments.

Based on the literature reviewed above, the following hypotheses were made:

• Hypothesis 1. Higher levels of re-experiencing symptoms in the prior week and month will be associated with flatter CAR trajectories.

• Hypothesis 2. Higher levels of emotional numbing symptoms in the prior week and month will be associated with flatter CAR trajectories.

• Hypothesis 3. Higher levels of arousal symptoms in the prior week and month will be associated with flatter CAR trajectories.

• Hypothesis 4. Lower levels of effortful avoidance symptoms in the prior week and month will be associated with flatter CAR trajectories.

2. Method

2.1. Participants

This study involved multi-site recruitment of women from battered women’s shelters across Northeastern Ohio. A total of 273 women were recruited and screened for PTSD diagnostic criteria stemming from IPV at the initial assessment. Participants had to be residents of one of the six participating shelters and had to report experiencing IPV in the month prior to admission to the shelter. Participants were excluded if they met diagnostic criteria for bipolar disorder or psychosis, substance dependence in the last month, and/or active suicidality with intent and plan. A total of 172 women presented with IPV-related PTSD, met criteria for the study and consented to provide saliva samples for cortisol assays. A total of 158 participants completed saliva samples at four different time points (waking, 30 min, 45 min, and 60 min) across two days. The majority of women identified as either African American (45.6 %) or Caucasian (45.3 %) and heterosexual (93 %). Participants ranged in age from 19 to 59 (M=35.46, SD = 9.16). The sample was highly traumatized with participants reporting experiencing an average of 3.59 (SD = 2.22) different types of traumatic events in their lifetime other than IPV. For further description of traumatic experiences and complete sample demographics, please see Table 1.

Table 1.

Demographic Characteristics of Participants.

Characteristic	n	%
Age (in years)	35 (9)
Sexual Orientation
Heterosexual	147	93 %
Bisexual	11	7%
Race
Caucasian	70	44.3%
African American	72	45.6 %
Native American	1	0.6%
Ethnicity
Non-Hispanic	152	96.2%
Hispanic	5	3.2%
Relationship Status
Single	82	51.9%
Married	17	10.8%
Separated	27	17.1%
Divorced	30	19%
Widowed	2	1.3%
Living with Abuser
No	25	15.8 %
Yes	133	84.2%
Children
No	10	6.3%
Yes	148	93.7%
Pregnant
No	136	91.9%
Yes	12	7.6%
Education
Junior High School	6	3.8%
Some High School	40	25.3%
Graduated from High School/ GED	43	27.2%
Some College	57	36.1%
Graduated from College	12	7.6%
Employed
No	136	86.1%
Part -Time	11	7%
Full-Time	11	7%
Personal Income
< 10,000	137	86.7%
10,000–15,000	12	7.6%
15,000–20,000	3	1.9%
20,000–25,000	2	1.3%
> 25,000	4	2.5%
Trauma Exposure
Beat-up or attacked outside of IPV	98	62%
Rape outside of IPV	93	59%
Robbery, mugging, being held up	63	40%
Death of a close friend or family member by accident, suicide, or homicide	77	49%
Experiencing a serious MVA	57	36%
Seeing someone seriously injured or killed	68	43%
Experiencing injury or property damage due to a fire	24	15%
Experiencing injury or property damage due to a natural or human-made disaster	14	9%
Serious physical injury as a result of a non-MVA related accident	33	21%
Other terrifying or shocking experience	28	18%

2.2. Procedure

The following procedures were approved by the University of Akron institutional review board (IRB), and all participants reviewed and signed informed consent prior to participation. Potential participants responded to posted fliers and brochures describing the research. Research staff also went to shelters during all-house meetings and other key times when residents tended to be home to provide information about the study and answer questions. Interested residents called a confidential research line and completed a brief phone screen for inclusion criteria (e.g., shelter resident and IPV experience during month prior to shelter). Potential participants were scheduled for an assessment with a trained research staff member in a private space within the shelter. All research staff were under the supervision of a licensed psychologist (fourth author). Participants who met inclusion criteria were provided instructions for saliva sample collection to measure cortisol during the first hour of waking across two consecutive days.

2.3. Measures

PTSD.

The Clinician-Administered PTSD Scale (CAPS) (Blake et al., 1995) is a semi-structured interview that assesses diagnostic criteria for PTSD from the Diagnostic and Statistical Manual of Mental Disorders (American Psychiatric & American Psychiatric, 2000) and provides a measure of PTSD symptom severity. DSM-IV TR (American Psychiatric Association, 2013) criteria for PTSD are characterized by three groups of symptoms: re-experiencing the traumatic event, avoidance of reminders associated with the traumatic event (including emotional numbing and effortful avoidance), and hyperarousal. Participants were asked about the frequency and intensity of re-experiencing, emotional numbing, effortful avoidance and arousal symptoms in the past week and the past month related to the IPV. Frequency and intensity of symptoms were assessed on a 5-point Likert scale (0=Never to 4=Daily or Almost Every Day; 0=Absent to 4=Extreme). The frequency and intensity scores were summed for each item and a combined score of three or greater was used to determine PTSD status. The CAPS has demonstrated good internal and concurrent validity with other PTSD measures (Blake et al., 1995). The total CAPS scores for this present study demonstrated strong reliability for past week (α = .94) and past month (α = .94). The four subscales also demonstrated adequate reliability in past week and past month with coefficient alphas respectively of 0.85 and .84 (re-experiencing), .84 and .79 (avoidance), .71 and .68 (effortful avoidance), .82 and .77 (emotional numbing), and 0.83 and .80 (hyperarousal).

IPV.

To be included in the study, participants had to endorse at least one incident of IPV during the month prior to shelter on one of three self-report measures of IPV: Severity of Violence Against Women Scale (SVAW; (Marshall, 1992), Psychological Maltreatment of Women Inventory-Short Version (PMWI-F; Tolman, 1999), or the Stalking Behavior Checklist (SBC;(Coleman, 1997)). When completing all three measures of IPV, participants reported the frequency that each event occurred in the month prior to shelter or indicated if the event did not occur in the past month but did occur in the past year. The PMWI is a 14-item self-report measure that assesses psychological IPV. It is scored on a 6-point Likert scale (1= never to 5= very frequently). Good internal consistency was found in the present study with a coefficient alpha of 0.88. The SBC is a 25-item self-report measure that quantifies exposure to stalking behavior and/or harassment. This measure uses a 6-item Likert scale (1 = never to 6 = once a day or more). The present study found that the SBC had good internal consistency with a coefficient alpha of .90. The SVAW is a self-report measure that assesses the occurrence and frequency of symbolic, mild, moderate, and serious threats of violence (19 questions); mild, minor, moderate, and serious violence (21 questions); as well as sexual aggression (6 questions). The SVAW is scored on a 4-item Likert scale ranging from 1(never) to 4 (4 or more times). The SVAW has established reliability, and the present study found the same (α = .97). These three measures of IPV were used to assure that all types of IPV (i.e., physical, sexual, emotional, threatening behaviors, and stalking) were assessed. Participants’ report of IPV also had to meet criterion A on the CAPS (Blake et al., 1995).

Trauma history.

The Traumatic Stress Schedule (TSS;(Norris, 2006) is a structured interview assessing lifetime history of a range of traumatic events other than IPV (i.e., being attacked, sexual violence or threatened sexual violence, motor vehicle accidents, and death of a loved one due to an accident, homicide, or suicide, etc.). The TSS assesses 10 possible types of traumatic events and outlines 12 follow-up questions when an event is endorsed. In this study, the number of types of events endorsed by participants was summed to assess the frequency of prior trauma experiences. Additionally, rates of trauma other than IPV were assessed with the TSS.

Cortisol.

Cortisol sampling procedures followed Pruessner et al.’s (Pruessner, Kirschbaum, Meinlschmid, & Hellhammer, 2003) recommendations for collecting waking cortisol samples. Saliva samples were collected with Salivette sampling devices (Sarstedt, Newton, NC) at four time points during the first hour after waking: upon waking and 30-, 45-, and 60-minutes after waking (Wust et al., 2000). Participants were asked to complete the saliva samples over two consecutive weekdays (not on Saturdays or Sundays) and to not eat, drink, smoke, or brush teeth before or during the sample period. Participants recorded the exact times they completed the samples, current medications, and other specifics about the collection (i.e. whether they smoked that morning or if they used an alarm upon waking). Saliva samples were collected and stored at −80 °C until sent to be assayed at the Center for Psychobiology and Psychosomatic Research in Trier, Germany. Analysis included a time-resolved immunoassay with fluorescence detection (Dressendörfer, Kirschbaum, Rohde, Stahl, & Strasburger, 1992).

2.4. Data preparation and analytic plan

The primary analyses examined whether the different PTSD clusters would differentially moderate awakening cortisol trajectories. We separately examined each PTSD symptom cluster (arousal, re-experiencing, emotional numbing, and effortful avoidance) reported for the past week and the past month from the CAPS assessment. Emotional numbing and effortful avoidance were examined separately, rather than grouping them as avoidance, to be more consistent with DSM-5 diagnostic criteria(American Psychiatric Association, 2013). Furthermore, a confirmatory factor analysis of DSM-IV criteria identified that the items in the avoidance cluster loaded onto two separate factors: emotional numbing and effortful avoidance (King, Leskin, King, & Weathers, 1998), providing additional justification for examining them separately.

With respect to data preparation, nine (of the 1011) cortisol results were identified as extreme outliers (over 10 standard deviations above the mean). These nine samples were provided by three participants. First, analyses were conducted after assigning these outliers the value of three standard deviations above the mean(Field, 2009). Second, to ensure that retaining these high values was not responsible for significant findings, we also repeated all analyses excluding all samples from the three participants with outlier values. Either way the findings did not change, indicating that results were not due to the presence of outliers.

Cortisol samples were excluded for two additional reasons: (a) if the participant reported smoking during saliva collection or (b) if saliva sampling took more than 2 h to collect. As cortisol samples were collected across two days, if a participant did not complete the saliva collection within 2 h or reported smoking for that day, that day of samples were excluded. However, we conducted analyses both excluding and replacing these samples. Again, results minimally varied; thus, for the following analyses, we report results of the most conservative iteration of exclusions– replacing outlier values and excluding smokers and samples collected outside the two-hour sampling window.¹ For a full description of exclusions, see Fig. 1.

Fig. 1.

Flow Chart of the Final Data Set.

For our analytic plan, multi-level models were conducted with the Stata 11.0 statistical software (StataCorp., 2009). The data were nested under two higher order levels: cortisol samples were nested under day (Day 1 and Day 2) and day was nested under individual yielding a three-level model. Initial analyses were conducted to identify the function of waking cortisol across the four time points (waking, 30-, 45-, and 60-minutes). These analyses were conducted to determine the function of cortisol without any moderators influencing the trajectory. The cortisol data were first examined as a linear model, which was not significant, β = .026, z = 0.18, p = .855. Once the quadratic term was included into the model, both the linear and quadratic terms had a significant effect on the trajectory of cortisol levels (linear function: β = 2.11, z = 4.39, p < .001 and quadratic function: β = −0.696, z = −0.4.53, p < .001). These results demonstrated that the inclusion of a quadratic function significantly improved model fit, which is consistent with prior research (Fries, Dettenborn, & Kirschbaum, 2009); (Clow, Thorn, Evans, & Hucklebridge, 2004). In general, cortisol levels peaked around 30–45 min after awakening and recovered (or declined) after that.

Next, we conducted moderation models applying the following formula:

$$Y={\beta }_0+{\beta }_{\text{x}}+{\beta }_{\text{x}}^2+{\beta }_{\text{w}}+{\beta }_{\text{xw}}+{\beta }_{\text{xw}}^2+e$$

Interaction terms were produced for both linear and quadratic functions. Y indicates the outcome of cortisol level, × describes time (waking, 30-, 45-, and 60-minutes) for the linear function, and x² allows for the incorporation of the quadratic function (i.e. curve over time). The moderator is labeled as w (PTSD cluster); xw is the linear interaction term, whereas x²w provides the quadratic interaction term. This allows for the interpretation of moderating effects for both the linear function (start and recovery) and the quadratic function (peak or drop) of cortisol levels. A stronger linear function suggests that cortisol levels are generally increasing over the sample collection window. A stronger negative quadratic effect reflects a stronger “peak” to cortisol levels combined with a stronger recovery or down-regulation. We also used the Benjamini-Hochberg Significance Test to adjust for multiple testing of symptom reports based on the past week and past month; the reported results reflect adjusted p-values (Thissen, Steinberg, & Kuang, 2002).

3. Results

For descriptive statistics see Table 1 and 2, and for correlations among variables see Table 3.

Table 2.

Descriptive Statistics.

	Min	Max	Mean	SD
CAPS_1wk	0.00	115.00	67.0000	24.40255
CAPS_1mo	29.00	121.00	74.8313	20.74591
RE_Exp_1wk	0.00	38.00	20.2188	8.87628
RE_Exp_1mo	3.00	40.00	22.4063	8.01359
Arous_1wk	0.00	38.00	21.4875	8.64971
Arous_1mo	6.00	38.00	24.2000	7.13553
EffAvoid_1wk	0.00	16.00	9.4313	3.84057
EffAvoid_1mo	0.00	16.00	9.8938	3.50130
EmoNumb_1wk	0.00	34.00	15.8625	8.14143
EmoNumb_1mo	4.00	37.00	18.3313	7.03694

Table 3.

Correlations of PTSD Clusters.

	1	2	3	4	5	6	7	8	9	10
1 CAPS tot-1wk	1	.864**	.859**	.768**	.885**	.776**	.665**	.523**	.807**	.626**
2 CAPS tot-1mo		1	.776**	.847**	.709**	.880**	.640**	.657**	.690**	.765**
3 ReExp- 1wk			1	.878**	.684**	.642**	.485**	.385**	.528**	.444**
4 ReExp- 1mo				1	.616**	.669**	.477**	.468**	.464**	.445**
5 Arous- 1wk					1	.777**	.520**	.412**	.600**	.395**
6 Arous-1mo						1	.565**	.525**	.535**	.556**
7 EffAvd-1wk							1	.797**	.440**	.375**
8 EffAvd-1mo								1	.335**	.374**
9 EmoN-1wk									1	.797**
10 EmoN-1mo										1

Our primary analyses examined whether PTSD symptom clusters (arousal, re-experiencing, emotional numbing and effortful avoidance) differentially moderated the CAR over the four time points (waking, 30-, 45-, and 60-minutes). Although the quadratic term was included in the model, none of the clusters significantly moderated the quadratic function; the only moderating effects were shown for the linear function (i.e. the direct line from waking to 1 -h time points). Post hoc analyses were also performed assessing the linear trajectory of cortisol between 45-and 60-minutes to further elucidate the recovery of the cortisol awakening response.

The linear interaction for arousal symptoms by time demonstrated a significant effect on cortisol levels, for the past week: β = −.124, z = −2.33, p = .028 and past month CAPS score: β = −0.147, z = −2.19, p = .028. The moderating effect was specific to the linear function of the data (i.e. the start and recovery of the data). Simple slopes analyses demonstrated that individuals with higher arousal symptom severity were more likely to have a flatter slope (past week β = 1.06; past month β = 1.11) at 1 -h in comparison to individuals with lower arousal symptom severity (past week β = 3.19; see Fig. 2, past month β = 3.21; see Fig. 3). Lower arousal symptom severity was associated with a steeper slope of cortisol levels from waking to recovery at 1 -h. The group with higher arousal symptom severity showed a flatter linear recovery trajectory (i.e., from 45 min to 60 min) (past week β = −1.59, past month β = −1.90) compared to the group with lower arousal symptoms (past week β = −2.42, past month β = −2.16).

Fig. 2.

CAR for high compared to low arousal symptoms reported over the past week. The upMod indicates one standard deviation above the mean and the dwnMod indicates one standard deviation below the mean on arousal symptoms in the past week.

Fig. 3.

CAR for high compared to low arousal symptoms reported over the past month. The upMod indicates one standard deviation above the mean and the dwnMod indicates one standard deviation below the mean on arousal symptoms in the past month.

Alternatively, emotional numbing symptoms for the past week demonstrated a linear moderating effect of cortisol across time that was trending in significance, β = −0.122, z = −2.07, p = .076. Decomposition of this interaction with simple slopes analysis indicated that individuals with higher emotional numbing symptoms also had a flatter cortisol trajectory upon waking (β = 1.10) than participants with lower emotional numbing (β = 3.09; see Fig. 4). Additionally, individuals with higher emotional numbing symptoms showed a flatter trajectory in their recovery between 45- and 60-minutes post-awaking (past week β = −1.66) compared to those with low emotional numbing symptoms (past week β = −2.35).

Fig. 4.

CAR for high compared to low emotional numbing symptoms reported over the past week. The upMod indicates one standard deviation above the mean and the dwnMod indicates one standard deviation below the mean on emotional numbing symptoms in the past week.

No other PTSD clusters significantly moderated cortisol trajectories [re-experiencing symptoms for the past week: β = −0.060, z = −1.11, p = .532, re-experiencing symptoms for the past month: β = −0.038, z = −0.63, p = .536, effortful avoidance symptoms for the past week: β = −.071 z = −0.57, p = .570, effortful avoidance symptoms for the past month: β = .140, z = 1.05, p = .570, and emotional numbing symptoms for the past month: β = −0.050, z = −0.69, p = .651]. Table 4 presents all moderating effects for the linear function across clusters.

Table 4.

Moderating effects for all PTSD clusters.

O		ß (SE)	z	p	LLCI	ULCI
Arousal	past week	−.124 (.053)	− 2.33	.028	−.228	−.020
	past month	−.147 (.067)	− 2.19	.028	−.279	−.016
Re-Experiencing	past week	−.060 (.054)	− 1.11	.532	−.166	.046
	past month	−.038 (.060)	− 0.63	.536	−.155	.079
Emotional Numbing	past week	−.122 (.059)	− 2.07	.076	−.238	−.007
	past month	−.050 (.072)	− 0.69	.651	−.191	.092
Effortful Avoidance	past week	−.071 (.126)	− 0.57	.570	−.318	.175
	past month	.139 (.133)	1.05	.570	−.122	.401
Total CAPS score	past week	−.039 (.019)	−2.03	.086	−.077	−.001
	past month	−.024 (.023)	− 1.02	.370	−.069	.022

Notes. Confidence intervals are included as an additional metric of statistical significance.

Although we hypothesized that different symptom clusters would differentially impact cortisol waking curves, we also examined the total CAPS score as a potential moderator. CAPS scores for past week symptom severity (β = −0.039, z = −2.03, p = .086) and past month (β = −0.024, z = −1.02, p = .307) did not significantly moderate the trajectory of cortisol.

Identical analyses without excluding participants who smoked during the sampling period and samples collected outside the sampling window produced similar results. Without exclusions, total CAPS scores for the past week and month continued to not significantly moderate cortisol levels across time. Similarly, the significant cluster findings from the initial analyses remained; arousal symptoms for the past week: β = −0.106, z = −2.04, p = .042, and the past month β = −0.133, z = −2.08, p = .038, continued to significantly moderate cortisol trajectory.

4. Discussion

The present study sought to examine the moderating effects of different PTSD clusters on the CAR in women presenting with PTSD from IPV. Results indicated that arousal symptom severity reported over the past week and past month significantly moderated the trajectory of cortisol levels. That is, women with higher arousal symptoms had a cortisol trajectory with a flatter slope and less fluctuation during the awakening response than women with lower arousal symptom severity scores. Emotional numbing moderated cortisol levels in a similar manner but at the level of a trend. Overall, these findings suggest that alterations in stress system functioning may be specific to arousal and possibly emotional numbing symptoms.

Although results have varied with PTSD symptom clusters and HPA activity, arousal and emotional numbing have had stronger associations with SNS activity compared to other PTSD symptom clusters in prior research (Feuer, Nishith, & Resick, 2005; Foa, Riggs, & Gershuny, 1995; Krause, Kaltman, Goodman, & Dutton, 2006; Taylor, Kuch, Koch, Crockett, & Passey, 1998). Arousal symptoms are central to fight and flight responses, while emotional numbing is believed to produce analgesic effects that prevent responses to pain while in the context of danger (Feuer et al., 2005). Additionally, these symptom clusters are repeatedly grouped together (Feuer et al., 2005; Taylor et al., 1998), with arousal being related to uncontrollable stress and emotional numbing being an automatic unconscious response to arousal (Foa, Zinbarg, & Rothbaum, 1992). Emotional numbing may be a way for people to disconnect from the arousal and cope physiologically to repeated abuse leading to the relationship between symptomatology and the psychogenesis of cortisol alterations (Krause et al., 2006). Furthermore, arousal symptoms have a stronger predictive association with emotional numbing than any other PTSD symptom cluster (Tull & Roemer, 2003). Purposeful avoidance and re-experiencing symptoms have also been related to emotional numbing; however, compared to hyperarousal symptoms, the impact was much weaker (Flack, Litz, Hsieh, Kaloupek, & Keane, 2000).

It is not surprising that we did not find a relationship between purposeful avoidance and re-experiencing symptoms and HPA axis functioning as these symptoms have been inconsistently related to cortisol levels in prior research (Feuer et al., 2005; Flack et al., 2000; Krause et al., 2006; Taylor et al., 1998). Inconsistencies in findings are likely explained by the complex interplay between symptom clusters; dysregulation of SNS is likely driven by arousal and numbing symptoms for the aforementioned reasons. Symptom clusters often increase and maintain other symptoms (i.e., purposeful avoidance, like emotional numbing, is hypothesized to occur in the context of high arousal with the goal of escaping high emotional states due to re-experiencing symptoms linked to the traumatic experience) (Asmundson, Stapleton, & Taylor, 2004). While this is an oversimplification of the interplay between symptoms, it helps explain how re-experiencing and purposeful avoidance may indirectly impact HPA functioning through arousal and emotional numbing. This would explain the inconsistent and weaker association between these clusters and alterations in stress hormone levels found in prior literature as well as in the current findings.

It is unclear, however, why we failed to find a relationship between emotional numbing symptoms reported over the prior month (versus prior week) and cortisol curves. As emotional numbing is believed to be a response to arousal it is possible that arousal had not persisted long enough to lead to emotional numbing. This explanation, however, is unlikely as the vast majority of the participants were experiencing repeated and prolonged abuse. Ultimately, it is unclear why emotional numbing symptoms during the last month were not related to CAR trajectories while symptoms during the last week trended towards significance.

Ultimately the current study found that arousal symptoms over the prior week and month significantly moderated CAR, and that emotional numbing symptoms in the prior week were trending towards significant; however, avoidance and re-experiencing symptoms were not related. These results support the view that different clusters are differentially associated with CAR, which is consistent with the previous literature and may help explain prior mixed findings.

Our interest in the differential impact of PTSD clusters was to provide further evidence highlighting the heterogeneity of PTSD presentations. Arousal (and marginally emotional numbing) symptom severity uniquely impacted the trajectory of cortisol levels, suggesting that heterogeneity is present in the relationship between PTSD symptom clusters and the CAR. Given the current findings, it is likely that assessing PTSD symptom clusters in relation to cortisol may better inform future interventions compared to studies that assess PTSD globally. Our findings also suggest that arousal symptoms may represent a subtype of PTSD patients that experience more significant alterations in HPA axis functioning. The current results help explain prior mixed findings in PTSD and cortisol literature; if replicated, these findings may aid in the creation of more targeted interventions based on individual symptom presentations. Given the heterogeneity of PTSD, determining if biological abnormalities are associated only with specific symptom presentations may aid in targeting those individuals with different pharmacological agents ultimately leading to an improvement in pharmacological interventions in the future.

The present findings should be interpreted with caution given that the sample consisted of women seeking shelter from IPV. It is of note that this sample is not only exposed to IPV but is likely experiencing distress due to living in a shelter and possible insecurity and uncertainty given their current living conditions. Additionally, this sample was highly traumatized, with most participants reporting additional Criteria A qualifying traumas other than IPV. Therefore, it is unclear whether the current findings would generalize to other samples; PTSD symptom clusters may manifest differently across other populations, especially in the context of immediate crisis. Additional research in other trauma samples is needed to determine the extent to which these findings generalize. It is also important to consider that while much research has found group differences in CARs between those with PTSD and controls, the exact function of the CAR and the utility of the CAR as a diagnostic marker is still unclear and speculative (Kudielka & Wüst, 2010). More research is needed to understand the role and implications of alterations in CARs. Despite this limitation, the present study underscores the importance of examining PTSD symptom clusters in studies of the biology of PTSD, as the unique clusters appear to be differentially related to CAR.

Future research should also consider the possibility that repeatedly traumatized samples may meet criteria for Complex PTSD. It is possible that the pattern of salivary cortisol alterations found in this study may be a characteristic of Complex PTSD, which is different from PTSD in that it occurs in the context of repeated traumatization. Further, in addition to the traditional PTSD symptom clusters, complex PTSD also presents with marked self-regulation problems (i.e., emotion regulation difficulties, disturbances in perception of the abuser, alterations in consciousness, etc.)(Jongh et al., 2016). While many of the participants in the present study would likely meet criteria for Complex PTSD, this specific diagnosis was not assessed; future research would benefit from exploring CARs in individuals who meet diagnostic criteria for Complex PTSD.

Ultimately, future research would benefit from a closer examination of specific symptom presentations of PTSD with the goal of better understanding biological risk factors and disease markers leading to improved matching of treatment approaches.