Introduction
Posttraumatic stress disorder (PTSD) is a common, debilitating disorder (Kilpatrick et al., Reference Kilpatrick, Resnick, Milanak, Miller, Keyes and Friedman2013) linked to significant functional impairment and economic burden (Bothe, Jacob, Kröger, & Walker, Reference Bothe, Jacob, Kröger and Walker2020). Women exposed to interpersonal violence (IPV) are at an especially high risk of developing PTSD (Shalev et al., Reference Shalev, Gevonden, Ratanatharathorn, Laska, Van Der Mei, Qi and Delahanty2019). Identifying biomarkers associated with PTSD in this population may aid in reducing the public health burden through early identification and intervention for those at greatest risk of developing PTSD.
Dysfunction in stress response systems, including the hypothalamic–pituitary–adrenal (HPA) axis, sympathetic nervous system (SNS), and parasympathetic nervous system has been linked to PTSD (Agorastos, Nicolaides, Bozikas, Chrousos, & Pervanidou, Reference Agorastos, Nicolaides, Bozikas, Chrousos and Pervanidou2020; Morris, Compas, & Garber, Reference Morris, Compas and Garber2012; Pole, Reference Pole2007), and may represent a biomarker for PTSD risk. The current study focuses on HPA and SNS dysfunction. Prior research has found lower daily cortisol output in individuals with PTSD compared to controls (Meewisse, Reitsma, De Vries, Gersons, & Olff, Reference Meewisse, Reitsma, De Vries, Gersons and Olff2007; Miller, Chen, & Zhou, Reference Miller, Chen and Zhou2007; Morris et al., Reference Morris, Compas and Garber2012; Speer, Semple, Naumovski, D'Cunha, & McKune, Reference Speer, Semple, Naumovski, D'Cunha and McKune2019; Yehuda, Reference Yehuda2002). Moreover, greater time since posttraumatic stress symptom onset is associated with lower daily cortisol output (Morris et al., Reference Morris, Compas and Garber2012). Theoretical models propose trauma exposures progressively attenuate cortisol secretion, which increases risk for developing PTSD following subsequent traumas (Steudte-Schmiedgen, Kirschbaum, Alexander, & Stalder, Reference Steudte-Schmiedgen, Kirschbaum, Alexander and Stalder2016) via hypothesized effects on trauma memory consolidation (de Quervain, Aerni, Schelling, & Roozendaal, Reference de Quervain, Aerni, Schelling and Roozendaal2009), context processing (Liberzon & Abelson, Reference Liberzon and Abelson2016), and circadian rhythm (Agorastos et al., Reference Agorastos, Nicolaides, Bozikas, Chrousos and Pervanidou2020). Consistent with theory, lower circulating cortisol levels are associated with increased risk for developing PTSD (Morris & Rao, Reference Morris and Rao2013; Morris, Hellman, Abelson, & Rao, Reference Morris, Hellman, Abelson and Rao2016), and glucocorticoid administration in recent trauma survivors reduces risk for PTSD onset (Sijbrandij, Kleiboer, Bisson, Barbui, & Cuijpers, Reference Sijbrandij, Kleiboer, Bisson, Barbui and Cuijpers2015). However, whether and in what manner blunted cortisol contributes to PTSD onset in recent IPV survivors remains unclear.
In contrast to decreased HPA activity, research suggests individuals with PTSD exhibit increased SNS activity relative to non-traumatized controls (NTCs) and trauma-exposed individuals without PTSD (Buckley & Kaloupek, Reference Buckley and Kaloupek2001; O'Donnell, Hegadoren, & Coupland, Reference O'Donnell, Hegadoren and Coupland2004; Pole, Reference Pole2007). Heightened SNS activity has been implicated in trauma memory consolidation models of PTSD (de Quervain et al., Reference de Quervain, Aerni, Schelling and Roozendaal2009), although this role has not always been supported by prospective (Morris & Rao, Reference Morris and Rao2013) or pharmacological intervention studies (Sijbrandij et al., Reference Sijbrandij, Kleiboer, Bisson, Barbui and Cuijpers2015). Nevertheless, meta-analytic evidence indicates elevated heart rate within 72 h of trauma exposure corresponds to higher PTSD risk (Morris et al., Reference Morris, Hellman, Abelson and Rao2016). Emerging evidence suggests individuals with PTSD exhibit higher alpha-amylase levels, an indicator of SNS activity (Nater & Rohleder, Reference Nater and Rohleder2009), than NTCs (Nicholson, Bryant, & Felmingham, Reference Nicholson, Bryant and Felmingham2014). Moreover, greater intrusive and hyperarousal symptoms are associated with higher morning alpha-amylase levels in adolescent trauma survivors (Keeshin, Strawn, Out, Granger, & Putnam, Reference Keeshin, Strawn, Out, Granger and Putnam2015). Together, findings suggest alpha-amylase is associated with PTSD, but it remains unclear whether alpha-amylase represents a biomarker for subsequent PTSD onset and/or maintenance.
Few studies have evaluated diurnal variation in both HPA and SNS secretion in relation to PTSD. Cross-sectional research suggests individuals with PTSD exhibit flatter diurnal cortisol slope compared to controls, whose cortisol levels rise after waking and drop throughout the day (Nicolson & Ponnamperuma, Reference Nicolson and Ponnamperuma2019; Thompson et al., Reference Thompson, Weissbecker, Cash, Simpson, Daup and Sephton2015). Meta-analyses have demonstrated a pattern of lower waking (Miller et al., Reference Miller, Chen and Zhou2007) or daily (Morris et al., Reference Morris, Compas and Garber2012) cortisol levels and higher afternoon/evening levels in individuals exposed to trauma compared to controls. This is consistent with research demonstrating flatter diurnal cortisol slope corresponds to negative physical and mental health outcomes (see Adam et al., Reference Adam, Quinn, Tavernier, McQuillan, Dahlke and Gilbert2017 for a review). One study, which examined diurnal variation in alpha-amylase, found individuals with PTSD exhibited a sharp increase in alpha-amylase levels after waking (compared to a decrease after waking in controls), with positive associations between alpha-amylase levels and PTSD symptoms (Thoma, Joksimovic, Kirschbaum, Wolf, & Rohleder, Reference Thoma, Joksimovic, Kirschbaum, Wolf and Rohleder2012). Adam et al. (Reference Adam, Quinn, Tavernier, McQuillan, Dahlke and Gilbert2017) theorize that psychosocial stressors can both trigger and maintain dysregulation in circadian processes (Hall et al., Reference Hall, Vasko, Buysse, Ombao, Chen, Cashmere and Thayer2004; Sadeh, Keinan, & Daon, Reference Sadeh, Keinan and Daon2004; Van Reeth et al., Reference Van Reeth, Weibel, Spiegel, Leproult, Dugovic and Maccari2000), which can have cascading effects on many regulatory systems (e.g. biological and behavioral). These changes can, in turn, negatively affect physical and mental health (Adam et al., Reference Adam, Quinn, Tavernier, McQuillan, Dahlke and Gilbert2017). However, the majority of this research has been cross-sectional and has not accounted for other important determinants of dysfunction in stress response symptoms following trauma exposure.
Several risk factors may influence diurnal HPA and SNS secretion in PTSD. Trauma history has been linked to both altered HPA functioning and risk of PTSD onset (Delahanty & Nugent, Reference Delahanty and Nugent2006). Moreover, meta-analytic findings highlight the need to examine the effect of co-occurring depressive symptoms given findings that depression and PTSD exert opposing effects on cortisol levels (Morris et al., Reference Morris, Compas and Garber2012). Research also suggests age may moderate the direction of the relationship between stress response system dysfunction and PTSD (Morris et al., Reference Morris, Hellman, Abelson and Rao2016). Few studies have examined these factors simultaneously. Elucidating the relations between these variables and diurnal cortisol/alpha-amylase secretion could assist in early identification and intervention for those at greatest risk of developing PTSD or other posttraumatic psychopathology.
Using an innovative, longitudinal design, the current study examined the dynamics of diurnal cortisol and alpha-amylase rhythms in women following recent IPV compared to healthy NTCs. The primary aim was to understand how PTSD, depressive symptoms, and childhood trauma are independently associated with diurnal cortisol and alpha-amylase rhythms. We hypothesized that flatter diurnal cortisol slope and steeper alpha-amylase slope would correspond to an increased risk of concurrent and subsequent PTSD onset and symptom severity.
Method
Participants
The present sample comprised of 98 women aged 18–30 (inclusive). Of these, 57 reported exposure to IPV, defined as any incident involving physical and/or sexual assault, within approximately 3 months of their baseline assessment (range of days since index trauma at baseline: 5–101, M = 45.32, s.d. = 24.74). A comparison group of 41 NTCs included healthy women with no IPV exposure. For the IPV group, exclusion criteria were: current major depressive disorder (MDD) that preceded their index trauma; PTSD resulting from an index trauma that occurred prior to the 3-month window; current substance use disorder; or history of bipolar or psychotic disorder. All participants were administered the Structured Clinical Interview (SCID) for DSM-IV (First, Spitzer, Gibbon, & Williams, Reference First, Spitzer, Gibbon and Williams2005) to assess for mood and substance use disorders at baseline. PTSD was evaluated in the IPV group at each assessment via the Clinician Administered PTSD Scale (CAPS) for DSM-IV (Blake et al., Reference Blake, Weathers, Nagy, Kaloupek, Gusman, Charney and Keane1995). For NTCs, exclusion criteria were: lifetime IPV exposure, any trauma exposure in the previous year, and history of mood or anxiety disorder or PTSD. Additional exclusion criteria for all participants were: serious health conditions which influence HPA or SNS activity, pregnancy, or current use of prescription or non-prescription drugs which affect HPA or SNS activity
Protocol
Details of the study protocol have been described previously (Morris et al., Reference Morris, Bailey, Hellman, Williams, Lannon, Kutcher and Rao2020). Briefly, participants were recruited through online advertisements, and the IPV group was also recruited from a local hospital and local agencies coordinating services for domestic violence and sexual assault survivors. Participants completed assessments and provided saliva samples at baseline and at 1-, 3-, and 6-month follow-up. At each assessment, the CAPS interview (Blake et al., Reference Blake, Weathers, Nagy, Kaloupek, Gusman, Charney and Keane1995) and the SCID mood modules (First et al., Reference First, Spitzer, Gibbon and Williams2005) were administered. Participants also completed questionnaires via the secure web-based Research Electronic Data Capture (REDCap) platform (Harris et al., Reference Harris, Taylor, Thielke, Payne, Gonzalez and Conde2009) and provided saliva samples over 2 consecutive days at each time point.
Measures
Prior trauma exposure
The Childhood Trauma Questionnaire (CTQ) (Bernstein et al., Reference Bernstein, Fink, Handelsman, Foote, Lovejoy, Wenzel and Ruggiero1994), a 28-item self-report measure, was used to assess exposure to childhood abuse and neglect in all participants. Coefficient alpha for the CTQ in this sample was 0.87.
Depressive symptoms
The Beck Depression Inventory, second edition (BDI-II) (Beck, Steer, Ball, & Ranieri, Reference Beck, Steer, Ball and Ranieri1996), a 21-item self-report measure, was used to assess depressive symptoms in all participants. Coefficient alphas in this sample were excellent (range: 0.92–0.94).
Posttraumatic stress symptoms and disorder
The clinician-administered CAPS interview for DSM-IV (Blake et al., Reference Blake, Weathers, Nagy, Kaloupek, Gusman, Charney and Keane1995) was used to assess overall posttraumatic stress symptom severity, symptom cluster severity (i.e. re-experiencing, avoidance, and hyperarousal), and PTSD diagnostic status (using the ‘F1/I2’ rule) at each assessment (Weathers, Ruscio, & Keane, Reference Weathers, Ruscio and Keane1999) in the IPV group. Inter-rater reliability in this sample was excellent for CAPS total (intraclass correlation coefficient; ICC = 0.997) and CAPS subscale (ICC's >0.993) scores.
Cortisol and alpha-amylase
Cortisol and alpha-amylase levels were determined by using saliva samples collected using acid-free cotton swabs (Sarstedt Inc., Newton, NC). Participants were instructed to provide five saliva samples (upon awakening, +30 min after awakening, before lunch, ~3 pm, and ~9 pm) per day over 2 consecutive weekdays. Participants were instructed to refrain from brushing their teeth, eating, drinking caffeine, or engaging in rigorous exercise within 30 min of each sample, and not to drink alcohol or use tobacco products during collection days. Participants recorded the timing of each saliva sample and stored samples in their home freezer before returning them at their next visit. Cortisol levels were determined by commercial chemiluminescence immunoassay (Kirschbaum & Hellhammer, Reference Kirschbaum and Hellhammer1989). Alpha-amylase levels were determined by a quantitative enzyme kinetic method (Nater, Rohleder, Schlotz, Ehlert, & Kirschbaum, Reference Nater, Rohleder, Schlotz, Ehlert and Kirschbaum2007). Intra- and inter-assay coefficients of variation for these assays were below 6%. Participants also completed paper-and-pencil measures, including an item which asked participants to report the number of hours they slept the night before each saliva sample collection day.
Data analysis
Distributional properties were examined for all variables, and cases were screened for outliers. The findings reported below were unchanged after winsorizing cortisol and alpha-amylase outliers (⩾3 s.d.). Salivary cortisol and alpha-amylase levels were natural log-transformed due to their right-skewed distributions. The Statistical Package for the Social Sciences (SPSS; Chicago, IL, version 27) was used for data management; descriptive statistics; and the calculation of total-R 2, a measure of the total proportion reduction in the total outcome variance across all levels of analysis. Missing cortisol and alpha-amylase data were handled using maximum-likelihood estimation. Multilevel models (MLMs) were specified using hierarchical linear models (v. 8) to examine how within-individual changes in diurnal cortisol or alpha-amylase levels varied across time of day (level 1), across assessments (level 2), and between individuals (level 3) (Raudenbush, Bryk, Cheong, Congdon, & du Toit, Reference Raudenbush, Bryk, Cheong, Congdon and du Toit2019). MLMs in the full sample separately examined the influence of age (level 3), childhood trauma (level 3), mean depressive symptoms (level 3), days since baseline (level 2), assessment-level depressive symptoms (level 2), PTSD diagnostic status at each assessment (level 2), and minutes from waking (level 1) on diurnal cortisol or alpha-amylase levels. Of note, the continuous minutes from waking variable, rather than an ordinal variable indexing each sample (i.e. 1–5), was entered as a level-1 predictor in order to examine the diurnal cortisol and alpha-amylase slopes because time since waking strongly affects neuroendocrine markers (Nicolson & Ponnamperuma, Reference Nicolson and Ponnamperuma2019; Thompson et al., Reference Thompson, Weissbecker, Cash, Simpson, Daup and Sephton2015). Predictors from these separate MLMs that were significantly associated with changes in diurnal cortisol or alpha-amylase were then entered into a combined model to examine their independent effects. The combined model including diagnostic status, depressive symptoms, age, and childhood trauma as predictors of diurnal cortisol is presented below.
Level 1 model (within day): Cortisol = π 0 + π 1 Minutes (from waking) + π 2 Collection Day + e
Level 2 model (between assessments): π 0 = β 00 + β 01Diagnostic Status + β 02Depression + β 03Days (since baseline) + r 0π 1 = β 10 + β 11Diagnostic Status + β 12Depression + β 13Days (since baseline) + r 1π 2 = β 20
Level 3 model (between person): β 00 = ϒ000 + ϒ001Childhood Trauma + ϒ002Mean Depression + ϒ003Age + u 0β 01 = ϒ010β 02 = ϒ020β 10 = ϒ100 + ϒ101Childhood Trauma + ϒ102Mean Depression + ϒ103Ageβ 11 = ϒ110β 12 = ϒ120β 20 = ϒ200
In the above equation, Cortisol denotes natural log-transformed cortisol levels for each participant on a specific collection day at a specific assessment time point. Predictors of waking cortisol (intercept) and linear change in cortisol across the day (slope) included: Age; Diagnostic Status (determined at each assessment: either PTSD, IPV without PTSD, or NTC); assessment-level Depression (to assess the within-individual effect of depressive symptoms); mean Depression (averaged across assessments to assess the between-person effect of depressive symptoms); and Childhood Trauma (CTQ total score). Of primary interest in the full sample were the cross-level interactions between Diagnostic Status and Minutes (β 11), assessment-level Depression and Minutes (β 12), mean Depression and Minutes (ϒ102), and Childhood Trauma and Minutes (ϒ101). Diagnostic status could change from one assessment to the next within the IPV group.
Next, MLMs within the IPV group simultaneously examined the influence of PTSD symptom severity (level 2), assessment-level depressive symptoms (level 2), and mean depressive symptoms (level 3) on diurnal cortisol or alpha-amylase levels. Of primary interest within the IPV group were the depressive symptoms (BDI-II) × Minutes from waking and PTSD symptom severity (CAPS) × Minutes from waking interactions.
Identical MLMs in the full sample and within the IPV group were specified to evaluate predictors of waking alpha-amylase and linear change in alpha-amylase across the day.
Finally, the analyses were repeated using time-lagged diagnostic and symptom measures to investigate whether cortisol and alpha-amylase predicted PTSD diagnostic status and/or depressive symptom severity at the following assessment using the approach described by Shirtcliff and Essex (Reference Shirtcliff and Essex2008).
For all MLMs, simple slopes analysis was used to probe significant interactions (Preacher, Curran, & Bauer, Reference Preacher, Curran and Bauer2006), and interaction patterns are presented for higher (+1 s.d.) and lower (−1 s.d.) values of continuous independent variables. To account for multiple hypothesis tests, we used the Benjamini–Hochberg false-discovery rate correction to control for type I errors by adjusting the p value based on the number of significant results in a family of tests (Benjamini & Hochberg, Reference Benjamini and Hochberg1995).
Results
Participant characteristics
Demographic and clinical characteristics are presented in Table 1 separately for women within the IPV group who met PTSD criteria at any point during the study (IPV + PTSD), women within the IPV group who never met criteria for PTSD during the study (IPV without PTSD), and NTC women. The IPV + PTSD group reported greater depressive symptoms and childhood trauma exposure compared to the IPV without PTSD group. The IPV without PTSD group reported greater depressive symptoms and childhood trauma exposure than the NTC group. Racial/ethnic differences between groups revealed a higher proportion of non-Hispanic White/Caucasian participants in the NTC compared to the IPV groups. Additionally, the NTC group reported greater educational attainment than either IPV group. The groups did not differ in age or marital status.
Descriptive and clinical characteristics for recent IPV survivors who met diagnostic criteria for PTSD at any point during the study (IPV + PTSD), those who did not (IPV without PTSD), and the NTC group
IPV, interpersonal violence; PTSD, posttraumatic stress disorder; NTC, non-traumatized control.
***p < 0.001; **p < 0.01; *p < 0.05. Within rows, values with different superscripts differ significantly at p < 0.05.
Preliminary analyses of diurnal cortisol and alpha-amylase secretion
Salivary cortisol levels (averaged across collection days) are presented separately for each group in Fig. 1 at baseline (panel A) and 1-month follow-up (panel B). MLMs revealed that cortisol levels declined significantly throughout the day (b = −0.0023, s.e. = 0.0001, p < 0.001) across all participants, controlling for days since baseline. Age moderated changes in cortisol levels throughout the day (b = −0.0001, s.e. = 0.00002, p = 0.002); older participants demonstrated a steeper diurnal decline in cortisol than younger participants. As time (days from baseline) elapsed, waking cortisol levels declined (b = −0.0041, s.e. = 0.0009, p < 0.001), and diurnal cortisol slopes flattened (b = 0.000004, s.e. = 0.000002, p = 0.045). All subsequent MLMs examining diurnal change in cortisol levels included age and days since baseline.
Salivary cortisol levels over time.
Note. Spaghetti plots depict assessment time point 1 (panel A) and assessment time point 2 (panel B) within-person changes in mean salivary cortisol levels for each participant, averaged across both collection days (depicted in gray). The diurnal cortisol trajectory defined by the mean parameters for each group are depicted in green (NTCs), blue (IPV without PTSD), and red (IPV with PTSD). Shaded area surrounding the trajectory defined by the mean parameters represents the standard error. NTC, non-traumatized control; IPV, interpersonal violence; PTSD, posttraumatic stress disorder.
MLMs revealed that alpha-amylase levels increased throughout the day (b = 0.0011, s.e. = 0.0001, p < 0.001) across all participants, controlling for days since baseline assessment. Age moderated changes in alpha-amylase levels throughout the day (b = 0.00004, s.e. = 0.00002, p = 0.019); older participants demonstrated a steeper diurnal increase in alpha-amylase than younger participants. No other main or interactive effects were significant (ps ⩾ 0.171). All subsequent MLMs examining diurnal changes in alpha-amylase levels included age and days since baseline as moderators.
Predictors of diurnal cortisol secretion: childhood trauma, depressive symptoms, and PTSD diagnostic status
Concurrent prediction
Results of MLMs examining the predictors of diurnal cortisol secretion in the full sample are presented in the online Supplementary material. Childhood trauma exposure was significantly associated with both waking cortisol levels and diurnal cortisol slopes. As depicted in Fig. 2, simple slope analysis revealed that women with greater childhood trauma exposure exhibited lower waking cortisol levels and flatter diurnal cortisol slopes (b = −0.0019, s.e. = 0.0001, p < 0.001) compared to women with lower childhood trauma exposure (b = −0.0025, s.e. = 0.0001, p < 0.001). The effect of childhood trauma on waking cortisol levels (b = −0.0137, s.e. = 0.0042, p = 0.002) and diurnal cortisol slope (b = 0.0002, s.e. = 0.000004, p < 0.001) remained significant when controlling for sample collection day, days from baseline, and age. In contrast, within-person (i.e. assessment level) depressive symptoms were not associated with waking cortisol levels (b = −0.0001, s.e. = 0.0074, p = 0.989) or diurnal cortisol slopes (b = −0.00001, s.e. = 0.00002, p = 0.716). However, between-person mean depressive symptoms were associated with waking cortisol (b = −0.0162, s.e. = 0.0068, p = 0.019) but not diurnal cortisol slope (b = 0.00001, s.e. = 0.00001, p = 0.357). PTSD diagnostic status was associated with waking cortisol levels and diurnal cortisol slopes. Simple slope analysis revealed lower waking cortisol levels and flatter diurnal cortisol slopes for women with current PTSD (b = −0.0017, s.e. = 0.0002, p < 0.001) compared to IPV survivors without PTSD (b = −0.0021, s.e. = 0.0001, p < 0.001) and NTCs (b = −0.0025, s.e. = 0.0001, p < 0.001). The effect of diagnostic status on waking cortisol level (b = −0.3394, s.e. = 0.0857, p < 0.001) and diurnal cortisol slope (b = 0.0004, s.e. = 0.0001, p < 0.001) remained significant when controlling for sample collection day, days from baseline, and age.
Effect of childhood trauma on diurnal cortisol slope.
Note. For illustrative purposes, graph depicts the simple diurnal cortisol slopes (collapsed across all groups) for individuals with lower (−1 s.d.) and higher (+1 s.d.) levels of childhood trauma, based on the Childhood Trauma Questionnaire. Additionally, values were defined at 0 and 960 min from waking to illustrate the effect of childhood trauma on the diurnal cortisol slope. CTQ = Childhood Trauma Questionnaire.
Summary of concurrent cortisol findings. In the combined model (total-R 2 = 0.321), childhood trauma was no longer a significant predictor of waking cortisol levels (b = −0.0076, s.e. = 0.0054, p = 0.162), but the effect on diurnal cortisol slopes remained significant (b = 0.00002, s.e. = 0.00001, p = 0.001). Within-person depressive symptom levels again did not predict waking cortisol (b = 0.0049, s.e. = 0.0072, p = 0.496) or diurnal cortisol slope (b = −0.00001, s.e. = 0.00002, p = 0.445). Similarly, between-person mean depressive symptoms did not predict waking cortisol (b = −0.0017, s.e. = 0.0086, p = 0.844), and the relationship between mean depressive symptoms and diurnal cortisol slope was no longer significant (b = −0.00002, s.e. = 0.00001, p = 0.052). Although PTSD diagnostic status continued to predict waking cortisol levels (b = −0.2164, s.e. = 0.1015, p = 0.034), it was no longer significantly associated with diurnal cortisol slopes (b = 0.0001, s.e. = 0.0001, p = 0.304) when simultaneously examining the effect of childhood trauma and depressive symptoms. To summarize, when entering diagnostic status, depressive symptoms, and childhood trauma in the same model, a diagnosis of PTSD was associated with lower concurrent waking cortisol, and higher childhood trauma history was associated with a flatter concurrent diurnal cortisol slope.Footnote †Footnote 1 There were no other significant findings.
Longitudinal prediction
Results of MLMs examining the relationship between diurnal cortisol secretion and subsequent diagnostic status and depressive symptom severity in the full sample are presented in Table 2. PTSD diagnostic status at the subsequent assessment was significantly associated with waking cortisol levels and diurnal cortisol slopes. Simple slope analysis revealed lower waking cortisol levels and flatter diurnal cortisol slopes in women with subsequent PTSD (b = −0.0018, s.e. = 0.0002, p < 0.001) compared to IPV survivors without PTSD (b = −0.0021, s.e. = 0.0001, p < 0.001) and NTCs (b = −0.0025, s.e. = 0.0001, p < 0.001). The relationship between waking cortisol level (b = −0.5038, s.e. = 0.0942, p < 0.001) and diurnal cortisol slope (b = 0.0003, s.e. = 0.0001, p = 0.005) and PTSD diagnostic status at the subsequent assessment remained significant when controlling for sample collection day, days from baseline, and age. There was no relationship between waking cortisol levels (b = −0.0087, s.e. = 0.0085, p = 0.308) or diurnal cortisol slopes (b = −0.0002, s.e. = 0.00002, p = 0.408) and depressive symptoms at the subsequent assessment (i.e. within-person depressive symptoms).
Longitudinal diurnal cortisol MLMs
Day, sample collection day (i.e. day 1 or day 2); minutes, minutes from waking; BDI, Beck Depression Inventory; CTQ, Childhood Trauma Questionnaire; diagnostic status (i.e. current PTSD, IPV exposure without current PTSD, NTC) was determined at the subsequent assessment.
*p < 0.05.
Summary of longitudinal cortisol findings. In the longitudinal model which controlled for childhood trauma and within- and between-person depressive symptom levels (total-R 2 = 0.384), the relationship between waking cortisol level and diagnostic status at the subsequent assessment remained significant (b = −0.5737, s.e. = 0.1303, p < 0.001), but the relationship between diurnal cortisol slope and diagnostic status at the subsequent assessment was no longer significant (b = 0.0002, s.e. = 0.0002, p = 0.222). In summary, when controlling for childhood trauma and depressive symptoms, lower waking cortisol level was associated with PTSD diagnosis at the subsequent assessment, but diurnal cortisol slope was not.Footnote 2
Predictors of diurnal cortisol secretion within the IPV group
Concurrent prediction
An MLM examined the independent effects of PTSD and depressive symptom severity on diurnal cortisol secretion, controlling for age, days from index trauma, and collection day. Assessment-level PTSD symptom severity was not associated with waking cortisol levels (b = 0.0047, s.e. = 0.0044, p = 0.287) or diurnal cortisol slope (b = −0.00001, s.e. = 0.00001, p = 0.529). In contrast, mean PTSD symptom severity was negatively associated with waking cortisol level (b = −0.0167, s.e. = 0.0064, p = 0.012) but not diurnal cortisol slope (b = 0.00001, s.e. = 0.00001, p = 0.109). Assessment-level depressive symptom severity was not associated with waking cortisol (b = −0.0009, s.e. = 0.0089, p = 0.915) or diurnal cortisol slope (b = 0.00002, s.e. = 0.00002, p = 0.920). Similarly, mean depressive symptom severity was not associated with waking cortisol (b = 0.0211, s.e. = 0.0120, p = 0.086) or diurnal cortisol slope (b = −0.00002, s.e. = 0.00001, p = 0.206).
Longitudinal prediction
An MLM examined the relationship between diurnal cortisol secretion and PTSD and depressive symptom severity at the subsequent assessment, controlling for age, days from index trauma, mean PTSD symptom severity, mean depressive symptom severity, and collection day within the IPV group. There was no significant relationship between waking cortisol levels (b = −0.0057, s.e. = 0.0061, p = 0.352) or diurnal cortisol slopes (b = −0.000001, s.e. = 0.00001, p = 0.967) and subsequent PTSD symptom severity. Similarly, there was no relationship between waking cortisol levels (b = 0.0155, s.e. = 0.0107, p = 0.153) or diurnal cortisol slopes (b = −0.0001, s.e. = 0.00002, p = 0.748) and subsequent depressive symptoms.
Summary of cortisol findings in the IPV group. To summarize, within the IPV group, greater PTSD symptoms were associated with lower concurrent waking cortisol levelsFootnote 3 (total-R 2 = 0.247). There was no significant relationship between waking cortisol or diurnal cortisol slope and PTSD or depressive symptoms at the subsequent assessment.
Predictors of alpha-amylase secretion: childhood trauma, depressive symptoms, and PTSD diagnostic status
Concurrent prediction
Results of MLMs examining the predictors of diurnal alpha-amylase secretion in the full sample are presented in Table 3. Childhood trauma exposure was significantly associated with both waking alpha-amylase levels and diurnal alpha-amylase slopes. Simple slope analysis revealed women with greater childhood trauma exposure exhibited higher waking alpha-amylase levels and slower diurnal alpha-amylase increase (b = 0.0008, s.e. = 0.0001, p < 0.001) compared to women with lower childhood trauma exposure (b = 0.0012, s.e. = 0.0001, p < 0.001). The effect of childhood trauma on waking alpha-amylase (b = 0.0093, s.e. = 0.0043, p = 0.033) and diurnal alpha-amylase slope (b = −0.00001, s.e. = 0.000003, p = 0.001) remained significant when controlling for sample collection day, days from baseline, and age. Within-person (assessment-level) depressive symptoms were not associated with waking alpha-amylase (b = 0.0016, s.e. = 0.0068, p = 0.810) or diurnal alpha-amylase slopes (b = −0.000004, s.e. = 0.00001, p = 0.754). Between-person mean depressive symptoms were not associated with waking alpha-amylase (b = 0.0067, s.e. = 0.0069, p = 0.332) or diurnal alpha-amylase slopes (b = −0.00001, s.e. = 0.00001, p = 0.059). PTSD diagnostic status was not significantly associated with waking alpha-amylase levels (b = 0.1372, s.e. = 0.0877, p = 0.119) or diurnal alpha-amylase slopes (b = −0.0001, s.e. = 0.0001, p = 0.165).
MLMs predicting concurrent diurnal alpha-amylase levels in the full sample
Day, sample collection day (i.e. day 1 or day 2); minutes, minutes from waking; BDI, Beck Depression Inventory; CTQ, Childhood Trauma Questionnaire; diagnostic status (i.e. current PTSD, IPV exposure without current PTSD, NTC) was determined at each assessment.
*p < 0.05.
Summary of concurrent alpha-amylase findings. To summarize, greater childhood trauma exposure was associated with higher waking alpha-amylase and slower diurnal alpha-amylase increase compared to lower childhood trauma exposureFootnote 4 (total-R 2 = 0.103). Depressive symptoms and PTSD diagnosis were not related to waking alpha-amylase or diurnal alpha-amylase slopes.
Longitudinal prediction
There was no relationship between waking alpha-amylase levels (b = 0.1619, s.e. = 0.1113, p = 0.148) or diurnal alpha-amylase slopes (b = −0.0001, s.e. = 0.0001, p = 0.316) and PTSD diagnostic status at the subsequent assessment. Similarly, there was no relationship between waking alpha-amylase levels (b = 0.0038, s.e. = 0.0089, p = 0.671) or diurnal alpha-amylase slopes (b < 0.000001, s.e. = 0.00002, p = 0.995) and depressive symptoms at the subsequent assessment.
Discussion
This longitudinal study of recent IPV survivors examined whether patterns of diurnal cortisol and alpha-amylase secretion were associated with concurrent PTSD diagnosis or symptom severity and predicted subsequent PTSD development. We also examined the independent effects of depressive symptoms and childhood trauma exposure on diurnal HPA/SNS secretion. MLMs revealed a pattern of lower waking cortisol levels and slower diurnal cortisol decline that distinguished women with concurrent PTSD from IPV survivors without PTSD and healthy NTCs. Moreover, lagged effects models demonstrated this diurnal cortisol secretion pattern predicted PTSD diagnostic status at the next assessment, which supports interpretation of this pattern as a risk marker for – rather than simply a consequence of – PTSD. Thus, a pattern of lower waking cortisol levels and flatter diurnal cortisol slope may serve as a predictive marker for early clinical course and may be important in PTSD prevention efforts. For example, if findings are replicated, it is possible that lower waking cortisol and flatter diurnal cortisol slope may be useful in identifying those who may be most likely to benefit from early, modified prolonged exposure therapy, which has shown promise for prevention of posttraumatic stress and depressive symptoms (Rothbaum et al., Reference Rothbaum, Kearns, Price, Malcoun, Davis, Ressler and Houry2012).
Results support a growing body of literature suggesting a neurobiological pathway for the development of PTSD symptoms via hypothesized effects on trauma memory consolidation (de Quervain et al., Reference de Quervain, Aerni, Schelling and Roozendaal2009), context processing (Liberzon & Abelson, Reference Liberzon and Abelson2016), and circadian rhythm (Agorastos et al., Reference Agorastos, Nicolaides, Bozikas, Chrousos and Pervanidou2020). Lower waking cortisol levels were associated with greater depressive symptoms. Moreover, lower waking cortisol levels and flatter diurnal cortisol slopes also characterized women with greater childhood trauma exposure. When simultaneously examining the effect of childhood trauma exposure, and depression, lower waking cortisol levels –but not flatter diurnal cortisol slopes – remained significantly associated with concurrent and subsequent PTSD diagnosis, while higher childhood trauma continued to predict flatter diurnal cortisol slopes. Notably, lower waking cortisol levels were also associated with higher mean posttraumatic stress symptom severity among IPV survivors. Together, these findings demonstrate distinct contributions of childhood trauma exposure and PTSD to diurnal cortisol secretory patterns and highlight the potential utility of blunted waking cortisol levels as a prognostic biomarker of PTSD risk and flatter diurnal cortisol slope as an index of childhood trauma among recent IPV survivors.
Exposure to past trauma is associated with altered HPA functioning, including flatter diurnal cortisol slopes compared to controls (Nicolson & Ponnamperuma, Reference Nicolson and Ponnamperuma2019; Thompson et al., Reference Thompson, Weissbecker, Cash, Simpson, Daup and Sephton2015), and increased risk of PTSD onset following subsequent trauma exposures (Delahanty & Nugent, Reference Delahanty and Nugent2006). Consistent with prior research, we found evidence of flatter diurnal cortisol slopes in individuals with greater childhood trauma history and an association between lower waking cortisol and concurrent and subsequent PTSD diagnosis. Higher mean posttraumatic stress symptom severity was associated with lower waking cortisol levels among IPV survivors, consistent with prior research showing blunted cortisol awakening responses in individuals with PTSD compared to trauma-exposed and healthy individuals (Rauch et al., Reference Rauch, King, Kim, Powell, Rajaram, Venners and Liberzon2020; Wessa, Rohleder, Kirschbaum, & Flor, Reference Wessa, Rohleder, Kirschbaum and Flor2006). These findings support theoretical models which posit that prior trauma exposures progressively attenuate cortisol secretion, thereby increasing risk for PTSD onset (Steudte-Schmiedgen et al., Reference Steudte-Schmiedgen, Kirschbaum, Alexander and Stalder2016).
Contrary to expectations, diurnal cortisol slopes were not significantly related to concurrent or subsequent PTSD diagnosis when simultaneously examining the effect of childhood trauma. One interpretation is that flatter diurnal cortisol slopes are more strongly tied to childhood trauma exposure than to PTSD diagnosis. Another possibility is that diurnal cortisol slopes are more closely related to the daily experiences of trauma survivors than to monthly variation in posttraumatic stress symptoms captured by the CAPS interview. Momentary changes in mood are known to influence daily fluctuations in cortisol levels (Smyth et al., Reference Smyth, Ockenfels, Porter, Kirschbaum, Hellhammer and Stone1998). Future studies are needed to capture daily covariation between diurnal cortisol rhythms and emotional state (Adam, Hawkley, Kudielka, & Cacioppo, Reference Adam, Hawkley, Kudielka and Cacioppo2006) among recent IPV survivors.
Regarding SNS activity, MLMs revealed women with greater childhood trauma exposure displayed higher waking alpha-amylase levels and slower diurnal alpha-amylase increase compared to women with lower childhood trauma exposure. This is partly consistent with evidence of heightened SNS activity among individuals with PTSD compared to controls (Buckley & Kaloupek, Reference Buckley and Kaloupek2001; O'Donnell et al., Reference O'Donnell, Hegadoren and Coupland2004; Pole, Reference Pole2007), including elevated waking alpha-amylase levels (Keeshin et al., Reference Keeshin, Strawn, Out, Granger and Putnam2015). PTSD diagnostic status and depressive symptoms were not associated with alpha-amylase.
Notably, within-person changes in depressive symptoms from assessment to assessment were not associated with waking cortisol or diurnal cortisol slopes. Although mean depressive symptoms (i.e. between-person differences in depressive symptoms) were associated with waking cortisol, the direction of the effect was counter to expectations; higher depressive symptoms were associated with lower waking cortisol. Furthermore, the relationship between depressive symptoms and waking cortisol was no longer significant when simultaneously examining PTSD diagnostic status. One possible explanation for these discrepant findings is that PTSD diagnosis better explains participants’ low mood, rather than major depression. Moreover, elevated cortisol levels are more likely to be found in melancholic than non-melancholic depression (Juruena, Bocharova, Agustini, & Young, Reference Juruena, Bocharova, Agustini and Young2018). Individuals with melancholic depression are more likely to be found in inpatient, compared to outpatient, settings and are more likely to display severe depressive symptoms. In contrast, participants in the present sample were recruited from hospitals and community settings and displayed moderate depressive symptoms. Additionally, there was no relationship between waking alpha-amylase levels or diurnal alpha-amylase slopes and concurrent or subsequent PTSD diagnostic status or depressive symptoms. Future studies are needed to replicate these findings and to evaluate plausible mechanisms linking posttraumatic psychopathology to waking levels of cortisol. For example, research suggests that whereas increased engagement in daily social activities is associated with more pronounced, normative declines in cortisol levels for non-depressed individuals, daily social activities are uncoupled from diurnal cortisol secretion for depressed individuals (Stetler, Dickerson, & Miller, Reference Stetler, Dickerson and Miller2004). It is unknown whether this pattern of uncoupling of daily social activities and decline in cortisol levels is present in individuals with PTSD.
Findings regarding diurnal neuroendocrine secretion in IPV survivors complement prior research examining changes in HPA/SNS stress response patterns following IPV exposure in an overlapping sample (Morris et al., Reference Morris, Bailey, Hellman, Williams, Lannon, Kutcher and Rao2020). Recent IPV survivors who develop PTSD are characterized by blunted diurnal cortisol secretion (i.e. lower waking cortisol levels and flatter diurnal cortisol slopes), heightened cortisol reactivity to initial presentation of a social-evaluative stress task compared to NTC, and rapid habituation of cortisol responses to subsequent stressors. In contrast, recent IPV survivors who do not develop PTSD exhibit steeper diurnal cortisol declines than those with PTSD and blunted cortisol reactivity to social-evaluative stressors compared to NTC. Although it is tempting to speculate that relatively preserved diurnal cortisol declines and blunted cortisol responses to social-evaluative threat together comprise an early biomarker of IPV resilience, these neuroendocrine features may yet confer increased vulnerability for developing PTSD in the context of future traumatic events (Steudte-Schmiedgen et al., Reference Steudte-Schmiedgen, Kirschbaum, Alexander and Stalder2016).
Limitations of the current study can guide future research on early predictors of post-IPV PTSD development. The rigorous saliva collection protocol for capturing within-person changes in diurnal HPA/SNS rhythms over time precluded a large sample size. Larger studies of IPV survivors are needed to identify diurnal neuroendocrine secretion components associated with membership in distinct posttraumatic and depressive symptom change trajectories. Additionally, the current study excluded male participants. Future research should examine whether gender moderates the relationship between diurnal cortisol and alpha-amylase secretion and symptomatology following IPV exposure. The ‘post–post’ longitudinal design cannot shed light on whether alterations in daily cortisol and alpha-amylase secretion were present before IPV exposure – a distinct possibility given results demonstrating a key role for childhood trauma exposure as a determinant of waking levels and diurnal slopes. Moreover, there is evidence for transgenerational transmission of PTSD vulnerability from mother to offspring. This is thought to occur through maternal epigenetic programing, which leads to lower cortisol levels in offspring (Yahyavi, Zarghami, & Marwah, Reference Yahyavi, Zarghami and Marwah2014). Future research is needed to fully understand the relationship between generational and childhood trauma, cortisol secretion, and PTSD. Finally, sleep, circadian rhythm, and neuroendocrine markers are closely linked and all play a role in the neurobiology of PTSD (Agorastos et al., Reference Agorastos, Nicolaides, Bozikas, Chrousos and Pervanidou2020). Although participants were asked to self-report their sleep, the current study did not carefully investigate the role of sleep and circadian rhythm in participants' clinical presentation. Future studies should include objective measures of sleep/wake cycles (e.g. actigraphy) to clarify the role of sleep and circadian rhythms in the development of PTSD symptoms. Nevertheless, the longitudinal design and lagged effects analytic approach were in line with recommendations for interpreting the causal direction of effects for diurnal cortisol secretion on health outcomes (Adam et al., Reference Adam, Quinn, Tavernier, McQuillan, Dahlke and Gilbert2017).
In summary, lower waking cortisol was associated with concurrent PTSD diagnostic status and symptom severity, as well as PTSD diagnosis at the subsequent assessment, while greater childhood trauma was associated with a slower diurnal cortisol decline, higher waking alpha-amylase, and slower diurnal alpha-amylase increase. Findings highlight the potential utility of blunted waking cortisol as an early biomarker of PTSD symptoms and risk of developing PTSD. Future studies employing daily diary or ecological momentary assessment paradigms are needed to examine how daily experiences shape the diurnal cortisol secretion patterns associated with risk for subsequent PTSD. This knowledge could inform early cognitive-behavioral interventions designed to promote resilience in the aftermath of IPV (Rothbaum et al., Reference Rothbaum, Kearns, Price, Malcoun, Davis, Ressler and Houry2012). Childhood trauma was an important correlate of both HPA and SNS activity, with greater exposure associated with blunted diurnal cortisol and alpha-amylase slopes and higher waking alpha-amylase. Childhood trauma exposure should be assessed in research and clinical care of recent trauma survivors, as results suggest it is associated with stress response dynamics in those at risk for PTSD.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291721004050.
Acknowledgements
This study was supported, in part, by grants from the National Institutes of Health (K01 MH101403, U54 MD007593, U54MD007586, R01MH108155, R01MD010757, and R01DA040966). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. Additionally, the authors are grateful for the comments of Dr Chrystyna Kouros and Dr Lesa Hoffman.
Conflict of interest
The authors report no conflicts of interest.
