Mild Traumatic Brain Injury and Posttraumatic Stress Disorder and Their Associations With Health Symptoms

Article Outline

• Abstract
• Methods
  • Participants
  • Groups
  • Symptom Complaints
  • Covariate Measures
    • Demographic variables
    • Medical conditions
    • Current psychiatric difficulties
    • Posttraumatic stress disorder before mild traumatic brain injury
  • Data Analytic Strategy
• Results
  • Effects of Mild Traumatic Brain Injury on Current Psychiatric Disorders
  • Effects of Mild Traumatic Brain Injury on Postconcussion Syndrome Symptoms
  • Effects of Mild Traumatic Brain Injury on Nonpostconcussion Syndrome Physical Health Symptoms
  • Moderator Analyses
  • Effects of Mild Traumatic Brain Injury on Recovery From Posttraumatic Stress Disorder
  • Mediator Analyses
• Discussion
  • Mediation or Interactive Effects between Posttraumatic Stress Disorder and Mild Traumatic Brain Injury
  • Current Study Compared With Hoge et al5
  • Postconsussive Syndrome Symptoms, Mild Traumatic Brain Injury, Posttraumatic Stress Disorder, and Treatment Implications
• Conclusions
• References
• Copyright

Abstract 

Vanderploeg RD, Belanger HG, Curtiss G. Mild traumatic brain injury and posttraumatic stress disorder and their associations with health symptoms.

Objective

To determine the association of various symptoms and psychiatric diagnoses with a remote history of mild traumatic brain injury (MTBI) and a current diagnosis of posttraumatic stress disorder (PTSD).

Design

Cross-sectional cohort study.

Setting

Nonclinical.

Participants

Three groups of randomly selected community dwelling male U.S. Army Vietnam-era veterans: healthy control (n=3218), those injured in a motor vehicle collision (MVC) but without a head injury (MVC injury control; n=548), and those who had an MTBI (n=278).

Interventions

None.

Main Outcome Measures

Prevalence of psychiatric diagnoses, physical, cognitive, and emotional symptoms, and course of PTSD across time.

Results

Logistic regression procedures were used to determine group association with symptoms and psychiatric diagnosis after controlling for demographic variables, combat intensity, medical disorders, and other current psychiatric conditions. MTBI was associated with headaches, memory problems, sleep problems, and fainting even after controlling for current psychiatric problems (including PTSD), as well as demographic variables, combat intensity, and comorbid medical conditions. MTBI also was associated with a current diagnosis of PTSD even controlling for other demographic, psychiatric, and medical covariates. MTBI did not moderate or mediate the relationship between PTSD and current symptomatology. However, MTBI did adversely affect long-term recovery from PTSD (odds ratio=1.59, 95% CI, 1.07–2.37). PTSD also was associated with physical, cognitive, and emotional symptoms, and had a larger effect size than MTBI.

Conclusions

MTBI, even in the chronic phase years postinjury, is not a benign condition. It is associated with increased rates of headaches, sleep problems, and memory difficulties. Furthermore, it can complicate or prolong recovery from preexisting or comorbid conditions such as PTSD. Similarly, PTSD is a potent cocontributor to physical, cognitive, and emotional symptoms.

Key Words: Brain concussion, Mental disorders, Neurobehavioral manifestations, Rehabilitation, Social support

List of Abbreviations: CI, confidence interval, DSM-IV, Diagnostic and Statistical Manual of Mental Disorders, 4th Edition, LOC, loss of consciousness, MTBI, mild traumatic brain injury, MVC, motor vehicle collision, OIF, Operation Iraqi Freedom, OR, odds ratio, PCS, postconcussion syndrome, PTSD, posttraumatic stress disorder, TBI, traumatic brain injury

TBI HAS BEEN IDENTIFIED as the signature injury of the Iraq and Afghanistan conflicts. In the current conflicts, the number of survived brain injury cases is second only to orthopedic injuries.¹, ² There is concern that concussion or MTBI may go undiagnosed and untreated because attention is focused on more visible injuries from improvised explosive devices and other blast mechanisms that have increased substantially in modern warfare.³, ⁴ As a result, both the Department of Defense and the Department of Veterans Affairs have implemented new screening measures designed to identify MTBI. In a recent survey study of returning soldiers, Hoge et al⁵ found that approximately 15% reported injuries with LOC and/or altered mental status suggestive of MTBI, most caused by blasts. The RAND Corporation recently reported similar findings; 19.5% reported experiencing a probable MTBI during deployment.⁶

In the civilian sector, the majority of cases with MTBI recover completely within 7 to 30 days, but a minority continue to complain of cognitive difficulties and other distressing symptoms for months⁷, ⁸, ⁹, ¹⁰ or years postinjury.⁹, ¹¹, ¹² Depending on whether DSM-IV¹³ or International Classification of Diseases, 10th Revision symptom criteria are used, the prevalence of persistent symptoms over injury control or population base rates varies across studies from 4%¹⁴ to a high of 18% to 24%.¹², ¹⁴ Frequently, complaints involve a constellation of somatic, emotional, and cognitive symptoms known collectively as PCS. Common symptom complaints include headaches, balance problems, dizziness, fatigue, depression, anxiety, irritability, and memory and attention difficulties, often without demonstrable structural changes to the brain¹⁵ or neuropsychologic dysfunction.⁸, ¹⁶

Risk factors identified in the development and maintenance of PCS symptoms include litigation status,¹⁷ age, educational and intellectual level, female sex, prior head injury, socioeconomic status, history of alcohol abuse, social difficulties, premorbid psychiatric disorders, environmental stress, and multiple trauma.¹⁸, ¹⁹, ²⁰, ²¹ Concomitant psychiatric difficulties also are important moderators of PCS. King²² reported that depression, anxiety, and stress within 1 week of mild to moderate head injury were highly predictive of severity of postconcussion symptoms 3 months postinjury. Furthermore, preexisting psychiatric disturbance and psychosocial problems are often found in symptomatic MTBI patients and tend to complicate recovery from MTBI.²³ Luis et al²⁴ found that limited social support, early life psychiatric difficulties, and lower intelligence predicted persistent postconcussion symptoms an average of 8 years post-MTBI. Finally, pain (including headaches) and sleep problems are significantly associated with postconcussive symptoms.²⁵, ²⁶

The RAND study of post-Iraqi military deployment⁶ (OIF) reported a high rate of co-occurrence between a history of MTBI and PTSD and depression. They found that of those experiencing an MTBI, over a third also had overlapping PTSD or depression. In a recent survey of returning soldiers, Hoge et al⁵ also found that PTSD and MTBI were strongly associated. On examination of multiple potential predictors of PTSD including non–TBI-related injuries, only combat intensity and MTBI with LOC were significantly associated with PTSD once other variables were controlled. These findings suggest that MTBI might increase the likelihood of developing PTSD. In terms of what etiologic conditions account for postdeployment physical and other health symptoms, preliminary analyses found that MTBI with LOC was associated with multiple PCS and other health symptoms. However, after adjusting for PTSD and depression, MTBI with LOC was no longer associated with 18 out of 19 symptoms. The 1 symptom that remained significantly associated with LOC was headaches. The authors conclude that the high rates of physical health problems reported by OIF soldiers with MTBI are mediated largely or entirely by PTSD or depression (ie, psychologic disorders rather than the MTBI). Data regarding timing of the concussion in relation to PTSD were not available; consequently, conclusions about causality were not possible. Similarly, because only information about current comorbidities were available, statistical evaluation of hypothesized mediational models was not possible.

These findings and conclusions in an OIF sample stand in contrast to a similar cross-sectional cohort study by Vanderploeg et al¹² who found that MTBI was a significant predictor of persistent postconcussive symptoms and of a number of specific physical (headaches and sleep problems) and cognitive (memory problems) postconcussive symptoms many years after MTBI in a sample of Vietnam-era veterans, even after controlling for demographic, psychiatric, and medical comorbidity variables. This study differed from Hoge et al⁵ in military era (OIF vs Vietnam) and in that it controlled for psychiatric conditions that predated the MTBI, reasoning that current psychiatric conditions could be an adverse outcome associated with MTBI. Consistent with this hypothesis was that current major depression was significantly elevated in both an MTBI group and an other-injury control group compared with a healthy control group, even after controlling for preinjury psychiatric conditions. In contrast, Hoge⁵ controlled for current depression and PTSD, thereby confounding potentially important health outcomes with covariate variables.

One goal of the current study was to determine if discrepant findings between Vanderploeg et al¹² and Hoge et al⁵ were caused by differences in methodology (ie, use of preinjury vs current psychiatric conditions as covariate controls and use of healthy control group vs non-TBI injury control group for OR comparisons, respectively). In the current study, we reanalyzed the Vanderploeg et al¹² data using current rather than preinjury psychiatric variables as covariates and “other” injury control group for OR comparisons. We reasoned that if the symptom outcomes previously associated with MTBI were no longer significant, this would suggest that current psychiatric conditions share important variance with MTBI that the statistical analyses allotted to PTSD rather than MTBI. We also controlled for other potentially potent covariates (eg, alcohol and drug abuse/dependence, premorbid intellectual functioning, level of education, race, presence of medical diseases).

Additional goals were to examine potential moderator effects between MTBI and mental health conditions (PTSD and/or depression) and PCS symptoms; that is, do co-occurring PTSD and/or depression interact with MTBI in influencing the frequency of PCS symptoms? Similarly, we examined whether MTBI had an adverse impact on the course or resolution of preinjury PTSD. Although our data did not allow examination of the potential mediator effect of PTSD on the relationship between prior MTBI and PCS symptoms, an additional goal was to examine a potential mediator effect of MTBI on the relationship between prior PTSD and PCS. Finally, we considered the relative contribution of comorbid PTSD to the symptom presentation.

Back to Article Outline

Methods 

Participants 

The data used in the current investigation were those of the Vietnam Experience Study.²⁷, ²⁸ The data were collected in the mid-1980s as part of a study investigating the effects of the Vietnam experience on veterans. The total sample consisted of 4462 male U.S. Army veterans randomly selected from those who served during the Vietnam era and who had (1) first entered the military between January 1965 and December 1971, (2) been on active duty for at least 4 months, (3) served only 1 tour of duty, (4) obtained a military occupational specialty, and (5) achieved a pay grade no higher than E-5 (sergeant) on discharge. Approximately half the sample had served in Vietnam, whereas the other half had served elsewhere (ie, United States, Korea, Germany).

Participants were flown from their city of residence to a city in the southwestern United States for a 3-day comprehensive evaluation involving extensive medical, psychologic, and neuropsychologic examinations. Details of the evaluation are published elsewhere.²⁷, ²⁸, ²⁹ Psychologic evaluation involved a semistructured psychiatric interview and a self-report measure of psychologic status. A medical history questionnaire was administered to gather information regarding health-related events that may have occurred during the time interval from military discharge to study date, which was approximately 16 years postmilitary discharge. Questions pertaining to current physical (eg, headaches, dizziness, fainting), emotional (eg, depression, anxiety), and cognitive (eg, memory and concentration problems) symptoms along with past history of various medical disorders were surveyed.

During the interview, participants were asked the following 3 questions: (1) Since your discharge from active duty, have you been injured in an MVC? (2) Since your discharge from active duty, have you injured your head? and (3) Did you lose consciousness as a result of the head injury? If participants were unclear if they had lost consciousness, they were asked if they had “blacked out” in the accident. Thus, this question captured both posttraumatic amnesia (ie, memory gaps) and LOC. Although individuals may have seen a physician or gone to the emergency department of a hospital, those individuals who required hospitalization after their head injury (n=40) were excluded from the present analyses in order to capture only those with milder injuries.

Groups 

Based on these questions, each participant was categorized into 1 of 3 groups: (1) those who had not been injured in an MVC nor had a head injury (healthy control; n=3218; 72.1%), (2) those who had been injured in an MVC but did not have a head injury (MVC injury control; n=548; 12.8%), and (3) those who had a head injury with altered consciousness, that is, MTBI (n=278; 6.2%). Those who could not be categorized into these 3 groups were excluded from further analyses (8.9%).

The 2 groups of primary interest were the MVC injury control and the MTBI groups, with the MVC control group serving as an injury control reference group for MTBI comparative analyses. Demographic characteristics of these 2 groups can be found in table 1. As can be seen, the MTBI group was significantly more likely to be of lower military rank and unmarried.

Table 1. Demographic Characteristics of the Sample by Group

Characteristic	MTBI	MVC Injury	P
	(n=278)	(n=548)
	n (%)	n (%)
Age, <40y	217(78)	402(73)	.080
Racial minority (not white)	45(16)	114(21)	.070
Marital status			.008
Single	38(14)	49(9)
Married	165(59)	383(70)
Divorced, separated	75(27)	116(21)
High school education or less	146(53)	288(53)	.530
Current IQ estimate			.310
<100	88(32)	143(26)
101–115	61(22)	128(23)
116–130	89(32)	179(33)
>130	40(14)	98(18)
Rank
≤E4	201(72)	354(65)	.030
Quartile of combat intensity
1	151(54)	292(53)	.750
2	32(12)	78(14)
3	48(17)	89(16)
4	47(17)	89(16)

NOTE. Approximately half of the sample did not serve in Vietnam and therefore had no combat exposure. MVC injury refers to those injured in MVC with no TBI.

Abbreviation: IQ, intelligence quotient.

Symptom Complaints 

Our objective was to examine current symptom complaints similar to the Hoge et al study⁵ in veterans who either did or did not have an MTBI. We examined the same symptoms as Hoge (eg, headaches, fainting, rapid heart beating, chest pain) except we did not have data on pain during sexual intercourse or balance problems. Unlike Hoge, our back pain variable was exclusively lower back pain, rather than general back pain. Also, our leg pain variable was specific to pain in the calf region of the leg.

The symptoms were gleaned from a medical history interview and questionnaire. To illustrate, the presence of headache was determined by a “yes” response to “During the past year have you had unusually frequent or severe headaches?” from the medical history questionnaire. More detailed descriptions of these variables have been published elsewhere.²⁷, ²⁸, ²⁹

For purposes of analyzing summary PCS symptom clusters, individual symptoms associated with PCS were categorized into somatic (headaches, sleep problems, dizziness, and fatigue), cognitive (memory and concentration problems), and emotional (irritability, depression/sadness, and anxiety) if 1 or more of those symptoms were present. We also examined to see if participants met the DSM-IV criteria for PCS (ie, 3 of more of the following symptom areas: fatigue; sleep problems; headaches; dizziness; irritability; and either anxiety, depression, or mood lability). Participants were also asked to indicate whether overall they considered their health status to be good versus poor.

Covariate Measures 

Variables associated with the presence of physical, cognitive, or emotional complaints in previous research studies were selected from the Vietnam Experience Study dataset. Three domains were represented: (1) demographic factors, (2) medical conditions, and (3) psychiatric conditions present within the past month.

Demographic variables, as presented in table 1, included age, rank, combat intensity, level of education, race, current level of intellectual ability, and marital status. Combat intensity was computed by summing across 12 items of combat experiences rated on a 5-point scale from a score of 0 (never) to 4 (very often). The level of intellectual ability was defined as performance on the General Technical Test³⁰ at the time of evaluation. The General Technical Test is a verbal/arithmetic aptitude test administered at enlistment in the Army and again at the time of data collection (the latter variable was used for the current covariate analyses). Results are reported in the same metric as a standard intelligence quotient score with a mean ± SD of 100±15. The General Technical Test has been shown to be a measure of general intellectual ability.³⁰ For statistical analyses, variables were coded by category as shown in table 1.

Responses on the medical history questionnaire were used to determine prior or current medical conditions. Specifically, the presence or absence of the following medical problems was examined: history of meningitis, malaria, diabetes, hypertension, angina, myocardial infarction, heart murmur, and peripheral vascular disease.

Responses on the psychiatric Diagnostic Interview Schedule, Version III-A³¹ were used to determine the presence of current psychiatric difficulties. The Diagnostic Interview Schedule, Version III-A is a standardized questionnaire used to assess the prevalence of psychiatric disorders in the general population. The instrument consists of modules that are grouped together in diagnostic categories (eg, anxiety, mood, substance use). Diagnoses on this version of the instrument were made according to Diagnostic and Statistical Manual of Mental Disorders, 3rd Edition,³² the current nomenclature at the time. Inclusion of a graphic timeline and redundant questions regarding age of onset and duration within each Diagnostic Interview Schedule, Version III-A criterion scale assisted recall of whether a symptom or disorder began within or before the asked-for timeframe. To be classified as having current psychiatric problems in the present study, participants had to endorse the pertinent criteria within the last month. Unlike the other psychiatric disorders, we could only examine somatization disorder within the past year (rather than the past month). Within the 2 groups compared in the current study, 7.7% had current PTSD, 9.6% had major depressive disorder, 13.9% reported alcohol abuse/dependence, 8.0% had drug abuse/dependence, and 3.0% had somatization disorder. The presence or absence of current problems was used in regression analyses.

To examine potential mediation effects of MTBI on the relationship between preexisting PTSD and PCS, a pre-MTBI PTSD variable was derived. Participants had to meet criteria for PTSD within 1 year of discharge from the military and continue to experience PTSD up to and through the time the MTBI occurred. Because the data did not specify the date of the MTBI, other than it occurred sometime between military discharge and the time of data collection (a 16-year window), PTSD symptoms had to continue to some degree up through the month before data collection. We also calculated the rate of resolution of PTSD from the year after military discharge to the present (again, a 16-year window of time).

Data Analytic Strategy 

By using the MVC injury control group as the reference group, logistic regression analyses were used to model ORs and 95% CIs for the association between group membership and outcome variables while ruling out spurious relationships attributable to confounding variables. Three models were used: (1) a model with the symptom of interest (eg, headaches) as the dependent measure and group membership as the independent measure with no covariates entered, (2) a model with Hoge et al⁵–similar covariate control variables entered before the symptom of interest, and (3) a model that included the Hoge–similar covariates and added additional covariate control variables.

Hoge et al⁵ controlled for age, military rank, sex, hospitalization, mechanism of injury (blast or other mechanism), level of combat intensity, current PTSD, and current depression. We used the same control variables as Hoge for model 2 (ie, age, rank, combat intensity). However, we did not use mechanism of injury, injury-related hospitalization, or sex as data because mechanism of injury was not available and our entire sample was male. In addition, we did not have data on whether or not hospitalization was required for injuries sustained in the MVC (ie, the control reference group). We categorized our covariate control variables to mimic those used by Hoge (see table 1 for classification categories).

The final model (model 3) included additional covariate control variables. We added level of education, current level of intellectual ability, race, marital status, history or presence of select medical conditions, and other current psychiatric disorders (ie, alcohol abuse/dependence, drug abuse/dependence, somatization) in addition to PTSD and depression. This allowed us to determine whether, after controlling for other variables, MTBI and various outcome measures were associated.

Moderator effects between MTBI and mental health conditions (PTSD and/or depression) and PCS symptoms were examined by including possible 2- and 3-way interactions between group (MVC injury control vs MTBI) and PTSD; group and depression; and the 3-way interaction between group, PTSD, and depression. For these moderator analyses, variables were entered in blocks, with the main effects entered first followed by the 2-way and then 3-way interactions.

Similarly, we examined whether MTBI had an adverse impact on the course or resolution of pre-MTBI PTSD. Logistic regression analyses were used to examine the likelihood that persons with PTSD in the year after their military discharge who subsequently sustained an MTBI were more likely to still meet diagnostic criteria for PTSD an average of 16 years later.

Finally, we sought to examine the potentially mediating effects of MTBI on the relationship between PTSD and current symptoms. We tested for a mediator effect by using the Sobel test.³³ In addition to testing whether the indirect effect is different from 0, this test determines the significance of the indirect effect of the mediator (MTBI) by testing the hypothesis of no difference between the total effect and the direct effect (PTSD predicting PCS) and also if the indirect effect is in the predicted direction. Because the Sobel test assumes a normal sampling distribution, bootstrapping was used to construct 3000 random resamples to approximate the sampling distribution of the average Sobel z value.

Back to Article Outline

Results 

Effects of Mild Traumatic Brain Injury on Current Psychiatric Disorders 

We examined the relationship between MTBI and current psychiatric conditions (table 2). Similar to Hoge et al,⁵ PTSD was significantly related to MTBI. Overall, 11.5% of those who had sustained MTBI met criteria for current PTSD as compared with 5.8% of MVC injury controls (χ²_1,n=826=8.30, P<.01). Significant group differences remained after including Hoge–similar covariate controls (OR=1.47; 95% CI, 1.11–1.96). This remained the case even after additional covariates were controlled, which included the demographic variables, medical disorders, and all other current psychiatric conditions (ie, alcohol/drug dependence, depression, and somatization) (OR=1.41; 95% CI, 1.05–1.91).

Table 2. ORs and 95% CIs for Current Psychiatric Disorders by Group

Symptom	Variables Controlled	MVC Injury (n=548)	%	MTBI (n=278)	%
PTSD
Model 1	None	1.0	5.8	1.45⁎(1.12–1.87)	11.5
Model 2	Demographic, depression	1.0		1.47⁎(1.11–1.96)
Model 3	Other psychiatric and medical disease	1.0		1.41⁎(1.05–1.91)
Major depression
Model 1	None	1.0	7.8	1.32⁎(1.05–1.67)	12.9
Model 2	Demographic, PTSD	1.0		1.22(0.95–1.56)
Model 3	Other psychiatric and medical disease	1.0		1.16(0.89–1.51)
Alcohol abuse/dependence
Model 1	None	1.0	12.0	1.25⁎(1.02–1.53)	17.6
Model 2	Demographic, PTSD, depression	1.0		1.20(0.97–1.47)
Model 3	Other psychiatric and medical disease	1.0		1.20(0.97–1.48)
Drug abuse/dependence
Model 1	None	1.0	7.5	1.10(0.85–1.43)	9.0
Model 2	Demographic, PTSD, depression	1.0		1.05(0.81–1.37)
Model 3	Other psychiatric and medical disease	1.0		1.06(0.80–1.39)
Somatization disorder
Model 1	None	1.0	2.7	1.15(0.77–1.73)	3.6
Model 2	Demographic, PTSD, depression	1.0		1.00(0.64–1.56)
Model 3	Other psychiatric and medical disease	1.0		0.97(0.61–1.57)

NOTE. The MVC-injured control group served as the reference group. ORs are first presented without controlling any potential confounds (model 1). Next, they are presented adjusting for the demographic factors of age, rank, combat intensity, and PTSD and depression, unless either of these was the dependent measure (model 2). Finally, in model 3, ORs are presented adjusting for all demographic factors (age, rank, combat intensity, education, current intelligence quotient, estimate, race, marital status), presence of other current psychiatric disorder (ie, PTSD, depression, alcohol abuse/dependence, drug abuse/dependence, somatization disorder), and presence of medical disease (see text for list of conditions included). Values in brackets represent the frequency for that condition.

Both depression and alcohol abuse/dependence were significantly associated with group but not once covariate controls were included in the model. Drug abuse/dependence and somatization disorder were not associated with MTBI (see table 2 for a summary of these analyses).

Effects of Mild Traumatic Brain Injury on Postconcussion Syndrome Symptoms 

The percentage of participants who endorsed each PCS symptom is shown in table 3 and is broken down by group. In terms of individual symptoms, the unadjusted analyses revealed significant relationships between several PCS symptoms and MTBI including memory problems, headaches, sleep problems, and irritability. After adjustment for variables similar to those used in the Hoge et al study,⁵ these relationships remained with the exception of irritability. After further adjusting for the presence of other current psychiatric disorders and medical disease, memory problems (OR=1.31; 95% CI, 1.09–1.58), headaches (OR=1.30; 95% CI, 1.09–1.57), and sleep problems (OR=1.21; 95% CI, 1.02–1.43) remained significantly related to MTBI.

Table 3. ORs and 95% CIs for Postconcussion Symptom Complaints Across Groups

Symptom	Variables Controlled	MVC Injury (n=548)	%	MTBI (n=278)	%
PCS symptoms
Memory problems
Model 1	None	1.0	17.2	1.40⁎(1.18–1.66)	28.9
Model 2	Demographic, PTSD, depression	1.0		1.33⁎(1.11–1.59)
Model 3	Other psychiatric and medical disease	1.0		1.31⁎(1.09–1.58)
Headaches
Model 1	None	1.0	16.5	1.37⁎(1.15–1.63)	27.0
Model 2	Demographic, PTSD, depression	1.0		1.31⁎(1.09–1.56)
Model 3	Other psychiatric and medical disease	1.0		1.30⁎(1.09–1.57)
Sleep problems
Model 1	None	1.0	32.4	1.32⁎(1.13–1.53)	45.3
Model 2	Demographic, PTSD, depression	1.0		1.24⁎(1.05–1.45)
Model 3	Other psychiatric and medical disease	1.0		1.21⁎(1.02–1.43)
Irritability
Model 1	None	1.0	31.8	1.20⁎(1.04–1.40)	40.2
Model 2	Demographic, PTSD, depression	1.0		1.16(0.99–1.36)
Model 3	Other psychiatric and medical disease	1.0		1.16(0.99–1.37)
Dizziness
Model 1	None	1.0	5.7	1.31(1.00–1.72)	9.4
Model 2	Demographic, PTSD, depression	1.0		1.18(0.89–1.57)
Model 3	Other psychiatric and medical disease	1.0		1.16(0.86–1.55)
Fatigue
Model 1	None	1.0	32.7	1.14(0.99–1.33)	38.8
Model 2	Demographic, PTSD, depression	1.0		1.11(0.95–1.30)
Model 3	Other psychiatric and medical disease	1.0		1.09(0.92–1.28)
Concentration problems
Model 1	None	1.0	19.9	1.11(0.93–1.32)	23.4
Model 2	Demographic, PTSD, depression	1.0		1.02(0.84–1.23)
Model 3	Other psychiatric and medical disease	1.0		0.99(0.81–1.20)
Depression/sadness
Model 1	None	1.0	15.3	1.05(0.86–1.27)	16.5
Model 2	Demographic, PTSD, depression	1.0		0.92(0.74–1.15)
Model 3	Other psychiatric and medical disease	1.0		0.89(0.71–1.12)
Anxiety
Model 1	None	1.0	18.4	1.02(0.85–1.23)	19.1
Model 2	Demographic, PTSD, depression	1.0		0.92(0.75–1.13)
Model 3	Other psychiatric and medical disease	1.0		0.88(0.71–1.09)
PCS symptoms clusters
Somatic PCS symptoms
Model 1	None	1.0	53.1	1.43⁎(1.22–1.67)	69.8
Model 2	Demographic, PTSD, depression	1.0		1.36⁎(1.16–1.60)
Model 3	Other psychiatric and medical disease	1.0		1.34⁎(1.13–1.58)
Cognitive PCS symptoms
Model 1	None	1.0	25.4	1.23⁎(1.05–1.44)	33.9
Model 2	Demographic, PTSD, depression	1.0		1.15(0.97–1.36)
Model 3	Other psychiatric and medical disease	1.0		1.13(0.95–1.35)
Emotional PCS symptoms
Model 1	None	1.0	44.2	1.13(0.98–1.30)	50.2
Model 2	Demographic, PTSD, depression	1.0		1.06(0.90–1.23)
Model 3	Other psychiatric and medical disease	1.0		1.04(0.88–1.22)
DSM PCS symptom complex
Model 1	None	1.0	25.2	1.45⁎(1.24–1.68)	41.4
Model 2	Demographic, PTSD, depression	1.0		1.22(0.99–1.49)
Model 3	Other psychiatric and medical disease	1.0		1.37⁎(1.15–1.63)

NOTE. The MVC-injured control group served as the reference group. ORs are first presented without controlling any potential confounds (model 1). Next, they are presented adjusting for the demographic factors of age, rank, combat intensity, and PTSD and depression, unless either of these was the dependent measure (model 2). Finally, in model 3, ORs are presented adjusting for all demographic factors (age, rank, combat intensity, education, current intelligence quotient, estimate, race, marital status), presence of other current psychiatric disorder (ie, PTSD, depression, alcohol abuse/dependence, drug abuse/dependence, somatization disorder), and presence of medical disease (see text for list of conditions included). Values in brackets represent the frequency for that condition.

When looking at the presence or absence of symptoms according to conceptual groupings of PCS symptoms, it was found that MTBI was associated with somatic and cognitive symptoms but not emotional symptoms. After adjusting for demographic, psychiatric, and medical disorders, however, only somatic symptoms were significantly associated with MTBI (OR=1.34; 95% CI, 1.13–1.58). By using the DSM-IV symptom-based diagnostic criteria for PCS, MTBI was also associated with the full symptom complex even after including all covariate controls (OR=1.37; 95% CI, 1.15–1.63).

Effects of Mild Traumatic Brain Injury on Nonpostconcussion Syndrome Physical Health Symptoms 

Table 4 shows the association between various non-PCS physical symptoms and group. The MTBI group was no more likely to complain of poor health overall (χ²_1,n=826=1.72, P=.19). In terms of individual non-PCS health symptoms, the unadjusted analyses revealed significant relationships between MTBI and fainting, calf pain, and rapid heart beat. After adjustment for variables similar to those used in the Hoge et al study,⁵ however, the MTBI group differed significantly from the MVC injury control group only on fainting. Fainting remained a significant group difference even after adjusting for additional demographic, psychiatric, and medical disorders (OR=1.80; 95% CI, 1.19–2.73).

Table 4. ORs and 95% CIs for Non-PCS Physical Symptom Complaints Across Groups

Symptom	Variables Controlled	MVC Injury (n=548)	%	MTBI (n=278)	%
Non-PCS physical symptoms
Poor health
Model 1	None	1.0	17.5	1.13(0.94–1.35)	21.3
Model 2	Demographic, PTSD, depression	1.0		1.06(0.88–1.28)
Model 3	Other psychiatric and medical disease	1.0		1.04(0.84–1.27)
Fainting
Model 1	None	1.0	2.0	2.05⁎(1.41–2.96)	7.9
Model 2	Demographic, PTSD, depression	1.0		1.92⁎(1.31–2.81)
Model 3	Other psychiatric and medical disease	1.0		1.80⁎(1.19–2.73)
Calf pain
Model 1	None	1.0	1.8	1.57⁎(1.02–2.40)	4.4
Model 2	Demographic, PTSD, depression	1.0		1.51(0.97–2.34)
Model 3	Other psychiatric and medical disease	1.0		1.48(0.94–2.33)
Rapid heart beating
Model 1	None	1.0	12.8	1.23⁎(1.01–1.49)	18.1
Model 2	Demographic, PTSD, depression	1.0		1.16(0.95–1.43)
Model 3	Other psychiatric and medical disease	1.0		1.14(0.92–1.41)
Shortness of breath
Model 1	None	1.0	6.6	1.24(0.96–1.61)	9.7
Model 2	Demographic, PTSD, depression	1.0		1.15(0.88–1.51)
Model 3	Other psychiatric and medical disease	1.0		1.10(0.83–1.46)
Nausea
Model 1	None	1.0	23.3	1.21(0.85–1.70)	30.7
Model 2	Demographic, PTSD, depression	1.0		1.21(0.84–1.73)
Model 3	Other psychiatric and medical disease	1.0		1.23(0.84–1.81)
Low back pain
Model 1	None	1.0	30.1	1.15(0.99–1.34)	36.3
Model 2	Demographic, PTSD, depression	1.0		1.11(0.95–1.29)
Model 3	Other psychiatric and medical disease	1.0		1.09(0.93–1.28)
Loose stool
Model 1	None	1.0	10.9	1.14(0.92–1.42)	13.8
Model 2	Demographic, PTSD, depression	1.0		1.07(0.86–1.35)
Model 3	Other psychiatric and medical disease	1.0		1.04(0.83–1.32)
Ringing in the ears
Model 1	None	1.0	10.2	1.09(0.87–1.37)	11.9
Model 2	Demographic, PTSD, depression	1.0		0.96(0.75–1.22)
Model 3	Other psychiatric and medical disease	1.0		0.89(0.69–1.15)
Chest pain
Model 1	None	1.0	7.7	1.07(0.82–1.39)
Model 2	Demographic, PTSD, depression	1.0		0.96(0.73–1.27)
Model 3	Other psychiatric and medical disease	1.0		0.89(0.67–1.20)
Abdominal pain
Model 1	None	1.0	11.5	0.93(0.74–1.18)	10.1
Model 2	Demographic, PTSD, depression	1.0		0.87(0.68–1.12)
Model 3	Other psychiatric and medical disease	1.0		0.85(0.66–1.10)

NOTE. The MVC-injured control group served as the reference group. ORs are first presented without controlling any potential confounds (model 1). Next, they are presented adjusting for the demographic factors of age, rank, combat intensity, and PTSD and depression, unless either of these was the dependent measure (model 2). Finally, in model 3, ORs are presented adjusting for all demographic factors (age, rank, combat intensity, education, current intelligence quotient, estimate, race, marital status), presence of other current psychiatric disorder (ie, PTSD, depression, alcohol abuse/dependence, drug abuse/dependence, somatization disorder), and presence of medical disease (see text for list of conditions included). Values in brackets represent the frequency for that condition.

Moderator Analyses 

The logistic regression analyses revealed no significant 2- or 3-way interactions between group and PTSD; group and depression; or group, PTSD, and depression on PCS symptoms. Figure 1 graphically displays the lack of any interactions or moderator effects for group and PTSD. The effects of PTSD and MTBI on PCS symptoms are simply additive. As reported earlier, MTBI increased the likelihood of having somatic PCS symptoms as well as the overall PCS symptom complex. However, MTBI did not significantly increase the likelihood of having the set of cognitive or emotional PCS symptoms. In contrast, PTSD increased the likelihood of having any type of PCS symptom set (somatic: χ²_1,n=826=24.03, P<.001; cognitive: χ²_1,n=826=29.94, P<.001; emotional: χ²_1,n=826=18.43, P<.001) as well as the entire PCS symptom complex (χ²_1,n=826=39.33, P<.001). As shown in figure 1, the effects of PTSD on symptom presentation were consistently and substantially larger than the effects of MTBI, by about 2.2-fold for somatic symptoms and up to 7.5-fold for emotional symptoms.

• 
  • View Large Image
  • Download to PowerPoint

• Fig 1. 

  The lack of moderator effects of PTSD and MTBI on PCS symptoms and for the DSM-IV symptom clusters for PCS.

Effects of Mild Traumatic Brain Injury on Recovery From Posttraumatic Stress Disorder 

A related question is in persons with PTSD, does having a subsequent MTBI adversely influence recovery from PTSD? As seen in table 5, MVC injury control and MTBI groups were comparable in the rate of PTSD in the year after military discharge, 18.6% and 21.9%, respectively (χ²_1,n=826=1.29, P=.26). Rates were significantly higher in those who served in Vietnam (29.1%) than in those who had not served in Vietnam (6.2%; χ²_1,n=826=66.03, P<.001). However, over the subsequent 16 years, 68.6% of the MVC injury control group no longer met criteria for PTSD. In contrast, in the MTBI group, only 47.5% resolved. Logistic regression analyses revealed that the OR for continuing to have PTSD 16 years after the initial diagnosis, even adjusting for all other current psychiatric covariates, was significantly higher in the MTBI group compared with the MVC injury control (OR=1.71; 95% CI, 1.15–2.53).

Table 5. Percentage of Subjects With PTSD Across Time by Group

Group	PTSD in First Year After Military Discharge	PTSD Past Year	PTSD Past Month	Whose PTSD Resolved
Uninjured controls (n=3218)	13.1	5.5	4.7	64.4
Vietnam service (n=1752)	21.2	8.8	7.5	64.8
No Vietnam service (n=1466)	3.3	1.6	1.3	61.2
MVC injury controls (n=548)	18.6	7.5	5.8	68.6
Vietnam service (n=333)	27.6	12.0	9.6	65.2
No Vietnam service (n=215)	4.7	0.5	0.0	100.0
MTBI (n=278)	21.9	12.9	11.5	47.5
Vietnam service (n=155)	32.3	20.0	17.4	46.0
No Vietnam service (n=123)	8.9	4.1	4.1	54.5

NOTE. Values are in percents. The timeframe represents an average of 16 years from the time of military discharge post-Vietnam era service to the time data were collected.

Mediator Analyses 

The mediator analysis did not reveal any mediating effect of MTBI on the relationship between PTSD and somatic, cognitive, or emotional PCS symptoms nor did it mediate the relationship between PTSD and full DSM symptom criteria for PCS. Figure 2 shows this relationship for the DSM PCS symptom criteria.

• 
  • View Large Image
  • Download to PowerPoint

• Fig 2. 

  No mediation effect of MTBI on the relationship between PTSD and the full DSM symptom criteria for PCS. NOTE: Primary pathways are correlation coefficients. The value in parentheses represents the relationship between PTSD and Diagnostic and Statistical Manual of Mental Disorders, 3rd Edition symptom criteria for PCS with MTBI as a mediator. The Sobel z value (1.22) was not significant, P>.05.

Back to Article Outline

Discussion 

The current study showed that a number of somatic PCS symptoms and memory problems are associated with a history of MTBI. This was true even after controlling for demographic factors, medical disease, and current psychiatric problems. Self-reported fainting was the only non-PCS physical health symptom that was associated with MTBI. These findings are consistent with previous analyses of these data¹² in which psychiatric conditions present before sustaining an MTBI, rather than current comorbid psychiatric disorders, were controlled. The same set of symptoms was again identified as significantly higher after an MTBI than after other types of bodily injury, even though the control criteria were more stringent (ie, current comorbid mental health conditions rather than mental health conditions that predated the MTBI). The current findings stand in partial contrast to the Hoge et al study⁵ in which only headache was associated with MTBI once demographic and current psychiatric conditions were controlled.

In the current study, military deployment potentially resulting in trauma exposures and PTSD occurred before a subsequent civilian-related MTBI, in many cases years before. Under these circumstances, a significant relationship between these 2 health conditions would not necessarily be expected. However, these 2 variables, PTSD and subsequent civilian-related MTBI, were significantly related, even after controlling for demographic, combat exposure, and comorbid medical and psychiatric conditions. Similarly, Hoge et al⁵ investigated the relationship between PTSD and MTBI in soldiers who sustained their MTBI during the same deployment that provided trauma exposures resulting in PTSD. Symptom status and PTSD issues were assessed shortly after this deployment. Under these circumstances, a relationship between these 2 conditions might be expected and in fact was found. The interpretation offered by Hoge et al was that the MTBI reflects “exposure to a very intense traumatic event that threatens loss of life and significantly increases the risk of PTSD”^5(p461) and that the high rates of physical health problems reported by soldiers with MTBI after deployment were mediated by PTSD or depression. Such an interpretation of the results of the present study cannot apply, given that PTSD preexisted the MTBI. Rather, the results of the present study suggest that in persons with PTSD subsequently sustaining an MTBI adversely impacts potential recovery from their PTSD.

Mediation or Interactive Effects between Posttraumatic Stress Disorder and Mild Traumatic Brain Injury 

The present study was also unable to test the mediational model proposed by Hoge et al⁵ but was able to test whether MTBI potentially mediates the presentation of PCS and other symptoms in persons with PTSD. No mediational effects were found. Instead, PTSD and MTBI had independent additive contributions to the presentation of somatic PCS symptoms and memory complaints, with PTSD having a substantially larger effect on current symptom presentation.

However, the timing of the PTSD and subsequent MTBI permitted an investigation of any potential adverse effects of MTBI on recovery from PTSD. Having an MTBI resulted in significantly fewer persons recovering from PTSD over time; OR for continued PTSD was 1.71 in the MTBI group compared with the MVC injury control group. These findings indicate that MTBI, even in the chronic phase years after injury, is not a benign condition. It is associated with increased rates of headaches, sleep problems, and memory difficulties. Furthermore, it can complicate or prolong recovery from preexisting or comorbid conditions such as PTSD.

Current Study Compared With Hoge et al⁵ 

There are many important differences between the current study and the Hoge et al study⁵ that should be taken into account when attempting to understand these results. Hoge et al separated their overall MTBI sample into 2 groups: those with an LOC (n=124) versus those reporting being only dazed or not recalling the injury (n=260). Their 1 positive finding of headache was associated only with LOC. Our MTBI sample was similar to their LOC subgroup and substantially larger (n=278). Therefore, the current study had greater power to detect group differences. So, for instance, although we found a significant association between MTBI and various symptoms while controlling for similar variables, the ORs were often similar and in many cases smaller in the current study. This suggests that the Hoge study may have been underpowered. Also, we controlled for additional extraneous variables by including more covariate control variables. However, this typically decreased the association between MTBI and current symptoms rather than enhancing these relationships, making the current findings a conservative estimate.

Although Hoge et al⁵ did not report the average time since injury of their sample, their sample was evaluated 3 or 4 months after returning from deployment and, therefore, their subjects were likely to be more recently injured and acutely PTSD symptomatic. The current sample was an average of 8 years post-MTBI and 16 years after the initial diagnosis of PTSD. These study differences may account for the discrepancy in the prevalence of current comorbid PTSD (43.9% of sample with MTBI in the Hoge study vs 11.5% of current MTBI sample, 17.4% for those with MTBI who served in Vietnam). A strength of the current study was the use of a structured formal diagnostic interview to define PTSD, whereas Hoge used a cut score from a PTSD checklist. Unlike Hoge, this study did not find an association between MTBI and depression. This discrepancy may represent sample-specific differences in that acute MTBI and depression may be more likely to co-occur.

Postconsussive Syndrome Symptoms, Mild Traumatic Brain Injury, Posttraumatic Stress Disorder, and Treatment Implications 

It is important to note that MTBI is only 1 factor in predicting the presence of PCS symptoms. In this study and that of Hoge et al,⁵ PTSD was clearly associated with multiple physical symptoms including those of PCS. Similarly in a previous study,²² we showed multiple factors, including demographic, psychiatric, and social support variables and their interactions, had significant contributions to ongoing PCS symptoms in persons with MTBI. In both the current study and that of Hoge, PTSD is a more potent predictor of physical, cognitive, and emotional symptoms than is MTBI. In fact, PTSD and MTBI have independent and additive roles.

Back to Article Outline

Conclusions 

These findings suggest the need to intervene early after military postdeployment with emotional and social adjustment interventions, including the development of stress management and relaxation skills, improved sleep hygiene, and education regarding substance use/abuse and alternative coping supports. Such early adjustment, stress management, and symptom-based interventions may minimize the development or prolongation of PTSD and additionally may serve to reduce residual symptoms associated with MTBI. Current findings also suggest that PTSD treatment likely should be prioritized after combat or other types of injury, regardless of MTBI status, to decrease symptom complaints and enhance outcomes. However, the current findings also indicate that in persons with PTSD who subsequently sustain an MTBI there may be a more complicated and extended recovery from PTSD. There are no empirically validated therapies to treat comorbid PTSD and MTBI postconcussive disorders, which may be confounded by alcohol misuse, abuse, or dependence. Combined approaches in which both mental health/PTSD providers and TBI rehabilitation professionals cotreat the variety of emotional/behavioral and cognitive health consequences may be needed.

Back to Article Outline

References 

1. Kraus JF, Nourjah P. The epidemiology of mild, uncomplicated brain injury. J Trauma. 1988;28:1637–1643
   • View In Article
   • MEDLINE
2. Okie S. Traumatic brain injury in the war zone. N Engl J Med. 2005;352:2043–2047
   • View In Article
   • CrossRef
3. Scott BA, Fletcher JR, Pulliam MW, Harris RD. The Beirut terrorist bombing. Neurosurgery. 1986;18:107–110
   • View In Article
   • MEDLINE
   • CrossRef
4. Scott SG, Vanderploeg RD, Belanger HG, Scholten J. Blast injuries: evaluating and treating the postacute sequelae. Fed Practitioner. 2005;22:67–75
   • View In Article
5. Hoge CW, McGurk D, Thomas JL, Cox AL, Engel CC, Castro CA. Mild traumatic brain injury in U.S. soldiers returning from Iraq. N Engl J Med. 2008;358:453–463
   • View In Article
   • CrossRef
6. Invisible wounds of war: psychological and cognitive injuries, their consequences, and services to assist recovery. In:  Tanielian T,  Jaycox LH editor. Santa Monica: RAND Corp; 2008;
   • View In Article
7. Alves W, Macciocchi S, Barth J. Postconcussive symptoms after uncomplicated mild head injury. J Head Trauma Rehabil. 1993;8:48–59
   • View In Article
   • CrossRef
8. Dikmen S, McLean A, Temkin N. Neuropsychological and psychosocial consequences of minor head injury. J Neurol Neurosurg Psychiatry. 1986;49:1227–1232
   • View In Article
   • MEDLINE
   • CrossRef
9. Hartlage LC, Durant-Wilson D, Patch PC. Persistent neurobehavioral problems following mild traumatic brain injury. Arch Clin Neuropsychol. 2001;16:561–570
   • View In Article
   • MEDLINE
   • CrossRef
10. Powell TJ, Collin C, Sutton K. A follow-up study of patients hospitalized after minor head injury. Disabil Rehabil. 1996;18:231–237
    • View In Article
    • MEDLINE
    • CrossRef
11. Deb S, Lyons I, Koutzoukis C. Neurobehavioural symptoms one year after a head injury. Br J Psychiatry. 1999;174:360–365
    • View In Article
    • MEDLINE
    • CrossRef
12. Vanderploeg RD, Curtiss G, Luis CA, Salazar AM. Long-term morbidities following self-reported mild traumatic brain injury. J Clin Exp Neuropsychol. 2007;29:585–598
    • View In Article
    • CrossRef
13. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 4th ed.. Washington (DC): American Psychiatric Association; 1994;
    • View In Article
14. Boake C, McCauley SR, Levin HS, et al. Diagnostic criteria for postconcussional syndrome after mild to moderate traumatic brain injury. J Neuropsychiatry Clin Neurosc. 2005;17:350–356
    • View In Article
15. Eisenberg HM, Levin HS. Computed tomography and magnetic resonance imaging in mild to moderate head injury. In:  Levin HS,  Eisenberg HM,  Benton AL editor. Mild head injury. New York: Oxford Univ Pr; 1989;p. 133–141
    • View In Article
16. Levin HS, Mattis S, Ruff RM, et al. Neurobehavioral outcome following minor head injury: a three-center study. J Neurosurg. 1987;66:234–243
    • View In Article
    • MEDLINE
    • CrossRef
17. Binder LM, Rohling ML. Money matters: a meta-analytic review of the effects of financial incentives on recovery after closed-head injury. Am J Psychiatry. 1996;153:7–10
    • View In Article
    • MEDLINE
18. Kibby MY, Long CJ. Minor head injury: attempts at clarifying the confusion. Brain Inj. 1996;10:159–186
    • View In Article
    • MEDLINE
    • CrossRef
19. Gouvier WD, Cubic B, Jones G, Brantley P, Cutlip Q. Postconcussion symptoms and daily stress in normal and head injured college populations. Arch Clin Neuropsychol. 1992;7:193–211
    • View In Article
    • MEDLINE
    • CrossRef
20. Wood F, Novak TA, Long CJ. Post-concussion symptoms: cognitive, emotional and environmental aspects. Int J Psychiatry Med. 1984;4:277–283
    • View In Article
21. Meares S, Shores EA, Taylor AJ, et al. Mild traumatic brain injury does not predict acute postconcussion syndrome. J Neurol Neurosurg Psychiatry. 2008;79:300–306
    • View In Article
    • CrossRef
22. King NS. Emotional, neuropsychological, and organic factors: their use in the prediction of persisting postconcussion symptoms after moderate and mild head injuries. J Neurol Neurosurg Psychiatry. 1996;61:75–81
    • View In Article
    • MEDLINE
    • CrossRef
23. Fann JR, Katon WJ, Uomoto JM, Esselman PC. Psychiatric disorders and functional disability in outpatients with traumatic brain injuries. Am J Psychiatry. 1995;152:1493–1499
    • View In Article
    • MEDLINE
24. Luis CA, Vanderploeg RD, Curtiss G. Predictors for a postconcussion symptom complex in community dwelling male veterans. J Int Neuropsychol Soc. 2003;9:1001–1015
    • View In Article
    • MEDLINE
25. Nicholson K. Pain, cognition and traumatic brain injury. NeuroRehabilitation. 2000;14:95–103
    • View In Article
26. Perlis ML, Artiola L, Giles DE. Sleep complaints in chronic postconcussion syndrome. Percept Mot Skills. 1997;84:595–599
    • View In Article
    • MEDLINE
    • CrossRef
27. Centers for Disease Control (CDC). Health status of Vietnam veterans: I. Psychosocial characteristics (The Centers for Disease Control Vietnam Experience Study). JAMA. 1988;18:2701–2707
    • View In Article
28. Centers for Disease Control. Health status of Vietnam veterans: vol. IV (Psychological and neuropsychological evaluation). Atlanta: Centers for Disease Control; 1989;
    • View In Article
29. Centers for Disease Control. Health status of Vietnam veterans: II (Physical health). JAMA. 1988;18:2708–2714
    • View In Article
30. Montague E, William H, Lubin A, Gieseking C. Army tests for assessment of intellectual deficit. U S Armed Forces Med J. 1957;8:883–892
    • View In Article
    • MEDLINE
31. Robins LN, Helzer JE, Croughan J, Ratcliff KA. National Institute of Mental Health Diagnostic Interview Schedule (Its history, characteristics, and validity). Arch Gen Psychiatry. 1981;38:381–389
    • View In Article
    • MEDLINE
32. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 3rd ed.. Washington (DC): American Psychiatric Association; 1980;
    • View In Article
33. Sobel ME. Asymptotic confidence intervals for indirect effects in structural equation models. In:  Leinhardt S editors. Sociological methodology. Washington (DC): American Sociological Association; 1982;p. 290–312
    • View In Article