Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 May 15.
Published in final edited form as: Psychiatry Res. 2022 Nov 28;319:114994. doi: 10.1016/j.psychres.2022.114994

Factor structure of the posttraumatic stress disorder checklist (PCL-17) in 279,897 million veteran program participants

Cassie Overstreet a,b,*, Daniel F Levey a,b, Hang Zhou a,b, Kelly M Harrington c,d, Rachel Quaden c, Murray B Stein e,f, Joel Gelernter a,b, Robert H Pietrzak a,g, Million Veteran Program
PMCID: PMC10184787  NIHMSID: NIHMS1876038  PMID: 36516638

Abstract

The Million Veteran Program (MVP) uses the posttraumatic stress disorder symptoms (PTSD) Checklist 17 (PCL- 17) self-report to assess PTSD. Existing literature suggests that the five-factor dysphoric arousal model best represents the PTSD symptom clusters; this can be tested within MVP, one of the largest samples collected with suitable data. We compared factor models within MVP across genetically defined subsamples (ancestry [European, African, admixed American, and East Asian], sex) via multi-group Confirmatory factor analyses in a sample of 279,897 participants. The five-factor dysphoric arousal model best fit the PCL-17 data, consistent with previous findings. The factor structure could also be imposed across all groups tested. Verifying the factor structure provides a framework for future phenotypic and genotypic analyses within MVP and other samples.

Keywords: Posttraumatic stress disorder, Five factor dysphoric arousal model, Multi-group confirmatory factor analysis

1. Introduction

Multiple large-scale studies assess symptoms of posttraumatic stress disorder (PTSD) using the PCL-17, the Diagnostic and Statistical Manual of Mental Disorders, 4th ed. (DSM-IV;American Psychiatric Association [APA], 1994) version of the Posttraumatic Stress Checklist (PCL; Weathers et al., 1991). The Million Veteran Program (MVP), a large-scale epidemiologic study of psychosocial and biological variables relevant to health in veterans, includes this self-report instrument in about half of its participants recruited to date. Significant associations between genetic loci and the re-experiencing symptom cluster of PTSD have been identified using these PCL-17 data in MVP (Gelernter et al., 2019). Differential findings based on multiple PTSD subphenotypes, and also electronic health record-based diagnosis vs. quantitative PCL trait definition, have also been demonstrated within the dataset (Stein et al., 2021).

A three-factor model of PTSD is implicit in the DSM-IV (re-experiencing, avoidance, hyperarousal), but alternative factor structures have been proposed (Harpaz-Rotem et al., 2014). The literature regarding the latent architecture of the PCL-17 includes models ranging from one to five factors (see Armour, Műllerová, and Elhai, 2016 for review), with differences largely present within the avoidance and hyperarousal clusters. Significant support has emerged for the five-factor dysphoric arousal model, which includes separate clusters of re-experiencing, avoidance, emotional numbing, dysphoric arousal (e.g., sleep disturbance) and anxious arousal (e.g., hypervigilance), as this model has demonstrated superior fit to symptom-level PTSD data relative to other models (Elhai et al., 2011; Armour et al., 2012; Pietrzak et al., 2012; Harpaz-Rotem et al., 2014). The sample sizes included in these studies range from 294 to 323,903 subjects. The five-factor dysphoric arousal model has not been empirically evaluated in MVP.

The most empirically supported factor models of the PCL-17 include the three-factor DSM-IV model, the four-factor dysphoria and numbing models, and the five-factor dysphoric arousal model. The three-factor DSM-IV model reflects the reexperiencing, avoidance, and hyperarousal symptom clusters per the DSM-IV diagnostic criteria. The four-factor dysphoria model maintains the same re-experiencing items as the three-factor model, while the majority of the remaining items fall under a dysphoria factor with two items pertaining to avoidance and two pertaining to hyperarousal (Simms et al., 2002). In the four-factor numbing model, the re-experiencing and hyperarousal items derived from the DSM-IV are retained while the avoidance subcluster is divided into avoidance and numbing items (King et al., 1998). Finally, in the five-factor dysphoric arousal model each of the re-experiencing items load onto a single factor, consistent with the DSM-IV. The avoidance items broadly load onto two separate factors reflecting avoidance and emotional numbing symptom clusters. One item pertaining to sense of foreshortened future (falling under the avoidance cluster per the DSM-IV) loads onto the dysphoric arousal cluster along with many of the hyperarousal items while the remaining two items within the hyperarousal cluster load onto an anxious arousal factor rather than a single hyperarousal factor per the DSM-IV (Elhai et al., 2011).

Inferences regarding PCL-17 factor structure as it relates to demographic variables including self-reported race and sex have been mixed. Although the factor structure was determined to be consistent across non-Hispanic African Americans and Caucasians [sic] in a sample of 413 participants, severity of symptoms within specific clusters varied by self-reported race, with non-Hispanic African Americans reporting greater symptoms of re-experiencing (Coleman et al., 2019). Other research supports statistical invariance across groups when the five-factor dysphoric arousal model was examined in a sample of 6,248 subjects (Contractor et al., 2015). Different rates of exposure to certain types of potentially traumatic events (e.g., men more likely to be exposed to combat while women are at greater risk of exposure to sexual violence) and symptom severity (e.g., women often present with greater severity of PTSD symptoms (Breslau et al., 1997)) suggest potential variability regarding latent factor structure. Although there is some support for sex differences in PCL-17 factor structure (Hall et al., 2012), a systematic review (Armour, Műllerová, and Elhai, 2016) demonstrated that the five-factor dysphoric arousal model provided superior fit across multiple studies. But arguably, all studies to date have had substantial power limitations. Well-powered statistical comparisons are needed to determine if the same factor structure can be applied across groups.

To address these gaps, we sought to determine the best fitting structural model of PTSD symptoms by comparing the three-factor DSM-IV model, the four-factor dysphoria and numbing models, and the five-factor dysphoric arousal model in a large cohort of more than 279,000 veterans in the MVP. Taking a nuanced approach to examining the latent factor structure of the PCL-17 may aid in improving PTSD gene finding efforts which would otherwise be missed when PCL-17 items are summed to yield a single total score and/or separated into the standard three symptom cluster solution per the DSM-IV. Thus, the aims of the present study were to: (1) compare factor models of PTSD within the MVP dataset; and (2) compare the factor structure across ancestry (European, African, Admixed American, and East Asian ancestry) and sex (men and women) via multi-group confirmatory factor analyses (CFA).

2. Methods

2.1. Subjects

The MVP has been previously described (Gaziano et al., 2016). To date, the MVP Lifestyle survey, which includes the PCL-17 was completed by 285,062 participants. Following removal of any participants with missingness on the PTSD Checklist described below or information regarding ancestry and/or sex (as determined by genotype data), the sample consisted of 279,897 participants (92.2% male, Mage=65.7, SD=11.3, range=19-112). Reference population groups in the 1000 Genomes samples were used to define European (EUR), African (AFR), admixed American (AMR), and East Asian (EAS) ancestry groups genetically. The EUR sample consisted of 233,707 participants, AFR sample 31,690 participants, AMR sample 12,250 participants, and EAS sample 2,250 participants.

2.2. Measures

2.2.1. Posttraumatic Stress Disorder Checklist (PCL-C)

The 17-item PCL version utilized in MVP is similar to the civilian version (PCL-C) in that it asks respondents to report how much they have been bothered in the past 30 days by symptoms in response to “stressful experiences”. Responses were made on a Likert-based scale between 1 “Not at all” and 5 “Extremely.” In the current sample, Cronbach’s alpha on the PCL was 0.96.

2.3. Data Analysis

Multi-group CFAs were conducted using robust maximum likelihood estimation with the Satorra and Bentler (S–B) χ2 scaling correction (Satorra and Bentler, 2001) to compare the three-factor DSM-IV model, the four-factor dysphoria and numbing models, and the five-factor dysphoric arousal model (item mapping across models presented in Table 1) because PCL item scores were non-normally distributed (14/17 items; Table 2). This procedure estimates standard errors under conditions of multivariate non-normality and calculates other χ2-dependent fit statistics based on the S–B χ2 statistic. In the CFA models, PCL items were specified to load only on one of the proposed factors and all factors were allowed to correlate. Model fit was evaluated using the S–B χ2 comparative fit index (CFI), Tucker Lewis Index (TLI), Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR) values. Higher CFI and TLI values and lower S–B χ>2, AIC, BIC, RMSEA, and SRMR values indicate a better fitting model. The following conventions were also considered when evaluating model fit: CFI and TLI ≥ 0.90 indicate adequate fit and ≥ 0.95 an excellent fit; RMSEA ≤ 0.08 as an adequate fit and ≤ 0.06 as indicative of excellent fit; and SRMR ≤ 0.08 suggesting good fit (Hu and Bentler, 1999).

Table 1.

Item Mapping across DSM-IV, Dysphoria, Numbing, and Five-factor Models.

PCL Items Model
DSM-
IV		 Dysphoria Numbing 5-
factor
1. Intrusive thoughts of trauma R R R R
2. Recurrent dreams of trauma R R R R
3. Flashbacks R R R R
4. Emotional reactivity to trauma cues R R R R
5. Physiological reactivity to trauma cues R R R R
6. Avoiding thoughts of trauma A A A A
7. Avoiding reminders of trauma A A A A
8. Inability to recall aspects of trauma A D N N
9. Loss of interest A D N N
10. Detachment A D N N
11. Restricted affect A D N N
12. Sense of foreshortened future A D N DA
13. Sleep disturbance H D H DA
14. Irritability H D H DA
15. Difficulty concentrating H D H DA
16. Hypervigilance H H H AA
17. Exaggerated startle response H H H AA
Open in a new tab

Note. DSM-IV = Diagnostic and Statistical Manual of Mental Disorders, 4th edition; R = Re-experiencing; A = Avoidance; N = Emotional numbing; H=Hyperarousal; D = Dysphoria; DA = Dysphoric Arousal; AA = Anxious Arousal

Table 2.

Spearman Correlations across 17 items of the Posttraumatic Stress Disorder Checklist (N = 279,897).

Items Mean (SD) Skewness Kurtosis Spearman Correlation
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17.
1. Intrusive thoughts of trauma 1.97 (1.14) .99 −.05 -
2. Recurrent dreams of trauma 1.77 (1.09) 1.34 .82 .76 -
3. Flashbacks 1.58 (0.98) 1.75 2.27 .71 .72 -
4. Emotional reactivity to trauma cues 1.79 (1.10) 1.34 .85 .75 .69 .74 -
5. Physiological reactivity to trauma cues 1.61 (1.02) 1.71 2.07 .67 .67 .73 .73 -
6. Avoiding thoughts of trauma 1.91 (1.21) 1.16 .17 .71 .66 .66 .72 .68 -
7. Avoiding reminders of trauma 1.51 (1.17) 1.51 1.17 .67 .65 .69 .71 .70 .76 -
8. Inability to recall aspects of trauma 1.59 (1.01) 1.75 2.27 .55 .54 .56 .57 .57 .58 .59 -
9. Loss of interest 2.13 (1.23) .86 −.36 .54 .50 .52 .55 .52 .54 .55 .50 -
10. Detachment 2.00 (1.24) 1.05 −.09 .60 .54 .57 .61 .58 .60 .60 .52 .69 -
11. Restricted affect 1.77 (1.16) 1.43 .91 .56 .53 .56 .58 .57 .58 .58 .50 .60 .71 -
12. Sense of foreshortened future 1.81 (1.19) 1.38 .74 .56 .53 .56 .57 .56 .56 .56 .49 .60 .64 .62 -
13. Sleep disturbance 2.13 (1.31) .68 −.76 .51 .49 .48 .50 .50 .49 .48 .41 .50 .52 .48 .50 -
14. Irritability 1.93 (1.12) 1.14 .45 .55 .52 .54 .58 .55 .55 .54 .47 .53 .59 .58 .54 .51 -
15. Difficulty concentrating 1.90 (1.09) 1.18 .59 .57 .53 .56 .58 .56 .56 .57 .56 .62 .63 .59 .57 .53 .61 -
16. Hypervigilance 2.01 (1.28) 1.04 −.15 .60 .58 .60 .61 .60 .60 .60 .50 .51 .56 .54 .53 .48 .55 .53 -
17. Exaggerated startle response 1.83 (1.18) 1.33 .68 .63 .61 .64 .64 .65 .62 .63 .54 .54 .59 .57 .56 .50 .60 .60 .70 -
Open in a new tab

Note. Items separated to reflect DSM-IV clusters of re-experiencing (items 1-5), avoidance (items 6-12), and hyperarousal (items 13-17)

Additionally, given use of the S–B χ2 statistic, χ2 difference tests were calculated for nested models with a correction factor to compare the relative fit of the five-factor model to the three-factor DSM-IV model and four-factor dysphoria and numbing models (Fan and Sivo, 2009). Non-nested models (i.e., DSM-IV vs. dysphoria model; numbing vs. dysphoria model) were compared using BIC (Schwarz, 1978) whereby models with a lower BIC value suggest better fit (difference of 6–10 indicates strong support and a difference >10 indicates very strong support; Raftery, 1995). Analyses were conducted in R (R Core Team, 2020) using the “lavaan” package (Rosseel, 2012).

We tested measurement invariance for both ancestry (EUR, AFR, AMR, EAS) and sex (men, women) in the following manner: (1) determine if the best fitting model identified by the CFA procedure described above could be constrained to equality across groups (similar number of factors and general pattern of loadings or configural invariance), (2) constrain model so that loadings are equal across groups (metric invariance), (3) constrain model so that loadings and intercepts are equal across groups (scalar invariance), (4) constrain model so that the loadings, intercepts, and residuals are equal across groups (strict invariance). Multi-group CFA is an iterative process which stops at any step that is noninvariant. Because there was a large sample size, fit statistics were evaluated with a particular focus on decreases in CFI (>.01) and RMSEA (>.015) to determine invariance across groups (Chen, 2007; Hirschfeld and Von Brachel, 2014).

3. Results

The mean, standard deviation, correlations, and additional descriptive information for each PCL item are provided in Table 2.

3.1. Confirmatory Factor Analyses

In the full sample, the CFA results for the three-factor DSM-IV model, the four-factor dysphoria and numbing models, and the five-factor dysphoric arousal model are presented in Table 3. Results indicated that the five-factor dysphoric arousal model provided a better fit to the data relative to the three-factor DSM-IV model and four-factor dysphoria and numbing models as demonstrated by lower S–B χ2, AIC, BIC, RMSEA, and SRMR values, as well as higher CFI and TLI values. Per the CFI and TFI results, the four-factor dysphoria, four-factor numbing, and five-factor dysphoric arousal models provided excellent fit (CFI and TLI ≥ 0.95) while the three-factor DSM-IV model demonstrated only adequate fit. Item loadings and factor covariances for the five-factor dysphoric arousal model are presented in Fig. 1.

Table 3.

Fit Statistics for the DSM-IV, Dysphoria, Numbing, and Five-factor Models in the Full Sample (n =279,897).

Model S–B χ2 χ 2 df CFI TLI RMSEA AIC BIC SRMR
DSM-IV 140113.742 253422.261 116 .939 .929 .088 10842188.714 10842578.775 .034
Dysphoria 68382.580 121689.164 113 .971 .965 .062 10710461.317 10710883.004 .025
Numbing 79099.212 140885.299 113 .966 .960 .067 10729657.452 10730079.139 .029
5-factor 59837.644 106163.370 109 .975 .968 .059 10694943.523 10695407.379 .025
Open in a new tab

Note. S–B χ2 = Satorra-Bentler Scaled Correction Chi-square, χ2 = Chi-square, df = degrees of freedom, CFI = comparative fit index, TLI = Tucker Lewis Index, RMSEA = root square error of approximation, AIC = Akaike Information Criterion, BIC = Bayesian Information Criterion, SRMR = standardized root mean square residual, EUR = European Ancestry, AFR = African Ancestry

Fig. 1.

Fig. 1.

Open in a new tab

Loadings for Five-factor Model in Full Sample (N = 279,897).

Additionally, the five-factor dysphoric arousal model’s CFI difference from the three-factor and numbing models was greater than 0.01, indicating a better fit (Fan and Sivo, 2009).The five-factor model was the only model demonstrating excellent fit per the RMSEA (RMSEA ≤ 0.06). χ2 difference tests further demonstrated that the five-factor model fit significantly better than the three-factor DSM-IV model, Δχ2(7) = 114743.86, p < .001; the four-factor dysphoria model, Δχ2(3) = 83352.77, p < .001; and the four-factor emotional numbing model, Δχ2(3) = 72467.98, p < .001. For non-nested models (DSM-IV vs. dysphoria model; numbing vs. dysphoria model), the dysphoric arousal model had a lower BIC (>10 indicating very strong support) when compared to the DSM-IV model and the numbing model (Table 3).

3.2. Multi-Group Confirmatory Factor Analyses

Following identification of the best-fitting factor model (Fig. 1), multi-group invariance models (Table 4) were examined to identify potential differences across groups (ancestry [EUR, AFR, AMR, EAS] and sex). Although χ2 difference tests were each significant (p <.001), this was not surprising considering the large sample size, and greater consideration was placed on differences in CFI and RMSEA. Multi-group CFA results for ancestry demonstrated configural, metric, scalar, and strict invariance. The same pattern of findings was also seen for sex. Standardized estimates and standard errors for the five-factor dysphoric arousal model in each group are provided in Tables 5 (ancestry) and 6 (sex).

Table 4.

Results of Multi-group Confirmatory Factor Models by Ancestry and Sex.

Model χ 2 df CFI TLI RMSEA AIC BIC SRMR
5-Factor Model 106163.4 109 .975 .968 .059 10694944 10695407 .025
Ancestry
Configural Invariance 106590.5 436 .974 .968 .059 10652808 10655380 .025
Metric Invariance 109313.6 572 .973 .969 .057 10655459 10657651 .027
Scalar Invariance 114450.0 508 .972 .970 .056 10660523 10662336 .028
Strict Invariance 129828.4 559 .968 .969 .057 10675799 10677075 .029
Sex
Configural Invariance 106978.5 218 .974 .968 .059 10672487 10673773 .024
Metric Invariance 108486.2 230 .974 .969 .058 10673971 10675131 .026
Scalar Invariance 114855.6 242 .973 .969 .058 10680316 10681349 .026
Strict Invariance 123639.4 259 .970 .969 .058 10689066 10689920 .027
Open in a new tab

Note. Configural Invariance = groups constrained to equality, Metric Invariance = loadings constrained to equality, Scalar Invariance = loadings and intercepts constrained to equality, Strict Invariance = loadings, intercepts, and residuals constrained to equality

Table 5.

Factor Loadings for Five-factor Model for each Ancestry Groups.

Items Standardized Factor Loadings
EUR (n =
233,707)		 AFR (n =
31,690)		 AMR (n =
12,250)		 EAS (n
= 2,250)
Factor 1: Re-experiencing
1. Intrusive thoughts of trauma .867 (.001) .889 (.001) .889 (.002) .889 (.005)
2. Recurrent dreams of trauma .849 (.001) .885 (.001) .884 (.002) .868 (.006)
3. Flashbacks .869 (.001) .900 (.001) .901 (.002) .909 (.004)
4. Emotional reactivity to trauma cues .890 (.001) .907 (.001) .906 (.002) .911 (.004)
5. Physiological reactivity to trauma cues .855 (.001) .868 (.002) .882 (.002) .880 (.005)
Factor 2: Avoidance
6. Avoiding thoughts of trauma .878 (.001) .904 (.001) .895 (.002) .913 (.005)
7. Avoiding reminders of trauma .890 (.002) .912 (.001) .915 (.002) .922 (.004)
Factor 3: Emotional Numbing
8. Inability to recall aspects of trauma .667 (.002) .696 (.003) .714 (.005) .740 (.010)
9. Loss of interest .793 (.001) .829 (.002) .842 (.003) .834 (.007)
10. Detachment .873 (.001) .880 (.002) .887 (.002) .882 (.006)
11. Restricted affect .819 (.001) .845 (.002) .843 (.003) .858 (.007)
Factor 4: Dysphoric Arousal
12. Sense of foreshortened future .769 (.001) .809 (.002) .816 (.003) .814 (.008)
13. Sleep disturbance .666 (.001) .732 (.003) .743 (.004) .746 (.010)
14. Irritability .768 (.001) .832 (.002) .822 (.003) .829 (.007)
15. Difficulty concentrating .793 (.001) .847 (.002) .846 (.003) .851 (.007)
Factor 5: Anxious Arousal
16. Hypervigilance .828 (.001) .804 (.002) .835 (.003) .824 (.008)
17. Exaggerated startle response .882 (.001) .897 (.002) .915 (.002) .918 (.006)
Open in a new tab

Note. EUR = European Ancestry, AFR = African Ancestry, AMR = Ad Mixed ancestry, EAS = East Asian ancestry

Table 6.

Factor Loadings for Five-factor Model for Men and Women.

Items Standardized Factor Loadings
Men (n = 257,986) Women (n = 21,911)
Factor 1: Re-experiencing
1. Intrusive thoughts of trauma .873 (.001) .869 (.002)
2. Recurrent dreams of trauma .864 (.001) .835 (.002)
3. Flashbacks .882 (.001) .869 (.002)
4. Emotional reactivity to trauma cues .895 (.001) .906 (.001)
5. Physiological reactivity to trauma cues .861 (.001) .884 (.002)
Factor 2: Avoidance
6. Avoiding thoughts of trauma .884 (.001) .882 (.002)
7. Avoiding reminders of trauma .897 (.001) .907 (.002)
Factor 3: Numbing
8. Inability to recall aspects of trauma .682 (.001) .656 (.004)
9. Loss of interest .800 (.001) .843 (.002)
10. Detachment .875 (.001) .883 (.002)
11. Restricted affect .827 (.001) .830 (.002)
Factor 4: Dysphoric Arousal
12. Sense of foreshortened future .787 (.001) .733 (.003)
13. Sleep disturbance .685 (.001) .670 (.004)
14. Irritability .782 (.001) .781 (.003)
15. Difficulty concentrating .805 (.001) .817 (.003)
Factor 5: Anxious Arousal
16. Hypervigilance .826 (.001) .876 (.002)
17. Exaggerated startle response .886 (.001) .880 (.002)
Posttraumatic Stress Disorder Checklist (Million Veteran Program)
How much have you been bothered by that problem in the past month?
Not at all A little bit Moderately Quite a bit Extremely
1. Repeated, disturbing memories, thoughts or images of a stressful experience from the past?
2. Repeated, disturbing dreams of a stressful experience from the past?
3. Suddenly acting or feeling as if a stressful experience were happening again (as if you were reliving it)?
4. Feeling very upset when something reminded you of a stressful experience from the past?
5. Having physical reactions (e.g., heart pounding, trouble breathing, sweating) when something reminded you of a stressful experience from the past?
6. Avoiding thinking about or talking about a stressful experience from the past or avoiding having feelings related to it?
7. Avoiding activities or situations because they reminded you of a stressful experience from the past?
8. Trouble remembering important parts of a stressful experience from the past?
9. Loss of interest in activities you used to enjoy?
10. Feeling distant or cutoff from other people?
11. Feeling emotionally numb or being unable to have loving feelings for those close to you?
12. Feeling as if your future will somehow be cut short?
13. Trouble falling or staying asleep?
14. Feeling irritable or having angry outbursts?
15. Having difficulty concentrating?
16. Being ’super alert’ or watchful or on guard?
17. Feeling jumpy or easily startled?
Open in a new tab

4. Discussion

The present study aimed to (1) compare factor models of PTSD within the MVP dataset and (2) compare the factor structure across ancestry (EUR, AFR, AMR, EAS) and sex (men, women). Consistent with the literature (Elhai et al., 2011; Armour et al., 2012; Pietrzak et al., 2012) and our hypotheses, the five-factor dysphoric arousal model provided the best fit for the data and was consistent across groups (ancestry and sex). Each of the re-experiencing items loaded onto a single factor, consistent with the DSM-IV. However, rather than a single avoidance factor as suggested by the DSM-IV, the majority of avoidance items loaded onto two separate factors reflecting avoidance and emotional numbing symptom clusters with the exception of the item regarding foreshortened future which loaded onto the dysphoric arousal factor along with many of the hyperarousal items as defined by the DSM-IV. Hyperarousal items were also divided into separate factors, representing dysphoric arousal and anxious arousal rather than a single hyperarousal factor per the DSM-IV. The factor structure was invariant across groups, indicating that the same general pattern of item loadings onto the five latent factors per the five-factor dysphoric arousal model did not differ significantly by ancestry or sex.

These findings highlight the utility of a nuanced five-factor dysphoric arousal model when examining PTSD symptoms using the DSM-IV version of the PTSD Checklist, rather than the three-factor model explicit in the DSM-IV criteria and implicit in the PCL-17’s conception. The five-factor dysphoric arousal model splits items typically subsumed within the avoidance cluster per the DSM-IV, as well as the items falling under the DSM-IV hyperarousal cluster. Collapsing these items together to create avoidance and hyperarousal scores as per the DSM-IV may obscure identification of phenotypic and biological associations that may otherwise be identified if the factors were divided into more phenotypically distinct avoidance, emotional numbing, dysphoric arousal, and anxious arousal subdomains as suggested by the five-factor dysphoric arousal model. This five factor solution may be particularly useful in research settings that continue to utilize the PCL-17 such as the MVP, a very large sample that continues to ascertain PTSD via the PCL-17. Abbreviated measures are also frequently utilized in large datasets, for example the Post-Traumatic Checklist – 6-item Civilian Version (PCL-6, Lang and Stein, 2005). The PCL-6 taps into each of the three DSM-IV domains (two from re-experiencing, one pertaining to avoidance, and three from hyperarousal). The PCL-6 also includes items that reflect the re-experiencing, avoidance, dysphoria, and hyperarousal domains of the dysphoric arousal five-factor model; however, it does not include any items reflecting the emotional numbing factor present in the five-factor model. Additional research is warranted to investigate whether abbreviated PTSD measures may better capture symptoms when reflecting the five-factor dysphoric arousal model compared to the three-factor DSM-IV model. It is noteworthy that although the PCL-6 does not tap into each domain reflected by the five-factor dysphoric arousal model, many studies demonstrate that it effectively serves its purpose of screening for PTSD (Lang et al., 2012) and inclusion of each domain may not be necessary given the purpose of the measure. Multiple studies have been published using MVP data, including genetic analyses focused on PTSD symptom clusters as defined by the DSM-IV (Pathak et al., 2022; Stein et al., 2021; Gelernter et al., 2019 [focused on the re-experiencing cluster alone]). A more refined approach to examination of PTSD could reduce heterogeneity and reveal unique and shared genetic and biological features associated with each symptom cluster. Notably, the present analyses utilized genotypic definitions of ancestry and sex, future studies may also want to consider excluding related individuals to further enhance phenotypic precision if the traits will be used for genetic studies.

Multi-group measurement invariance model results indicated invariance at a configural level, demonstrating that the five-factor dysphoric arousal model could be imposed across groups. This is consistent with the literature suggesting that the five-factor model could be applied across some demographic groups with limited concern regarding number of factors and pattern of loading varying greatly across groups (Armour et al. 2016). To control for population stratification, genetic analyses (e.g., genome-wide association studies) are conducted by ancestry group and then meta-analyzed. If the multi-group models were non-invariant this would have critical implications on genetic analyses (e.g., the avoidance cluster may include different items depending on group). Demographic features remain important factors to be considered within the context of PTSD given variable prevalence rates of certain types of trauma and PTSD symptom severity.

There are a number of noteworthy limitations of this study. First, the present analyses focus on the PCL-17, which corresponds to DSM-IV PTSD criteria. Alterations in the criteria made in the DSM-5 (e.g., inclusion of items pertaining to negative cognitions and mood) are not reflected in the PCL-17. However, large datasets including MVP continue to use the DSM-IV version; and CFA studies have largely supported a separation of avoidance symptoms, as well as dysphoric and anxious arousal (Armour et al., 2015; Lee et al., 2019; Wang et al., 2017). Future research investigating an empirical “crosswalk” between DSM-IV and DSM-5 responses may be useful, even if current methods only focus on total scores (Moshier et al., 2019). Exploration of potential methods of translating DSM-IV PTSD symptoms to DSM-5 may also prove beneficial. Second, the sample is focused on military veterans and the specific trauma histories were not assessed. Trauma type and cumulative trauma load (i.e., number of traumatic events one is exposed to within one’s lifetime) have been shown to have an impact on PTSD symptom presentation (Kessler et al., 2017; Scott et al., 2013) and additional research evaluating invariance of the latent structure of PTSD in relation to specific index traumas is needed. Future inclusion of measures such as the Life Events Checklist for DSM-5 (Weathers et al., 2013) would also be useful in providing a thorough assessment and consideration of trauma history. Additionally, the predominantly male sample may also limit generalizability of the findings to female veterans. Other variables (e.g., cognition, inhibition, level of social support, schemas regarding the world, themselves, and others) may also influence results and could be helpful to consider in future studies of the five-factor dysphoric arousal model. Finally, although there is sufficient power to examine these variables on a phenotypic level, genetic analyses within non-EUR and female subsamples may prove difficult due to relatively lower sample sizes. Many prior factor analytic studies included relatively small sample sizes; however, Harpaz-Rotem and colleagues (2014) compared the factor structure in a large sample of individuals receiving services at VA medical centers (N=323,903). Given that this study was conducted among veterans utilizing services at the VA, it is possible that there is some overlap across participants (VA and MVP). This concern, which cannot be evaluated empirically, is somewhat mitigated by the ongoing nature of data collection in MVP and that PCL data were collected from each sample independently: i.e. even if there is overlap in subjects there is no overlap in the phenotypic measures. MVP data collection is ongoing which will tend to address the power issue over time.

Despite these limitations, these findings aid in refining understanding of the PTSD phenotype assessed using the DSM-IV version of the PTSD Checklist. In addition to adding to the literature examining the factor structure in large samples, this detailed approach to phenotyping may improve our understanding of the biological factors underlying PTSD symptoms by allowing us to define evidence-based diagnosis subphenotypes that may be sued in future GWAS analyses. Previous phenotypic research suggests that there are differences across symptom clusters (Rusch et al., 2019) and although high genetic correlations were demonstrated across the three DSM-IV PTSD subdomains in a previous study conducted within MVP (Stein et al., 2021), it is possible that a re-analysis focused on the five factor model symptom clusters (with a larger sample size) may identify important differences and improve gene-finding efforts.

Funding and Acknowledgements

C.O. effort supported by the Yale T32 Biological Sciences Training Program. R.P. effort supported by Clinical Neurosciences Division of the U.S. Department of Veterans Affairs National Center for Posttraumatic Stress Disorder. J.G. NIH grant R01DA054869 and funding from the Department of Veterans Affairs Office of Research and Development grant I01CX001849, and the VA Cooperative Studies Program study, no. 575B; and the West Haven VA Mental Illness Research, Education and Clinical Center. D.F.L. was funded by VA grant 1IK2BX005058-01A2 and is a BBRF NARSAD PandS Fund Investigator and Aimee Mann Fellow of Psychiatric Genetics. M.B.S. has a received research support from NIH, Department of Veterans Affairs, and the Department of Defense.

Footnotes

CRediT authorship contribution statement

Cassie Overstreet: Conceptualization, Formal analysis, Writing – original draft. Daniel F. Levey: Writing – review & editing. Hang Zhou: Writing – review & editing. Kelly M. Harrington: Data curation, Writing – review & editing. Rachel Quaden: Data curation, Writing – review & editing. Murray B. Stein: Writing – review & editing, Supervision. Joel Gelernter: Conceptualization, Writing – review & editing, Supervision. Robert H. Pietrzak: Conceptualization, Writing–review & editing, Supervision.

Declaration of Competing Interest

C.O., D.L., H.Z., K.H., R.Q., J.G., and R.P. have no conflicts of interest to declare. M.B.S has in the past 3 years received consulting income from Acadia Pharmaceuticals, Aptinyx, atai Life Sciences, Boehringer Ingelheim, Bionomics, BioXcel Therapeutics, Clexio, Eisai, EmpowerPharm, Engrail Therapeutics, Janssen, Jazz Pharmaceuticals, Neuro-Trauma Sciences, PureTech Health, Sumitomo Pharma, and Roche/Genentech. Dr. Stein has stock options in Oxeia Biopharmaceuticals and EpiVario. He is paid for his editorial work on Depression and Anxiety (Editor-in-Chief), Biological Psychiatry (Deputy Editor), and UpToDate (Co-Editor-in-Chief for Psychiatry). He is on the scientific advisory board for the Brain and Behavior Research Foundation and the Anxiety and Depression Association of America.

References

  1. American Psychiatric Association, 1994. Diagnostic and statistical manual of mental disorders, 4th ed. Author, Washington, DC. [Google Scholar]
  2. Armour C, Elhai JD, Richardson D, Ractliffe K, Wang L, Elklit A, 2012. Assessing a five factor model of PTSD: is dysphoric arousal a unique PTSD construct showing differential relationships with anxiety and depression? J. Anxiety Disord 26 (2), 368–376. [DOI] [PubMed] [Google Scholar]
  3. Armour C, Tsai J, Durham TA, Charak R, Biehn TL, Elhai JD, Pietrzak RH., 2015. Dimensional structure of DSM-5 posttraumatic stress symptoms: support for a hybrid Anhedonia and Externalizing Behaviors model. J. Psychiatr. Res 61, 106–113. [DOI] [PubMed] [Google Scholar]
  4. Armour C, Műllerovα J, Elhai JD, 2016. A systematic literature review of PTSD’s latent structure in the Diagnostic and Statistical Manual of Mental Disorders: DSM-IV to DSM-5. Clin. Psychol. Rev 44, 60–74. [DOI] [PubMed] [Google Scholar]
  5. Breslau N, Davis GC, Andreski P, Peterson EL, Schultz LR, 1997. Sex differences in posttraumatic stress disorder. Arch. Gen. Psychiatry 54, 1044–1048. [DOI] [PubMed] [Google Scholar]
  6. Coleman JA, Ingram KM, Sheerin CM, 2019. Racial differences in posttraumatic stress disorder symptoms among African American and Caucasian male veterans. Traumatology 25 (4), 297. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Chen FF, 2007. Sensitivity of goodness of fit indexes to lack of measurement invariance. Struct. Eq. Model 14 (3), 464–504. [Google Scholar]
  8. Contractor AA, Claycomb MA, Byllesby BM, Layne CM, Kaplow JB, Steinberg AM, Elhai JD, 2015. Hispanic ethnicity and Caucasian race: Relations with posttraumatic stress disorder’s factor structure in clinic-referred youth. Psycholog. Trauma Theor. Res. Pract. Policy 7 (5), 456. [DOI] [PubMed] [Google Scholar]
  9. Elhai JD, Biehn TL, Armour C, Klopper JJ, Frueh BC, Palmieri PA, 2011. Evidence for a unique PTSD construct represented by PTSD’s D1–D3 symptoms. J. Anxiety Disord 25 (3), 340–345. [DOI] [PubMed] [Google Scholar]
  10. Fan X, Sivo SA, 2009. Using Δgoodness-of-fit indexes in assessing mean structure invariance. Struct. Equ. Model 16 (1), 54–69. [Google Scholar]
  11. … and Gaziano JM, Concato J, Brophy M, Fiore L, Pyarajan S, Breeling J, O’Leary TJ, 2016. Million Veteran Program: A mega-biobank to study genetic influences on health and disease. J. Clin. Epidemiol 70, 214–223. [DOI] [PubMed] [Google Scholar]
  12. … and Gelernter J, Sun N, Polimanti R, Pietrzak R, Levey DF, Bryois J, Stein MB, 2019. Genome-wide association study of post-traumatic stress disorder reexperiencing symptoms in>165,000 US veterans. Nat. Neurosci 22 (9), 1394–1401. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Hall BJ, Elhai JD, Grubaugh A, Tuerk P, Magruder K, 2012. Examining the factor structure of PTSD between male and female veterans in primary care. J. Anxiety Disord 26 (3), 409–415. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Harpaz-Rotem I, Tsai J, Pietrzak RH, Hoff R, 2014. The dimensional structure of posttraumatic stress symptomatology in 323,903 US veterans. J. Psychiatr. Res 49, 31–36. [DOI] [PubMed] [Google Scholar]
  15. Hirschfeld G, Von Brachel R, 2014. Improving Multiple-Group confirmatory factor analysis in R-A tutorial in measurement invariance with continuous and ordinal indicators. Pract. Assess. Res. Eval. 19 (1), 7. [Google Scholar]
  16. Hu LT, Bentler PM, 1999. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Struct. Equ. Model 6 (1), 1–55. [Google Scholar]
  17. Kessler RC, Aguilar-Gaxiola S, Alonso J, Benjet C, Bromet EJ, Cardoso G, Degenhardt L, de Girolamo G, Dinolova RV, Ferry F, Florescu S, Koenen KC, 2017. Trauma and PTSD in the WHO world mental health surveys. Eur. J. Psychotraumatol 8 (sup5), 1353383. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. King DW, Leskin GA, King LA, Weathers FW, 1998. Confirmatory factor analysis of the Clinician-Administered PTSD Scale: evidence for the dimensionality of posttraumatic stress disorder. Psychol. Assess 10 (2), 90. [Google Scholar]
  19. Lang AJ, Stein MB, 2005. An abbreviated PTSD checklist for use as a screening instrument in primary care. Behav. Res. Ther 43 (5), 585–594. [DOI] [PubMed] [Google Scholar]
  20. … and Lang AJ, Wilkins K, Roy-Byrne PP, Golinelli D, Chavira D, Sherbourne C, Stein MB, 2012. Abbreviated PTSD Checklist (PCL) as a guide to clinical v response. Gen. Hosp. Psychiatry 34 (4), 332–338. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Lee DJ, Bovin MJ, Weathers FW, Palmieri PA, Schnurr PP, Sloan DM, Keane TM, Marx BP, 2019. Latent factor structure of DSM-5 posttraumatic stress disorder: Evaluation of method variance and construct validity of novel symptom clusters. Psychol. Assess 31 (1), 46. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. … and Moshier SJ, Lee DJ, Bovin MJ, Gauthier G, Zax A, Rosen RC, Marx BP, 2019. An empirical crosswalk for the PTSD checklist: Translating DSM-IV to DSM-5 using a veteran sample. J. Trauma Stress 32 (5), 799–805. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. … and Pathak GA, Singh K, Wendt FR, Fleming TW, Overstreet C, Koller D, Polimanti R, 2022. Genetically regulated multi-omics study for symptom clusters of posttraumatic stress disorder highlights pleiotropy with hematologic and cardiometabolic traits. Mol. Psychiatry 1–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Pietrzak RH, Tsai J, Harpaz-Rotem I, Whealin JM, Southwick SM, 2012. Support for a novel five-factor model of posttraumatic stress symptoms in three independent samples of Iraq/Afghanistan veterans: a confirmatory factor analytic study. J. Psychiatr. Res 46 (3), 317–322. [DOI] [PubMed] [Google Scholar]
  25. Raftery AE, 1995. Bayesian model selection in social research. Sociol. Methodol 25, 111–163. [Google Scholar]
  26. R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/. [Google Scholar]
  27. Rosseel Y, 2012. lavaan: An R package for structural equation modeling. J. Stat. Softw 48, 1–36. [Google Scholar]
  28. Rusch HL, Robinson J, Yun S, Osier ND, Martin C, Brewin CR, Gill JM, 2019. Gene expression differences in PTSD are uniquely related to the intrusion symptom cluster: a transcriptome-wide analysis in military service members. Brain Behav. Immun 80, 904–908. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Satorra A, Bentler PM, 2001. A scaled difference chi-square test statistic for moment structure analysis. Psychometrika 66 (4), 507–514. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Schwarz G, 1978. Estimating the dimension of a model. Ann. Stat 6, 461–464. [Google Scholar]
  31. Scott KM, Koenen KC, Aguilar-Gaxiola S, Alonso J, Angermeyer MC, Benjet C, Bruffaerts R, Caldas-de-Almeida JM, De Girolamo G, Florescu S, Iwata N, Levinson D, Lim C, Murphy S, Ormel J, Posado-Villa J, Kessler RC, 2013. Associations between lifetime traumatic events and subsequent chronic physical conditions: a cross-national, cross-sectional study. PLoS One 8 (11), e80573. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Simms LJ, Watson D, Doebbelling BN, 2002. Confirmatory factor analyses of posttraumatic stress symptoms in deployed and nondeployed veterans of the Gulf War. J. Abnorm. Psychol. 111 (4), 637. [DOI] [PubMed] [Google Scholar]
  33. … and Stein MB, Levey DF, Cheng Z, Wendt FR, Harrington K, Pathak GA, Gelernter J, 2021. Genome-wide association analyses of post-traumatic stress disorder and its symptom subdomains in the Million Veteran Program. Nat. Genet 53 (2), 174–184. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Wang L, Cao X, Cao C, Fang R, Yang H, Elhai JD, 2017. Factor structure of DSM-5 PTSD symptoms in trauma-exposed adolescents: Examining stability across time. J. Anxiety Disord 52, 88–94. [DOI] [PubMed] [Google Scholar]
  35. Weathers FW, Blake DD, Schnurr PP, Kaloupek DG, Marx BP, Keane TM, 2013. The Life Events Checklist for DSM-5 (LEC-5). National Center for PTSD. Instrument available from theat. http://www.ptsd.va.gov. [Google Scholar]
  36. Weathers FW, Huska JA, Keane TM, 1991. PCL-C for DSM-IV. National Center for PTSD-Behavioral Science Division, Boston. [Google Scholar]

RESOURCES