Neuropsychological Outcomes of Army Personnel Following Deployment to the Iraq War FREE

Since early 2003, significant numbers of military personnel have deployed in support of Operation Iraqi Freedom (OIF). Although contemporary battlefield measures¹ have improved war-zone survival, success in preventing fatalities has not eliminated adverse physical or mental^2- 3 health consequences. One major war-related health risk is brain dysfunction.

Brain dysfunction is often indicated by neuropsychological (ie, cognitive and emotional) impairment. In past military conflicts, cognitive impairment figured prominently among veteran health complaints, ranking fourth among 1991 Gulf War veterans in government health registries.⁴ Because of its potential negative impact on occupational and psychosocial functioning^5- 7 in a predominantly young population, war-related neuropsychological impairment has significant public health implications.

Yet, the consequences of war-zone deployment on neuropsychological health remain poorly understood. Knowledge gaps stem largely from a lack of baseline (predeployment) health information, reliance in large studies on subjective outcome indices, assessments conducted long (sometimes years) after war-zone exposure, and infrequent use of appropriate nondeployed comparison samples.

Our study objective was to examine neuropsychological outcomes following Iraq deployment. The study incorporated a prospective, cohort-controlled design measuring subjective and objective neuropsychological outcomes in US Army soldiers deploying to Iraq. Army soldiers with similar military characteristics from units not deploying overseas comprise the comparison group. Based on the anticipation that Iraq deployment would involve risks of neuropsychological compromise (eg, environmental exposures, prolonged physiological arousal associated with survival responses, head injury), we hypothesized that deployment would be associated with adverse neuropsychological outcomes.

Human subjects approval was obtained from human subjects research review boards of the Army, Tulane University Health Sciences Center, and the Department of Veterans Affairs. All participants provided written informed consent prior to participation.

The target population was male and female active duty US Army soldiers serving between April 2003 and June 2005. Participants were categorized by their deployment status during the study period: those deployed to Iraq and those not deployed overseas. Military units at high likelihood of deployment during the study period were assessed prior to deployment to Iraq (time 1, between April and December 2003) and again following their return (time 2, between January and May 2005). Although military unit deployment status during the study period could be anticipated, each unit's and participant's deployment was subject to evolving military operational requirements and could not be verified until time 2. Units at low likelihood of Iraq deployment during the study were also assessed twice, at periods timed to be as close as possible to their deploying counterparts. At time 1, most deployers belonged to units that were anticipated to deploy to Iraq within 75 days and were functioning under conditions of increased operational demands. Because nondeployers were preparing for extended intensive desert training within the continental United States, they were also functioning at increased operational tempo.

To capture heterogeneous deployment experiences and geographic separation within the war zone, unit selection was based on a modified categorization procedure.⁸ Deploying and nondeploying units represented combat, combat support, and combat service support functions and were well matched in these attributes. Battalion-level units originated from Fort Hood, Texas, and Fort Lewis, Washington. Battalion leaders were asked to refer potential participants at random (eg, every third name on the unit roster) to facilitate a sample representative of the battalion.

Potential participants consented individually and were provided with a way to exit the study area unobserved if they declined to participate. Study volunteers were excluded if pending separation from service or reassignment to another installation at time 1 or if unable to complete the study protocol because of physical limitations (eg, a broken hand). In addition, time 1 participants no longer at their originating military installations were invited to complete the survey portion of the protocol via mail but are not included in the analyses because of the infeasibility of collecting primary performance-based neurobehavioral outcome measures without in-person administration.

Sample size determinations were calculated taking into consideration statistical power and possible attrition from time 1 to time 2. Estimated attrition (20%) was based on anticipation of atypical deployment durations and military discharges. Using attentional data from a previous deployment study, calculations determined that a sample of 600 deployed and 300 nondeployed soldiers (adjusted for attrition) would provide 80% power to detect average change between the 2 groups corresponding to a small to medium effect size of 0.29 at the .05 significance level after Bonferroni adjustment for 10 comparisons (P≤.005).

Comprehensive description of primary assessment data and secondary data obtained from automated military databases has been published elsewhere.⁹ Measures relevant to hypotheses addressed in this report follow.

Demographic, Neuromedical, and Historical Information. Each assessment documented current demographic and military information (eg, age, rank), risk factors for neuropsychological disorders (eg, history of neurodevelopmental disorders, psychiatric disorders, brain injury), and situational factors (eg, recent sleep and alcohol use) potentially affecting neuropsychological performance. Self-reported race/ethnicity data were gathered to help gauge the representativeness of the sample. At time 2, deployed participants were interviewed about their locations while in Iraq.

Performance-Based Neuropsychological Tests. Although neuropsychological measures applied in clinical contexts are typically interpreted using deviations from normative values to form localized or syndromal diagnoses.^10- 11 However, epidemiological studies use neuropsychological measures as continuous outcomes to identify relationships in populations between exposures and performance patterns indicative of brain dysfunction,^12- 13 documenting subtle population shifts at scores frequently falling short of the range of clinical impairment.

Test battery selection emphasized continuous outcome measures and construct domains (sustained attention, working memory/executive functioning, fine motor speed, verbal and visual learning and memory, reaction time, and cognitive efficiency) sensitive to stress-related disorders and neurotoxicant exposures (Table 1).

Computer-assisted tasks were derived from the Automated Neuropsychological Assessment Metric (ANAM)¹⁹ and the Neurobehavioral Evaluation System, third edition (NES3)^20- 21 and required button-press responses. For the ANAM, scores reflecting accuracy and response time (“throughput”) were created to measure a reaction time variable and cognitive efficiency across other neuropsychological domains. Motor speed was measured by mean taps per 10-second interval on ANAM tapping. The NES3 Continuous Performance Task is a sustained (approximately 8-minute) attention task requiring detection of targets from a random sequence of distractor stimuli.

Non–computer administered tasks included the Trailmaking Test,²² the Wechsler Memory Scale, third edition (WMS3)¹⁶ Verbal Paired Associates (requiring learning and subsequent recall of unrelated word pairs), and the WMS²³ Visual Reproductions (requiring reproduction of 2-dimensional geometric designs from memory immediately after their presentation and after a delayed interval). For the Trailmaking Test, time to complete Part A (drawing lines between numerals in sequential order) was subtracted from time to complete Part B (drawing lines between sequential numbers and letters in alternation). The subtraction procedure parcels out basic attentional, speed, and visual tracking skills, resulting in a better measure of working memory and cognitive flexibility. For the WMS3 Verbal Paired Associates and WMS Visual Reproductions, the percentage of retention (Verbal Paired Associates: delayed recall/trial 4 recall × 100; Visual Reproductions: delayed recall/immediate recall × 100;) reflects how well information was remembered over time.

All scores were free of subjective judgment except for WMS Visual Reproductions, in which designs drawn from memory were scored by a rater according to set criteria. Although the primary rater was aware of deployment status, 10% to 15% of a randomly selected sample of drawings from each assessment episode were also scored by a second rater blinded to deployment status. Intraclass correlations (0.78-0.95) indicated high interrater reliability.

Deployment Experiences, Emotional Distress, and Functional Neurocognitive Health Perception. Deployment experiences were quantified by a modified version of the Deployment Risk and Resilience Inventory (DRRI).²⁴ State affect, commonly affected by neurotoxicant exposure^10,25 was measured with the Profile of Mood States (POMS).¹⁵ Persistent stress and depression symptom severity, assessed as potential covariates in outcome analyses, were quantified by the PTSD Checklist (PCL)^26- 27 and the Center for Epidemiological Studies Depression Inventory, 9-item version (CES-D),^28- 29 respectively. The 4-item version of the Medical Outcomes Study Cognitive Functioning Scale (MOS-CF)¹⁴ assessed functional neurocognitive health perception. The DRRI, POMS, PCL, CES-D, and MOS-CF are all psychometric self-report inventories yielding continuous variables. Although cut-point scores can be applied to the PCL and CES-D as crude screening estimates, neither instrument yields clinical diagnoses.

Assessment of Response Validity. Validity of response profiles on questionnaires was assessed via inspection of scales with bidirectional items (eg, a score of 5 endorses pathological functioning on some items and intact functioning on others). If a respondent provided all extreme responses in the same direction on a scale with bidirectional items, that respondent's data were not analyzed. The Test of Memory and Malingering,³⁰ trial 1, was administered to assess cognitive engagement. Data from participants scoring below 38, a cutoff found to show reasonable sensitivity and specificity in detecting insufficient effort on neurobehavioral tasks,³¹ were also excluded from analyses.

Assessments were conducted at military installations by a civilian examiner team. All performance-based neuropsychological measures were individually administered according to scripted, standardized instructions. Participants completed the paper-and-pencil surveys in small groups. Examiners and participants were typically aware of each participant's anticipated deployment status at time 1 and actual deployment status at time 2.

When data distributions departed significantly from normal, raw scores were normalized via logarithmic transformation. POMS summary scores were converted to sex-based T scores. Missing values for specific items on questionnaires (occurring in <3% of cases) were replaced by the mean value of the individual's completed items for that measure if the participant responded to at least 50% of the items. If fewer than 50% of the items on a measure were completed, summary scores were not computed. Outliers were truncated at 3 SDs from the mean.

Baseline characteristics and differences between time 2 respondents and nonrespondents were examined via t test or χ² test, as appropriate. To examine primary hypotheses, we used SAS software, version 8 (SAS Institute Inc, Cary, NC) to fit a generalized estimating equation linear regression model for each time 2 outcome variable. The study incorporated a cluster-sampling design, with participants sampled within battalion-level military units. The generalized estimating equation regression accounts for correlation in responses among participants from the same battalions to adjust for the multilevel structure of the sampling plan. Deployment status (deployed vs nondeployed) served as the independent variable of interest. To account for initial levels of outcomes, the time 1 value for the time 2 outcome measure of interest was entered as a covariate in each model. Age at time 1, sex, years of education, average hours of sleep per day in the week prior to time 2 assessment, and average number of standardized alcoholic drinks consumed per week during the month prior to time 2 assessment were also included as covariates because of their potential influence on cognitive performances. The model resulting from this covariate set is the core model.

Significance levels were adjusted via Bonferroni corrections to avoid type I error. Sixteen neurobehavioral outcome and 7 subjective outcome measures were considered, resulting in an adjusted significance level of P = .003 (.05/16) for neurobehavioral data and P = .007 (.05/7) for subjective data.

Because of the potential for stress-related symptoms and head injury to modify deployment-related outcomes, the core outcome analyses were repeated in 3 sets with PCL summary scores, CES-D summary scores, or head injury with loss of consciousness incurred between time 1 and time 2 included as a covariate.

At time 1, approximately 94% (n = 1368) of the 1457 invited soldiers volunteered participation. At time 2, soldiers assessed at time 1 who remained assigned to units located at the same military installation were again invited to participate in the full study protocol. Of the 1368 soldiers assessed at time 1, approximately 75% (72% from deployed units and 80% from nondeployed units) participated in the on-site assessment at time 2. The predominant reason for nonparticipation at time 2 was separation from service (Table 2). Of the 1028 time 2 participants, 26 completed questionnaires but did not complete performance tasks because of scheduling conflicts and were excluded from the analyses. Twenty-six participants were excluded for invalid questionnaire responses and 15 for questionable cognitive effort, resulting in a final sample of 961. All but 23 deployers examined predeployment and postdeployment served a 12-month OIF rotation.

In the final sample, 654 participants were categorized as deploying and 307 as nondeploying. Postdeployment assessments occurred a mean of 73.4 (SD, 19.8) days (median, 75 days; interquartile range, 58-84 days) from each participant's return from Iraq, except for 19 soldiers who returned early.

Participants in the final sample (Table 3) generally reflected the broader OIF-deployed Army population. Women were slightly underrepresented compared with the expected proportion of contemporaneously deployed Army women. Although enlisted personnel constitute the majority of deployers, commissioned officers were nonetheless underrepresented in the sample. At time 1, 11% had participated in a prior major overseas operational deployment (3% in 2001 or later). The most prevalent military occupational categories were infantry/gun crew (35%), communication/intelligence (19%), electrical/mechanical equipment repair (13%), and service supply (9%).

Nondeployers (80%) were somewhat more likely than deployers (72%) to participate at time 2 (P = .002). Among both deployers and nondeployers, time 2 respondents and nonrespondents did not differ at time 1 in age, marital status, years of formal education, years served in the Army, self-reported race/ethnicity, or most baseline values of subjective and objective outcome measures. Among deployers, nonrespondents at time 2 were more likely at time 1 to be female (18% vs 8%; P<.001), to be officers (6% vs 2%; P<.001), to report more fatigue on the POMS (P = .04), and to perform less proficiently on simple reaction time (P = .02) but more proficiently on the WMS Visual Reproductions immediate recall (P = .002). Among nondeployers, nonrespondents at time 2 were more likely at time 1 to be female (24% vs 9%; P<.001) and to describe themselves as racial/ethnic minorities (48% vs 35%; P = .02). In sum, there were few differences between respondents and nonrespondents, especially on time 1 outcome measures.

Deployed and nondeployed participants did not differ at time 1 on demographic or contextual variables (eg, sleep, developmental disorders, alcohol consumption) (Table 3), with the exception that nondeployers more frequently identified themselves as racial/ethnic minorities (P = .008). The 2 groups did not differ at time 1 on neurobehavioral or emotional measures (Table 4) except that deployers performed more poorly on the WMS Visual Reproductions immediate recall (P<.001) than nondeployers.

The interval between time 1 and time 2 for deployers was greater than for nondeployers (mean, 16.9 [SD, 3.1] months vs 8.3 [SD, 2.2] months; P<.001). This was attributable to scheduling time 2 testing for nondeployers early enough to avoid possible early deployment and an unanticipated delay in deployment for 1 of the deployed units. To assess the influence of duration of test-retest interval on outcomes while holding unit membership and deployment status constant, partial correlations controlling for core covariates (age, sex, education, alcohol use, sleep, and time 1 performance) were conducted within the 3 largest contributing brigade-level military units (2 deployed and 1 nondeployed with ≥288 personnel in each unit study sample) between test-retest interval duration and all primary outcomes. Of a possible 69 correlations, only 6 were significant at P<.05 and showed inconsistency across units, across variables, and in the direction of the association.

Deployed soldiers reported being located primarily within 3 regions of Iraq during their deployment, although some participants reported considerable movement throughout Iraq. Deployers reported significant combat activity, even when assigned to support roles. Frequently reported combat experiences included receiving hostile incoming small arms–type fire (98%), participating in a support convoy (95%), and going on combat patrols/missions (91%). Many deployers reported witnessing Americans/allies (55%) or enemy combatants (61%) being seriously wounded or killed. Numerous soldiers reported receiving hostile incoming fire (35%) or participating in a combat mission (49%) daily. Seeing people begging for food (98%), observing homes or villages destroyed (77%), and seeing Americans or allies after they had been severely wounded or disfigured in combat (63%) ranked among the most frequently reported of other potentially stressful war-zone events. Ninety-eight percent of deployed participants reported exposure to at least 1 potential environmental agent (eg, air pollution, pesticides, other routinely used chemicals), although less than 1% reported exposure to chemical or biological weapons. Fourteen percent of deployed participants reported being wounded or injured in combat; 7.6% of deployers (vs 3.9% of nondeployers) specifically reported experiencing head injury with related loss of consciousness between time 1 and time 2. Following their deployment, 11.6% of deployed participants screened positive for likely PTSD, as determined by the “strict” screening criteria outlined by Hoge et al³; 25.0% scored above a CES-D cutoff value,²⁹ suggesting a heightened probability of clinically significant depression symptoms.

Generalized estimating equations revealed significant deployment effects on WMS3 Verbal Paired Associates I (initial) recall, WMS Visual Reproductions percentage retention, NES3 Continuous Performance Task omission errors, ANAM simple reaction time throughput scores, and POMS confusion and tension subscales (Table 5).

Deployers showed a greater decline from time 1 to time 2 on the WMS Visual Reproductions retention than did nondeployers (Table 4, Table 5, and Table 6). In addition, nondeployers showed anticipated practice effects from time 1 to time 2 on WMS3 Verbal Paired Associates I and NES3 Continuous Performance Task omissions, whereas deployers showed little or no improvement (Table 4, Table 5, and Table 6). The absence of practice effects on certain cognitive tasks reflects impairment. In contrast, ANAM simple reaction time performance changed little over time for deployers but declined from time 1 to time 2 for nondeployers.

Deployment was associated with longitudinal increases in confusion and tension (Table 4, Table 5, and Table 6). There were no significant deployment effects in subjective estimates of cognitive impairment on the MOS-CF.

The individual inclusion of time 2 PCL scores, CES-D scores, and head injury incurred between time 1 and time 2 as covariates to the core models revealed specific associations between the covariates and several outcome measures. However, taking into account variance attributable to these covariates revealed that deployment continued to exert a significant effect for all neurobehavioral outcomes found to be significant using the core model (Table 7). Post hoc analyses taking into account change in PCL scores from time 1 to time 2 as a covariate likewise indicated that deployment exerted a significant effect for neurobehavioral outcomes found to be significant using the core model, independent of any worsening of PTSD symptoms. Additional post hoc analyses that repeated the generalized estimating equations using the core model but excluded from the sample the 63 participants who reported incurring a head injury with loss of consciousness between time 1 and time 2 revealed the identical pattern of results to those generated using the entire sample.

This, to our knowledge, is the first controlled cohort study to incorporate prospective examination of objective neuropsychological outcomes associated with war-zone deployment. The design included primary data collection both prior to and shortly after deployment and a nondeployed comparison sample. Results suggest that OIF deployment is associated at least transiently with subtle alterations in neural functioning, as indicated by population shifts in the neuropsychological performance of deployed vs nondeployed soldiers. These shifts include reduced proficiency in sustained attention and memory, heightened negative state affect reflecting increased feelings of confusion and tension, and an advantage in reaction time. Conversely, there were no significant effects of deployment on fine motor speed, executive aspects of attention, cognitive efficiency, or state measures of irritability, depression, fatigue, and vigor. Previous reports have described significant negative mental health consequences associated with OIF deployment.^2- 3 The findings of this study suggest that mental status changes related to Iraq War participation extend beyond psychiatric concerns to circumscribed adverse neuropsychological consequences, an outcome domain with high relevance to occupational and psychosocial functioning and highly sensitive to brain dysfunction.

The memory and attention problems commonly reported by Gulf War veterans highlighted neuropsychological dysfunction as an area of concern among deploying military personnel. However, results of studies examining the objective neuropsychological performances of Gulf War veterans yielded a mixed pattern of results that has been difficult to interpret because of the absence of baseline data and long intervals between war-zone return and neuropsychological assessment.³² Our findings indicating deployment effects on sustained attention, learning, and memory suggest that negative neuropsychological outcomes following Iraq deployment cannot be attributed to preexisting dysfunction and that it is unlikely that intervening variables influenced performances significantly due to the relatively abbreviated interval between war-zone return and assessment. Consistent with a recent report of British military personnel deployed to Iraq,³³ we did not find a deployment effect for subjective indices of neuropsychological compromise as pronounced as that revealed by some Gulf War findings^34- 36; however, deployment to Iraq in this study was associated with increased self-report of confusion.

Our findings also suggest that deployment is associated with a neurobehavioral advantage in reacting quickly and efficiently to simple targets. This finding seemingly contradicts the decrements in memory and attentional outcomes revealed by this study. However, when considered within an evolutionary framework, the pattern of findings is consistent with neurobiological responses directed toward survival. That is, when confronted with life threat, physiological responses occur in preparation for life-preserving action. Among the array of neurobiological events encompassed by the “flight or fight” response, neurotransmitter systems associated with increased arousal (eg, noradrenergic system) become activated, while neuroendocrine responses become altered via the hypothalamic-pituitary-adrenal axis.^37- 39

Such neurobiological alterations can result in heightened behavioral reactivity (eg, quickened response times) but dampened attention, learning, and memory for non–threat-relevant stimuli and events.^37,40- 41 Most of the participants in this study faced prolonged exposure to significant war-zone stressors while deployed, many of which would be categorized as imminently life-threatening. Such physiologically based responses could arguably have continued to affect cognitive functions into the period in which postdeployment assessments were conducted.

As predicted by previous research,^42- 44 higher levels of PTSD and depression symptoms were associated with relative performance deficits on several neuropsychological measures. However, emotional symptoms did not fully account for associations between deployment status and neurobehavioral outcomes. Thus, although deployment-related neuropsychological decrements may be in part related to emotional status, neuropsychological dysfunction also occurred independent of emotional responses. Such dissociations between self-reported emotional symptoms and neuropsychological performances may reflect at least transient desynchrony between subjective and physiological components of the stress response.⁴⁵ Alternatively, it may be that participants underendorsed emotional symptoms.

These findings could not be explained by contextual or demographic variables such as formal educational attainment or recent sleep and alcohol consumption patterns. It is possible that other contextual variables, such as differing perceptions about the significance of the study and changes in motivation, contributed to a deployment effect. However, there were no differences between groups on a cognitive test of effort, and all participants performing below a threshold on the effort task were excluded from analyses. Furthermore, qualitative remarks made by participants suggested that deployment increased the participants' understanding of the study, unlikely leading to a deployment-related decrease in motivation.

It is also possible that other attributes of the deployment resulted in neuropsychological compromise. A subset of both deployed and nondeployed study participants reported mild concussive injury, but the presence of an intervening concussive head injury failed to exert significant impact on objective neuropsychological outcomes. This finding should be interpreted cautiously, given that the measure of head injury included in the analyses was quite general (potentially including a range of severities), that the small number of participants with head injury may have resulted in reduced statistical power to detect a head injury effect, and that we did not measure repetitive nonconcussive blast exposures.

In addition, many participants reported exposure to potential environmental hazards, although most were consistent with exposures typical of modern urban life (eg, air pollution, use of insect repellant), and deployment did not show an effect for some measures typically thought to be sensitive to some types of neurotoxicants (eg, fine motor speed, executive functioning). Nonetheless, follow-up efforts will necessarily include examination of potential mechanisms, such as more detailed head injury characteristics, extent and type of stress exposure, and objective environmental monitoring data, as they become available.

Regardless of the mechanism involved, these findings point to a possible negative health consequence of war-zone deployment: neuropsychological compromise. The levels of such compromise were relatively mild and circumscribed. That is, the neuropsychological disadvantages associated with deployment include a range of scores that overall do not reach absolute clinical thresholds of significant impairment akin to advanced neurological disease state; however, even small declines in the ability to sustain attentional focus and learn and remember new information may reflect subtle neural dysfunction, lead to problems in day-to-day life, and negatively affect performance in high-pressure contexts such as subsequent war-zone participation. Such subtle alterations may also represent a prodrome or surrogate for disease.⁴⁶ In this case, it could be speculated that the stress-consistent pattern of deficits, if sustained, represents a surrogate for biological stress responsivity and a prodrome for eventual development of stress-related somatic (eg, cardiovascular) and mental (eg, posttraumatic stress disorder) health disorders.

Epidemiological investigations of exposure-outcome relationships frequently document subtle brain dysfunction in which an affected group includes a few members with optimal outcomes but more members with poor outcomes, indicating a shift away from normal health. For example, scores of as little as 2.5 IQ-point equivalents have been found to differentiate children exposed to methlymercury and lead from those not exposed.^47- 48 These “average” scores may not appear to be clinically significant but represent a population shift in which the risk for the population is significantly altered by the exposure variable.⁴⁶ Clinical implications of our findings include implementation of neuropsychological screening among military personnel returning from war-zone deployment and attention to the cognitive complaints of military personnel returning from deployment (even when medical workups do not support a clinical diagnosis).

Because we included only active-duty Army soldiers in this report, generalization of results to other military branches or to National Guard and Reserve personnel activated for deployment may be limited. In addition, our assessment of brain dysfunction was restricted to behavioral indices. Future efforts will benefit from inclusion of other measures of brain integrity, such as neuroimaging. Moreover, because this report includes only 1 postdeployment assessment, it is unclear whether changes in neuropsychological functioning endure or whether they are predictive of subsequent somatic or mental health problems. Nonetheless, this effort addresses many of the more significant limitations of prior deployment health outcome studies, including the lack of baseline data, small or regionally recruited convenience samples, reliance on subjective appraisals of neuropsychological health, and prolonged intervals between war-zone return and assessment. We have continued to follow this cohort longitudinally, broadened the assessment to include occupational outcomes, and included a sample of National Guard personnel, all necessary steps in determining the course of deployment-related neuropsychological decrements and understanding the longer-term public health impact of war-zone deployment.