Nociceptive stress interferes with neural processing of cognitive stimuli in Gulf War Veterans with chronic musculoskeletal pain

Abstract

Aims:

Disrupted cognition and chronic musculoskeletal pain (CMP) are prevalent experiences among Gulf War Veterans (GWV). A negative association between CMP and cognition (i.e., chronic pain-related cognitive interference) has been observed in some chronic pain populations but has not been evaluated in GWV. Additional research suggests that disrupted cognition in GWV with CMP may be exacerbated by stressing the nociceptive system. Therefore, we compared cognitive performance and related neural activity between CMP and healthy control (CO) GWV in the absence and presence of experimental pain.

Main methods:

During functional magnetic resonance imaging (fMRI), Veterans (CMP=29; CO=27) completed cognitive testing via congruent and incongruent conditions of a modified Stroop task (Stroop-only). A random subset (CMP=13; CO=13) also completed cognitive testing with experimental pain (Pain+Stroop). Yuen’s modified t-test and robust mixed-model analysis of variance (ANOVA) models were used for analyzing cognitive performance data. Independent t-tests and repeated-measures ANOVA models were employed for fMRI data with thresholding for multiple-comparisons (p<0.005) and cluster size (> 320 mm³).

Key findings:

Functional MRI analysis revealed significant between-group differences for the incongruent but not congruent-Stroop run. Neither correct responses nor reaction time differed between groups in either Stroop condition (all p≥0.21). Significant group (CMP, CO) by run (Stroop-only, Pain+Stroop) interactions revealed greater neural responses in CMP Veterans during Pain+Stroop runs. No significant interactions were observed for correct responses or reaction time (p≥0.31).

Significance:

GWV with CMP require a greater amount of neural resources to sustain cognitive performance during nociceptive stress.

1. Introduction

Following deployment to the Persian Gulf War, a significant proportion of Veterans began reporting multiple chronic debilitating symptoms, an affliction now recognized as Gulf War Illness (GWI).¹ Among the most prevalent GWI symptoms are memory problems (39.6%), difficulty concentrating (33.3%), and trouble finding words when speaking (22.8%),² confirming earlier assertions that cognitive issues are a component of GWI symptomology.³ Notably, cognitive symptoms may be compounded by chronic musculoskeletal pain (CMP) which occurs in 22.8–33.3% of Gulf War Veterans (GWV).² Negative associations between CMP and cognition (i.e., chronic pain-related cognitive interference) have been observed in other chronic pain populations such as fibromyalgia (FM), chronic low back pain, and chronic whiplash associated disorder.⁴ Although cognitive function has been investigated from a general perspective of symptomatic versus non-symptomatic GWV^5,6, to our knowledge chronic pain-related cognitive interference and associated neural substrates have not been evaluated in GWV.

Controlled laboratory studies demonstrate the value of using physical or mental stressors in identifying physiological and behavioral differences between ill and healthy participants. For instance, Falvo and colleagues revealed blunted cerebral autoregulation and blood flow velocity in GWI Veterans following orthostatic challenge.⁷ Wylie and colleagues reported greater increases in fatigue and higher activity in parietal, inferior frontal, and cerebellar brain regions for Veterans with GWI than healthy control Veterans during cognitive testing.⁸ Further, interactive effects between physical and mental stressors have also been reported such as in Cook et al. who showed that exercise challenge induced greater neural responses to a working memory task in people with Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS)⁹. Thus, chronic pain-related cognitive interference may be detected in GWVs by studying isolated and interactive effects of physical and mental stressors.

Prior functional magnetic resonance imaging (fMRI) studies suggest that stressing the nociceptive system with experimental pain stimuli is a viable approach to provoking physiology and behavior during cognitive testing. For example, Gopinath and colleagues found that GWI subgroups with cognitive (i.e., mild cognitive impairment) or neurological (i.e., confusion-ataxia) symptomology showed significantly greater activity than healthy control Veterans in several pain-relevant brain regions during administration of pain threshold calibrated thermal stimuli.¹⁰ Also, Cook and colleagues showed that people with FM differed from healthy controls in response to absolute (47°C) but not perceptually equivalent (temperature corresponding to 5 “strong pain” on 0–10 scale) thermal stimuli, and similar findings were reported in a study using pressure pain stimuli.¹² Moreover, experiments involving healthy adults¹³ and chronic pain patients^14,15 show that neural responses to cognitive stimuli are affected by nociceptive stress.

Chronic pain may interfere with cognition in GWV, especially during elevated nociceptive stress. Hence, we conducted an fMRI-based, case-control study comparing performance on a modified Stroop task (herein referred to as cognitive testing or performance) in Veterans with CMP to otherwise healthy control Veterans (CO) in the absence (Aim 1) and presence (Aim 2) of experimental pain. We hypothesized that Veterans with CMP would present with chronic pain-related cognitive interference, as indicated by greater neural responses to cognitive stimuli and lower cognitive performance than CO Veterans, and that these differences would be amplified by stressing the nociceptive system with experimental pain stimuli.

2. Methods

The methods below describe an fMRI study investigating neural responses to experimental pain and cognitive stimuli as part of a larger investigation comparing brain structure and function between Veterans with CMP and CO.^16,17 Aim 1 includes the full sample who completed cognitive testing during fMRI (CMP=29; CO=27) and Aim 2 includes a subset who were randomized to receive painful thermal stimuli in addition to cognitive testing during scanning (CMP=13; CO=13).

2.1. Participants

Sixty‐one Veterans (CMP = 30; CO = 31) deployed to the Persian Gulf region (e.g., Iraq, Kuwait, Saudi Arabia) during the Persian Gulf War (some of whom also served in the more recent Operation Iraqi Freedom). Groups were matched on age, sex, education, and deployment status. CMP was operationally defined as moderate pain lasting at least 6 months and occurring in three or more quadrants of the body. Additionally, reported pain symptoms could not be explained by an acute injury or other known chronic pain condition (e.g., rheumatoid arthritis). Presence or absence of CMP was confirmed via a clinical assessment and medical chart review by a local Department of Veterans Affairs (VA) rheumatologist. Veterans who did not endorse the presence of a chronic illness or pain were designated as CO. All participants gave their informed written consent and were compensated $75 per study visit. The study was approved by the University of Wisconsin-Madison Health Sciences Institutional Review Board and Madison VA Research and Development Committee.

2.1.1. Exclusion criteria

Potential participants were excluded for MRI contraindications (e.g., claustrophobia, ferrous metal in the body, planned or current pregnancy). Individuals weighing greater than 136 kg were excluded due to inherent weight limitations of the MRI scanning bed. To reduce variability in brain structure/function and to control for potential confounds, individuals were also excluded if they had one or more of the following: bipolar disorder, schizophrenia, major depressive disorder, post‐traumatic stress disorder, medical or neurologic disorders, colorblindness, diabetes, an episode of unconsciousness lasting longer than 5 min, current use of illegal drugs, substance abuse or dependence (within 2 years), or current use of muscle relaxants, anticonvulsant medications, or any prescribed narcotic pain medications. Absence of exclusionary criteria was confirmed via medical chart review and urine screening.

2.1.2. Recruitment

A letter of invitation to participate was sent to Veterans in the Madison, Milwaukee, and Tomah VA medical center patient databases who met our most basic inclusion criteria, that is, age (30–55) and deployment (Persian Gulf region, 1990–1992, 2003–2011). In addition, participants were recruited through postings of study information at local Veteran organizations, on approved Veteran websites, and newspaper advertisements.

2.2. Procedures

Procedures included an initial phone screening for eligibility and two visits to the laboratory, separated by approximately 1 week. Following confirmation of study eligibility, participants were randomized to receive either painful (‘pain condition’) or non-painful (‘no pain condition’) thermal heat stimuli as part of the larger fMRI protocol.

2.2.1. Day 1 familiarization visit

Day 1 consisted of a physical exam, medical chart review, urine screening, questionnaires (described in section 2.3.1), and a familiarization trial in a simulated MRI environment designed to mimic the fMRI test planned for Day 2. Participants randomized to the pain condition were informed they would receive a range of painful temperatures and that some of these stimuli could be considered extremely painful, whereas participants in the no pain condition were reassured that only nonpainful warm stimuli would be delivered during scanning.

Next, participants were introduced to a range of painful (43–49°C) or nonpainful (34–40°C) thermal stimuli on the palm of their left hand. Each temperature within the assigned range for the pain or no pain conditions was presented twice in a random order for a total of 14 stimulus presentations, with an inter-stimulus interval of 1 min. Participants also practiced rating pain intensity and unpleasantness of each stimulus (see section 2.3.3 for pain scale description). As participants became more familiar with the scales, they were encouraged to provide their ratings within the 10-s time period that would be allotted during the Day 2 scan.

After orientation to thermal stimuli, participants in the pain condition were informed that those same stimuli would be administered during the Day 2 experimental visit. Participants in the no pain condition were given the same instructions unless they rated a given thermal stimulus higher than 0, in which case they were informed that only stimuli they rated as 0 (i.e., no pain sensation) would be administered during the Day 2 experimental visit. Finally, participants practiced congruent (CStroop) and incongruent (IStroop) versions of the Stroop color word task (see section 2.3.4 for Stroop task description). This task was also practiced while receiving either painful (pain condition) or nonpainful warm (no pain condition) stimuli.

2.2.2. Day 2 experimental visit

On Day 2, participants arrived at the laboratory, were given an overview of the study procedures for the experimental scan and were led to the MRI room for instrumentation and positioning in the scanner. Next, functional imaging data were acquired over six separate runs in the pain condition and five separate runs in the no pain condition using a block design with total duration of each run lasting 230 s. In both pain and no pain conditions, run 1 (Warm only) consisted of five blocks of non-painful warm stimuli corresponding to a rating of 0 (“no pain”) on the 0–20 Gracely pain intensity scale. This run tested the effect of pain anticipation on perceptual and brain responses to non-painful stimuli and the results from this experiment are summarized elsewhere.¹⁷

The pain condition consisted of five additional runs presented in randomized order and involved (i) thermal pain stimuli corresponding to a rating of 11–13 (“slightly intense-strong”) on the 0–20 Gracely pain intensity subscale as determined during familiarization (Pain-only run), (ii) thermal pain stimuli combined with the congruent (Pain+CStroop run) or incongruent (Pain+IStroop run) versions of the Stroop task, and (iii) congruent (CStroop-only run) or incongruent (IStroop-only run) versions of the Stroop task alone (Figure 1). The no pain condition consisted of four additional runs, also presented in randomized order and involved (i) warm stimuli combined with the congruent (Warm+CStroop run) or incongruent (Warm+IStroop run) version of the Stroop task, and (ii) congruent (CStroop-only run) or incongruent (IStroop-only run) versions of the Stroop task alone.

Figure 1.

Graphic display of the functional magnetic resonance imaging protocol for participants randomized to the pain condition.

Note: Each run consists of an initial rest period followed by five experimental blocks and lasts 230 s in duration. Run 1 consists of non-painful warm stimuli (Warm-only run). Runs 2 through 6 are presented in randomized order and consist of (i) thermal pain stimuli alone (Pain-only run), (ii) thermal pain stimuli combined with the congruent (Pain+CStroop run) or incongruent (Pain+IStroop run) version of the Stroop task, and (iii) congruent (CStroop-only run) or incongruent (IStroop-only run) versions of the Stroop task alone.

Pain rating scales were not administered during the CStroop-only and IStroop-only runs.

For runs involving thermal stimuli alone or in combination with cognitive stimuli, each started with a 30-s off-period (no stimulation) followed by five blocks. Blocks consisted of a 3-s countdown, 20-s thermal stimulus and/or cognitive task, and 17-s period in which participants were given a maximum of 10 s to provide perceptual ratings of the thermal stimuli. The remaining time was spent waiting for the countdown period of the next block to begin (Figure 1). Runs involving cognitive stimuli only (i.e. CStroop-only and IStroop-only) were similar to those involving thermal stimuli, except perceptual rating scales were not presented during the 17-s period that followed the 20-s stimulus period.

2.3. Instruments

2.3.1. Questionnaires

Several questionnaires were administered on Day 1 to characterize demographic information and self-reported depression, anxiety, general physical and mental health, pain catastrophizing, and physical activity behavior of the participants, including: the Beck Depression Inventory¹⁸, the International Physical Activity Questionnaire,¹⁹ the Pain Catastrophizing Scale,²⁰ the 36-Item Short-Form Health Survey,²¹ Spielberger’s State-Trait Anxiety Inventory,²² a 0–10 pain anticipation scale,¹⁷ and the Mcgill Pain Questionnaire.²³

2.3.2. Thermal stimuli

Thermal stimuli were administered to the palm of the left hand via a Pathway Pain & Sensory Evaluation System with a 900 mm² Peltier thermode (Medoc Advanced Medical Systems, Ramat Yishai, Israel). Baseline temperature for the fMRI scan was maintained at 32°C and increased to the stimulus target temperature at a rate of 8°C/s. The duration of each thermal stimulus upon reaching the target temperature was 20 s.

2.3.3. Pain rating scales

On both study visits, thermal stimuli were rated with the Gracely Pain Scale, a self-report measure with two 0–20 category-ratio scales that separately assess the sensory and affective components of pain.²⁴ The Gracely Pain Scale has established psychometric properties²⁵ and has previously been used to measure ratings of experimental pain stimuli in people with CMP.^15,26 The pain intensity (sensory) scale was always presented first and was followed by the pain unpleasantness (affective) scale after each thermal stimulus administration. Participants were instructed to provide a 0 rating on both scales if they did not experience pain. All participants were given identical detailed instructions for the use of the scales on Day 1, and those directions were reiterated before the experimental MRI scan on Day 2.

Rating scales were presented using E-Prime software (Psychology Software Tools, Pittsburg, PA) viewed with MRI-compatible goggles (Avotec, Inc., Stuart, FL). Each participant made their ratings on the presented scales within 10 s. after each thermal stimulus using a scanner-compatible button press response unit (Current Designs, Philadelphia, PA), which was operated by their right hand. All participants, regardless of condition assignment, were instructed to use the scales only to rate pain experienced from the thermal stimuli and avoid rating other pain sensations they might be feeling (e.g., musculoskeletal pain symptoms).

2.3.4. Stroop color-word task

Cognitive stimuli were administered either alone or in combination with thermal stimuli using a modified version of the Stroop color-word task,^15,26 which has been used in our prior work involving people with FM^15,26 and is a recommended common data element in neuropsychological research involving Veterans with GWI.⁶ The Stroop task is an executive function/attentional test that presents the words “red”, “green”, “yellow” and “blue” in the font colors red, green, yellow and blue. Words are presented in either congruent (e.g., the word “red” appears in red font) or incongruent (e.g., the word “red” appears in blue font) fashion and participants are asked to indicate the correct font color and ignore the word.²⁷ Incongruent presentations require longer processing time and more sustained attention than congruent presentations, a process referred to as “interference”.²⁸

Responses were made using the button-press unit described above which contained four separate buttons. The color of each button corresponded to the color of the Stroop task stimuli (i.e., red, yellow, blue, green). Participants were instructed to: (1) identify only the visual color of the word and not read the actual word, (2) press the button that corresponds to the correct response color (e.g., the index finger for a word that is colored blue), and (3) respond to each stimulus as quickly as possible but not to sacrifice accuracy for speed. Performance on both tasks was measured via the number of correct responses and reaction time for correct responses (ms).

Similar to pain rating scales, congruent and incongruent words were viewed with a set of MRI-compatible goggles (Avotec, Inc., Stuart, FL) and presented with the use of E-Prime software (Psychology Software Tools, Pittsburg, PA). Each word was presented for 1500 ms, positioned center-screen, with an inter-stimulus interval of 500 ms. Ten stimuli were presented over each 20-second block, identical to that which occurred on the familiarization visit.

2.3.5. fMRI acquisition and processing

All anatomical and functional magnetic resonance images were collected on a 3-Tesla GE SIGNA MRI scanner (GE Health Systems, Waukesha, WI) with a whole-head transmit-receive coil. A vacuum pillow and/or foam padding was used to limit head motion within the coil. Participants were fitted with MRI-compatible headphones for communications to and from the experimenter and to minimize scanner noise during acquisition.

Anatomical acquisitions were collected using 3D IR-prepped fast-gradient echo-pulse sequence, which consisted of 124 (1–2 mm thick), T1-weighted (repetition time 9,000 ms, echo time 93 ms, field of view 24 cm, flip angle 30/90°), axial images with a matrix of 256 × 256 × 64. High-resolution functional images were obtained using echoplanar imaging (EPI) with a gradient echo EPI sequence (repetition time 2,000 ms, echo time 30 ms, flip angle 90°) and consisted of 30 4‐mm thick (1‐mm gap) sagittal slices. The acquisition matrix was 64 × 64 mm and the field of view 24 cm, delivering an in‐plane voxel resolution of 3.75 × 3.75 × 4 mm.

Data processing was conducted using Statistical Parametric Mapping version 12 (SPM 12; Wellcome Department of Imaging Neuroscience, London, UK) and MATLAB (MathWorks, Natick, MA) software. Functional images were motion corrected, field‐map corrected, normalized, smoothed with an 8‐mm Gaussian filter, and registered to the Montreal Neurological Institute’s (MNI) 152 template using an affine transformation. To avoid saturation effects, the first three volumes collected were removed from the functional data analyses. Anatomical locations of significant MNI coordinates observed from statistical analyses of functional data (section 2.5) were confirmed via the Harvard-Oxford cortical and subcortical atlases.

2.4. Statistical analysis of participant characteristics and behavioral data

Prior to analysis of participant characteristics and fMRI behavioral task data, Kolmogorov-Smirnov and Levene’s tests, and visual inspection of violin plots were conducted to examine normality and homogeneity of variance. Because these assumptions were violated in a number of instances, we used the WRS2 package in R²⁹ to conduct Yuen’s modified t-test for independent trimmed means and robust mixed-model analysis of variance (ANOVA). These tests are considered more robust to violations of normality and homoscedasticity than conventional independent samples t-test and repeated measures (RM) ANOVAs.^30–32 To quantify between-group differences, we used Wilcox and Tian’s explanatory measure of effect size (ξ) with percentile bootstrapped 95% confidence intervals because it is more robust to violations of homoscedasticity than parametric effect sizes such as Hedges’ d.³³ Values of 0.10, 0.30, and 0.50 were considered to be small, medium, and large, respectively.³¹

2.4.1. Participant characteristics

Yuen’s modified t-tests compared CMP and CO for demographics and responses to questionnaires completed during the familiarization visit (Day 1). These analyses were conducted for the full sample investigated in Aim 1 and the pain condition subset investigated in Aim 2.

2.4.2. Behavioral data from fMRI scan

Prior to group comparisons, behavioral data collected from each participant’s fMRI scan were averaged across the five blocks comprising each run (i.e., Pain-only, CStroop-only, IStroop-only, Pain+CStroop, Pain+IStroop).

Chronic pain-related cognitive interference (Aim 1):

Yuen’s modified t-tests compared correct responses and reaction time for correct responses (ms) between CMP and CO during CStroop-only and IStroop-only runs. Chronic pain-related cognitive interference was operationalized as lower Stroop Task performance in CMP relative to CO during Stroop-only runs. This analysis included the full sample, regardless of whether they were randomized to the pain or no pain conditions. The remaining behavioral analyses pertain to the subset of 26 participants randomized to the pain only condition.

Effect of experimental pain on chronic pain-related cognitive interference (Aim 2):

To confirm that Veterans with CMP and CO received perceptually equivalent experimental pain stimuli, Yuen’s modified t-tests compared pain intensity and unpleasantness ratings of thermal stimuli administered during the Pain+CStroop and Pain+IStroop runs. We also tested whether the absolute intensity of thermal stimuli differed between CMP and CO.

Next, the effect of experimental pain on cognitive performance was examined by comparing changes in correct responses and reaction time for correct responses (ms) from the Stroop-only run to the Pain+Stroop run. To examine whether these changes differed between groups, we used robust mixed-model ANOVAs with Group (CMP, CO) as the between-subjects factor and Run (Stroop-only, Pain+Stroop) as the within-subjects factor. Four separate models examined correct responses and reaction time for the CStroop and IStroop tasks.

2.5. fMRI data analysis

Data from five participants were excluded from fMRI analyses because of technical difficulties with the thermal stimulus administration equipment (n=1) or E-Prime software (n=3), and poor signal-to-noise ratio from motion artifact (n=1). Thus, a total of 56 participants (CMP = 29; CO = 27) were included in the analytic sample.

In line with our prior work involving Veterans with CMP¹⁷ and people with FM,^26,34 this study examined activity in brain regions involved in pain encoding and processing.³⁵ Therefore, we performed a region-of-interest analysis with a study-specific mask based on experimental pain research in healthy participants and people with FM.³⁵ The following regions were included in the mask: pre- and post-central gyri, superior parietal lobule, cingulate cortices, brainstem, frontal medial cortex, frontal and parietal opercula, frontal pole, insula, thalamus, and middle frontal and orbital frontal gyri.

To account for multiple comparisons, the statistical map was thresholded at a voxelwise p <.005, and clusters of activity of fewer than 98 contiguous voxels (< 320 mm³) were ignored. The cluster threshold was calculated using 3dClustSim, an Analysis of Functional NeuroImages program with the uncorrected per-voxel p value set at .005, the corrected cluster alpha value set at .05, and the full width at half maximum Gaussian filter value set at 5.6 for only those voxels included in the region-of-interest mask.

Chronic pain-related cognitive interference (aim 1):

To address aim 1, we used independent samples t-tests in SPM12 to test for differences between CMP and CO during CStroop-only, and IStroop-only runs.

Effect of experimental pain on chronic pain-related cognitive interference (aim 2):

To determine whether neural responses associated with effects of experimental pain on cognition differed between CMP and CO, we used a two-way RM-ANOVA with Group (CMP, CO) as the between-subjects factor and Run (Pain+Stroop, Stroop-only) as the within-subjects factor to test for a group-by-run interaction. Two separate RM-ANOVA models were used, one for CStroop and one for IStroop analyses.

RM-ANOVAs were conducted using the Sandwich Estimator (SwE) toolbox, an SPM12 extension developed by Guillaume and colleagues.³⁶ The SwE toolbox was developed for analysis of longitudinal neuroimaging data, but is also appropriate for cross-sectional studies involving repeated-measures. Additional advantages of SwE include robustness to (i) misspecification of the covariance model, (ii) heterogenous variances across time and groups, and (iii) small samples.³⁶ Main effects and interactions were decomposed with one-sample t-tests to determine the direction of significant findings.

3. Results

3.1. Participant characteristics

Participant characteristics are summarized in Table 1. In the full sample, CMP Veterans reported significantly greater symptom severity for pain, depression, trait anxiety, pain catastrophizing, and overall physical health (all p≤0.04). In the pain condition subset, CMP Veterans had significantly higher body mass index (p=0.04) and reported significantly greater symptom severity for pain, depression, trait anxiety, and overall physical and mental health (all p≤0.04).

Table 1.

Descriptive comparison of mean (SD) participant characteristics between the full sample of Gulf War Veterans with chronic musculoskeletal pain (CMP) and healthy control Gulf War Veterans (CO) and those who were randomized to the pain condition

	Full sample				Pain condition subset
	CMP (n=29)	CO (n=27)	p	Effect size (95% CI)	CMP (n=13)	CO (n=13)	p	Effect size (95% CI)
Age (years)	45.52 (6.53)	44.78 (6.91)	0.75	0.06 (0, 0.41)	44.85 (7.24)	47.15 (5.57)	0.33	0.29 (0, 0.75)
Sex (Male/Female)	27/2	23/4	0.41d	n/a	11/2	11/2	1.0d	n/a
Body Mass Index (kg/m2)	28.57 (3.75)	27.47 (3.7)	0.11	0.32 (0, 0.63)	28.97 (3.3)	26.44 (3.36)	0.04	0.51 (0.08, 0.87)
McGill Pain Questionnaire Total Scorea	7.9 (6.71)	0.3 (0.54)	<0.001	0.86 (0.77, 0.98)	8.15 (5.06)	0.15 (0.38)	0.01	0.88 (0.78, 0.99)
Medical Outcomes Survey Short-Form (bodily pain subscale)b	53.55 (16.43)	94.22 (6.55)	<0.001	0.94 (0.87, 0.99)	53.19 (13.15)	97.23 (3.11_	<0.001	0.92 (0.88, 0.99)
Medical Outcomes Survey Short-Form (physical health component score)b	40.93 (7.22)c	56.7 (2.52)	<0.001	0.92 (0.87, 0.99)	40.42 (5.67)	57.81 (1.6)	<0.001	0.95 (0.9, 0.99)
Medical Outcomes Survey Short-Form (mental health component score)b	53.2 (7.96)	56.73 (3.68)	0.13	0.31 (0, 0.68)	50.03 (8.49)	57.32 (3.22)	0.01	0.77 (0.26, 0.98)
Beck Depression Inventory (total score)a	6.05 (4.4)	0.93 (1.3)	<0.001	0.87 (0.71, 0.98)	7.54 (5.28)	0.69 (1.03)	0.005	0.82 (0.71, 0.99)
State Trait Anxiety Inventory (form Y-2 score)a	31.5 (8.28)	26.44 (4.64)	0.04	0.41 (0.07, 0.74)	34.67 (7.85)	25.38 (3.38)	0.01	0.79 (0.46, 0.97)
Pain Catastrophizing Scale (total score)a	9.52 (8.39)	5.74 (8.07)	0.03	0.47 (0.12, 0.78)	7.69 (6.96)	3.77 (3.85)	0.2	0.47 (0, 0.93)
IPAQ total score (MET mins/week)b	7637.28 (11066.4)	7974.67 (10048.32)	0.9	0.03 (0, 0.5)	9039.00 (15716.04)	8158.00 (7669.47)	0.5	0.26 (0, 0.84)
Pain anticipation (0–10 scale)	2.35 (2.26)	3.24 (2.93)	0.38	0.19 (0, 0.59)	4.19 (1.77)	5.15 (2.38)	0.36	0.26 (0, 0.78)

Note.

^aHigher values indicate worse health;

^bLower values indicate worse health;

^cMissing data for n=1 participant;

^dFisher’s exact test; n/a= not applicable. Effect sizes reflect Tian’s explanatory measure of effect size (ξ) with percentile bootstrapped 95% CI. Confidence interval boundaries that overlap zero are reported as 0. Values of 0.10, 0.30, and 0.50 were considered small, medium, and large, respectively.

IPAQ = International physical activity questionnaire

3.2. Chronic pain-related cognitive interference (Aim 1)

Correct responses and reaction time for correct responses during CStroop-only and IStroop-only runs are detailed in Table 2 and supplemental digital content 1. No significant between-group differences were observed (all p≥0.21). Independent samples t-tests revealed no between-group differences in BOLD signal for the CStroop-only run. For the Istroop-only run, increases in BOLD signal were significantly greater in CO than CMP in the left precentral gyrus (Table 3, Figure 2).

Table 2.

Congruent (CStroop) and incongruent (IStroop) Stroop task performance for Gulf War Veterans with chronic musculoskeletal pain (CMP) and healthy control Gulf War Veterans (CO) during CStroop-only and IStroop-only runs

	CMP (n=29)	CO (n=27)	p	Effect size (95% CI)
CStroop-only run
Correct responses (%)a	95 (7.8)	97 (6.4)	0.27	0.21 (0, 0.69)
Reaction time (ms)b	804.02 (127.47)	746.57 (79.15)	0.21	0.27 (0, 0.66)
IStroop-only run
Correct responses (%)a	95 (6.5)	95 (9.3)	0.49	0.11 (0, 0.48)
Reaction time (ms)b	878.47 (103.97	832.9 (99.06)	0.32	0.21 (0, 0.58)

Note. Data are presented as averages across 5 total administration blocks.

^aHigher values indicate better performance;

^bLower values indicate better performance. Effect sizes reflect Tian’s explanatory measure of effect size (ξ) with percentile bootstrapped 95% CI. Confidence interval boundaries that overlap zero are reported as 0. Values of 0.10, 0.30, and 0.50 were considered small, medium, and large, respectively.

Table 3.

Results from functional magnetic resonance imaging analysis comparing neural responses between CMP and CO during the IStroop-only run

Region in focus point	Hemisphere	Peak MNI coordinates			Number of voxels in cluster
		x	y	z
Precentral gyrus	Left	−42	0	38	162

Note: Findings reported here survived corrections for multiple comparisons (voxelwise p <.005) and cluster thresholding (> 98 contiguous voxels). CMP: chronic musculoskeletal pain; CO: healthy control.

Figure 2.

Between-group comparison of neural responses during IStroop-only run revealed greater blood oxygen level dependent signal changes in the healthy control group than the chronic musculoskeletal pain group in the left precentral gyrus.

3.3. Effect of experimental pain on chronic pain-related cognitive interference (Aim 2)

Mean±SD thermal stimulus temperatures (CMP=48.2±0.68; CO=47.8±0.8) did not significantly differ between groups (p=0.02; ξ=0.35). During Pain+Stroop runs, neither pain intensity nor unpleasantness ratings differed between groups (all p>0.76, all ξ<0.09). Changes in correct responses and reaction time for correct responses from CStroop-only to Pain+CStroop runs and IStroop-only to Pain+IStroop runs are detailed in Table 4 and supplemental digital content 2. Two-way robust mixed model ANOVAs revealed a significant main effect of run for CStroop correct responses (p<0.001), IStroop correct responses (p<0.001), and IStroop reaction time (p=0.01) (all indicating better performance in the Stroop-only run relative to the Pain+Stroop runs), but not CStroop reaction time (p=0.32). Significant group-by-run interaction effects were not observed for any model (all p≥0.31).

Table 4.

Cognitive performance for Gulf War Veterans with chronic musculoskeletal pain (CMP) and healthy control Gulf War Veterans (CO) during Stroop-only runs and Pain+Stroop runs

	CMP (n=13)	CO (n=13)	p	Effect size (95% CI)
CStroop-only run
Correct responses (%)a	98.2 (2.9)	98.2 (4)	0.84	0
Reaction time (ms)b	777.58 (109.36)	733.92 (78.25)	0.56	0.22 (0, 0.83)
Pain+CStroop run
Correct responses (%)a	87.5 (4.2)	86.8 (4.9)	0.51	0.23 (0, 0.8)
Reaction time (ms)b	796.66 (125.21)	747.15 (70.28)	0.45	0.36 (0, 0.83)
IStroop-only run
Correct responses (%)a	96.8 (3.8)	95.5 (7.6)	0.81	0.08 (0, 0.78)
Reaction time (ms)b	868.6 (93.7)	838.02 (80.47)	0.52	0.21 (0, 0.74)
Pain+IStroop run
Correct responses (%)a	86 (4.6)	86.5 (3.1)	0.99	0 (0, 0.64)
Reaction time (ms)b	914.61 (118.27)	866.79 (143.08)	0.79	0.08 (0, 0.75)

Note. Data are presented as averages across 5 total administration blocks.

^aHigher values indicate better performance;

^bLower values indicate better performance. Mean (SD) stimulus temperatures (C°) for each group are: CMP = 48.15 (0.68) and Control = 47.76 (0.80). Effect sizes reflect Tian’s explanatory measure of effect size (ξ) with percentile bootstrapped 95% CI. Confidence interval boundaries that overlap zero are reported as 0. Values of 0.10, 0.30, and 0.50 were considered small, medium, and large, respectively.

Repeated measures ANOVAs revealed a significant group-by-run interaction showing that BOLD signal changes from the CStroop-only to Pain+CStroop runs differed between CMP and CO. Additionally, a significant group-by-run interaction revealed that BOLD signal changes from the IStroop-only to Pain+IStroop runs also differed between CMP and CO. After reviewing results from 1-group t-tests for Pain+Stroop and Stroop-only runs, these interactions reflected greater neural activity for the CMP group during the Pain+Stroop run. These results are further illustrated in Tables 5–6 and Figure 3.

Table 5.

Results from repeated measures Group (CMP, CO) by Run (Pain+CStroop, CStroop-Only) functional magnetic resonance imaging analysis examining the effect of experimental pain during completion of the congruent Stroop task

Region in focus point	Hemisphere	Peak MNI coordinates			Number of voxels in cluster
		x	y	z
Insula	Right	39	−8	0	197
Frontal pole	Right	42	36	15	756
		20	44	30
		33	44	21
Planum temporale	Right	57	−26	14	404
		63	−33	20
Post-central gyrus		58	−9	22
Frontal pole	Right	27	38	−14	171
		39	38	−9
Frontal orbital cortex		34	32	−12
Middle frontal gyrus	Right	30	4	50
Precentral gyrus		36	4	30	662
		46	−2	38
Paracingulate gyrus	Right	12	16	36	140
Cingulate gyrus (anterior division)		12	26	26
		10	4	34

Note: All findings reported here survived corrections for multiple comparisons (voxelwise p <.005) and cluster thresholding (> 98 contiguous voxels). CMP: chronic musculoskeletal pain; CO: healthy control.

Table 6.

Results from repeated measures Group (CMP, CO) by Condition (Pain+IStroop, IStroop-Only) functional magnetic resonance imaging analysis examining the effect of experimental pain during completion of the incongruent stroop task

Region in focus point	Hemisphere	Peak MNI coordinates			Number of voxels in cluster
		x	y	z
Frontal pole	Right	32	45	24	1815
Inferior frontal gyrus		50	33	12
Frontal pole		20	46	38
Precentral gyrus	Right	54	2	39	1043
		62	12	20
		60	10	6
Inferior frontal gyrus	Right	52	21	26	353
Superior temporal gyrus, posterior division	Right	48	−27	−2	372
		62	−33	2
Thalamus	Left	−8	−28	6	176
Brainstem		−3	−32	−2
Middle frontal gyrus	Right	34	4	58	414
		30	18	52
		30	2	50
Frontal pole	Left	−20	54	21	196
		−21	39	24
		−21	48	27
Paracingulate gyrus	Right	15	36	18	239
		12	26	30
		12	16	39
Middle frontal gyrus	Right	34	12	28	112
Precentral gyrus		42	8	33

Note: All findings reported here survived corrections for multiple comparisons (voxelwise p <.005) and cluster thresholding (> 98 contiguous voxels). CMP: chronic musculoskeletal pain; CO: healthy control.

Figure 3.

Significant group-by-run interaction effects for chronic musculoskeletal pain (CMP=13) and healthy control (CO=13) Veterans during CStroop (panel a) and IStroop tasks (panel b)

Note. Panel A shows neural responses in the model comparing CStroop-only to Pain+CStroop runs. Panel B shows neural responses in the model comparing IStroop-only to Pain+IStroop runs. Panel C is a combination of panels A and B. Neural responses unique to Congruent and Incongruent Stroop models are shown in red and yellow, respectively. Neural responses that overlap between Congruent and Incongruent models are shown in green.

4. Discussion

A large proportion (22.8–33.3%) of GWV live with CMP,² highlighting an urgent need to better understand the pathophysiology of this condition and its effect on various aspects of health-related quality of life, including cognition. Underscoring the lower health-related quality of life in Veterans with CMP, we observed moderate-to-large effect size differences between CMP and CO Veterans (ξ=0.41–0.92) across multiple self-reported outcomes of mental and physical health (Table 1). Cognitive testing during fMRI revealed greater activity in the left precentral gyrus of CO Veterans (Aim 1) whereas combining experimental pain with cognitive testing revealed greater brain activity across multiple regions in CMP Veterans (Aim 2). Our findings are interpreted below with an emphasis on chronic pain-related cognitive interference in GWV.

4.1. Cognitive performance did not differ between groups, but CO veterans showed greater activity in the left precentral gyrus during the IStroop-only run

Neuropsychological evidence indicates that symptomatic GWV exhibit lower cognitive performance than asymptomatic GWV;^5,6 however, the relationship between CMP and cognition in these Veterans is not well understood. To address this knowledge gap, we studied chronic pain-related cognitive interference in GWV by comparing performance on a modified Stroop color-word task and associated neural activity between Veterans with and without CMP. Counter to our hypothesis and what has been observed in several other chronic pain populations,⁴ non-significant differences in accuracy and reaction time between groups revealed that CMP was not associated with lower cognitive performance (Table 2, Figure 2). Therefore, in the absence of a nociceptive stressor, there was not strong evidence of chronic pain-related cognitive interference in this sample of Veterans with CMP.

There are several potential explanations for why cognitive performance was not lower in CMP Veterans than CO Veterans. Considering the extended period of time passing since the Persian Gulf War with minimal progress made toward establishing effective treatments for GWI,³⁷ one possibility is that the long-term experience of CMP symptoms has led these Veterans to become more resilient to chronic pain-related cognitive interference than other populations experiencing CMP.³⁸ Another possibility is that tests other than the Stroop task are more sensitive to chronic pain-related cognitive interference. For example, a recent meta-analysis reported lower performance in tests of visuospatial ability, attention, or learning and memory in symptomatic GWV compared to asymptomatic GWV.⁵ Third, longer testing durations may be required to detect cognitive performance differences between ill and healthy Veterans, as seen in ME/CFS.³⁹ Here, each fMRI run involving cognitive testing lasted less than four 4 min, thus cognitive resources may not have been challenged to the extent required to affect performance.

Concerning neural responses to the Stroop task, minimal differences were found between groups although activity in the left precentral gyrus was greater in CO than CMP during the IStroop-only but not CStroop-only run. Given that the left precentral gyrus contains a region involved in contralateral voluntary motor behavior (i.e., the primary motor cortex)⁴⁰ and neural activity occurred contralateral rather than ipsilateral to the hand operating the button press unit, this activity may be associated with motor responses to the Stroop task. Between-group differences only occurred during the IStroop-only run, possibly because it is more challenging than the CStroop task.²⁸ However, why these neural responses would differ between CMP and CO Veterans is less clear as there is not compelling evidence for altered primary motor cortex function in chronic pain conditions.^41,42 Taking these fMRI findings together with the non-significant between-group differences in Stroop performance, there was minimal evidence for chronic pain-related cognitive interference in our sample of GWV with CMP during normal testing conditions.

4.2. Experimental pain did not interfere with cognitive performance, but did elicit substantial neural differences between CMP and CO Veterans

The second aim of this study examined whether cognitive function among CMP and CO Veterans was differentially affected by a nociceptive stressor. To that end we examined the effect of experimental pain stimuli on Stroop task performance and associated BOLD signal responses. For both CStroop and IStroop models, we found significant group-by-run interactions suggesting differential response patterns to the Stroop task when combined with experimental pain. In partial support of our hypothesis, these interactions revealed greater neural activity for CMP Veterans during Pain+Stroop runs. After also considering the non-significant group-by-run interactions for CStroop or IStroop performance, we interpret these fMRI findings as follows – when stress to the nociceptive system is elevated, Veterans with CMP require a greater amount of neural resources to sustain their cognitive performance.

The effect of experimental pain on cognition has primarily been studied in young healthy adults in non-fMRI settings,^43–46 but two fMRI studies also testing the effect of experimental pain on cognition help contextualize our findings. Forkmann and colleagues examined the effect of thermal pain during a visual encoding task and memory task in 28 healthy adults (mean age=26.5).¹³ Compared to cognitive testing without pain, pain slowed reaction time for visual encoding and decreased accuracy for memory. Moreover, activity in several pain-relevant cortical and subcortical regions differed between pain and visual encoding versus visual encoding-only conditions. Thus, experimental pain affected some aspects of cognitive performance and potentially increased burden on brain regions involved in encoding of visual stimuli.

More recently and most applicable to our study design, Mathur and colleagues examined flanker task performance during painful and non-painful thermal stimulation in 14 migraine patients (10 chronic, 4 episodic) and healthy controls.¹⁴ Similar to the present study, a group-by-run interaction was observed for fMRI data but not cognitive performance, indicating that experimental pain during the flanker task differentially affected brain activity in migraine patients (decreased activity in the superior frontal gyrus) versus healthy controls (increased activity in the posterior cingulate cortex). These findings concur with the present study, suggesting that people who typically experience pain may require greater recruitment of cognitive resources in order to perform as well as pain-free individuals, particularly during increased stress to the nociceptive system.

Interestingly, compared to what we observed in the absence of experimental pain (Aim1), neural differences between CMP and CO were more robust when an experimental pain stressor was added to cognitive testing (Aim 2). An analogous example of this finding comes from an earlier study by our laboratory which examined the effect of a physical stressor on cognitive function in people with ME/CFS.⁹ We measured neural and behavioral responses to two simple cognitive tasks (i.e., finger tapping and auditory monitoring) and one more difficult task (i.e., Paced Auditory Serial Addition Task) before and 24-hours after 30 minutes of moderate intensity cycling. Compared to healthy control participants, people with ME/CFS exhibited significantly greater changes from pre- to post-exercise in the inferior and superior parietal cortices and the cingulate cortex during the Paced Auditory Serial Addition Task. Notwithstanding methodological differences between the present study and our prior work in terms of the timing (simultaneous stress versus non-simultaneous stress) and type (experimental pain versus aerobic exercise) of stressor used to affect cognition, these findings converge to support using physiological challenges to study physiology and behavior in GWI.

4.3. Limitations and future directions

Several potential shortcomings of this study should be considered. The (i) restriction of cognitive testing outcomes to one specific domain of cognition versus exploring chronic pain-related cognitive interference across a variety of tasks, (ii) inclusion of Veterans who served in the more recent Iraq War in our participant sample (although all served in the 1990–1991 Persian Gulf), and (iii) the relatively small sample size used to investigate Aim 2 may hinder the generalizability of our findings as well as our statistical power for detecting between-group differences in Stroop performance. Additionally we did not test for GWI status with a standardized instrument such as the Kansas Symptom Questionnaire, thus limiting our ability to determine whether cognitive symptoms that have been reported in other GWV studies (e.g., difficulty with memory and concentration)² were present in our sample of GWVs.

Concerning future directions, studies examining associations between CMP and performance on tasks measuring other domains of cognitive performance (e.g., visuospatial ability, attention, learning, memory) will provide a more comprehensive understanding of how cognitive function in GWVs is affected by CMP. Additionally, studying relationships between functional and structural indices of brain health such as grey matter volume and white matter integrity would bring perspective to the discussion of central nervous system dysregulation as a pathophysiological mechanism of CMP in these Veterans. Lastly, the robust neural differences observed between CMP and CO Veterans when cognitive testing was combined with nociceptive stress may signify the value of this method in gauging treatment effectiveness in clinical trials. However, higher relative stimulus intensities may be required to elicit differences in cognitive performance. Future work from our laboratory may explore this possibility in further detail.

5. Conclusion

We found more robust neural differences between CMP and CO Veterans during cognitive testing combined with thermal heat pain, but no differences in cognitive performance. These findings suggest that compared with CO Veterans, Veterans with CMP require a greater level of neural resources to sustain cognitive performance during increased nociceptive stress, suggesting dysregulation of central pain processing. Establishing the external validity of these findings requires further investigation, but a better understanding of the impact of nociceptive stressors occurring in real-world settings may help clarify why some GWV report cognitive symptoms despite showing normal neuropsychological function in controlled laboratory environments.