Journal of Abnormal Psychology 2015, Vol. 124, No. 1, 155-171
© 2014 American Psychological Association 0021-843X715/$ 12.00 http://dx.doi.org/10.1037/abn0000020
Posttraumatic Stress in Deployed Marines: Prospective Trajectories of Early Adaptation William P. Nash
Alyssa M. Boasso
Boston VA Research Institute
VA Boston Healthcare System and Massachusetts Veterans Epidemiology Research and Information Center
Maria M. Steenkamp
Jonathan L. Larson and Rebecca E. Lubin
VA Boston Healthcare System, Massachusetts Veterans Epidemiology Research and Information Center, and Boston University School of Medicine
VA Boston Healthcare System and Massachusetts Veterans Epidemiology Research and Information Center
Brett T. Litz VA Boston Healthcare System, Massachusetts Veterans Epidemiology Research and Information Center, and Boston University School of Medicine We examined the course of PTSD symptoms in a cohort of U.S. Marines (N = 867) recmited for the Marine Resiliency Study (MRS) from a single infantry battalion that deployed as a unit for 7 months to Afghanistan during the peak of conflict there. Data were collected via structured interviews and self-report questionnaires 1 month prior to deployment and again at 1, 5, and 8 months postdeployment. Second-order growth mixture modeling was used to disaggregate symptom trajectories; multinomial logistic regression and relative weights analysis were used to assess the role of combat exposure, prior life span trauma, social support, peritraumatic dissociation, and avoidant coping as predictors of trajectory membership. Three trajectories best fit the data: a low-stable symptom course (79%), a new-onset PTSD symptoms course (13%), and a preexisting PTSD symptoms course (8%). Comparison in a separate MRS cohort with lower levels of combat exposure yielded similar results, except for the absence of a new-onset trajectory. In the main cohort, the modal trajectory was a low-stable symptoms course that included a small but clinically meaningful increase in symptoms from predeployment to 1 month postdeployment. We found no trajectory of recovery from more severe symptoms in either cohort, suggesting that the relative change in symptoms from predeployment to 1 month postdeployment might provide the best indicator of first-year course. The best predictors of trajectory membership were peritraumatic dissociation and avoidant coping, suggesting that changes in cognition, perception, and behavior following trauma might be particularly useful indicators of first-year outcomes. Keywords: military, trajectory, PTSD, dissociation, coping Supplemental materials: http://dx.doi.org/10.1037/abn0000020.supp
Posttraumatic stress disorder (PTSD) is a psychopathological condition for which the course of symptoms, as they evolve over time, is of particular theoretical and practical importance. It is one of a small group of mental disorders, termed trauma- and stressorrelated disorders in the fifth edition of the Diagnostic and the Statistical Manual o f Mental Disorders (DSM-5; American Psy
chiatric Association, 2013), whose onset is temporally linked to a specific triggering event. To be diagnosed, symptoms must persist beyond the first 30 days following trauma exposure, and one of the two subtypes of PTSD, delayed PTSD, is characterized by symp toms that first appear 6 months after the event. Moreover, the time course of PTSD symptoms is central to the study of adaptation to
This article was published Online First November 24, 2014. William P. Nash, Boston VA Research Institute; Alyssa M. Boasso, VA Boston Healthcare System and Massachusetts Veterans Epidemiology Re search and Information Center; Maria M. Steenkamp and Brett T. Litz, VA Boston Healthcare System, Massachusetts Veterans Epidemiology Re search and Information Center, and Boston University School of Medicine; lonathan L. Larson and Rebecca E. Lubin, VA Boston Healthcare System and Massachusetts Veterans Epidemiology Research and Information Cen ter.
This study was funded by VA Health Service Research and Develop ment (SDR 09-0128) and by the U. S. Marine Corps and Navy Bureau of Medicine and Surgery. The authors acknowledge the Marine Resiliency Study (MRS) team, General lohn M. Paxton Ir., USMC, and Debbie Paxton, RN, who made this work possible. We also thank Kevin Grimm, who provided statistical feedback and suggestions. Correspondence concerning this article should be addressed to Brett Litz, VA Boston Healthcare System, 150 South Huntington Avenue, 13B74, lamaica Plain, MA 02130. E-mail: [email protected]
155
156
NASH, BOASSO, STEENKAMP, LARSON, LUBIN, AND LITZ
trauma, and key concepts such as risk, resilience, and recovery have clear meanings only when defined in terms of changes in symptoms and functioning over time (Layne, Warren, Watson, & Shalev, 2007). Tracking the course of PTSD symptoms over time might identify the processes that determine long-term outcomes, discriminating between normative and pathological responses to experiences of trauma and recognizing the points at which early interventions might positively influence outcomes. The theoretical and practical implications of the time course of PTSD symptoms might be especially salient for the military, which bears primary responsibility for preventing negative psychological health outcomes such as PTSD in service members whose occu pations place them at high risk for exposure to potentially trau matic events. For military prevention and early intervention pro grams, one key question can only be answered in terms of the time course of posttraumatic stress symptoms: among service members exposed to potentially traumatic events, at which point in their symptom courses can those at greatest risk for developing chronic PTSD be identified so that their further exposure to potentially traumatic events can be limited and early interventions to promote recovery can be provided? Early cross-sectional and retrospective research on PTSD in both military and nonmilitary populations divided groups of peo ple who had been exposed to potentially traumatic events into those who exceeded diagnostic thresholds for the disorder and those who did not, ignoring potential heterogeneity within those two categories. More recent longitudinal studies in many different populations suggest that posttraumatic stress symptoms might trace a number of distinct trajectories over time, which can be grouped into clusters of approximately similar patterns (Bonanno, 2004). The research question addressed by these studies is whether between-individual differences in adaptation to exposure to poten tially traumatizing events generate groups of people following similar within-individual courses. A number of statistical tools can be used to answer this question, including growth models, growth mixture models, and latent profile analysis; each procedure has strengths and weaknesses. PTSD researchers have chosen to use growth mixture modeling (GMM), a data-driven statistical proce dure that groups participants on the basis of their intra-individual change patterns (e.g., McArdle & Epstein, 1987; Muthen, 2004). In this article, we report the results of GMM analyses of posttrau matic stress trajectories in two cohorts of U.S. Marines enrolled in the Marine Resiliency Study (MRS; see Baker et al., 2012), as sessed approximately 1 month prior to their 7-month deployments to Afghanistan, then reexamined at three time points during the 8 months immediately following their return from deployment. We chose GMM chiefly to maximize the comparability of the results with prior studies of deployed service members, all of which employed GMM. To date, four longitudinal studies have used GMM to identify latent PTSD trajectories in military populations: U.S. Army sol diers followed for 9 months after they deployed to Kosovo on a 6-month NATO-led peacekeeping mission in 2002 (Dickstein et al., 2010); U.S. Army soldiers assessed within 5 days of their return from the Gulf War in 1991 and again at 1.5 and 6 years later (Orcutt, Erickson, & Wolfe, 2004); Danish soldiers followed for 7 months after their return from a 6-month deployment to Afghan istan in 2009 (Bemtsen et al., 2012); and a large and heterogeneous cohort of military service members who deployed one or more
times to Iraq or Afghanistan between 2001 and 2008 (Bonanno et al., 2012). All of these studies found evidence for qualitatively distinct trajectories of PTSD symptoms, the exact nature of which depended on the sample, methodology, and context. Trajectories characterized by low symptom levels at all time points were modal in all four studies, comprising between 57% and 84% of each sample. Chronic symptoms or new-onset symptoms were uncom mon, each representing less than 10% of the sample, when present. Currently, there are no studies of the course of PTSD symptoms in a sample of U.S. service members from a single military unit deployed together for combat duties in Iraq or Afghanistan. Bo nanno et al.’s (2012) sample was more heterogeneous, comprising members of all service branches and all occupational fields, many of whom did not serve in combat roles. No existing studies have reported PTSD symptom trajectories in a cohort of U.S. service members deployed to a war zone specifically to engage in ground combat, a population at high risk for combat-related PTSD and, therefore, of great interest to leaders of military PTSD prevention, screening, and treatment programs. In this study, we chose five sets of self-reported predictor variables relevant to the military that have been repeatedly found to correlate with posttraumatic stress outcomes: combat-related stressor exposures experienced during deployment, lifetime stres sor exposures experienced outside the index deployment, per ceived social support during and after deployment, peritraumatic dissociation, and avoidant coping. Combat exposure has consis tently been shown to be a leading risk factor for PTSD in military personnel, typically in a dose-response fashion (e.g., Dohrenwend et al., 2006; Foy, Sipprelle, Rueger, & Carroll, 1984; Green, Grace, Lindy, Gleser, & Leonard, 1990; King, King, Foy, Keane, & Fairbank, 1999). As predicted by diathesis-stress models of PTSD (e.g., McKeever & Huff, 2003), greater prior life span trauma exposure has been found to confer heightened risk for combat-related PTSD, and high rates of predeployment trauma are present in military personnel (e.g., Clancy et al., 2006; Vogt, Pless, King, & King, 2005; Vogt et al., 2011; Zaidi & Foy, 1994). Because we are interested in the relationship between these two variables, we examined the influence of the interaction between prior trauma and combat exposure on PTSD symptom course. Consistent with the diathesis-stress model, we expected prior life span trauma to moderate the relationship between combat expo sure and new-onset or chronic PTSD symptom trajectories. We expected social support during and after deployment to buffer adverse psychological outcomes because it mitigates distress and promotes shared meaning making (e.g., Brailey, Vasterling, Proc tor, Constans, & Friedman, 2007), and veterans with PTSD con sistently report lower unit support and postdeployment social support (e.g., Keane, Scott, Chavoya, Lamparski, & Fairbank, 1985; Pietrzak et al., 2010). Peritraumatic dissociation, which entails transient alterations in the normal integration of cognitive, emotional, somatic, and be havioral processes during or immediately after a potentially trau matic event, was included as a predictor variable because it is a marker for stressors experienced by a given person at a given point in time that exceeded their current adaptive capacity, as predicted by the stress injury model of PTSD (Nash, 2007; Nash, Silva, & Litz, 2009; Nash et al., 2010). The stress injury model does not predict which variables confer risk or are protective per se, but focuses on the relationship between moment-to-moment stress
POSTTRAUMATIC STRESS IN DEPLOYED MARINES
levels and ideographic stress breaking points determined by fluc tuating biological, psychological, and social functional capacities. According to this model, stress outcomes that follow superthresh old stressor exposures, including subclinical stress injuries and mental disorders such as PTSD, are more likely to be pathological than outcomes that follow less extreme stressor experiences, which are more likely to be normative. Previous studies have found peritraumatic dissociation to be associated with more adverse postevent outcomes, and trauma-exposed people with PTSD are more likely to report having experienced peritraumatic dissocia tion than did those without PTSD (e.g., Bremner & Brett, 1997; Marmar et al., 1994; O’Toole, Marshall, Schureck, & Dobson, 1999; Tichenor, Marmar, Weiss, Metzler, & Ronfeldt, 1996). Avoidant coping, characterized as utilizing distraction, denial, or disengagement as mechanisms to manage problems, is hypothe sized to increase risk for PTSD in two ways: habitual avoidant coping leading up to exposure to a traumatic stressor might con tribute to vulnerability for PTSD, and, in the aftermath of trauma, overgeneralized avoidant coping and self-soothing repertoires might lessen the likelihood of corrective recovery-promoting ex periences. Avoidant coping has repeatedly been found to correlate with PTSD symptom severity (e.g., Pietrzak, Harpaz-Rotem, & Southwick, 2011; Sharkansky et al., 2000; Solomon, Mikulincer, & Benbenishty, 1989). Service members with PTSD have also been shown to be more likely to use avoidant rather than non avoidant coping strategies (e.g., Sutker, Davis, Uddo, & Ditta, 1995), whereas decreased use of avoidant coping over time has been associated with recovery from combat stress (Solomon, Mi kulincer, & Avitzur, 1988). On the basis of previous longitudinal studies of PTSD symptom courses, we predicted that six trajectories of PTSD symptom severity would best describe the data: (1) a quadratic recovery course characterized by low predeployment symptoms, followed by high initial postdeployment symptoms, and then a marked decrease in symptoms toward baseline; (2) a relatively flat low stable course with low symptom levels across all time points; (3) a new- onset course characterized by high and relatively unremit ting symptoms across all postdeployment time points that follow a low level of PTSD symptoms prior to deployment; (4) a preexisting-improving course characterized by high levels of PTSD prior to deployment followed by a decrease in symptom levels postdeployment; (5) a preexisting-chronic course charac terized by high levels of PTSD prior to deployment that do not decrease during the postdeployment period; and (6) a delayed course characterized by low symptom levels before and immedi ately after deployment, but an increase in PTSD-symptom burden during the 8 months following return from deployment. Given the high frequency of significant warzone stressors expected in Af ghanistan and the high levels of resilience expected in highly trained Marines, we predicted that the recovery trajectory would be most prevalent. We expected the next most prevalent courses to be low-stable and new-onset courses. Given prior warzone deploy ments and other predeployment stressor exposures in our cohort, we expected the two preexisting PTSD courses to also occur in significant percentages of participants. We predicted that combat-related experiences during the index deployment would best predict membership in the new-onset, delayed, and preexisting-chronic trajectories and that prior life span stressor exposures would best predict membership in the
157
preexisting-improving and preexisting-chronic trajectories. We expected prior life span trauma and combat experiences during the index deployment to interact to increase vulnerability for the new-onset, chronic, and delayed PTSD courses. That is, we pre dicted that Marines with extensive trauma histories would be affected by relatively lower doses of combat exposure, tracing worse PTSD outcomes over time than Marines with similar levels of combat exposure but no prior trauma. We expected peritrau matic dissociation and avoidant coping to confer risk for all persistently negative outcomes, including the new-onset, delayed, and preexisting-chronic trajectories. Conversely, we expected per ceived social support to serve a protective function, with Marines in the recovery, low-stable, and preexisting-improving trajectories reporting greater perceived social support than Marines in the new-onset, delayed, and preexisting-chronic trajectories. Assess ment of the relative importance of these predictors, using relative weights analysis, was largely exploratory.
Method Design and Participants The data source for this study was the MRS, a longitudinal field study of four consecutive all-male cohorts (named Cohorts 1, 2, 3, and 4) of active-duty ground-combat Marines, each recruited pri marily from a single infantry battalion scheduled to deploy to Iraq or Afghanistan between 2008 and 2011 from either Marine Corps Base Camp Pendleton or Marine Corps Air Ground Combat Cen ter, Twenty-Nine Palms, both in California (Baker et al., 2012). Four assessment time points were planned for each cohort: 1 month prior to its 7-month deployment, and 1 week, 3 months, and 6 months postdeployment. Overall, 2,593 Marines completed the Time-0 (TO) predeployment assessment; 2,317 (89.3%) completed the Time-1 (Tl) assessment; 1,901 (73.3%) completed the Time-2 (T2) assessment; and 1,634 (63.0%) completed the Time-3 (T3) assessment. Participation at each assessment was voluntary and individual informed consent was obtained before enrollment at baseline with no senior unit leaders present. For this study, full analyses focused exclusively on Cohort 4, whereas Cohort 3 was used for a post hoc comparison of latent trajectory patterns. Cohorts 1 and 2 were excluded because their PTSD scores at baseline were indexed exclusively to military events, whereas their postdeployment PTSD scores were indexed to any currently distressing lifetime events; this threat to internal validity made an examination of PTSD symptom trajectories prob lematic across time in Cohorts 1 and 2. In Cohorts 3 and 4, PTSD symptoms were indexed at all time points to any currently dis tressing lifetime event. Cohorts 3 and 4 were analyzed separately to avoid two other internal validity problems. The first of these arose because modal postdeployment assessment times differed by as much as 3 months between these cohorts. The second reason we analyzed Cohorts 3 and 4 separately was because these cohorts predominantly comprised members of two distinct Marine infantry battalions that trained, deployed, and then returned as units from two very different sets of warzone challenges. Combining them into a single larger sample might introduce a number of uncon trolled between-unit variances. Cohort 4 deployed to Helmand Province in Afghanistan in late 2010, when U.S. forces sustained their highest causality rates. Cohort 3, having deployed earlier than
158
NASH, BOASSO, STEENKAMP, LARSON, LUBIN, AND LITZ
Outcome Measures
Herman, Huska, & Keane, 1993). The CAPS was used at TO, T2, and T3, but was not used at T l to minimize participant burden in the early weeks postdeployment. The PCL was used at all time points. At every time point, CAPS and PCL assessments were indexed to any lifetime traumatic event endorsed by the participant as currently most distressing. Consequently, index events were allowed to change during the course of the study, ensuring the capture of maximum symptom burden at each time point. The CAPS assesses the frequency and intensity of PTSD symp toms, each rated on a Likert-type scale, ranging from 0 (“Never” or “None”) to 4 (“Daily or almost daily” or “Extreme, incapaci tating distress, cannot dismiss memories, unable to continue ac tivities”). Total CAPS PTSD symptom severity was calculated by summing the frequency and intensity scales for each item (yielding a range of 0 to 136; Blake et al., 1995). Raters were systematically trained and certified doctoral-level personnel. All CAPS inter views were audio recorded and 15% were randomly selected and co-rated to determine interrater reliability (intraclass correlation coefficient [ICC] = .99; Shrout & Fleiss, 1979). The PCL assesses the severity of PTSD symptoms on a 1 (not at all bothersome) to 5 (extremely bothersome) Likert-type scale. Total PTSD symptom burden was calculated by summing across all 17 symptoms (yield ing a range of 17 to 85). The CAPS and the PCL have been shown to have excellent psychometric properties in numerous studies with varied populations (see Weathers et al., 2001). Convergent outcome indicators. To substantiate the class solutions generated by the GMM, we compared membership in the PTSD trajectories we found with four classes of outcomes we believed would covary with PTSD trajectory: full or sub threshold PTSD caseness based on diagnostic criteria, depres sion, anxiety, and overall functioning. Using the CAPS at TO, T2, and T3, we defined a PTSD diagnosis as meeting the minimum type and number of symptoms required by DSM-IV criteria (American Psychiatric Association, 2000), each rated at least at a frequency of 1 and a severity of 2 (Weathers et al., 1999). We defined subthreshold PTSD conservatively, by re quiring a participant to meet the DSM-IV criteria for Category B symptoms and either Category C or D symptoms (e.g., Blanchard et al., 1995). Because the CAPS was not adminis tered at T l, the PCL was used to determine full and subthresh old PTSD caseness at the initial postdeployment time point; a required symptom was considered present if it was endorsed on the PCL at a severity level of moderately (a value of 3) or above. We assessed depression using the Beck Depression Inventory-II (BDI-II; Beck, Steer, & Brown, 1996), a 21-item questionnaire that assesses symptoms of depression. The inter nal consistency of the BDI-II in our study was uniformly high (TO: a = .90; T l: a = .89; T2: a = .91; T3: a = .90). We assessed anxiety using the Beck Anxiety Inventory (BAI; Beck, Epstein, Brown, & Steer, 1988), which was also found to have high levels of internal consistency in the MRS (TO: a = .90; T l : a = .92; T2: a = .92; T3: a = .94). Summary scores for both scales at each time point were created by summing across all 21 items. We assessed overall functioning and self-reported dis-
PTSD. PTSD symptom severity was assessed using both a structured clinical interview, the Clinician Administered PTSD Scale (CAPS; Blake et al., 1995), and a self-report questionnaire, the Posttraumatic Stress Disorder Checklist (PCL; Weathers, Litz,
1 A one-way ANOVA and a post hoc Tukey’s test revealed that Cohort 4 had significantly higher combat exposure than had all other cohorts, F(3, 2205) = 371.87, p < .001.
did Cohort 4, before the heaviest fighting began, reported signif icantly lower combat exposure than Cohort 4, t(1,926) = 14.27, p < .001. For this study, Cohort 4 (N = 892) offered the best opportunity to examine the course of PTSD in highly combatexposed U.S. service members. We used Cohort 3 (N = 673) to compare the GMM results of Cohort 4 with a sample of similarly assessed Marines with less overall combat exposure. For this study, we removed Marines who did not deploy (9 in Cohort 3 and 4 in Cohort 4) and those who died during deployment (2 in Cohort 3 and 17 in Cohort 4). To address variability around the modal postdeployment assessment times, which differed be tween Cohorts 3 and 4, we used the following procedures, outlined by King et al. (2006). For the three postdeployment assessments, scores on all measures were assigned to three follow-up date ranges determined by the count of days since the date of return from deployment. We aimed to minimize the dispersion of days within each date range and to maximize the number of included participants. This was done separately for Cohorts 3 and 4. The date ranges across Cohorts 3 and 4 differed only with respect to the second postdeployment assessment (T2). The ranges of days that best fit the data were 20 to 40 days for T1 (Cohort 3: M = 30, SD = 6; Cohort 4: M = 30, SD = 4), 80 to 100 days for T2 for Cohort 3 (M = 84, SD = 3), but 140 to 160 days for T2 for Cohort 4 (M = 153, SD = 4), and 240 to 260 days for T3 (Cohort 3: M = 251, SD = 2; Cohort 4: M = 249, SD = 5). In other words, on average, assessments for Cohort 4 occurred 1 month predeployment (TO) and 1-month (Tl), 5-months (T2), and 8-months (T3) postde ployment. For Cohort 3, on average, assessments occurred 1 month predeployment (TO) and 1-month (Tl), 2-months (T2), and 8-months (T3) postdeployment. Once the data were redistributed according to actual date ranges, 4 Marines in Cohort 4 and 4 Marines in Cohort 3 were missing data at all time points and were excluded from analyses. For Cohort 4,1 the final subsample of responders consisted of 867 Marines: 859 at baseline, 554 at T l, 328 at T2, and 287 at T3. For Cohort 3, the final subsample of responders consisted of 658 Marines: 653 at baseline, 377 at T l, 382 at T2, and 215 at T3. Table 1 displays statistical comparisons at baseline (TO) be tween Marines whose PTSD symptom severity data were available at each subsequent time point (responders) and Marines missing PTSD outcome data at those time points (nonresponders). Nonre sponders at T l, T2, and T3 were more likely to have previously deployed. In addition, nonresponders at T2 were more functionally impaired and had more prior lifetime trauma. Nonresponders at T3 were more educated, and nonresponders at T2 and T3 were older. In the final Cohort 4 sample, Marines were primarily Caucasian (83.1%). At baseline, participants had served an average of 3.10 (SD = 3.15) years in the military and 51.54% had deployed at least once before. Participants’ ages at baseline ranged from 18 to 43 (M = 23.16, SD = 3.67); 68.1% had no more than a high school diploma, and 41.1% were married (see Table 2). Sample information for Cohort 3 can be found in Tables 1 and 2 of the supplemental materials.
159
POSTTRAUMATIC STRESS IN DEPLOYED MARINES
vo q oo O nO \n ^ovooNO vr'• h —- ^ >n oo ^ f'l nv. r - o— in • ■“ ' ^ O O ooonoO '-'OO'*''ov d
d
o
o- d
I V o o d o o
I
hoonOMnoiOcviooOhvooovoo in n n in n n ^ n , vo n n n,
OOOOOOOOOOOOIOOO—HOOOOOOOOOOOO
X X X X X
C/5 a> □ jg C QC/3 § a 0 c 04 H
# Ov o O' CO VO 04 q ov o 1 CO o II II II II
* * * O' 04 00 01 q CO o o -H co 1 II II II
# o 04 04 1 II
00 o o II
* CO O' O' 04 00 q d q 01 CO 1 o o d 1 d II II II II II II
r - 00 CO ON 01 OV vO 04 o o O' vO o OV CO i/ i U1 U1 >/i 01 Oi "1- 01 vO VO 01 Oi Oi q 00 00 00 00 00 00 00 OO 00 OO 00 00 00 00 00 _' - *• - - Z' (N^ rq n
© 2014 American Psychological Association 0021-843X715/$ 12.00 http://dx.doi.org/10.1037/abn0000020
Posttraumatic Stress in Deployed Marines: Prospective Trajectories of Early Adaptation William P. Nash
Alyssa M. Boasso
Boston VA Research Institute
VA Boston Healthcare System and Massachusetts Veterans Epidemiology Research and Information Center
Maria M. Steenkamp
Jonathan L. Larson and Rebecca E. Lubin
VA Boston Healthcare System, Massachusetts Veterans Epidemiology Research and Information Center, and Boston University School of Medicine
VA Boston Healthcare System and Massachusetts Veterans Epidemiology Research and Information Center
Brett T. Litz VA Boston Healthcare System, Massachusetts Veterans Epidemiology Research and Information Center, and Boston University School of Medicine We examined the course of PTSD symptoms in a cohort of U.S. Marines (N = 867) recmited for the Marine Resiliency Study (MRS) from a single infantry battalion that deployed as a unit for 7 months to Afghanistan during the peak of conflict there. Data were collected via structured interviews and self-report questionnaires 1 month prior to deployment and again at 1, 5, and 8 months postdeployment. Second-order growth mixture modeling was used to disaggregate symptom trajectories; multinomial logistic regression and relative weights analysis were used to assess the role of combat exposure, prior life span trauma, social support, peritraumatic dissociation, and avoidant coping as predictors of trajectory membership. Three trajectories best fit the data: a low-stable symptom course (79%), a new-onset PTSD symptoms course (13%), and a preexisting PTSD symptoms course (8%). Comparison in a separate MRS cohort with lower levels of combat exposure yielded similar results, except for the absence of a new-onset trajectory. In the main cohort, the modal trajectory was a low-stable symptoms course that included a small but clinically meaningful increase in symptoms from predeployment to 1 month postdeployment. We found no trajectory of recovery from more severe symptoms in either cohort, suggesting that the relative change in symptoms from predeployment to 1 month postdeployment might provide the best indicator of first-year course. The best predictors of trajectory membership were peritraumatic dissociation and avoidant coping, suggesting that changes in cognition, perception, and behavior following trauma might be particularly useful indicators of first-year outcomes. Keywords: military, trajectory, PTSD, dissociation, coping Supplemental materials: http://dx.doi.org/10.1037/abn0000020.supp
Posttraumatic stress disorder (PTSD) is a psychopathological condition for which the course of symptoms, as they evolve over time, is of particular theoretical and practical importance. It is one of a small group of mental disorders, termed trauma- and stressorrelated disorders in the fifth edition of the Diagnostic and the Statistical Manual o f Mental Disorders (DSM-5; American Psy
chiatric Association, 2013), whose onset is temporally linked to a specific triggering event. To be diagnosed, symptoms must persist beyond the first 30 days following trauma exposure, and one of the two subtypes of PTSD, delayed PTSD, is characterized by symp toms that first appear 6 months after the event. Moreover, the time course of PTSD symptoms is central to the study of adaptation to
This article was published Online First November 24, 2014. William P. Nash, Boston VA Research Institute; Alyssa M. Boasso, VA Boston Healthcare System and Massachusetts Veterans Epidemiology Re search and Information Center; Maria M. Steenkamp and Brett T. Litz, VA Boston Healthcare System, Massachusetts Veterans Epidemiology Re search and Information Center, and Boston University School of Medicine; lonathan L. Larson and Rebecca E. Lubin, VA Boston Healthcare System and Massachusetts Veterans Epidemiology Research and Information Cen ter.
This study was funded by VA Health Service Research and Develop ment (SDR 09-0128) and by the U. S. Marine Corps and Navy Bureau of Medicine and Surgery. The authors acknowledge the Marine Resiliency Study (MRS) team, General lohn M. Paxton Ir., USMC, and Debbie Paxton, RN, who made this work possible. We also thank Kevin Grimm, who provided statistical feedback and suggestions. Correspondence concerning this article should be addressed to Brett Litz, VA Boston Healthcare System, 150 South Huntington Avenue, 13B74, lamaica Plain, MA 02130. E-mail: [email protected]
155
156
NASH, BOASSO, STEENKAMP, LARSON, LUBIN, AND LITZ
trauma, and key concepts such as risk, resilience, and recovery have clear meanings only when defined in terms of changes in symptoms and functioning over time (Layne, Warren, Watson, & Shalev, 2007). Tracking the course of PTSD symptoms over time might identify the processes that determine long-term outcomes, discriminating between normative and pathological responses to experiences of trauma and recognizing the points at which early interventions might positively influence outcomes. The theoretical and practical implications of the time course of PTSD symptoms might be especially salient for the military, which bears primary responsibility for preventing negative psychological health outcomes such as PTSD in service members whose occu pations place them at high risk for exposure to potentially trau matic events. For military prevention and early intervention pro grams, one key question can only be answered in terms of the time course of posttraumatic stress symptoms: among service members exposed to potentially traumatic events, at which point in their symptom courses can those at greatest risk for developing chronic PTSD be identified so that their further exposure to potentially traumatic events can be limited and early interventions to promote recovery can be provided? Early cross-sectional and retrospective research on PTSD in both military and nonmilitary populations divided groups of peo ple who had been exposed to potentially traumatic events into those who exceeded diagnostic thresholds for the disorder and those who did not, ignoring potential heterogeneity within those two categories. More recent longitudinal studies in many different populations suggest that posttraumatic stress symptoms might trace a number of distinct trajectories over time, which can be grouped into clusters of approximately similar patterns (Bonanno, 2004). The research question addressed by these studies is whether between-individual differences in adaptation to exposure to poten tially traumatizing events generate groups of people following similar within-individual courses. A number of statistical tools can be used to answer this question, including growth models, growth mixture models, and latent profile analysis; each procedure has strengths and weaknesses. PTSD researchers have chosen to use growth mixture modeling (GMM), a data-driven statistical proce dure that groups participants on the basis of their intra-individual change patterns (e.g., McArdle & Epstein, 1987; Muthen, 2004). In this article, we report the results of GMM analyses of posttrau matic stress trajectories in two cohorts of U.S. Marines enrolled in the Marine Resiliency Study (MRS; see Baker et al., 2012), as sessed approximately 1 month prior to their 7-month deployments to Afghanistan, then reexamined at three time points during the 8 months immediately following their return from deployment. We chose GMM chiefly to maximize the comparability of the results with prior studies of deployed service members, all of which employed GMM. To date, four longitudinal studies have used GMM to identify latent PTSD trajectories in military populations: U.S. Army sol diers followed for 9 months after they deployed to Kosovo on a 6-month NATO-led peacekeeping mission in 2002 (Dickstein et al., 2010); U.S. Army soldiers assessed within 5 days of their return from the Gulf War in 1991 and again at 1.5 and 6 years later (Orcutt, Erickson, & Wolfe, 2004); Danish soldiers followed for 7 months after their return from a 6-month deployment to Afghan istan in 2009 (Bemtsen et al., 2012); and a large and heterogeneous cohort of military service members who deployed one or more
times to Iraq or Afghanistan between 2001 and 2008 (Bonanno et al., 2012). All of these studies found evidence for qualitatively distinct trajectories of PTSD symptoms, the exact nature of which depended on the sample, methodology, and context. Trajectories characterized by low symptom levels at all time points were modal in all four studies, comprising between 57% and 84% of each sample. Chronic symptoms or new-onset symptoms were uncom mon, each representing less than 10% of the sample, when present. Currently, there are no studies of the course of PTSD symptoms in a sample of U.S. service members from a single military unit deployed together for combat duties in Iraq or Afghanistan. Bo nanno et al.’s (2012) sample was more heterogeneous, comprising members of all service branches and all occupational fields, many of whom did not serve in combat roles. No existing studies have reported PTSD symptom trajectories in a cohort of U.S. service members deployed to a war zone specifically to engage in ground combat, a population at high risk for combat-related PTSD and, therefore, of great interest to leaders of military PTSD prevention, screening, and treatment programs. In this study, we chose five sets of self-reported predictor variables relevant to the military that have been repeatedly found to correlate with posttraumatic stress outcomes: combat-related stressor exposures experienced during deployment, lifetime stres sor exposures experienced outside the index deployment, per ceived social support during and after deployment, peritraumatic dissociation, and avoidant coping. Combat exposure has consis tently been shown to be a leading risk factor for PTSD in military personnel, typically in a dose-response fashion (e.g., Dohrenwend et al., 2006; Foy, Sipprelle, Rueger, & Carroll, 1984; Green, Grace, Lindy, Gleser, & Leonard, 1990; King, King, Foy, Keane, & Fairbank, 1999). As predicted by diathesis-stress models of PTSD (e.g., McKeever & Huff, 2003), greater prior life span trauma exposure has been found to confer heightened risk for combat-related PTSD, and high rates of predeployment trauma are present in military personnel (e.g., Clancy et al., 2006; Vogt, Pless, King, & King, 2005; Vogt et al., 2011; Zaidi & Foy, 1994). Because we are interested in the relationship between these two variables, we examined the influence of the interaction between prior trauma and combat exposure on PTSD symptom course. Consistent with the diathesis-stress model, we expected prior life span trauma to moderate the relationship between combat expo sure and new-onset or chronic PTSD symptom trajectories. We expected social support during and after deployment to buffer adverse psychological outcomes because it mitigates distress and promotes shared meaning making (e.g., Brailey, Vasterling, Proc tor, Constans, & Friedman, 2007), and veterans with PTSD con sistently report lower unit support and postdeployment social support (e.g., Keane, Scott, Chavoya, Lamparski, & Fairbank, 1985; Pietrzak et al., 2010). Peritraumatic dissociation, which entails transient alterations in the normal integration of cognitive, emotional, somatic, and be havioral processes during or immediately after a potentially trau matic event, was included as a predictor variable because it is a marker for stressors experienced by a given person at a given point in time that exceeded their current adaptive capacity, as predicted by the stress injury model of PTSD (Nash, 2007; Nash, Silva, & Litz, 2009; Nash et al., 2010). The stress injury model does not predict which variables confer risk or are protective per se, but focuses on the relationship between moment-to-moment stress
POSTTRAUMATIC STRESS IN DEPLOYED MARINES
levels and ideographic stress breaking points determined by fluc tuating biological, psychological, and social functional capacities. According to this model, stress outcomes that follow superthresh old stressor exposures, including subclinical stress injuries and mental disorders such as PTSD, are more likely to be pathological than outcomes that follow less extreme stressor experiences, which are more likely to be normative. Previous studies have found peritraumatic dissociation to be associated with more adverse postevent outcomes, and trauma-exposed people with PTSD are more likely to report having experienced peritraumatic dissocia tion than did those without PTSD (e.g., Bremner & Brett, 1997; Marmar et al., 1994; O’Toole, Marshall, Schureck, & Dobson, 1999; Tichenor, Marmar, Weiss, Metzler, & Ronfeldt, 1996). Avoidant coping, characterized as utilizing distraction, denial, or disengagement as mechanisms to manage problems, is hypothe sized to increase risk for PTSD in two ways: habitual avoidant coping leading up to exposure to a traumatic stressor might con tribute to vulnerability for PTSD, and, in the aftermath of trauma, overgeneralized avoidant coping and self-soothing repertoires might lessen the likelihood of corrective recovery-promoting ex periences. Avoidant coping has repeatedly been found to correlate with PTSD symptom severity (e.g., Pietrzak, Harpaz-Rotem, & Southwick, 2011; Sharkansky et al., 2000; Solomon, Mikulincer, & Benbenishty, 1989). Service members with PTSD have also been shown to be more likely to use avoidant rather than non avoidant coping strategies (e.g., Sutker, Davis, Uddo, & Ditta, 1995), whereas decreased use of avoidant coping over time has been associated with recovery from combat stress (Solomon, Mi kulincer, & Avitzur, 1988). On the basis of previous longitudinal studies of PTSD symptom courses, we predicted that six trajectories of PTSD symptom severity would best describe the data: (1) a quadratic recovery course characterized by low predeployment symptoms, followed by high initial postdeployment symptoms, and then a marked decrease in symptoms toward baseline; (2) a relatively flat low stable course with low symptom levels across all time points; (3) a new- onset course characterized by high and relatively unremit ting symptoms across all postdeployment time points that follow a low level of PTSD symptoms prior to deployment; (4) a preexisting-improving course characterized by high levels of PTSD prior to deployment followed by a decrease in symptom levels postdeployment; (5) a preexisting-chronic course charac terized by high levels of PTSD prior to deployment that do not decrease during the postdeployment period; and (6) a delayed course characterized by low symptom levels before and immedi ately after deployment, but an increase in PTSD-symptom burden during the 8 months following return from deployment. Given the high frequency of significant warzone stressors expected in Af ghanistan and the high levels of resilience expected in highly trained Marines, we predicted that the recovery trajectory would be most prevalent. We expected the next most prevalent courses to be low-stable and new-onset courses. Given prior warzone deploy ments and other predeployment stressor exposures in our cohort, we expected the two preexisting PTSD courses to also occur in significant percentages of participants. We predicted that combat-related experiences during the index deployment would best predict membership in the new-onset, delayed, and preexisting-chronic trajectories and that prior life span stressor exposures would best predict membership in the
157
preexisting-improving and preexisting-chronic trajectories. We expected prior life span trauma and combat experiences during the index deployment to interact to increase vulnerability for the new-onset, chronic, and delayed PTSD courses. That is, we pre dicted that Marines with extensive trauma histories would be affected by relatively lower doses of combat exposure, tracing worse PTSD outcomes over time than Marines with similar levels of combat exposure but no prior trauma. We expected peritrau matic dissociation and avoidant coping to confer risk for all persistently negative outcomes, including the new-onset, delayed, and preexisting-chronic trajectories. Conversely, we expected per ceived social support to serve a protective function, with Marines in the recovery, low-stable, and preexisting-improving trajectories reporting greater perceived social support than Marines in the new-onset, delayed, and preexisting-chronic trajectories. Assess ment of the relative importance of these predictors, using relative weights analysis, was largely exploratory.
Method Design and Participants The data source for this study was the MRS, a longitudinal field study of four consecutive all-male cohorts (named Cohorts 1, 2, 3, and 4) of active-duty ground-combat Marines, each recruited pri marily from a single infantry battalion scheduled to deploy to Iraq or Afghanistan between 2008 and 2011 from either Marine Corps Base Camp Pendleton or Marine Corps Air Ground Combat Cen ter, Twenty-Nine Palms, both in California (Baker et al., 2012). Four assessment time points were planned for each cohort: 1 month prior to its 7-month deployment, and 1 week, 3 months, and 6 months postdeployment. Overall, 2,593 Marines completed the Time-0 (TO) predeployment assessment; 2,317 (89.3%) completed the Time-1 (Tl) assessment; 1,901 (73.3%) completed the Time-2 (T2) assessment; and 1,634 (63.0%) completed the Time-3 (T3) assessment. Participation at each assessment was voluntary and individual informed consent was obtained before enrollment at baseline with no senior unit leaders present. For this study, full analyses focused exclusively on Cohort 4, whereas Cohort 3 was used for a post hoc comparison of latent trajectory patterns. Cohorts 1 and 2 were excluded because their PTSD scores at baseline were indexed exclusively to military events, whereas their postdeployment PTSD scores were indexed to any currently distressing lifetime events; this threat to internal validity made an examination of PTSD symptom trajectories prob lematic across time in Cohorts 1 and 2. In Cohorts 3 and 4, PTSD symptoms were indexed at all time points to any currently dis tressing lifetime event. Cohorts 3 and 4 were analyzed separately to avoid two other internal validity problems. The first of these arose because modal postdeployment assessment times differed by as much as 3 months between these cohorts. The second reason we analyzed Cohorts 3 and 4 separately was because these cohorts predominantly comprised members of two distinct Marine infantry battalions that trained, deployed, and then returned as units from two very different sets of warzone challenges. Combining them into a single larger sample might introduce a number of uncon trolled between-unit variances. Cohort 4 deployed to Helmand Province in Afghanistan in late 2010, when U.S. forces sustained their highest causality rates. Cohort 3, having deployed earlier than
158
NASH, BOASSO, STEENKAMP, LARSON, LUBIN, AND LITZ
Outcome Measures
Herman, Huska, & Keane, 1993). The CAPS was used at TO, T2, and T3, but was not used at T l to minimize participant burden in the early weeks postdeployment. The PCL was used at all time points. At every time point, CAPS and PCL assessments were indexed to any lifetime traumatic event endorsed by the participant as currently most distressing. Consequently, index events were allowed to change during the course of the study, ensuring the capture of maximum symptom burden at each time point. The CAPS assesses the frequency and intensity of PTSD symp toms, each rated on a Likert-type scale, ranging from 0 (“Never” or “None”) to 4 (“Daily or almost daily” or “Extreme, incapaci tating distress, cannot dismiss memories, unable to continue ac tivities”). Total CAPS PTSD symptom severity was calculated by summing the frequency and intensity scales for each item (yielding a range of 0 to 136; Blake et al., 1995). Raters were systematically trained and certified doctoral-level personnel. All CAPS inter views were audio recorded and 15% were randomly selected and co-rated to determine interrater reliability (intraclass correlation coefficient [ICC] = .99; Shrout & Fleiss, 1979). The PCL assesses the severity of PTSD symptoms on a 1 (not at all bothersome) to 5 (extremely bothersome) Likert-type scale. Total PTSD symptom burden was calculated by summing across all 17 symptoms (yield ing a range of 17 to 85). The CAPS and the PCL have been shown to have excellent psychometric properties in numerous studies with varied populations (see Weathers et al., 2001). Convergent outcome indicators. To substantiate the class solutions generated by the GMM, we compared membership in the PTSD trajectories we found with four classes of outcomes we believed would covary with PTSD trajectory: full or sub threshold PTSD caseness based on diagnostic criteria, depres sion, anxiety, and overall functioning. Using the CAPS at TO, T2, and T3, we defined a PTSD diagnosis as meeting the minimum type and number of symptoms required by DSM-IV criteria (American Psychiatric Association, 2000), each rated at least at a frequency of 1 and a severity of 2 (Weathers et al., 1999). We defined subthreshold PTSD conservatively, by re quiring a participant to meet the DSM-IV criteria for Category B symptoms and either Category C or D symptoms (e.g., Blanchard et al., 1995). Because the CAPS was not adminis tered at T l, the PCL was used to determine full and subthresh old PTSD caseness at the initial postdeployment time point; a required symptom was considered present if it was endorsed on the PCL at a severity level of moderately (a value of 3) or above. We assessed depression using the Beck Depression Inventory-II (BDI-II; Beck, Steer, & Brown, 1996), a 21-item questionnaire that assesses symptoms of depression. The inter nal consistency of the BDI-II in our study was uniformly high (TO: a = .90; T l: a = .89; T2: a = .91; T3: a = .90). We assessed anxiety using the Beck Anxiety Inventory (BAI; Beck, Epstein, Brown, & Steer, 1988), which was also found to have high levels of internal consistency in the MRS (TO: a = .90; T l : a = .92; T2: a = .92; T3: a = .94). Summary scores for both scales at each time point were created by summing across all 21 items. We assessed overall functioning and self-reported dis-
PTSD. PTSD symptom severity was assessed using both a structured clinical interview, the Clinician Administered PTSD Scale (CAPS; Blake et al., 1995), and a self-report questionnaire, the Posttraumatic Stress Disorder Checklist (PCL; Weathers, Litz,
1 A one-way ANOVA and a post hoc Tukey’s test revealed that Cohort 4 had significantly higher combat exposure than had all other cohorts, F(3, 2205) = 371.87, p < .001.
did Cohort 4, before the heaviest fighting began, reported signif icantly lower combat exposure than Cohort 4, t(1,926) = 14.27, p < .001. For this study, Cohort 4 (N = 892) offered the best opportunity to examine the course of PTSD in highly combatexposed U.S. service members. We used Cohort 3 (N = 673) to compare the GMM results of Cohort 4 with a sample of similarly assessed Marines with less overall combat exposure. For this study, we removed Marines who did not deploy (9 in Cohort 3 and 4 in Cohort 4) and those who died during deployment (2 in Cohort 3 and 17 in Cohort 4). To address variability around the modal postdeployment assessment times, which differed be tween Cohorts 3 and 4, we used the following procedures, outlined by King et al. (2006). For the three postdeployment assessments, scores on all measures were assigned to three follow-up date ranges determined by the count of days since the date of return from deployment. We aimed to minimize the dispersion of days within each date range and to maximize the number of included participants. This was done separately for Cohorts 3 and 4. The date ranges across Cohorts 3 and 4 differed only with respect to the second postdeployment assessment (T2). The ranges of days that best fit the data were 20 to 40 days for T1 (Cohort 3: M = 30, SD = 6; Cohort 4: M = 30, SD = 4), 80 to 100 days for T2 for Cohort 3 (M = 84, SD = 3), but 140 to 160 days for T2 for Cohort 4 (M = 153, SD = 4), and 240 to 260 days for T3 (Cohort 3: M = 251, SD = 2; Cohort 4: M = 249, SD = 5). In other words, on average, assessments for Cohort 4 occurred 1 month predeployment (TO) and 1-month (Tl), 5-months (T2), and 8-months (T3) postde ployment. For Cohort 3, on average, assessments occurred 1 month predeployment (TO) and 1-month (Tl), 2-months (T2), and 8-months (T3) postdeployment. Once the data were redistributed according to actual date ranges, 4 Marines in Cohort 4 and 4 Marines in Cohort 3 were missing data at all time points and were excluded from analyses. For Cohort 4,1 the final subsample of responders consisted of 867 Marines: 859 at baseline, 554 at T l, 328 at T2, and 287 at T3. For Cohort 3, the final subsample of responders consisted of 658 Marines: 653 at baseline, 377 at T l, 382 at T2, and 215 at T3. Table 1 displays statistical comparisons at baseline (TO) be tween Marines whose PTSD symptom severity data were available at each subsequent time point (responders) and Marines missing PTSD outcome data at those time points (nonresponders). Nonre sponders at T l, T2, and T3 were more likely to have previously deployed. In addition, nonresponders at T2 were more functionally impaired and had more prior lifetime trauma. Nonresponders at T3 were more educated, and nonresponders at T2 and T3 were older. In the final Cohort 4 sample, Marines were primarily Caucasian (83.1%). At baseline, participants had served an average of 3.10 (SD = 3.15) years in the military and 51.54% had deployed at least once before. Participants’ ages at baseline ranged from 18 to 43 (M = 23.16, SD = 3.67); 68.1% had no more than a high school diploma, and 41.1% were married (see Table 2). Sample information for Cohort 3 can be found in Tables 1 and 2 of the supplemental materials.
159
POSTTRAUMATIC STRESS IN DEPLOYED MARINES
vo q oo O nO \n ^ovooNO vr'• h —- ^ >n oo ^ f'l nv. r - o— in • ■“ ' ^ O O ooonoO '-'OO'*''ov d
d
o
o- d
I V o o d o o
I
hoonOMnoiOcviooOhvooovoo in n n in n n ^ n , vo n n n,
OOOOOOOOOOOOIOOO—HOOOOOOOOOOOO
X X X X X
C/5 a> □ jg C QC/3 § a 0 c 04 H
# Ov o O' CO VO 04 q ov o 1 CO o II II II II
* * * O' 04 00 01 q CO o o -H co 1 II II II
# o 04 04 1 II
00 o o II
* CO O' O' 04 00 q d q 01 CO 1 o o d 1 d II II II II II II
r - 00 CO ON 01 OV vO 04 o o O' vO o OV CO i/ i U1 U1 >/i 01 Oi "1- 01 vO VO 01 Oi Oi q 00 00 00 00 00 00 00 OO 00 OO 00 00 00 00 00 _' - *• - - Z' (N^ rq n
