Predictors of PTSD Treatment Response Trajectories in a Sample of Childhood Sexual Abuse Survivors: The Roles of Social Support,Coping,and PTSD Symptom Clusters

Abstract

This study aimed to (a) identify posttraumatic stress disorder (PTSD) trajectories in a sample of Danish treatment-seeking childhood sexual abuse (CSA) survivors and (b) examine the roles of social support, coping style, and individual PTSD symptom clusters (avoidance, reexperiencing, and hyperarousal) as predictors of the identified trajectories. We utilized a convenience sample of 439 CSA survivors attending personalized psychotherapy treatment in Denmark. Four assessments were conducted on a six monthly basis over a period of 18 months. We used latent class growth analysis (LCGA) to test solutions with one to six classes. Following this, a logistic regression was conducted to examine predictors of the identified trajectories. Results revealed four distinct trajectories which were labeled high PTSD gradual response, high PTSD treatment resistant, moderate PTSD rapid response, and moderate PTSD gradual response. Emotional and detached coping and more severe pretreatment avoidance and reexperiencing symptoms were associated with more severe and treatment resistant PTSD. High social support and a longer length of time since the abuse were associated with less severe PTSD which improved over time. The findings suggested that treatment response of PTSD in CSA survivors is characterized by distinct patterns with varying levels and rates of PTSD symptom improvement. Results revealed that social support is protective and that emotional and detached coping and high pretreatment levels of avoidance and reexperiencing symptoms are risk factors in relation to PTSD severity and course. These factors could potentially identify patients who are at risk of not responding to treatment. Furthermore, these factors could be specifically addressed to increase positive outcomes for treatment-seeking CSA survivors.

Keywords

sexual abuse child abuse PTSD treatment

Introduction

There is a plethora of evidence suggesting that childhood sexual abuse (CSA) is a risk factor for a wide range of long-term negative effects, including substance and alcohol abuse (Cutajar et al., 2010), revictimisation (Arata, 2002), and physical and mental health disorders (Fergusson, McLeod, & Horwood, 2013). Posttraumatic stress disorder (PTSD) as a consequence of CSA has been widely studied, and it has been estimated that between 37% and 43% of individuals who experienced CSA meet the criteria for PTSD (Paolucci, Genuis, & Violato, 2001). Some individuals will require treatment for their trauma-related symptoms. A range of evidence-based psychological treatments, such as trauma-focused group therapy (Cole, Sarlund-Heinrich, & Brown, 2007), cognitive behavioral therapy (CBT; Cloitre, Koenen, Cohen, & Han, 2002), and cognitive processing therapy (CPT; Chard, 2005), have found to be effective in reducing trauma-associated symptoms. Indeed, one meta-analysis of 44 studies (comprising 59 treatment conditions) found moderate effect sizes relating to PTSD improvement among adults attending treatment for CSA-related psychopathology (g = 0.72-0.77; Taylor & Harvey, 2010).

Although there have been a number of studies and subsequent meta-analyses exploring changes in PTSD over the course of treatment among CSA survivors, many studies have examined how PTSD changes over time by using methods which assume a homogeneous population (Chard, 2005; Cloitre et al., 2002; Cole et al., 2007; Taylor & Harvey, 2010). These methods, however, do not reflect differential treatment response trajectories. Recently, there has been an increase in the number of studies which have identified distinct subgroups which differ in terms of levels and rates of treatment response. In addition, these studies have examined predictors of group membership. It has been suggested that identifying distinct subgroups could lead to treatment which is more tailored, as survivors who may require a more intensive or alternative therapy could be identified at an early stage (Elliott, Biddle, Hawthorne, Forbes, & Creamer, 2005). Furthermore, examining predictors of class membership could potentially identify risk or protective factors which could be targeted during treatment.

One study examining CPT treatment response in a sample of female victims of interpersonal violence found two distinct trajectories. One group (87%) was characterized by a reduction in PTSD symptoms and the second group (13%) was characterized by PTSD symptoms which did not significantly improve over time (Stein, Dickstein, Schuster, Litz, & Resick, 2012). Participants who received only a cognitive component of CPT or a written account component of CPT (compared with those who received full CPT), as well as individuals with major depression or severe hyperarousal symptoms, were more likely to be in the class which did not significantly improve over time (Stein et al., 2012). Another study examined treatment response, using a sample of 805 veterans who completed a residential program for the treatment of PTSD symptoms. Three trajectories were identified: one group (48.8%) of participants demonstrated significant reductions in PTSD symptoms which were maintained at the follow-up, the second group (41%) had high levels of PTSD which did not improve over time, and the third group (10.2%) had low levels of PTSD symptoms which remained stable over time (Currier, Holland, & Drescher, 2014). This study also revealed that individuals who had symptoms which responded to treatment had intermediate levels of symptom severity, mental and physical health, and combat exposure when compared with individuals who had symptoms which did not respond to treatment (Currier et al., 2014). More recently, Steine et al. (2017) examined posttraumatic symptom trajectories among adult CSA survivors who had attended a support center (offering support including information and the opportunity to attend support groups). Two trajectories were identified: one was characterized by subclinical PTSD which decreased over time and another was characterized by clinical PTSD which only decreased slightly over time. This study also found that individuals in the clinical PTSD class had experienced more severe abuse and had higher levels of relationship difficulties and lower levels of perceived social support (Steine et al., 2017). The studies described above have highlighted that there are distinct unobservable subgroups relating to the longitudinal course of PTSD. Although one study (Steine et al., 2017) has identified PTSD trajectories among CSA survivors, there are no known studies which have identified PTSD treatment response trajectories within this population.

Research has also evidenced that factors such as social support (Karstoft, Armour, Elklit, & Solomon, 2013; Steine et al., 2017), coping style (Elklit, 2015; Filipas & Ullman, 2006; Karstoft, Armour, Elklit, & Solomon, 2015; Spaccarelli, 1994), and PTSD symptom clusters (Stein et al., 2012) can influence PTSD course and severity. It is plausible that protective and risk factors associated with PTSD development and maintenance are also able to explain variance in treatment response trajectories among CSA survivors. To the best of our knowledge, these factors have not been studied specifically in relation to PTSD treatment response trajectories among CSA survivors. This gap in the literature will be addressed in the current study.

Coping style has been described by Lazarus and Folkman (1984) as the behavioral and cognitive efforts used to manage both external and internal stressors and demands. Emotion-focused coping aims to reduce internal stress via methods such as using distraction drugs, alcohol, and reappraisal. On the contrary, problem-focused coping (or rational coping) aims to minimize external stress via directly addressing the problem (Lazarus & Folkman, 1984). Other coping styles which have been conceptualized include detached coping (e.g., feeling independent from the circumstances) and avoidant coping (e.g., daydreaming about when things were better; Roger, Jarvis, & Najarian, 1993). Coping style has consistently been found to explain variance in long-term functioning in trauma populations (e.g., Elklit, 2015; Karstoft et al., 2015). Research has shown that emotional and avoidance coping are associated with more severe and chronic PTSD (Karstoft et al., 2015). One recent longitudinal study found that less use of emotional coping in veterans was associated with lower odds of being in the chronic PTSD trajectory (Karstoft et al., 2015). Similar results have also been evidenced in CSA survivors (Elklit, 2015). In contrast, problem-focused coping has been found to be protective (Coffey, Leitenberg, Henning, Turner, & Bennett, 1996). Recently, Karstoft et al. (2015) found that veterans who had higher levels of problem solving coping had decreased odds of membership in both the chronic and worsening PTSD classes. Problem-focused coping has also been found to be associated with less psychological distress in adult CSA survivors (Coffey et al., 1996). The evidence above suggests that emotional and avoidant coping is associated with increased risk and rational coping is protective. It has been argued that individuals who use emotional and avoidance coping methods find it more difficult to process the negative trauma-related emotions, thus maintaining PTSD and making it more difficult to recover (Ehlers & Clark, 2000). On the contrary, individuals who use problem-focused coping methods are more able to process such emotions (Ehlers & Clark, 2000).

Social support has also been found to influence the development and maintenance of PTSD (Karstoft et al., 2013). Social support has been described as material and psychological resources which can increase an individual’s capacity to cope with stress (Cohen & Wills, 1985). Two meta-analyses both concluded that low levels of perceived social support was one of the strongest predictors of PTSD development (Brewin, Andrews, & Valentine, 2000; Ozer, Best, Lipsey, & Weiss, 2003). Consistent with this, higher perceived social support has been shown to predict PTSD recovery in veteran samples (Karstoft et al., 2013; Koenen, Stellman, Stellman, & Sommer, 2003). Furthermore, studies of CSA populations have identified social support as a protective factor against PTSD symptom development (Elklit, 2015; Tremblay, Hebert, & Piche, 1999). The stress buffering hypothesis asserts that social support buffers high stress levels which can protect against maladaptive behavior and symptom development (Cohen & Wills, 1985). Tremblay, Hébert, and Piché (1999) argued that social support can influence the cognitive evaluation of the traumatic event and this in turn can reduce the reaction of the victim. Consistent with this theory, Spaccarelli and Kim (1995) found that parental support was associated with higher social competence, lower psychopathology and stress, and negative appraisals in female victims of CSA. The evidence described suggests that social support may protect against PTSD symptom development; furthermore, it has been found to be associated with recovery. Alternatively, it has been suggested that PTSD can have a negative impact on social relationships. For example, one study found that more severe PTSD predicted marital problems and low relationship satisfaction (Campbell & Renshaw, 2013).

In the Diagnostic and Statistical Manual of Mental Disorders (4th ed., text rev.; DSM-IV-TR; American Psychiatric Association [APA], 2000), PTSD was characterized by three symptom clusters. These are avoidance and numbing, reexperiencing, and hyperarousal. The criteria have recently been updated and a fourth symptom cluster (negative alterations in cognition and mood) has been added (Diagnostic and Statistical Manual of Mental Disorders [5th ed.; DSM-5; APA, 2013]). The roles of individual PTSD symptom clusters have been implicated in the overall severity of PTSD. However, there have been some inconsistencies in the findings. For example, Schell, Marshall, and Jaycox (2004) found that arousal symptoms predicted severity of all other symptoms at 3- and 12-month assessments. Consistent with these results, Stein et al. (2012) revealed that hyperarousal symptom severity predicted the trajectory characterized by little improvement when attending treatment. The authors suggested that the arousal symptoms are distracting and that this could interfere with treatment engagement (Stein et al., 2012). However, other studies have suggested that reexperiencing symptoms are the most clinically relevant (Creamer, Burgess, & Pattison, 1992). It has been postulated that high levels of reexperiencing and hyperarousal reinforce the learning that occurred during the traumatic experience, contributing to the development and maintenance of PTSD (Shalev, 2002). Contrary to the above, there is also evidence which suggests that avoidance and numbing predict overall PTSD (Hyland et al., 2016). Indeed, it has been argued that avoidance is used as an attempt to reduce the stress associated with hyperarousal and reexperiencing, however, in the long-term it maintains the PTSD (Clark & Beck, 2010). Overall, the results with regard to PTSD symptom clusters and PTSD severity and course have been inconsistent and further research is required. Further understanding of which symptoms are important in predicting treatment response would have important theoretical and clinical implications. Finally, trauma characteristics (e.g. Steine et al., 2017) and demographic factors (e.g. Ullman & Filipas, 2005) have been found to influence PTSD symptomology. Therefore, the current study will also include trauma (length of time since the trauma occurred and whether the participant experienced an additional trauma) and demographic (age, sex and education) variables in the analysis.

Taken together, the extant literature suggests that there are qualitatively distinct patterns of longitudinal PTSD and that coping, social support, and PTSD symptoms clusters are important predictors of long-term PTSD course and severity. To the best of our knowledge, these factors have not been examined as predictors of PTSD treatment response trajectories in CSA survivors. The current study aims to (a) identify PTSD trajectories in a sample of Danish treatment-seeking CSA survivors and (b) examine predictors of the identified trajectories. Based on previous research, we predict that there would be an overall decrease in PTSD symptom scores. We also predicted that there would be multiple and differing trajectories relating to treatment response. Furthermore, we predict that emotional coping will be associated with treatment resistance and social support and rational coping will be associated with treatment response. No specific predictions were made with respect to PTSD symptom clusters.

Method

Participants

Participants (N = 456) were outpatients attending treatment centers for survivors of CSA in Denmark. After exclusions due to missing data (discussed below), the effective sample size was 439. The mean age of the sample was 36.46 years (range = 15-77; SD = 10.83); all participants were Caucasian and the majority (85.8%) were female. The mean number of years spent in education was 13.44 (SD = 3.51). The mean age that the abuse started was 6.62 years (SD = 4.26), and the mean age the abuse ended was 13.32 years (SD = 7.38). Participants presenting under the influence of drugs or alcohol, or with a personality disorder characterized mainly by perpetrating traits, self-destructive behavior, psychosis, or receiving treatment elsewhere were excluded and referred to another relevant agency.

Procedure

Personalized psychotherapy was conducted by psychologists. This method of treatment can involve multiple interventions (including cognitive, psychodynamic, and behavioral treatments) that are matched specifically to the patient depending on underlying personality features thought to be related to the problematic symptoms (Millon, 1999). There is no common treatment manual, and treatment plans are based on the scores derived from the initial assessments. Treatment was free, individual, and weekly and there was no limit to the number of sessions offered. During the second therapy session (T1), all participants attending completed a number of questionnaires. These questionnaires were repeated at 6 months (T2), 12 months (T3), and 18 months (T4). Continuous variables for the duration or number of treatment sessions were not available; however, only those who had been regularly participating in therapy received the assessments.

Measures

Demographics

The following sociodemographic characteristics were included in the current analysis: total education and age (both continuous scores measured in years) and sex (male reference group).

Trauma characteristics

Participants were also asked about other traumas they had experienced, including rape, physical assault, life threatening accident, fire, flood or natural disaster, physically abused as a child, neglected as a child, witnessed another person being seriously injured or killed, threatened, held captive or kidnapped. Each question had a yes or no response option and a single dichotomous variable was created. Individuals were also asked the length of time in years since the end of the abuse.

Harvard Trauma Questionnaire (HTQ)

PTSD symptoms were measured using part IV of the Danish HTQ (Mollica et al., 1992). This section is comprised of 30 items, the first 16 of which correspond to the PTSD symptoms in the DSM-IV-TR (APA, 2000). Each item is scored on a 4-point Likert-type scale ranging from ‘not at all’ to ‘all the time.’ The first 16 items were summed together giving a score ranging from 0 to 64. This score of overall PTSD severity was used to estimate the trajectories. The standard cutoff score for a probable PTSD diagnosis is a mean score of 2.5 or a total score of 40 (Mollica et al., 1992). In addition, there are three subscales: hyperarousal, reexperiencing, and avoidance. These were included in the regression analysis. Part IV of the HTQ has been validated in Denmark (Bach, 2003) and has been found to have good reliability and criterion validity (Mollica et al., 1992).

The Coping Styles Questionnaire (CSQ)

The CSQ (Roger et al., 1993) was used to measure coping strategies used. The version used in the current study has 37 items, all of which are scored on a 4-point Likert-type scale ranging from never to always. Validation of the CSQ has confirmed there are four clusters (Elklit, 1996). These are emotional, avoidance, rational, and detached coping. O’Connor and Elklit (2008) reported the following Cronbach’s alpha coefficients: rational = 0.7, emotional = 0.75, avoidance = 0.65, detached = 0.43.

Crisis Support Scale (CSS)

The CSS (Joseph, Andrews, Williams, & Yule, 1992) was used to measure perceived social support both during the time of the trauma and at baseline. There are seven items each of which is measured on a 7-point Likert-type scale ranging from never to always. The scale has been found to have good internal consistency (Cronbach alphas range between 0.67 and 0.82) and good discriminatory power (Elklit, Pedersen, & Jind, 2001).

Missing Data and Attrition

Participants with over 20% of all baseline values missing were excluded from the analysis. We conducted multiple imputation to handle all missing HTQ scores over all four time points. Multiple imputation replaces the missing scores with plausible estimations based on the values of observed variables, with standard errors taking the uncertainty of each value into account (Rubin, 1987). It is based on the assumption that the data are missing at random (MAR). Missing data analysis revealed that 99.05% of values were complete at baseline. Of the initial 439 participants at T1, 70.84% of participants completed T2, 44.43% completed T3, and 25.06% completed T4. We imputed 100 data sets in SPSS version 23 and these were then exported to Mplus 7.3 (Muthén & Muthén, 2014). The results were pooled based on Rubin’s rules (Rubin, 1987). The imputation model utilized variables associated with HTQ scores at T2, T3, and T4 to increase the plausibility of the MAR assumption. Variables used included T1 HTQ scores, education, age, and social support. Finally, due to the small amount of missing data on all predictor variables (0.2% missing cases), we imputed values using the expectation maximization algorithm in SPSS version 22. Little’s (1988) missing completely at random (MCAR) test suggested that the missing baseline values were missing completely at random (χ² = 7424.65, df = 7322, p = .198).

Analytic Plan

The analysis was conducted using Mplus 7.2 (Muthén & Muthén, 2014). We conducted latent class growth analysis (LCGA; Muthén, 2001; Nagin, 1999). LCGA is a type of finite mixture modeling, which divides a heterogeneous sample into a number of latent subgroups based on different growth trajectories. The variance of latent slope and intercept are fixed to zero within classes and are allowed to vary only between classes. As there is no covariance between the slope and intercept, and there are less parameters to estimate, it is easier for the model to converge (Nagin, 1999; Roeder, Lynch, & Nagin, 1999). This method has been used by other studies which have examined trajectories of PTSD treatment response (e.g., Currier et al., 2014). We tested class solutions comprising between one to six classes. Optimal model fit was determined using a variety of fit indices: the Bayesian information criterion (BIC; Schwarz, 1978), the Akaike information criterion (AIC; Akaike, 1987), and the sample-size-adjusted Bayesian information criterion (adjusted BIC; Sclove, 1987). As multiple imputation was used, the Lo-Mendell-Rubin likelihood ratio test (LMR-LRT) and the bootstrapped likelihood ratio test (BLRT) were not available. It has been argued that the optimal group solution is characterized by the lowest BIC, adjusted BIC, and AIC and a high classification accuracy (the entropy value should be close to 1). There is evidence to suggest that the BIC performs best when deciding on the number of groups (Nylund, Asparoutiov, & Muthén, 2007). We tested the model with both linear and quadratic terms. The robust maximum likelihood estimator was used. Maximum likelihood estimators are suitable for data that do not meet the assumption of normality (Satorra & Bentler, 1994; Yuan & Bentler, 2000) and have been found to be asymptotically efficient and consistent in large samples (Bollen, 1989).

A multinomial logistic regression was then performed to examine predictors of class membership using the three-step approach (R3STEP function; Asparouhov & Muthén, 2014; Vermunt, 2010). We included age, sex, education, coping styles, PTSD symptom clusters, and social support in the analysis. A multinomial logistic regression and a latent class analysis can be combined in one step; however, the resulting classes will be based on the predictors as opposed to only the latent class indicators. This can lead to the classes losing their meaning (Asparouhov & Muthén, 2014). The current approach, however, estimates the latent class model using only the required indicators. Following this first step, a variable S (most likely class membership) is created; this is based on the latent class posterior distribution (obtained during step one), and the classification uncertainty rate is also taken into account. In the final step, the most likely class variable (which includes measurement error) is used as the dependant variable in a regression analysis (Asparouhov & Muthén, 2014; Vermunt, 2010).

Results

Examination of the mean scores revealed a reduction in PTSD symptoms over time. At T1, the mean PTSD score was 45.53 (SD = 8.07), at T2 the mean score was 40.53 (SD = 9.46), at T3 the mean score was 37.12 (SD = 10.84), and at T4 the mean score was 34.77 (SD = 10.54). We used LCGA to examine PTSD trajectories. The model fit improved with the addition of the quadratic term. Fit indices for each model (with the quadratic term) are displayed in Table 1. Each time a class was added, the results revealed an improvement in fit. When the fifth class was added, the reductions in the fit indices were small suggesting that the addition of this class did not greatly improve the model. The BIC leveled off (there was an increase of 0.08) and although the adjusted BIC and the AIC decreased, the reductions (12.61 and 16.25, respectively) were small. Therefore, based on the fit indices, theory, and parsimony the four-class solution was chosen as optimal for the data.

Table 1.

Fit Indices for Each LCGA Model.

No. of classes	Entropy	BIC	Adjusted BIC	AIC
1	—	3,238.507	3,216.292	3,209.916
2	0.815	2,637.802	2,602.893	2,592.872
3	0.823	2,459.844	2,412.241	2,398.576
4	0.777	2,435.271	2,374.974	2,357.665
5	0.763	2,435.355	2,362.365	2,341.412
6	0.744	2,435.507	2,349.822	2,325.225

Note. Cutoff score for probable PTSD diagnosis is >40. LCGA = latent class growth analysis; BIC = Bayesian information criterion; AIC = Akaike information criterion.

The trajectories are presented in Figure 1. The first trajectory (15.03%; high PTSD treatment resistant) was characterized by high clinical levels of PTSD at T1 which did not significantly change over time; the quadratic term was not significant for this class (Intercept = 3.32, SE = 0.1, p < .00; Slope = −0.00, SE = 0.1, p > .05; Quadratic = −0.04, SE = 0.03, p > .05). This class experienced the least changes in symptoms over time (decrease of 5.09). Clinical levels of PTSD remained even after 18 months of treatment. The second trajectory (15.71%; moderate PTSD rapid response) was characterized by moderate subclinical levels of PTSD at T1 which significantly improved over time; the quadratic term was also significant (Intercept = 2.27, SE = 0.18, p < .00; Slope = −0.72, SE = 0.13, p < .00; Quadratic = 0.13, SE = 0.04, p < .00). The largest improvement occurred between T1 and T2 (decrease of 9.43), and after T2 the symptoms continued to improve but at a more gradual pace. This group experienced the largest improvements in PTSD symptoms (15.59). The third pattern (36.22%; high PTSD gradual response) was characterized by high clinical levels of PTSD at T1 which significantly improved over time. At T4, the mean PTSD score was below the clinical cutoff score. The quadratic term was not significant (Intercept = 3.01, SE = 0.15, p < .00; Slope = −0.22, SE = 0.12, p = .05; Quadratic = 0.01, SE = 0.03, p > .05). The symptoms decreased in a linear fashion, and there was an overall decrease of 9.21 units. Trajectory four (33.03%; moderate PTSD gradual response) was characterized by moderate but clinical levels of PTSD at T1 which significantly improved over time; the quadratic term was significant (Intercept = 2.66, SE = 0.10, p < .001; Slope = −0.46, SE = 0.13, p < .001; Quadratic = 0.07, SE = 0.03; p < .05). In this group, the HTQ score decreased by 12.26 units.

Figure 1.

PTSD treatment response trajectories.

The demographic characteristics of the full sample and each class can be found in Table 2. The odds ratios (ORs) and 95% confidence intervals (CIs) for all predictors are presented in Table 3. A multinomial logistic regression was conducted to examine predictors of class membership. All classes were compared with each other. Higher social support at T1 (OR = 0.81, 95% CI: [0.74, 0.88]) and a longer time period since the abuse (OR = 0.93, 95% CI: [0.87, 0.98]) predicted a decrease in odds of being in the treatment resistant class when compared with the moderate PTSD gradual response class, whereas increased levels of reexperiencing symptoms (OR = 1.85, 95% CI: [1.36, 2.49]), avoidance symptoms (OR = 1.49, 95% CI: [1.2, 1.84]), arousal symptoms (OR = 1.38, 95% CI: [1.01, 1.87]), and emotion-focused coping (OR = 1.22, 95% CI: [1.02, 1.46]) were associated with increased odds of being in the treatment resistant class. When the moderate PTSD rapid response class was compared with the moderate PTSD gradual response class, increased social support at T1 (OR = 1.1, 95% CI: [1.01, 1.21]) and older age (OR = 1.63, 95% CI: [1.09, 2.42]) predicted an increase in odds of being in the rapid response class. A longer length of time since the abuse occurred was associated with a decrease in odds of membership in the rapid response class (OR = 0.93, 95% CI: 0.89-0.97). The odds of being in the high PTSD gradual response compared with moderate PTSD gradual response were decreased by being female (OR = 0.33, 95% CI: [0.12, 0.87]), having higher social support (OR = 0.91, 95% CI: [0.85, 0.97]) and a longer period of time since the abuse occurred (OR = 0.95, 95% CI: [0.9, 0.99]). The odds of being in the high class were increased with higher levels of reexperiencing (OR = 1.34, 95% CI: [1.1, 1.63]) and avoidance symptoms at T1 (OR = 1.22, 95% CI: [1.01, 1.45]).

Table 2.

Demographics and Descriptive Statistics for the Full Sample and for Each Trajectory.

Variable	Full Sample	Class 1	Class 2	Class 3	Class 4
Variable	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)
Female N (%)	376 (85.8)	65 (95.6)	66 (93)	110 (19.7)	135 (83.3)
Other trauma N (%)	372 (84.7)	60 (90.9)	50 (72.4)	143 (89.9)	119 (82.1)
Education	13.44 (3.51)	12.43 (3.59)	14.01 (3.27)	12.99 (3.49)	14.01 (3.48)
Age	36.46 (10.83)	38.59 (11.52)	33.37 (8.87)	37 (11.67)	36.46 (10.37)
Social support^a	11.97 (6.21)	9.56 (3.13)	12.67 (7.71)	11.38 (5)	13.3 (7.11)
Social support^b	30.13 (7.45)	24.76 (6.42)	35.01 (6.87)	28.45 (7.11)	31.74 (6.52)
Coping rational	24.21 (4.94)	23.43 (4.91)	25.33 (5.04)	23.41 (5)	24.72 (4.73)
Emotional	24.80 (5.82)	28.65 (4.89)	20.46 (4.76)	26.89 (5.39)	23.51 (5.3)
Detached	10.91 (2.82)	10.35 (2.57)	10.64 (2.72)	10.59 (2.72)	11.54 (2.95)
Avoidant	21.68 (4.24)	22.26 (4.34)	19.95 (4.13)	22.42 (4.17)	21.57 (4.12)
Reexperiencing	10.44 (2.84)	12.98 (2.22)	8.56 (2.59)	10.97 (2.51)	9.73 (2.56)
Avoidance	19.51 (4.09)	22.79 (2.47)	15.35 (3.92)	12.19 (3.36)	18.53 (3.34)
Hyperarousal	15.47 (2.94)	17.68 (1.6)	13.12 (3.5)	21.19 (2.24)	14.86 (2.72)
Time since abuse	22.2 (11.69)	20.61 (13.21)	21.28 (9.5)	22.11 (11.78)	25.29 (1.48)

Note. Class 1 = high PTSD treatment resistant, Class 2 = moderate PTSD rapid response, Class 3 = high PTSD gradual response, Class 4 = moderate PTSD gradual response.

During trauma.

At baseline.

Table 3.

ORs and 95% CIs for Predictors of Trajectory (Class) Membership.

Predictor	1 vs. 4	2 vs. 4	3 vs. 4	1 vs. 3	2 vs. 3	1 vs. 2
Predictor	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Age	1.32 [0.78, 2.26]	1.63 [1.09, 2.42]*	1.23 [0.91, 1.66]	1.07 [0.63, 1.81]	1.31 [0.86, 2.02]	0.81 [0.43, 1.51]
Female	0.51 [0.07, 3.59]	0.99 [0.18, 5.29]	0.33 [0.12, 0.87]*	1.55 [0.21, 11.02]	3.02 [0.57, 15.77]	0.51 [0.04, 5.76]
Education	0.93 [0.8, 1.08]	0.89 [0.75, 1.06]	0.94 [0.58, 1.05]	0.99 [0.85, 1.14]	0.94 [0.78, 1.36]	1.05 [0.83, 1.3]
Time since trauma	0.93 [0.87, 0.98]*	0.93 [0.89, 0.97]*	0.95 [0.9, 0.99]*	0.98 [0.9, 1.03]	0.99 [0.9, 1.04]	0.99 [0.93, 1.06]
Social support^a	0.9 [0.72, 1.11]	0.95 [0.87, 1.04]	0.97 [0.91, 1.03]	0.93 [0.74, 1.16]	0.98 [0.89, 1.08]	0.95 [0.75, 1.18]
Social support^b	0.81 [0.74, 0.88]***	1.1 [1.01, 1.21]*	0.91 [0.85, 0.97]**	0.89 [0.82, 0.96]*	1.21 [1.09, 1.34]***	0.73 [0.65, 0.83]***
Reexperiencing	1.85 [1.36,2.49]***	0.86 [0.68, 1.07]	1.34 [1.11, 1.63]**	1.37 [1.03, 1.81]*	0.64 [0.49, 0.83]**	2.15 [1.5, 3.1]***
Hyperarousal	1.38 [1.01,1.87]*	0.97 [0.83, 1.13]	1.13 [0.96, 1.33]	1.22 [0.9, 1.64]	0.86 [0.71, 1.05]	1.42 [1.02, 1.96]**
Avoidance	1.49 [1.2, 1.84]***	0.87 [0.71, 1.06]	1.22 [1.01, 1.45]*	1.23 [1.03, 1.45]*	0.72 [0.57, 0.91]**	1.71 [1.31, 2.24]***
Rational coping	0.96 [0.81, 1.12]	0.97 [0.87, 1.08]	1.05 [0.95, 1.15]	0.91 [0.77, 1.07]	0.93 [0.81, 1.06]	0.98 [0.81, 1.18]
Emotional coping	1.22 [1.02, 1.46]*	0.89 [0.79, 1.01]	1.09 [0.98, 1.19]	1.13 [0.94, 1.34]	0.82 [0.72, 0.94]**	1.37 [1.11, 1.68]**
Detached coping	1.28 (0.89-1.8)	0.84 [0.67, 1.04]	1 [0.93, 1.2]	1.28 [0.88, 1.85]	0.84 [0.65, 1.08]	1.52 [1.02, 2.27]*
Avoidant coping	0.83 [0.67, 1]	0.98 [0.83, 1.14]	0.95 [0.85, 1.05]	0.87 [0.71, 1.05]	1.03 [0.88, 1.21]	0.84 [0.66, 1.06]
Other trauma	0.63 [0.135, 2.9]	0.63 [0.24, 1.64]	1.43 [0.32, 6.35]	0.44 [0.12, 1.58]	0.44 [0.09, 2.07]	0.98 [0.19, 5.1]

Note. Class 1 = high PTSD treatment resistant, Class 2 = moderate PTSD rapid response, Class 3 = high PTSD gradual response, Class 4 = moderate PTSD gradual response. OR = odds ratio; CI = confidence interval.

During trauma.

At T1.

p < .05. **p < .01. ***p < .001.

Odds of being in the high PTSD treatment resistant class were increased with higher levels of reexperiencing (OR = 1.37, 95% CI: [1.03, 1.81]) and avoidance symptoms (OR = 1.23, 95% CI: [1.03, 1.45]) and decreased with higher levels of social support (OR = 0.89, 95% CI: [0.82, 0.96]) when compared with the high PTSD moderate response class. When the moderate PTSD rapid response class was compared with the high PTSD gradual response class, odds of being in the moderate class increased with higher levels of social support (OR = 1.21, 95% CI: [1.09, 1.34]) and decreased with higher reexperiencing (OR = 0.64, 95% CI: [0.49, 0.83]), avoidance symptoms (OR = 0.72, 95% CI: [0.57, 0.91]), and higher levels of emotional coping (OR = 0.82, 95% CI: [0.72, 0.94]). Finally, odds of being in the treatment resistant class decreased with higher levels of social support (OR = 0.73, 95% CI: [0.65, 0.83]) and increased with higher reexperiencing (OR = 2.15, 95% CI: [1.5, 3.1]), avoidance (OR = 1.71, 95% CI: [1.31, 2.24]), and hyperarousal symptoms (OR = 1.42, 95% CI: [1.02, 1.96]) and higher levels of emotion-focused (OR = 1.37, 95% CI: [1.11, 1.68]) and detached coping (OR = 1.52, 95% CI: [1.02, 2.27]) when compared with the moderate PTSD rapid responding group.

Discussion

Consistent with the extant literature, the results revealed that overall PTSD scores decreased over the course of treatment (Taylor & Harvey, 2010). At the first assessment, the mean PTSD score was over the cutoff score for a probable PTSD diagnosis (M = 45.53, SD = 8.07), and by T3, this had reduced to below the clinical cutoff score (M = 37.12, SD = 10.84). The first aim of this study was to examine whether multiple PTSD trajectories were present. The initial examination of the mean HTQ scores did not reflect the four distinct subgroups within this population which were identified by applying LCGA. Two classes characterized by moderate PTSD at baseline were identified. The moderate PTSD gradual response class (33.03%) entailed a moderate PTSD responding group in which symptoms improved steadily over a period of 18 months. The moderate PTSD rapid response class (15.71%) was characterized by a rapid improvement in symptoms within the first 6 months of treatment; this improvement continued but was more gradual over the remaining treatment period. The identification of this trajectory was in line with a study by Elliot et al. (2005) examining PTSD trajectories in veterans attending treatment. This suggests that the moderate PTSD rapid response trajectory is not exclusive to CSA trauma populations. It may be that this subgroup does not require long-term treatment as the majority of the improvement occurs within the earlier stages of the treatment. The identification of two groups with moderate PTSD that respond differently has not been found in previous studies. This finding suggests that individuals with moderate PTSD at T1 are likely to have a significant improvement in symptoms; however, they may improve at differing rates. Two groups with relatively high PTSD scores at baseline were also identified. The high PTSD gradual response group (36.22%) was characterized by clinical PTSD symptoms which decreased gradually to a nonclinical level over a period of 18 months. This finding was similar to a class found in a sample of female victims of interpersonal violence, characterized by high PTSD at baseline, which were below the clinical cutoff at the end of treatment (Stein et al., 2012). Although these results have suggested that CSA survivors with high PTSD at baseline can benefit from treatment, there was another class which showed no significant improvement in symptoms over time. The high PTSD treatment resistant class (15.03%) did not appear to benefit from the treatment, and after 18 months, the mean HTQ score was still above the clinical cutoff point. Treatment resistant trajectories have also been found among veterans attending inpatient PTSD treatment centers (Currier et al., 2014) and female victims of interpersonal violence (Stein et al., 2012).

The second aim of the current study was to identify predictors of PTSD treatment response trajectories. The extant literature has demonstrated that emotion-focused coping is a risk factor for the development and maintenance of PTSD (Karstoft et al., 2015). Consistent with this, our findings revealed that higher levels of emotional coping predicted membership in the treatment resistant class when compared with both moderate PTSD response classes and the high PTSD gradual responder when compared with the moderate PTSD rapid response class. This suggests that emotional coping is associated with more severe pretreatment PTSD which may not respond to treatment. This finding highlights the long-term maladaptive nature of emotion-focused coping. It has been suggested that emotional coping can be adaptive in the short term (Briere, 2002). For example, during a traumatic event (such as CSA), problem-focused coping (e.g., resistance) may lead to increased aggression from the perpetrator, whereas emotional coping would reduce the immediate distress for the victim (Spaccarelli, 1994). However, if this coping style is maintained, it can have harmful long-term effects. It is thought that it can interfere with the processing of the trauma memories and subsequently maintain PTSD (Briere, 2002; Ehlers & Clark, 2000). Interestingly, the results also revealed that detached coping was associated with membership in the treatment resistant class when compared with the moderate rapid response class. This suggests that feelings of being independent from the trauma can also have a long-term negative impact. It is possible that feeling detached from the traumatic event prevents the processing of the traumatic memories in a similar way to emotion-focused coping (Ehlers & Clark, 2000). This finding however is contrary to the view that detached coping is an adaptive response to stress (Roger et al., 1993). Notably, rational and avoidant coping were not associated with PTSD recovery. It is possible that the inconsistencies in the findings are due to the measurements used. Although these results require further investigation, they have some important clinical implications. First, individuals with high levels of emotional and detached coping styles should be considered as having an increased risk of severe PTSD which does not respond to psychotherapy. Second, it is possible that reducing emotion-focused and detached coping and teaching a more adaptive coping style would allow effective processing of the trauma memories and potentially reduce PTSD symptoms, thus improving outcomes for CSA survivors.

The results in the present study concurred with previous literature which has consistently evidenced the protective nature of social support in relation to PTSD (Brewin et al., 2000; Ozer et al., 2003; Steine et al., 2017). Social support (at T1) was associated with reduced odds of being in the high PTSD treatment resistant class when compared with both moderate PTSD classes. Furthermore, when both high PTSD classes were compared, social support was associated with treatment response, and when both moderate classes were compared, it was associated with a more rapid response. These results suggest that current social support is associated with less severe PTSD. Furthermore, individuals with severe PTSD and greater social support are more likely to respond to treatment, and individuals with moderate PTSD and those with greater social support are more likely to respond to treatment at a faster rate. Our findings lend support to the stress buffering theory which posits that social support improves stress regulation by influencing subjective appraisal and both internal and external stress responses (Cohen & Wills, 1985). In line with this theory, social support has been found to increase healthy behaviors and decrease risky behaviors (Holahan, Moos, Holahan, & Brennan, 1995). Moreover, it has been found to be associated with higher self-worth, and a sense of purpose, which has been thought to increase motivation for more positive self-care (Southwick & Charney, 2012). It is also possible the association between PTSD severity and low social support in the current study is due to the negative impact that PTSD symptomology can have on relationships (Campbell & Renshaw, 2013). Further studies are required to increase understanding of the mechanisms underlying this association. If social support influences PTSD severity and course, it is possible that targeting social skills during treatment could also improve outcomes for CSA survivors. In line with this, previous research has demonstrated the effectiveness of social skills interventions. For example, a review of over 100 studies examining the outcomes of interventions focusing on social and behavioral skills found that 83% of the studies evidenced positive effects such as decreased psychological distress (Hogan, Linden, & Najarian, 2002).

Previous research examining the role of symptoms clusters in overall PTSD severity has shown conflicting results. Some studies have highlighted the importance of hyperarousal symptoms (Stein et al., 2012) and others have evidenced the importance of avoidance (Hyland et al., 2016) or reexperiencing symptoms (Creamer et al., 1992). When examining the role of symptoms clusters, our results revealed that higher levels of reexperiencing, avoidance, and hyperarousal predicted membership in the treatment resistant class when compared with the moderate PTSD treatment response class. It is possible that individuals with higher levels of overall PTSD are less likely to improve over the course of treatment. Of note, no individual symptoms were predictive of class membership when both moderate classes were compared, suggesting that no individual symptom should take priority among individuals with moderate levels of PTSD. Interestingly, when both high classes were compared, treatment resistance was predicted by more severe reexperiencing and avoidance. This finding suggests that these symptoms may be a priority among individuals with higher levels of PTSD. Emotional processing theories suggest that avoidance is an attempt to reduce distressing symptoms, and it has been argued that this prevents effective processing of traumatic memory and therefore maintains PTSD (Foa, Steketee, & Rothbaum, 1989). Our findings are also consistent with Kleim, Ehlers, and Glucksman’s (2007) study which found that reexperiencing symptoms were an early predictor of later PTSD severity. Moreover, there is evidence suggesting that specifically targeting flashbacks in treatment lead to higher rates of PTSD improvement (Nijdam, Baas, Olff, & Gersons, 2013). Reexperiencing and arousal symptoms reinforce the learning that occurred during the traumatic event which can maintain PTSD (Shalev, 2002). It is possible that the individuals in the treatment resistant trajectory have PTSD which is maintained by the presence of high reexperiencing symptoms which continue to reinforce the learning that occurred during the trauma. However, due to the presence of avoidance symptoms, individuals are not able to effectively process these traumatic memories. Clinicians should be aware of these findings, and reexperiencing and avoidance symptoms should be specifically targeted among individuals with severe PTSD. This may potentially improve overall PTSD outcomes. Further research on the role of symptoms clusters should be conducted as this could have important theoretical and clinical implications for trauma survivors with PTSD.

In relation to abuse characteristics, the study found that experiencing another trauma in addition to CSA was not predictive of trajectory membership; this is contrary to other studies which have suggested that childhood cumulative trauma predicts more severe PTSD among CSA survivors (Steine et al., 2017). In the current study, the traumas included were not confined to childhood. It is possible that the results may have been different if childhood and adulthood traumas had been examined separately. Increased length of time since the abuse occurred was associated with membership in the moderate PTSD gradual response class when compared with all other classes. This contrasts with the results of another recent study which found that the time since the trauma occurred was not predictive of treatment outcomes (Murray, El-Leithy, & Billings, 2017). Given these inconsistencies, further research is warranted. Finally, being female was associated with membership in the moderate PTSD gradual responding class when compared with the high PTSD gradual responding class, and older age was associated with membership in the moderate PTSD rapid response class when compared with the moderate PTSD gradual responding class. These findings suggest that older age and being female may be associated with less severe PTSD and better treatment outcomes.

Overall, this study revealed four distinct trajectories, suggesting that not all CSA survivors respond to treatment in the same way. It has also highlighted that the examination of changes in group means is not sufficient for evaluating the effectiveness of treatment. The majority of participants in this sample did respond positively to treatment; however, 15.03% did not experience an improvement in PTSD symptoms. In addition, we have identified protective and risk factors associated with PTSD treatment response. Emotional and detached coping, reexperiencing and avoidance symptoms, and low social support have been found to increase risk of treatment resistance, and social support has been found to be associated with less severe symptoms which are more likely to improve over the course of treatment. These findings have important clinical implications. These factors could potentially identify participants who are at risk of not responding to treatment. Furthermore, the factors could be specifically addressed to increase positive outcomes for all participants.

The findings of the current study should be interpreted in the light of several limitations. First, in this study there was no control group used. Therefore, we cannot conclude that the changes in symptoms were a result of the psychotherapy and it could be suggested that the changes in symptoms were due to the passage of time. Although some of the participants had recent experiences of sexual abuse, the majority of the CSA experiences were historical (the mean length of time since the abuse was 22.2 years, with a range of 1-61 years). It is therefore reasonable to suggest that many of the participants may have had chronic PTSD prior to attending treatment and that at least some of the changes in symptomology were due to the treatment and not only due to the natural progression of PTSD over time. Future studies examining PTSD trajectories comparing multiple types of therapy or using a waitlist control group will be important to examine the specificity of the findings in the current study. Second, there have been recent changes to PTSD’s diagnostic criteria (DSM-5; APA, 2013). Future studies should utilize data reflecting these changes to determine if the results remain consistent. Second, attrition rates were high within this sample. Attrition is a common problem within longitudinal studies (Spratt et al., 2010), and it has been argued that biases associated with case wise deletion can be reduced by using methods such as multiple imputation for managing the missing data (Sterne et al., 2009). Finally, a number of individuals (including those presenting under the influence of drugs and alcohol or with current psychosis) were excluded from the treatment and referred for more appropriate treatment. This may have impacted the results and is likely to reduce the generalizability of the findings. Despite these limitations, this study was based on a relatively large treatment-seeking sample and the findings add to the literature by identifying differential treatment response in adult survivors of CSA. Furthermore, to the best of our knowledge, this was the first to examine the role of social support, coping style, and PTSD symptom clusters as predictors of heterogeneous trajectories of PTSD treatment response within the current sample.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Author Biographies

Shelley Fletcher, MRes, is a PhD student in the School of Psychology at Ulster University. Her research interests are in the area of childhood trauma and mental health outcomes in adulthood.

Ask Elklit, PhD, is a professor of clinical psychology and founder of the National Research Centre of Psychotraumatology at the Department of Psychology, University of Southern Denmark. His research interests focus broadly on the field of psychological trauma, crisis intervention, and crisis management in Denmark and internationally.

Mark Shevlin, PhD, is a professor of psychology at Ulster University and an Honorary Professor of psychological research methods at the University of Southern Denmark. His research interests are in the areas of trauma, psychosis, and posttraumatic stress disorder. He is also interested in the applications of latent variable models in mental health research. He has published widely and has also co-authored a popular student textbook on statistical analysis.

Cherie Armour, PhD, is associate dean of Research and Impact, Faculty of Life and Health Science, Ulster University. She has published in excess of 100 peer reviewed journal articles in the field of psychotraumatology and conducts research with many traumatized groups including survivors of interpersonal violence, survivors of childhood maltreatment, refugees, in service military personnel, military veterans, police, and emergency service personnel.

References

Akaike

(1987). Factor analysis and AIC. Psychometrika, 52(3), 317–332.

American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders (4th ed., text rev.). Washington, DC: Author.

American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Arlington, VA: Author.

Arata

C. M.

(2002). Child sexual abuse and sexual revictimization. Clinical Psychology: Science and Practice, 9, 135-164.

Asparouhov

Muthén

(2014). Auxiliary variables in mixture modeling: Three-step approaches using Mplus. Structural Equation Modeling: A Multidisciplinary Journal, 21, 329-341.

Bach

M. E.

(2003). En empirisk belysning og analyse af” emotional numbing” som eventuel selvstænding faktor i PTS. [An empirical illumination and analysis of emotional numbing as a potential independent factor in PTSD]. Psykologisk Studie Skriftserie, 6(1), 1–99.

Bollen

K. A.

(1989). Structural equations with latent variables. New York, NY: John Wiley.

Brewin

C. R.

Andrews

Valentine

J. D.

(2000). Meta-analysis of risk factors for posttraumatic stress disorder in trauma-exposed adults. Journal of Consulting and Clinical Psychology, 68, 748-766.

Briere

(2002). Treating adult survivors of severe childhood abuse and neglect: Further development of an integrative model. In Myers

J. E. B.

Berliner

Briere

Hendrix

C. T.

Reid

Jenny

(Eds.), The APSAC handbook on child maltreatment (2nd ed., pp. 175-202). Newbury Park, CA: SAGE.

10.

Campbell

S. B.

Renshaw

K. D.

(2013). PTSD symptoms, disclosure, and relationship distress: Explorations of mediation and associations over time. Journal of Anxiety Disorders, 27, 494-502.

11.

Chard

K. M.

(2005). An evaluation of cognitive processing therapy for the treatment of posttraumatic stress disorder related to childhood sexual abuse. Journal of Consulting and Clinical Psychology, 73, 965-971.

12.

Clark

D. A.

Beck

A. T.

(2010). Cognitive therapy of anxiety disorders. London, England: Guilford Press.

13.

Cloitre

Koenen

K. C.

Cohen

L. R.

Han

(2002). Skills training in affective and interpersonal regulation followed by exposure: A phase-based treatment for PTSD related to childhood abuse. Journal of Consulting and Clinical Psychology, 70, 1067-1074.

14.

Coffey

Leitenberg

Henning

Turner

Bennett

R. T.

(1996). Mediators of the long-term impact of child sexual abuse: Perceived stigma, betrayal, powerlessness, and self-blame. Child Abuse & Neglect, 20, 447-455.

15.

Cohen

Wills

T. A.

(1985). Stress, social support, and the buffering hypothesis. Psychological Bulletin, 98, 310-357.

16.

Cole

K. L.

Sarlund-Heinrich

Brown

L. S.

(2007). Developing and assessing effectiveness of a time-limited therapy group for incarcerated women survivors of childhood sexual abuse. Journal of Trauma & Dissociation, 8, 97-121.

17.

Creamer

Burgess

Pattison

(1992). Reaction to trauma: A cognitive processing model. Journal of Abnormal Psychology, 101, 452-459.

18.

Currier

J. M.

Holland

J. M.

Drescher

K. D.

(2014). Residential treatment for combat-related posttraumatic stress disorder: Identifying trajectories of change and predictors of treatment response. PLoS ONE, 9(7), e101741.

19.

Cutajar

M. C.

Mullen

P. E.

Ogloff

J. R.

Thomas

S. D.

Wells

D. L.

Spataro

(2010). Psychopathology in a large cohort of sexually abused children followed up to 43 years. Child Abuse & Neglect, 34, 813-822.

20.

Ehlers

Clark

D. M.

(2000). A cognitive model of posttraumatic stress disorder. Behaviour Research and Therapy, 38, 319-345.

21.

Elklit

(1996). Coping Styles Questionnaire: A contribution to the validation of a scale for measuring coping strategies. Personality and Individual Differences, 21, 809-812.

22.

Elklit

(2015). Treatment of Danish survivors of child sexual abuse—A cohort study. Behavioral Sciences, 5, 589-601.

23.

Elklit

Pedersen

S. S.

Jind

(2001). The Crisis Support Scale: Psychometric qualities and further validation. Personality and Individual Differences, 31, 1291-1302.

24.

Elliott

Biddle

Hawthorne

Forbes

Creamer

(2005). Patterns of treatment response in chronic posttraumatic stress disorder: An application of latent growth mixture modeling. Journal of Traumatic Stress, 18, 303-311.

25.

Fergusson

D. M.

McLeod

G. F.

Horwood

L. J.

(2013). Childhood sexual abuse and adult developmental outcomes: Findings from a 30-year longitudinal study in New Zealand. Child Abuse & Neglect, 37, 664-674.

26.

Filipas

H. H.

Ullman

S. E.

(2006). Child sexual abuse, coping responses, self-blame, posttraumatic stress disorder, and adult sexual revictimization. Journal of Interpersonal Violence, 21, 652-672.

27.

Foa

E. B.

Steketee

Rothbaum

B. O.

(1989). Behavioral/cognitive conceptualizations of post-traumatic stress disorder. Behavior Therapy, 20, 155-176.

28.

Hogan

B. E.

Linden

Najarian

(2002). Social support interventions: Do they work? Clinical Psychology Review, 22, 381-440.

29.

Holahan

C. J.

Moos

R. H.

Holahan

C. K.

Brennan

P. L.

(1995). Social support, coping, and depressive symptoms in a late-middle-aged sample of patients reporting cardiac illness. Health Psychology, 14, 152-163.

30.

Hyland

Shevlin

Hansen

Vallières

Murphy

Elklit

(2016). The temporal relations of PTSD symptoms among treatment-seeking victims of sexual assault: A longitudinal study. Journal of Loss and Trauma, 21, 492-506.

31.

Joseph

Andrews

Williams

Yule

(1992). Crisis support and psychiatric symptomatology in adult survivors of the Jupiter cruise ship disaster. British Journal of Clinical Psychology, 31, 63-73.

32.

Karstoft

K. I.

Armour

Elklit

Solomon

(2013). Long-term trajectories of posttraumatic stress disorder in veterans: The role of social resources. The Journal of Clinical Psychiatry, 74(12), e1163-e1168.

33.

Karstoft

K. I.

Armour

Elklit

Solomon

(2015). The role of locus of control and coping style in predicting longitudinal PTSD-trajectories after combat exposure. Journal of Anxiety Disorders, 32, 89-94.

34.

Kleim

Ehlers

Glucksman

(2007). Early predictors of chronic post-traumatic stress disorder in assault survivors. Psychological Medicine, 37, 1457-1467.

35.

Koenen

K. C.

Stellman

J. M.

Stellman

S. D.

Sommer

J. F.

Jr. (2003). Risk factors for course of posttraumatic stress disorder among Vietnam veterans: A 14-year follow-up of American Legionnaires. Journal of Consulting and Clinical Psychology, 71, 980-986.

36.

Lazarus

R. S.

Folkman

(1984). Stress, appraisal, and coping. New York, NY: Springer.

37.

Little

R. J.

(1988). A test of missing completely at random for multivariate data with missing values. Journal of the American Statistical Association, 83(404), 1198-1202.

38.

Millon

(1999). Personality-guided therapy. New York, NY: John Wiley.

39.

Mollica

R. F.

Caspi-Yavin

Bollini

Truong

Tor

Lavelle

(1992). The Harvard Trauma Questionnaire: Validating a cross-cultural instrument for measuring torture, trauma, and posttraumatic stress disorder in Indochinese refugees. The Journal of Nervous and Mental Disease, 180, 111-116.

40.

Murray

El-Leithy

Billings

(2017). Intensive cognitive therapy for post-traumatic stress disorder in routine clinical practice: A matched comparison audit. British Journal of Clinical Psychology, 56, 474-478.

41.

Muthén

(2001). Second-generation structural equation modeling with a combination of categorical and continuous latent variables: New opportunities for latent class–latent growth modeling. In Collins

Sayer

(Eds.), New methods for the analysis of change (pp. 291-322). Washington, DC: American Psychological Association.

42.

Muthén

L. K.

Muthén

B. O.

(2014). Mplus User’s Guide. Seventh Edition. Los Angeles, CA: Muthén & Muthén.

43.

Nagin

D. S.

(1999). Analyzing developmental trajectories: A semiparametric, group-based approach. Psychological Methods, 4, 139-157.

44.

Nijdam

M. J.

Baas

Olff

Gersons

B. P. R.

(2013). Hotspots in trauma memories and their relationship to successful trauma-focused psychotherapy: A pilot study. Journal of Traumatic Stress, 26, 38-44. doi:10.1002/jts.21771

45.

Nylund

K. L.

Asparouhov

Muthén

B. O.

(2007). Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo simulation study. Structural Equation Modelling, 14, 535-569.

46.

O’Connor

Elklit

(2008). Attachment styles, traumatic events, and PTSD: A cross-sectional investigation of adult attachment and trauma. Attachment & Human Development, 10, 59-71.

47.

Ozer

E. J.

Best

S. R.

Lipsey

T. L.

Weiss

D. S.

(2003). Predictors of posttraumatic stress disorder and symptoms in adults: A meta-analysis. Psychological Bulletin, 129, 52-73.

48.

Paolucci

E. O.

Genuis

M. L.

Violato

(2001). A meta-analysis of the published research on the effects of child sexual abuse. The Journal of Psychology, 135, 17-36.

49.

Roeder

Lynch

K. G.

Nagin

D. S.

(1999). Modeling uncertainty in latent class membership: A case study in criminology. Journal of the American Statistical Association, 94, 766-776.

50.

Roger

Jarvis

Najarian

(1993). Detachment and coping: The construction and validation of a new scale for measuring coping strategies. Personality and Individual Differences, 15, 619-626.

51.

Rubin

D. B.

(1987). The calculation of posterior distributions by data augmentation: Comment: A noniterative sampling/importance resampling alternative to the data augmentation algorithm for creating a few imputations when fractions of missing information are modest: The SIR algorithm. Journal of the American Statistical Association, 82, 543-546.

52.

Satorra

Bentler

P. M.

(1994). Corrections to test statistics and standard errors in covariance structure analysis. In von Eye

Clogg

C. C.

(Eds.), Latent variables analysis: Applications for developmental research (pp. 399-419). Thousand Oaks, CA: SAGE.

53.

Schell

T. L.

Marshall

G. N.

Jaycox

L. H.

(2004). All symptoms are not created equal: The prominent role of hyperarousal in the natural course of posttraumatic psychological distress. Journal of Abnormal Psychology, 113, 189-197.

54.

Schwarz

(1978). Estimating the dimension of a model. The Annals of Statistics, 6(2), 461–464.

55.

Sclove

S. L.

(1987). Application of model-selection criteria to some problems in multivariate analysis. Psychometrika, 52(3), 333–343.

56.

Shalev

A. Y.

(2002). Acute stress reactions in adults. Biological Psychiatry, 51, 532-543.

57.

Southwick

S. M.

Charney

D. S.

(2012). The science of resilience: Implications for the prevention and treatment of depression. Science, 338, 79-82.

58.

Spaccarelli

(1994). Stress, appraisal, and coping in child sexual abuse: A theoretical and empirical review. Psychological Bulletin, 116, 340-362.

59.

Spaccarelli

Kim

(1995). Resilience criteria and factors associated with resilience in sexually abused girls. Child Abuse & Neglect, 19, 1171-1182.

60.

Spratt

Carpenter

Sterne

J. A.

Carlin

J. B.

Heron

Henderson

Tilling

(2010). Strategies for multiple imputation in longitudinal studies. American Journal of Epidemiology, 172, 478-487.

61.

Stein

N. R.

Dickstein

B. D.

Schuster

Litz

B. T.

Resick

P. A.

(2012). Trajectories of response to treatment for posttraumatic stress disorder. Behavior Therapy, 43, 790-800.

62.

Steine

I. M.

Winje

Skogen

J. C.

Krystal

J. H.

Milde

A. M.

Bjorvatn

. . . Pallesen

(2017). Posttraumatic symptom profiles among adult survivors of childhood sexual abuse: A longitudinal study. Child Abuse & Neglect, 67, 280-293.

63.

Sterne

J. A.

White

I. R.

Carlin

J. B.

Spratt

Royston

Kenward

M. G.

. . . Carpenter

J. R.

(2009). Multiple imputation for missing data in epidemiological and clinical research: Potential and pitfalls. British Medical Journal, 338, Article 2393.

64.

Taylor

J. E.

Harvey

S. T.

(2010). A meta-analysis of the effects of psychotherapy with adults sexually abused in childhood. Clinical Psychology Review, 30, 749-767.

65.

Tremblay

Hébert

Piché

(1999). Coping strategies and social support as mediators of consequences in child sexual abuse victims. Child Abuse & Neglect, 23, 929-945.

66.

Ullman

S. E.

Filipas

H. H.

(2005). Gender differences in social reactions to abuse disclosures, post-abuse coping, and PTSD of child sexual abuse survivors. Child Abuse & Neglect, 29, 767-782.

67.

Vermunt

J. K.

(2010). Latent class modeling with covariates: Two improved three-step approaches. Political Analysis, 18, 450-469.

68.

Yuan

K. H.

Bentler

P. M.

(2000). Three likelihood-based methods for mean and covariance structure analysis with nonnormal missing data. Sociological Methodology, 30, 165-200.