Introduction
Post-traumatic stress disorder (PTSD) is associated with substantial medical morbidity (Schnurr et al. Reference Schnurr, Spiro and Paris2000; Ahmadi et al. Reference Ahmadi, Hajsadeghi, Mirshkarlo, Budoff, Yehuda and Ebrahimi2011; Bartoli et al. Reference Bartoli, Carrà, Crocamo, Caretta and Clerici2013; O'Donovan et al. Reference O'Donovan, Cohen, Seal, Bertenthal, Margaretten, Nishimi and Neylan2015), with striking effects observed for obesity (Bartoli et al. Reference Bartoli, Crocamo, Alamia, Amidani, Paggi, Pini, Clerici and Carrà2015), and cardiometabolic and cardiovascular conditions (Ahmadi et al. Reference Ahmadi, Hajsadeghi, Mirshkarlo, Budoff, Yehuda and Ebrahimi2011; Heppner et al. Reference Heppner, Lohr, Kash, Jin, Wang and Baker2012; Bartoli et al. Reference Bartoli, Carrà, Crocamo, Caretta and Clerici2013; Wentworth et al., Reference Wentworth, Stein, Redwine, Xue, Taub, Clopton, Nayak and Maisel2013; Roberts et al. Reference Roberts, Agnew-Blais, Spiegelman, Kubzansky, Mason, Galea, Hu, Rich-Edwards and Koenen2015; Rosenbaum et al. Reference Rosenbaum, Stubbs, Ward, Steel, Lederman and Vancampfort2015b ; Roy et al. Reference Roy, Foraker, Girton and Mansfield2015; Sumner et al. Reference Sumner, Kubzansky, Elkind, Roberts, Agnew-Blais, Chen, Cerdá, Rexrode, Rich-Edwards, Spiegelman, Suglia, Rimm and Koenen2015). The co-occurrence of PTSD with metabolic syndrome (MetS), as defined by three or more of central obesity, hypertension, dyslipidemia, and elevated blood sugars [National Cholesterol Education Program (NCEP), 2001; Grundy et al. Reference Grundy, Cleeman, Daniels, Donato, Eckel, Franklin, Gordon, Krauss, Savage, Smith, Spertus and Costa2005], is particularly high, with recent meta-analyses suggesting that MetS is prevalent in nearly 40% of those with PTSD (Bartoli et al. Reference Bartoli, Carrà, Crocamo, Caretta and Clerici2013; Rosenbaum et al. Reference Rosenbaum, Stubbs, Ward, Steel, Lederman and Vancampfort2015b ). The association between PTSD and MetS is intriguing given that stress is implicated in the pathogenesis and course of MetS (Vitaliano et al. Reference Vitaliano, Scanlan, Zhang, Savage, Hirsch and Siegler2002; see also Epel, Reference Epel2009) and that MetS may be part of the pathway linking PTSD to subsequent deleterious health conditions, such as cardiovascular disease (Roy et al. Reference Roy, Foraker, Girton and Mansfield2015; Sumner et al. Reference Sumner, Kubzansky, Elkind, Roberts, Agnew-Blais, Chen, Cerdá, Rexrode, Rich-Edwards, Spiegelman, Suglia, Rimm and Koenen2015), type 2 diabetes (Roberts et al. Reference Roberts, Agnew-Blais, Spiegelman, Kubzansky, Mason, Galea, Hu, Rich-Edwards and Koenen2015), decreased cortical thickness (Wolf et al. Reference Wolf, Sadeh, Leritz, Logue, Stoop, McGlinchey, Milberg and Millerin press), cognitive impairment (Green et al. Reference Green, Fairchild, Kinoshita, Noda and Yesavage2016), and premature mortality (Ahmadi et al. Reference Ahmadi, Hajsadeghi, Mirshkarlo, Budoff, Yehuda and Ebrahimi2011).
Although nearly all of the studies linking PTSD to MetS employed a cross-sectional design, many investigators hypothesize that the stress of PTSD influences MetS risk (e.g. Bartoli et al. Reference Bartoli, Carrà, Crocamo, Caretta and Clerici2013). This could occur through biological pathways, such as increased autonomic reactivity, immune and hypothalamic-pituitary-adrenal (HPA) axis system dysregulation (Kibler et al. Reference Kibler, Tursich, Ma, Malcolm and Greenbarg2014; Levine et al. Reference Levine, Levine and Levine2014), and/or oxidative stress processes (Grattagliano et al. Reference Grattagliano, Palmieri, Portincasa, Moschetta and Palasciano2008). Potential PTSD-related behavioral pathways include poor nutrition and sedentary lifestyle (Hall et al. Reference Hall, Hoerster and Yancy2015), cigarette and alcohol use (Dennis et al. Reference Dennis, Watkins, Calhoun, Oddone, Sherwood, Dennis, Rissling and Beckham2014), and poor sleep (Talbot et al. Reference Talbot, Rao, Cohen, Richards, Inslicht, O'Donovan, Maguen, Metzler and Neylan2015). It is also possible that MetS may negatively affect PTSD symptoms. For example, greater pre-deployment inflammation (C-reactive protein), which often co-occurs with MetS, was recently shown to predict subsequent post-deployment PTSD (Eraly et al. Reference Eraly, Nievergelt, Maihofer, Barkauskas, Biswas, Agorastos, O'Connor and Baker2014). PTSD and MetS may also exert bidirectional effects on the severity of each other, particularly in trauma-exposed samples. In support of this, bidirectional effects have been reported for depression and MetS (Pulkki-Råback et al. Reference Pulkki-Råback, Elovainio, Kivimäki, Mattsson, Raitakari, Puttonen, Marniemi, Viikari and Keltikangas-Järvinen2009), and this may generalize to PTSD given that PTSD is highly comorbid with depression (Pietrzak et al. Reference Pietrzak, Goldstein, Southwick and Grant2012) and both disorders may arise out of a shared underlying vulnerability towards internalizing psychopathology (Miller et al. Reference Miller, Fogler, Wolf, Kaloupek and Keane2008). Only longitudinal designs can address the question of directionality. To our knowledge, two such studies exist to date, and both had analytic concerns that limited the strength of the causal conclusions.
Specifically, Francis et al. (Reference Francis, Nikulina and Widom2015) followed 78 physically abused children and 349 non-abused children into middle age and found that childhood abuse was associated with PTSD symptoms during young adulthood, which, in turn, predicted obesity in middle age. Baseline obesity was not controlled for analytically, making it difficult to draw conclusions about the direction of this association. Farr et al. (Reference Farr, Ko, Joung, Zaichenko, Usher, Tsoukas, Thakkar, Davis, Crowell and Mantzoros2015) also suggested that PTSD was associated with increasing metabolic risk by showing that, among 55 urban-area community adults, greater PTSD severity was associated with increased obesity and systolic blood pressure over the course of 2.5 years, controlling for baseline body mass index (BMI). Unfortunately, results were difficult to interpret because of the small sample size, control for only baseline BMI, and the fact that PTSD symptoms were split into quartiles based on sample distribution (i.e. not evaluated per the DSM diagnostic definition or total severity). Neither study tested whether MetS predicted subsequent PTSD.
In light of these concerns, our aim was to evaluate potential bidirectional influences between PTSD and MetS using a cross-lagged panel model (Rosenthal & Rosnow, Reference Rosenthal and Rosnow1991), which simultaneously evaluates the longitudinal effect of each variable on the other while controlling for baseline levels of both PTSD and MetS. We hypothesized that PTSD would be associated with increasing MetS risk over time, and that if there was evidence for MetS influencing subsequent PTSD, that this effect would be weaker in magnitude than that for PTSD predicting MetS. This aim was evaluated in a large national cohort of US military veterans deployed to the wars in Iraq and/or Afghanistan and who completed two waves of assessments, separated by approximately 2.5 years (see Rosen et al. Reference Rosen, Marx, Maserejian, Holowka, Gates, Sleeper, Vasterling, Kang and Keane2012). As women were over-sampled and represented just over 50% of the cohort, we were also able to evaluate potential sex differences in the relationship between PTSD and MetS.
Method
Participants
Participants were U.S. Army or Marine Corps veterans enrolled between 2009 and 2012 in the baseline assessment of Project VALOR (Veterans’ After-Discharge Longitudinal Registry), a registry of VA mental healthcare users with and without PTSD who deployed in support of Operation Enduring Freedom or Operation Iraqi Freedom (see Supplementary material and Rosen et al. Reference Rosen, Marx, Maserejian, Holowka, Gates, Sleeper, Vasterling, Kang and Keane2012 for details). To be included veterans must have undergone a mental health evaluation at a VA facility. Veterans with probable PTSD according to VA medical records (i.e. at least two instances of a PTSD diagnosis by a mental health professional associated with two separate visits) were oversampled at a 3:1 ratio, and female veterans were oversampled to comprise ~50% of the cohort.
The current study included the largest possible subsample of n = 1355 participants from Project VALOR (out of 1649 total) who had data pertaining to, at least, time 1 (T1) PTSD severity and T1 MetS severity. Demographic characteristics of this sample are shown in Table 1. Time 2 (T2) PTSD severity was available for n = 1124 (83%) of the T1 sample and T2 MetS severity data were available for n = 971 (72%) of the T1 participants, yielding the final T2 total of 971 (see Supplementary material for comparisons of those with v. without T2 data).
Demographic and PTSD-related characteristics of the sample
T1, Time 1; T2, time 2; PTSD, post-traumatic stress disorder; Dx, diagnosis; s.d., standard deviation.
Demographic characteristics are based on T1 data (n = 1355). The sample size at T2 was 971 and T2 PTSD percentages are based on that total. Comparisons of demographic and other differences between those with and without T2 data are presented in the Supplementary material. PTSD severity is based on a symptom count of the number of endorsed items on the SCID PTSD module.
Measures
PTSD module of the Structured Clinical Interview for DSM
Doctoral-level diagnosticians assessed current (past month) PTSD via telephone using the PTSD module of the Structured Clinical Interview for DSM (SCID). The SCID for DSM-IV (First et al. Reference First, Spitzer, Williams and Gibbon2000) was administered at T1 and for DSM-5 (First et al. Reference First, Williams, Karg and Spitzer2015) at T2. Both have demonstrated excellent psychometric properties (Bovin & Weathers, Reference Bovin, Weathers, Beck and Sloan2012; Regier et al. Reference Regier, Narrow, Clarke, Kraemer, Kuramoto, Kuhl and Kupfer2013). The SCID was administered up to two times at each time point in relation to two, distinct index traumatic experiences. PTSD symptom severity was operationalized as the maximum score (number of PTSD symptoms endorsed) from either of the two SCID administrations at each time point. PTSD diagnosis was operationalized as meeting the DSM-IV (at T1)/DSM-5 (at T2) diagnostic criteria based on either of the two concurrent SCID administrations. Interviews were digitally recorded and 100 were randomly chosen for secondary independent ratings at T1 and T2 yielding excellent inter-rater agreement at T1 (κ = 0.91) and T2 (κ = 0.82).
Life events checklist for DSM-IV (LEC)
The LEC (Gray et al. Reference Gray, Litz, Hsu and Lombardo2004) is a self-report questionnaire of trauma exposure that comprises the PTSD Criterion A1 assessment on the Clinician Administered PTSD Scale (Blake et al. Reference Blake, Weathers, Nagy, Kaloupek, Gusman, Charney and Keane1995). Participants indicated if they experienced, witnessed, learned about, or were exposed to any of 16 potentially traumatic events.
Additional measures that were the focus of secondary analyses are described in the Supplementary material.
Procedure
At T1, participants provided informed consent verbally over the telephone in accordance with the research protocol approved by all institutional review boards and the Human Research Protection Office of the US Army Medical Research and Materiel Command. Study staff then invited participants to complete a self-administered survey either online or via mail. Once completed, participants underwent diagnostic interview by telephone and received $50 compensation. Approximately 2–4 years later, participants were re-contacted for the second phase of the study, which followed the same approach as T1. Participants were compensated $100 at T2.
Data pertaining to MetS features were extracted from the VA electronic medical record using laboratory values that were linked as closely as possible in time to the SCID-based PTSD assessment and were no more than ±6 months of the PTSD assessment. On average, there was well less than a month between the PTSD/biometric assessments (see Supplementary materials). The two PTSD assessments occurred, on average, ~2.5 years apart (range 18.80–56.50 months; see Supplementary material). Time difference variables were evaluated as covariates in preliminary analyses.
MetS was defined per the NCEP Adult Treatment Panel (ATP) III definition (NCEP, 2001; Grundy et al. Reference Grundy, Cleeman, Daniels, Donato, Eckel, Franklin, Gordon, Krauss, Savage, Smith, Spertus and Costa2006), as detailed in Table 2 Footnote 1 Footnote †. MetS severity was defined as the number of MetS criteria present (0–5). MetS diagnosis was defined as meeting three or more of the MetS criteria (NCEP, 2001; Grundy et al. Reference Grundy, Cleeman, Daniels, Donato, Eckel, Franklin, Gordon, Krauss, Savage, Smith, Spertus and Costa2006).
Metabolic syndrome criteria definitions
BMI, Body mass index; HDL, high-density lipoprotein.
Three out of five criteria (central obesity, low HDL, high triglycerides elevated glucose, and elevated systolic or diastolic blood pressure) were required for the metabolic syndrome diagnosis.
a See note 1.
Data analysis
We first examined the prevalenceFootnote 2 of PTSD diagnosis (on the SCID) and MetS diagnosis (and each MetS criterion) at each time point in the sample overall and then conducted χ2 analyses to evaluate MetS-related differences as a function of PTSD diagnosis at each time point. We also compared a population-based estimate of MetS among 20- to 39-year-olds (20.3%; Ervin, Reference Ervin2009) with the T1 MetS prevalence among veterans with PTSD in this same age group using a Z test for two population proportions. We then examined each dimensional MetS variable (raw laboratory values and criteria count) as a function of PTSD diagnosis using t tests for independent samples. We tested potential differences in metabolic profiles as a function of PTSD diagnosis and sex (and their interaction) using multivariate analysis of variance (MANOVA). Correlations between total lifetime trauma exposure, PTSD severity, and MetS severity at and across each time point and those between MetS, PTSD severity, and potential covariates were evaluated (see Supplementary material).
We then ran our primary cross-lagged panel model using the statistical modeling program Mplus 7.11 (Muthén & Muthén, Reference Muthén and Muthén2012). In the path model, the autoregressive effects of each variable on itself over time (e.g. T1 PTSD to T2 PTSD) were modeled, as were the cross-lagged paths (e.g. T1 PTSD to T2 MetS). These models focused on PTSD severity (symptom count on the SCID) and MetS severity (number of MetS criteria met). The concurrent correlation between the two variables at T1 was modeled as was their residual correlation at T2. Total lifetime trauma exposure (on the LEC) was included as a predictor of T1 MetS and PTSD severity and the indirect effects of trauma exposure on T2 PTSD and MetS severity via T1 PTSD and MetS severity were estimated using the ‘model indirect’ command. Significant covariates, based on the results of initial bivariate correlations, were included as predictors of T1 PTSD and MetS severity. The model employed the robust maximum likelihood estimator, which accounts for non-normality in the data by adjusting the standard errors to reduce the likelihood of Type I error. This estimator includes all available data using full information likelihood estimation, conditional on the presence of at least one exogenous variable. Due to missing covariate data, the final sample size for the cross-lagged model was 1341. Path models were evaluated using standard fit indices and guidelines (Hu & Bentler, Reference Hu and Bentler1999).
We then conducted a logistic regression in SPSS v. 21 (IBM Corp., USA) to test whether T1 PTSD severity predicted T2 MetS diagnosis, controlling for T1 MetS severity and demographic covariates. Analyses evaluating potential moderators, covariates, and confounders (including combat exposure, depression, substance use, and psychotropic medication use) of our main associations are detailed in the Supplementary material.
Ethical standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
Results
Prevalence and severity of PTSD, MetS, and their co-occurrence
Descriptive statistics pertaining to the presence and severity of current PTSD at T1 and T2 are listed in Table 1 and descriptive statistics for MetS are shown in Table 3. Among veterans with PTSD aged <40 years at T1, the prevalence of MetS was 29.0%, which was significantly greater than the 20.3% previously reported (Ervin, Reference Ervin2009) in an age-matched epidemiological sample (Z = −3.19, p = 0.001). In contrast, the prevalence of MetS among veterans without current PTSD in this age group (20.2%) was nearly identical to that reported by Ervin (Reference Ervin2009). In the full sample, the mean number of T1 MetS criteria was 2.00 (Table 3), with 89.4% meeting at least one MetS criterion and 70.0% meeting at least 2 MetS criteria. At T2, the mean number of MetS criteria was 2.10 (Table 3), with 90.5% above the threshold for at least one MetS criterion and 63.6% above the threshold for at least two criteria.Footnote 3 At T1, 17.4% were taking cholesterol-lowering medication, 27% anti-hypertensive medication, and 2.4% were taking diabetes-related medication. At T2, 15.1% were taking cholesterol-lowering medication, 23.6% were taking anti-hypertensive, and 3.2% were taking diabetes-related medications. This medication use was factored into the MetS definition (Table 2).
Metabolic syndrome diagnosis and features in the overall sample and as a function of PTSD diagnosis
MetS, Metabolic syndrome; s.d., standard deviation; PTSD+, positive post-traumatic stress disorder diagnosis; PTSD−, negative post-traumatic stress disorder diagnosis; T1, time 1; T2, time 2; BMI, body mass index; HDL, high-density lipoprotein; Dx, diagnosis. p values for dimensional variables are based on independent t tests as a function of current (T1 or T2) PTSD diagnosis. p values for categorical variables are based on Pearson χ2 tests.
The number of participants for each analysis varied due to missing data. Details are as follows: At time 1: (a) BMI total = 1303, PTSD+ = 864, PTSD− = 439; (b) diastolic BP total = 1325, PTSD+ = 881, PTSD− = 444; (c) systolic BP total = 1326, PTSD+ = 882, PTSD− = 444; (d) HDL cholesterol total = 900, PTSD+ = 594, PTSD− = 306; (e) triglycerides total = 888, PTSD+ = 589, PTSD− = 299; (f) glucose total = 1013, PTSD+ = 686, PTSD− = 327; (g) total no. of MetS features total = 1355, PTSD+ = 741, PTSD− = 407; (h) Mets Dx total = 1355, PTSD+ = 741, PTSD− = 407. Sample sizes at time 2: (a) total = 927, PTSD+ = 635, PTSD− = 292; (b) total = 944, PTSD+ = 649, PTSD− = 295; (c) total = 946, PTSD + 649, PTSD− = 297; (d) total = 669, PTSD+ = 461, PTSD− = 208; (e) total = 658, PTSD+ = 454, PTSD− = 204; (f) total = 760, PTSD+ = 532, PTSD− = 228; (g) total = 971, PTSD+ = 664, PTSD− = 307; (h) total = 971, PTSD+ = 664, PTSD− = 307. Using the higher BMI cut-point of 30 in the diagnostic algorithm yielded a MetS prevalence of 25.7% at T1 and 28.5% at T2 (of those with T2 data). A greater percentage of individuals with PTSD were diagnosed with MetS using this more stringent BMI criterion at both T1 and T2, per χ2 analysis. Using the higher BMI cut-point of 30 in the MetS criteria count revealed that 79% were above the threshold for at least 1 MetS criterion and 49% were above the threshold for at least 2 MetS criteria at T1. At T2, 80% were above the threshold on at least 1 MetS criterion and 52% were above the threshold on at least 2 MetS criteria. At both time points, t tests revealed that individuals with PTSD met the threshold for a greater number of MetS criteria than did those without PTSD.
As shown in Table 3, χ2 analyses revealed that the prevalence of T1 MetS diagnosis was greater among those with a concurrent PTSD diagnosis (36.6%) compared to those without (26.3%, p < 0.001). This held at T2, wherein the prevalence of T2 MetS diagnosis was 37.8% among those with a concurrent PTSD diagnosis and 30.9% among those without (p = 0.038). Individuals with PTSD at T1 also met criteria for a greater number of T1 MetS features compared to those without T1 PTSD (Table 3). χ2 analyses suggested that a greater percentage of individuals with T1 PTSD met the criteria for central obesity, hypertension, elevated blood sugars, and high triglycerides than those without PTSD (Table 3). Additionally, t tests revealed higher mean raw laboratory values for each T1 MetS component among this group (Table 3). Those with PTSD at T2 also met criteria for a greater number of T2 MetS features compared with those without T2 PTSD (Table 3); however, no group differences emerged in the mean T2 raw metabolic values and the only T2 criterion difference was for hypertension (Table 3).
MANOVAs examined sex, PTSD, and sex × PTSD differences in raw metabolic laboratory values at each time point. At T1, the multivariate test yielded main effects for sex (Pillai's trace = 0.163, F 6,800 = 25.96, p < 0.001) and PTSD (Pillai's trace = 0.030, F 6,800 = 4.14, p < 0.001), but no significant interaction between the two (Pillai's trace = 0.013, F 6,800 = 1.78, p = 0.10). All sex differences were in the direction of women evidencing less pathological laboratory values than the men (details available from first author). The main effect of sex held at T2 (Pillai's trace = 0.158, F 6,619 = 19.35, p < 0.001), but there were no significant multivariate main effects of T2 PTSD or of PTSD × sex. Based on this, sex was not included as a moderator in primary models, though it was included as a covariate and evaluated further in secondary analyses (see Supplementary material).
Cross-lagged panel models
Preliminary correlation-based analyses are detailed in the Supplementary material and Supplementary Table S1. We found that none of the time difference variables were correlated with their respective dependent variables, so they were excluded from path models. In contrast, race, sex, age, and education were associated with some or all of the PTSD and MetS variables (detailed in the Supplementary material) and were therefore included as covariates of T1 PTSD and MetS.Footnote 4
The cross-lagged panel model fit the data well: χ2(10, n = 1341) = 59.10, p < 0.001, root mean square error of approximation = 0.06, standardized root mean square residual = 0.02, confirmatory fit index = 0.97, Tucker–Lewis index = 0.91. As shown in Fig. 1, T1 PTSD severity was a strong predictor of T2 PTSD severity (β = 0.67, p < 0.001), and T1 MetS severity was strongly related to T2 MetS severity (β = 0.62, p < 0.001). After controlling for these autoregressive effects, we found a significant cross-lagged effect, such that T1 PTSD severity predicted T2 MetS severity (β = 0.08, p = 0.002), but T1 MetS did not predict T2 PTSD severity (β = 0.005, p = 0.82). The association between PTSD and MetS severity at T1 was significant; however, their residual correlation was not significant at T2 after controlling for the shared effects of T1 variables. Age, sex, and education were significant covariates of T1 MetS severity; age, race, and education were significant covariates of T1 PTSD severity (Fig. 1). Total lifetime trauma exposure at T1 was significantly associated with T1 PTSD severity but not with T1 MetS. Results suggested indirect effects of trauma on T2 MetS severity via T1 PTSD severity (β = 0.03, p = 0.002) and on T2 PTSD severity via T1 PTSD severity (β = 0.21, p < 0.001). In total, the model explained 40% of the variance in T2 MetS severity and 44% of the variance in T2 PTSD severity. Analyses of potential confounds of key associations are reported in the Supplementary material; none altered the primary pattern of results.
The figure shows the results of the cross-lagged panel model. Primary and significant paths of interest are bolded. Correlations are represented via double-headed arrows and regressive paths via single-headed arrows. MetS, Metabolic syndrome; PTSD, post-traumatic stress disorder; Edu, educational attainment. *p < 0.05, **p < 0.01, ***p < 0.001.
Effects of PTSD on subsequent MetS diagnosis
Finally, to evaluate the effects of PTSD severity on subsequent MetS diagnosis, we conducted a logistic regression with T1 PTSD severity and T1 MetS severity as predictors of T2 MetS diagnosis, controlling for age, race, sex, and education. We found that for every additional PTSD symptom at T1, the odds ratio (OR) for MetS diagnosis at T2 increased by 5.6% [95% confidence interval (CI) 1.9–9.4%, Wald χ2(1, n = 961) = 9.09, p = 0.003]. This means that for every 10 symptoms endorsed on the SCID at T1, the odds of a MetS diagnosis at T2 increased by 56%. Each increase in MetS criteria at T1 was associated with nearly three times the odds for a subsequent MetS diagnosis [OR 2.86, 95% CI 2.44–3.37, Wald χ2(1, n = 961) = 124.31, p < 0.001]. Nagelkerke's R 2 for the overall model was 0.40 (p < 0.001).Footnote 5
Discussion
This study adds to a growing chorus of concerns regarding substantial PTSD-related metabolic health decline among veterans of the conflicts in Iraq and Afghanistan. In contrast to prior epidemiological estimates of MetS in the US population (20.3% among 20- to 39-year-olds; Ervin, Reference Ervin2009), we found substantially more veterans (29.0%) with PTSD in this age range with MetS. Our estimates of the prevalence of MetS (36.6% at T1 and 37.8% at T2) among those with PTSD are remarkably similar to the near 40% that has been reported in two recent meta-analyses (Bartoli et al. Reference Bartoli, Carrà, Crocamo, Caretta and Clerici2013; Rosenbaum et al. Reference Rosenbaum, Stubbs, Ward, Steel, Lederman and Vancampfort2015b ). This study extended prior work by addressing a critical question that has, to date, gone unanswered regarding the temporal relationships between PTSD and MetS. Results indicated that PTSD increased MetS risk over the course of, on average, 2.5 years, after controlling for initial MetS features, but that MetS did not predict subsequent PTSD symptoms.
That PTSD longitudinally predicted MetS carries implications for conceptualizing the course and treatment of both conditions. MetS is considered a syndrome (as opposed to a disease) in part because there is no obvious biological process that connects the individual MetS features. Traumatic stress may be one pathogenic environmental factor that, through biological and behavioral pathways, simultaneously intensifies the degeneration of multiple physiological processes and links them together. For example, PTSD may lead to both cardiovascular and HPA axis system dysregulation (Kibler et al. Reference Kibler, Tursich, Ma, Malcolm and Greenbarg2014; Brudey et al. Reference Brudey, Park, Wiaderkiewicz, Kobayashi, Mellman and Marvar2015), which would be expected to increase blood pressure, circulating lipids, blood sugars, and inflammation (Epel, Reference Epel2009); together, these alterations can increase central fat deposits (Epel, Reference Epel2009). At the same time, PTSD-related increases in reactive oxygen species (Miller & Sadeh, Reference Miller and Sadeh2014; Gautam et al. Reference Gautam, D'Arpa, Donohue, Muhie, Chakraborty, Luke, Grapov, Carroll, Meyerhoff, Hammamieh and Jett2015; Atli et al. Reference Atli, Bulut, Bez, Kaplan, Özdemir, Uysal, Selçuk and Sirin press) may alter the expression of genes important for regulating metabolic processes, ultimately compounding metabolic dysregulation (Grattagliano et al. Reference Grattagliano, Palmieri, Portincasa, Moschetta and Palasciano2008). In addition, PTSD-related poor sleep (Gavrieli et al. Reference Gavrieli, Farr, Davis, Crowell and Mantzoros2015; Talbot et al. Reference Talbot, Rao, Cohen, Richards, Inslicht, O'Donovan, Maguen, Metzler and Neylan2015), unhealthy diet (Hall et al. Reference Hall, Hoerster and Yancy2015), insufficient exercise (Georgiades et al. Reference Georgiades, Sherwood, Gullette, Babyak, Hinderliter, Waugh, Tweedy, Craighead, Bloomer and Blumenthal2000; Hall et al. Reference Hall, Hoerster and Yancy2015), cigarette and alcohol use (Dennis et al. Reference Dennis, Watkins, Calhoun, Oddone, Sherwood, Dennis, Rissling and Beckham2014), and psychotropic medication use (Vancampfort et al. Reference Vancampfort, Stubbs, Mitchell, De Hert, Wampers, Ward, Rosenbaum and Correll2015) may exert effects on metabolic health that additively and/or synergistically further contribute to the cascade of broad metabolic dysfunction.
We suspect that PTSD-related MetS may reflect an underlying process wherein the stress and chronicity of PTSD symptoms contribute to accelerated cellular aging and premature disease onset (Miller & Sadeh, Reference Miller and Sadeh2014; Lohr et al. Reference Lohr, Palmer, Eidt, Aailaboyina, Mausbach, Wolkowitz, Thorp and Jeste2015; Wolf et al. Reference Wolf, Logue, Hayes, Sadeh, Schichman, Stone, Salat, Milberg, McGlinchey and Miller2016). The prevalence of MetS is strongly associated with age in the US population (Ervin, Reference Ervin2009); however, we found that PTSD was associated with MetS independent of age, with a prevalence that was greater than expected by age. Thus, MetS may occur prematurely among those with PTSD and may be a clinical manifestation of accelerated aging. Consistent with this, prior work suggests that: (a) PTSD is related to advanced cellular age compared to chronological age, as reflected in DNA methylation (Wolf et al. Reference Wolf, Logue, Hayes, Sadeh, Schichman, Stone, Salat, Milberg, McGlinchey and Miller2016) and telomere length (Tyrka et al. Reference Tyrka, Parade, Price, Kao, Porton, Philip, Welch and Carpenter2016); and (b) metabolic dysregulation is also associated with shortened telomere length (Epel, Reference Epel2009) and contributes to biological aging (Belsky et al. Reference Belsky, Caspi, Houts, Cohen, Corcoran, Danese, Harrington, Israel, Levine, Schaefer, Sugden, Williams, Yashin, Poulton and Moffitt2015). Moreover, in our prior work in an independent sample of veterans from the wars in Iraq and Afghanistan, we found that PTSD-related MetS was cross-sectionally associated with substantial and widespread decreases in cortical thickness across temporal, parietal, and frontal brain regions (Wolf et al. in press). Together, these findings suggest that PTSD-related accelerated cellular aging may be reflected in premature genomic, physical health, and neurocognitive decline, highlighting the need to identify those at greatest risk and develop effective interventions.
It may be prudent to closely monitor the metabolic profiles of individuals with PTSD, even among young adults, so that early indications of problems can be discussed with the patient, careful consideration paid to the potential for weight gain side effects in prescribed medications, lifestyle changes recommended, and an appropriate treatment plan aimed at reducing metabolic pathologies enacted. Early screening for other age-dependent health conditions (e.g. cardiovascular disease, type 2 diabetes) may also be warranted. Although we found sex-related differences in MetS features, we found no evidence that PTSD was differentially related to MetS as a function of sex; thus early MetS screening among individuals with PTSD should be conducted with both men and women.
With respect to treatment implications, a recent, if small, meta-analysis found that physical activity was an effective intervention for PTSD (Rosenbaum et al. Reference Rosenbaum, Vancampfort, Steel, Newby, Ward and Stubbs2015c ) and may also improve physical health parameters among individuals with PTSD (Rosenbaum et al. Reference Rosenbaum, Sherrington and Tiedemann2015a ). No study to date has evaluated if exercise intervention for PTSD can reverse MetS, making this an important area for future research. It is also important for future trials of PTSD treatments to evaluate if psychological interventions for PTSD have indirect beneficial effects on MetS.
Results should be interpreted in light of study limitations including that metabolic profiles were not directly measured but instead were extracted from the medical record. This undoubtedly added methodological variance to the measurement of MetS (e.g. time between assessments, laboratory procedures), which would be expected to attenuate the magnitude of our associations. This medical record approach also led to missing data that we addressed via our analytic design, but which may alter results compared with complete data. There are also a number of other potentially important covariates and health indicators (e.g. insulin, inflammation, waist-to-hip-ratio, waist circumference) that we were unable to reliably assess via medical record review and that could have allowed us to test the International Diabetes Federation's ethnicity-based MetS criteria (Alberti et al. Reference Alberti, Zimmet, Shaw and IDF2005). With respect to the longitudinal design of the study, we controlled for baseline PTSD symptoms, but the metabolic profiles of individuals prior to trauma exposure and PTSD onset were not available. We did not observe PTSD group differences in raw laboratory values at T2 and this may have been due to differences in sample characteristics (e.g. PTSD severity; see Supplementary material) among those who did v. did not complete T2. The DSM changed from version IV to 5 between T1 and T2, and this could have also lead to different patterns of results in group-based analyses at T2 compared to T1. However, this would not be expected to substantively alter our primary results, which were focused on PTSD severity evaluated via regression, as prior work comparing DSM-IV with DSM-5 PTSD severity suggests very strong correlations across the two definitions (Miller et al. Reference Miller, Wolf, Kilpatrick, Resnick, Marx, Holowka, Keane, Rosen and Friedman2013; Bovin et al. Reference Bovin, Marx, Weathers, Gallagher, Rodriguez, Schnurr and Keanein press).
The strengths of this study include that it is the first longitudinal evaluation of potential bi-directional associations between PTSD and MetS that controls for baseline effects and does so parsimoniously in a single analysis. Additional study strengths include the large sample size, inclusion of Iraq/Afghanistan veterans from across the country, the ability to evaluate sex-specific effects, and our use of a structured diagnostic interview to assess PTSD.
In conclusion, we found that young veterans of the conflicts in Iraq and Afghanistan with PTSD exhibited signs of substantial premature health decline. This should be of grave concern to mental health and primary-care clinicians alike and suggests the critical importance of developing interventions that reduce both psychiatric and metabolic pathology. MetS is hugely costly on its own (Sullivan et al. Reference Sullivan, Ghushchyan, Wyatt and Hill2007), and the economic, personal, and societal costs can only balloon if the condition gives rise to other associated diseases such as premature cardiovascular disease (Lakka et al. Reference Lakka, Laaksonen, Lakka, Niskanen, Kumpusalo, Tuomilehto and Salonen2002), type 2 diabetes (Wilson et al. Reference Wilson, D'Agostino, Parise, Sullivan and Meigs2005), cancer (Esposito et al. Reference Esposito, Chiodini, Colao, Lenzi and Giugliano2012), dementia (Yaffe et al. Reference Yaffe, Kanaya, Lindquist, Simonsick, Harris, Shorr, Tylavsky and Newman2004), and death (Lakka et al. Reference Lakka, Laaksonen, Lakka, Niskanen, Kumpusalo, Tuomilehto and Salonen2002). This is a major public health concern and addressing it in this population now has the potential to reduce preventable morbidity and mortality among the nation's newest cohort of veterans.
Supplementary material
For supplementary material accompanying this paper visit http://dx.doi.org/10.1017/S0033291716000817.
Acknowledgements
This work was supported by a Career Development Award to Erika J. Wolf from the United States Department of Veterans Affairs, Clinical Sciences Research and Development Program. Brian Marx and Terence Keane were supported by funding from the Department of Defense (W81XWH-08-2-0100, W81XWH-12-2-0117). Dr. Marx was additionally supported by funding from the National Institute of Mental Health (1R01MH095737-01A1), Defense Advanced Research Programs Agency (N66001-11-C-4006), and Department of Defense (W81XWH-10-2-0181), and the Department of Veterans Affairs (Cooperative Studies Program no. 591). Dr. Keane was additionally supported by funding from the Consortium to Alleviate PTSD and National Institute of Mental Health (5T32MH019836-16). Raymond Rosen was supported by funding from the Department of Defense, (W81XWH-12-1-0532; W81XWH-14-2-0139; W81XWH-08-2-0102, and W81XWH-12-2-0121). Karen Mitchell's contribution was partly supported by funding from the National Institute of Mental Health (K01MH093750). Lewina Lee was supported by funding from the National Institute on Aging (K08AG048221) and the National Center for Advancing Translational Sciences (BU-CTSI Grant Number 1UL1TR001430). The contents of this article do not represent the views of the U.S. Department of Veterans Affairs or the United States Government.
We would like to thank Gayatri Rangananthan of NERI for her contribution to this work, which included data management and manuscript review.
Declaration of Interest
None.
