Characterizing Extreme Phenotypes for Pain Catastrophizing in Persons with Chronic Pain following Mild to Severe Traumatic Brain Injury Requiring Inpatient Rehabilitation: A NIDILRR and VA TBI Model Systems Collaborative Project

Abstract

Objective:

Define and characterize extreme phenotypes of pain catastrophizing for persons with chronic pain following mild to severe Traumatic Brain Injury (TBI) requiring inpatient rehabilitation.

Setting:

18 TBI Model System (TBIMS) centers.

Participants:

1762 TBIMS participants 1–30 years post-injury reporting chronic pain.

Design:

Cross-sectional, secondary analyses.

Primary Measures:

Catastrophizing, sociodemographic, injury, functional outcome, pain, and treatment characteristics.

Results:

Participants were male predominantly (73%), White (76%), middle-aged (mean 46.5 years), injured in motor vehicle accidents (53%) or falls (20%). Extreme phenotypes were identified based on upper and lower 25^th percentiles to create low catastrophizing (N=434) and high catastrophizing (N=458) groups. Bivariate comparisons found significant differences (p<.001) on all measures of concurrent function and pain interference with those in the low catastrophizing phenotype experiencing better function and lower pain interference than those in the high catastrophizing phenotype. Combination Lasso and logistic regression identified multivariable predictors of phenotypes. Increased odds of high versus low catastrophizing extreme phenotypes were associated being younger (odds ratio [OR] = 1.24 for a 10-unit decrease), less than a college level of education (OR = 1.70–2.53), no military history (OR = 3.25), lower FIM motor (OR = 1.20 for a 5-unit decrease) and cognitive (OR = 1.53 for a 5-unit decrease) scores, increased pain intensity (OR = 1.22 for a 1 unit increase) and pain interference (OR = 1.93 for a 1-unit increase), neuropathic type pain (OR = 1.82–1.86), and migraine type pain versus no head pain (OR = 1.65).

Conclusion:

High pain catastrophizing phenotypes were associated with a greater degree of pain and functional disability and higher likelihood of neuropathic pain and migraine headache. Given pain catastrophizing’s contribution to pain-related disability and treatment outcomes, additional research is necessary to investigate its role in adjustment to chronic pain among individuals with TBI. Adapting evidence-based interventions for this population that specifically targets pain catastrophizing is warranted.

Introduction

Chronic pain, which persists for longer than 3 months, is a frequent problem for persons with traumatic brain injury (TBI) and is the most cited comorbid medical condition up to 10-years post injury.¹ A recent study showed chronic pain incidence of 60% in the 3 decades following TBI.² Importantly, comorbid chronic pain is associated with worse long-term rehabilitation outcomes following TBI including reduced quality of life,³ psychiatric diagnoses,⁴ opioid-related overdose,⁵ and sleep disturbance.⁶ Chronic pain comorbidities after TBI have significant impact on psychological health and well-being and may be moderated or mediated by pain coping behaviors, which are modifiable and a potential treatment target. Specifically, pain catastrophizing, an important psychosocial factor influencing adjustment to chronic pain, is understudied among those with TBI.

Pain catastrophizing is a negative cognitive-affective response to anticipated or actual pain associated with maladjustment to these experiences via helplessness and pessimism.⁷ Pain catastrophizing is one of the most robust correlates of adverse pain experiences⁸ including pain severity in every pain population it has been studied. Higher pain catastrophizing is predictive of reduced recovery from injury⁹ and return to work,¹⁰ poorer response to pain rehabilitation interventions,¹¹ increased care utilization¹² and opioid medication misuse.¹³ Clinical trials have demonstrated that reductions in pain catastrophizing are associated with improvement on a variety of pain-related outcomes^8,14 highlighting it as a therapeutic mechanism that reduces pain and improves functioning. Pain catastrophizing is conceptualized to play a role in the chronification of pain via a fear-avoidance model, whereby increased pain-related fear leads to hypervigilance and activity avoidance contributing to deconditioning, depressive symptoms, disability, and impaired functioning – ultimately amplifying pain experiences.¹⁵

Pain catastrophizing following mild TBI (mTBI) is associated with worse pain severity, greater functional interference, and post-concussive symptoms.^16,17 Among combat Veterans who have sustained multiple TBIs, pain catastrophizing is associated with reduced executive functioning and processing speed,¹⁸ as well disability following polytrauma.¹⁹ Pain catastrophizing limits post-traumatic headache improvement at 3- and 6-months post-injury²⁰ (i.e., daily headache frequency). Engagement in physical activity 1- to 2-weeks following mTBI is predictive of both greater endogenous inhibitory pain control and lower pain catastrophizing 1-month following injury,²¹ with deficient endogenous pain inhibitory capacity and catastrophizing being associated with persistent post-traumatic headache up to 4-months post-injury.²²

There is less information about pain catastrophizing among individuals with TBI requiring inpatient rehabilitation, so factors distinguishing their individual risk for pain catastrophizing are unknown. Recent work found that catastrophizing was related to worse participation outcomes and that Black individuals were more likely to cope by catastrophizing following TBI requiring inpatient rehabilitation.²³ Given the complex role of pain catastrophizing in adjustment to chronic pain in individuals with TBI, one novel and efficient methodology to identify those at risk for poor outcomes and facilitate precision medicine is extreme phenotyping.²⁴ Phenotypes allow for the classification of an observed patient, with a specific disease, and comparison of characteristics of the patient with regard to outcomes such as disease progression (natural history) and treatment responsiveness in the context of a treatment trial. Phenotyping has been used in prior research among patients with TBI,²⁵ including those with pain and post-traumatic headache,^26,27 examining impact of pain treatment,²⁸ and how pain may be related to other outcomes.²⁹ Extreme phenotypes represent those at the extreme ends of select characteristics (e.g., best and worst outcome) and may help with identifying those at risk for the worst outcomes, and therefore be able to target specific treatments for those in the at-risk group. Among those with pain following TBI requiring inpatient rehabilitation, recent epidemiological investigations underscore pain catastrophizing as a factor in distinguishing individuals with extreme high pain interference phenotypes from low pain interference counterparts.²⁶ To date, exploration of extreme pain catastrophizing phenotypes has not been examined. The aim of this secondary analysis was to identify extreme phenotypes and characterize participants in the extreme groups in order to inform future research and potentially consider pain catastrophizing extreme groups as a stratification factor in future clinical trials.

METHODS

Participants

Data were collected as part of the National Institute on Disability, Independent Living, and Rehabilitation Research TBI Model System (TBIMS) Collaborative study “The Characterization and Treatment of Chronic Pain after TBI”² (aka Chronic Pain study). The TBIMS research network is the largest number of individuals with TBI who are enrolled and followed longitudinally after inpatient rehabilitation across several civilian and VA centers. This well-characterized cohort consists of individuals with injuries ranging from motor vehicle accidents to falls. The TBIMS provides detailed data on functional outcomes and long-term recovery, making it a valuable resource for studying chronic pain in individuals with TBI.³⁰ Eligibility criteria for the NIDILLRR civilian and VA TBIMS cohort are outlined in Table 1. Participants in the Chronic Pain study were recruited from those already enrolled in the TBIMS National Database³⁰ at participating centers (17 civilian and 1 VA TBIMS) during one of their regularly scheduled follow-up interviews (years 1, 2, 5, 10, 15, 20, 25, 30) occurring between May 1, 2019 and Aug 31, 2022. TBIMS participants were eligible for the present study if they completed their TBIMS follow-up interview in English and without a proxy. Participants in the current study were 1762 participants who endorsed “persistent or recurring pain that lasts longer than three months. It includes headaches or pain anywhere in the body, which occurs more than half of the days over a three-month period” and completed a one-time survey within 8 weeks of their TBIMS follow-up interview that included assessment of their experience of pain and pain treatment since their TBI. All centers received approval from their institutional review board.

Table 1:

Summary of Inclusion and Exclusion criteria for NIDILRR and VA TBIMS.

NIDILRR Inclusion Criteria	1) patient sustained TBI;
	2) the TBI meets a complicated mild to severe threshold based on post-traumatic amnesia, neuroimaging, loss of consciousness duration, and Glasgow Coma Scale scores upon presentation at the emergency department;
	3) patient is at least 16 years old;
	4) patient presented to the TBIMS center within 72-hours of injury;
	5) patient received both acute care and inpatient rehabilitation at the TBIMS center.
NIDILRR Exclusion Criteria	1) care related to another neurological condition (e.g., encephalopathies, stroke, seizure disorders, intracranial surgery);
	2) having a 72-hour or greater absence from the TBIMS center prior to rehabilitation admission;
	3) being in law enforcement custody at admission or prior to discharge from rehabilitation;
VA Inclusion Criteria	1) medically documented or self-reported TBI at any severity level (including disorders of consciousness);
	2) age 18 or older at time of TBI;
	3) patient presents to a VA PRC for comprehensive inpatient TBI rehabilitation.
VA Exclusion Criteria	1) presenting for care related to another neurological condition;
	2) already being enrolled in TBIMS (VA or NIDILRR stud.
Note:	The VA TBIMS does not require PRC admission within a specified period of time or require exclusively moderate to severe TBI given the high prevalence of mild TBI in recent conflicts.

Measures

Pain Catastrophizing Outcome for Extreme Phenotyping

Pain catastrophizing was measured using the ordinal catastrophizing subscale of the Coping Strategies Questionnaire.³¹ Participants were asked to rate 6 questions regarding their coping responses to pain each on a scale of 0 (never do that) to 6 (always do that). A total pain catastrophizing score was computed for each individual as the sum of these six items and ranged from 0 to 36, with higher scores indicative of higher pain catastrophizing. Items included: It’s terrible and I feel it’s never going to get any better; It’s awful and I feel that it overwhelms me; I feel my life isn’t worth living; I worry all the time about whether it will end; I feel I can’t stand it anymore; I feel like I can’t go on.

Participant Characteristics for Characterizing Extreme Pain Interference Phenotypes

Sociodemographic characteristics included self-reported sex, race, ethnicity, military history, as well as information obtained at the time of TBIMS follow-up including year post-injury of follow-up (1, 2, 5 and every 5 years through 30 years), age, marital status, level of education, employment status, primary person living with, and substance use (alcohol and illicit drugs).

Injury-related variables included cause of injury and associated spinal cord injury (SCI). Function at the time of interview was measured by FIM^™ Motor and Cognitive³² scores (range 18–91 and 5–35, respectively; higher scores indicate more independence), Disability Rating Scale (DRS)³³ scores (range 0–29; higher scores indicate more disability), and Glasgow Outcome Scale – Extended³⁴ (GOS-E; 8-point ordinal scale; higher scores indicate better global functioning).

Pain characteristics collected included intensity, interference, location, frequency for the most bothersome location, and type (neuropathic and/or headache). Pain intensity was measured using the ordinal Brief Pain Inventory (BPI)³⁵ and rated as 0 (no pain) to 10 (pain as bad as you can imagine) on four items: when pain is at its “worst,” “least,” “average,” and “right now.” The composite intensity score was calculated as the average of these four items. Pain interference was measured using the ordinal average of the 7 BPI items that query the extent to which pain interferes with general activity, mood, walking ability, normal work, relations with other people, sleep, and enjoyment with life each rated as 0 (does not interfere) to 10 (completely interferes). The painDETECT Questionnaire,³⁶ a screening tool for determining likelihood of a neuropathic component to pain was used to classify participants as positive, negative, or unclear. This nine-item questionnaire consists of seven sensory symptom items consistent with neuropathic pain rated ordinally from 0= never to 5= very strongly. Headache type was classified based on survey items previously described³⁷ resulting in three categories: migraine/probable migraine, tension/cervicogenic/unclassifiable or none.

Pain treatment history was assessed by asking if participant had received any treatment for their chronic pain including medical services (medication, injection, surgery), exercise therapies (physical therapy, pool/aquatic therapy, occupational therapy, yoga, tai chi), psychological services (psychotherapy, biofeedback, guided imagery, mindfulness therapy, hypnotherapy), complementary and alternative therapies (chiropractic, acupuncture, massage), and comprehensive pain rehabilitation programs.

Statistical Analysis

Defining Extreme Pain Catastrophizing Phenotypes

Extreme phenotypes for pain catastrophizing were data driven and followed our prior approach^26,27 of using the 25^th and 75^th percentiles of the data. Cut-points for percentiles were selected to capture as close to 25% of the sample as possible. Total pain catastrophizing scores ≤25^th percentile defined the extreme low catastrophizing phenotype representing positive outcomes and total scores ≥75^th percentile defined the extreme high catastrophizing phenotype representing negative outcomes; the remaining 50% of the sample was defined as having neutral outcome. As there has been limited work on catastrophizing in TBI, explicit cut-points for this subscale are not available. Factor analytic investigation using multiple measures has demonstrated empirically derived groups with the “poorest adaptation to pain” cohort representing approximately 30% of a pain rehabilitation sample who were distinct from other groups by their elevated catastrophizing subscale.³⁸ Additionally, prior work has demonstrated that a mean CSQ catastrophizing subscale score of 13.1 among those with moderate to severe pain (i.e., >4/10 pain intensity ratings) distinguishes those with mild pain and that these differences persist at the item level.³⁹ Current study scores were determined to be consistent with both clinical interpretations of extreme low or high catastrophizing in the context of prior research efforts.⁴⁰

Characterization and Comparisons of Extreme Phenotypes

Participant characteristics (sociodemographic, injury, functional, pain, and pain treatment) were summarized separately for low and high catastrophizing phenotypes. Means, standard deviations (SDs), and percentiles were used to summarize continuous variables and frequency counts and percentages were used to summarize categorical variables. These characteristics were each compared bivariately between the groups using two-sample t-tests or Wilcoxon rank-sum tests for continuous variables and chi-square tests for categorical variables. For categorical characteristics with more than two levels, cell-chi square values >2 were used to identify specific levels contributing to large global differences.

Multivariable Predictive Model for Discriminating Extreme Pain Interference Phenotypes

A two-step modeling approach was employed to assess the multivariable relationships between the participant characteristics (sociodemographic, injury, functional outcome, pain, treatment) and extreme pain catastrophizing phenotype (extreme low, non-extreme neutral, and extreme high catastrophizing). In the first step we performed variable selection using Least Absolute Shrinkage and Selection Operator (LASSO) regression⁴¹ with 100-fold cross-validation using the glmnet package in R. LASSO regression was chosen for variable selection because it applies a regularization to reduce overfitting, manage multicollinearity, and identify predictors with strong signals. In the LASSO model, all predictor variables were dummy coded to ensure proper handling within the model framework. The optimal lambda was selected to be the value that gives the most regularized model such that the misclassification error was within one standard error of the minimum mean cross-validated error for lambda. All variables were considered for the model selection process with the exception of DRS, GOS-E, and head pain. Headache type was selected instead of head pain as these variables are dependent and headache type contains more information regarding head pain. As such, a total of 72 (dummy coded) predictor variables measuring sociodemographic, injury, functional outcome, pain, and pain treatment characteristics were assessed for the multivariable predictive model. In the second step, the subset of predictors identified by LASSO was entered into a multinomial logistic regression model to estimate odds ratios and confidence intervals which quantify the multivariable relationship among the predictors and pain catastrophizing outcome. While the LASSO regression and multinomial models include all three phenotype outcome categories (extreme low, non-extreme neutral, and extreme high catastrophizing) to correctly parametrize the model, inference focused on the comparison between extreme high and low catastrophizing phenotypes only. Bonferroni adjustments were used to control for multiple comparisons within factors with more than two levels. Area under the Receiver Operator Curve (AUC) was used to quantify the accuracy of the model at predicting pain catastrophizing outcome.⁴²

RESULTS

Summary of the Sample

A total of 1762 participants endorsed chronic pain; this sample was previously summarized in Harrison-Felix, PhD et al². Our analytic sample included 1699 (96.4%) of those with chronic pain who completed the pain catastrophizing measure. Participants in the analytic sample were predominately male (73%), White race (76%), middle-aged (mean age 46.5 years) participants who were injured primarily in motor vehicle accidents (53%) or falls (20%). Missing data rates were the highest for variables characterizing neuropathic pain type (5.9%) and DRS (5.3%), and otherwise less than 1.4%.

Defining Extreme Pain Catastrophizing Phenotypes

Distribution of pain catastrophizing scores is shown in Figure 1 and classification of these scores into extreme pain catastrophizing phenotypes is summarized in Table 2. The distribution of pain catastrophizing scores was skewed right, with a median score of 7 (interquartile range [IQR] =1, 13) representing a moderate level of pain catastrophizing; however, scores across the entire range from 0 to 36 were reported. Participants with pain catastrophizing score ≤ 1 were classified into the extreme low catastrophizing phenotype (n = 434, 25.5%) and those with pain catastrophizing score ≥ 13 into the extreme high catastrophizing phenotype (n = 458, 27.0%).

Figure 1:

Distribution of Pain Catastrophizing Scores

Table 2:

Classification of Pain Catastrophizing Scores into Extreme Phenotypes

Pain Catastrophizing Total Score (CSQ-PC)	N = 1699
Mean (SD)	9.0 (8.7)
[Minimum, 25th percentile, median, 75th percentile, maximum]	[0, 1, 7, 13, 36]
Pain Catastrophizing Phenotypes
Extreme High Catastrophizing Phenotype [CSQ-PC ≥ 13], N (%)	458 (27.0%)
Non-Extreme Neutral Phenotype [1 < CSQ-PC < 13], N (%)	807 (47.5%)
Extreme Low Catastrophizing Phenotype [CSQ-PC ≤ 1], N (%)	434 (25.5%)

CSQ-PC = Coping Strategies Questionnaire – Pain Catastrophizing; SD = standard deviation

Characteristics of Extreme Pain Catastrophizing Phenotypes

The distribution of the sociodemographic, injury, concurrent functional, pain, and pain treatment characteristics are summarized separately for extreme low and high catastrophizing phenotypes in Table 3, along with bivariate statistical comparison of these characteristics between extreme groups. Note, these bivariate comparisons do not adjust for any other participant characteristics (e.g., marital comparison does not control for age).

Table 3:

Characteristics of Extreme Pain Catastrophizing Phenotypes

	Low Catastrophizing Phenotype (N = 434)	High Catastrophizing Phenotype (N = 458)	p-value
Sociodemographic / Participant Characteristics
Years Post-Injury, Median (IQR)	9.5 (4.5, 14.6)	5.3 (1.9, 14.6)	0.1087 b
Age at Follow-Up, Mean (SD)	49.0 (15.8)	45.0 (14.5)	0.0001 a
18–29 Years, N (%)	52 (12.0%)	69 (15.1%)	-
30–39 Years, N (%)	85 (19.6%)	112 (24.5%)	-
40–49 Years, N (%)	96 (22.2%)	113 (24.7%)	-
50–59 Years, N (%)	83 (19.2%)	82 (17.9%)	-
60–69 Years, N (%)	66 (15.2%)	59 (19.9%)	-
70–79 Years, N (%)	38 (8.8%)	16 (3.5%)	*
80–99 Years, N (%)	13 (3.0%)	7 (1.5%)	-
Sex, N Male (%)	330 (76.0%)	317 (69.2%)	0.0225 c
Race†
White, N (%)	358 (82.5%)	300 (65.8%)	< 0.0001 c
Black / African American, N (%)	55 (12.7%)	121 (26.5%)	< 0.0001 c
Asian, N (%)	9 (2.1%)	10 (2.2%)	0.9021 c
American Indian / Alaskan Native, N (%)	13 (3.0%)	18 (3.9%)	0.4388 c
Native Hawaiian / Pacific Islander, N (%)	3 (0.7%)	7 (1.5%)	0.2325 c
Hispanic, Latino, or Spanish Ethnicity, N (%)	44 (10.1%)	70 (15.3%)	0.0207 c
Marital Status at Follow-Up			< 0.0001 c
Single/Other, N (%)	125 (28.9%)	191 (41.8%)	*
Married, N (%)	199 (46.0%)	129 (28.2%)	*
Separated/Divorced/Widowed, N (%)	109 (25.2%)	137 (30.0%)
Education Level at Follow-Up			< 0.0001 c
Less than High School, N (%)	44 (10.2%)	108 (23.7%)	*
High School, N (%)	82 (19.0%)	141 (30.9%)	*
Some College / Associate’s, N (%)	155 (35.9%)	156 (34.2%)	-
Bachelor’s or Higher, N (%)	151 (35.0%)	51 (11.2%)	*
Employment Status at Follow-Up			< 0.0001 c
Employed/Student, N (%)	216 (50.0%)	113 (24.9%)	*
Retired, N (%)	145 (33.6%)	222 (48.9%)	*
Unemployed, N (%)	49 (11.3%)	94 (20.7%)	*
Other, N (%)	22 (5.1%)	25 (5.1%)	-
Military History, N Yes (%)	121 (27.9%)	87 (19.0%)	0.0017 c
Living Situation at Follow-Up			< 0.0001 c
Alone, N (%)	82 (19.0%)	75 (16.4%)	-
Spouse / Significant Other, N (%)	232 (53.7%)	174 (38.1%)	*
Other Family, N (%)	91 (21.1%)	173 (37.9%)	*
Someone Else, N (%)	27 (6.3%)	35 (7.7%)	-
Illicit Drug Use at Follow-Up, N Yes (%)	59 (13.7%)	93 (20.5%)	0.0071 c
Drinking Category at Follow-Up			< 0.0001 c
Abstaining, N (%)	173 (40.1%)	263 (58.4%)	*
Light, N (%)	108 (25.1%)	90 (20.0%)	-
Moderate, N (%)	111 (25.8%)	68 (15.1%)	*
Heavy, N (%)	39 (9.0%)	29 (6.4%)	-
Injury Severity Characteristics
Cause of Injury			0.1483 c
Vehicular, N (%)	219 (50.7%)	242 (52.8%)	-
Falls, N (%)	106 (24.5%)	86 (18.8%)	-
Violence, N (%)	57 (13.2%)	76 (16.6%)	-
Sports / Other, N (%)	50 (11.6%)	54 (11.8%)	-
Associated SCI, N Yes (%)	27 (6.3%)	40 (8.7%)	0.1605 c
Concurrent Functional Outcomes
FIM Motor at Follow-Up, Median (IQR)	90 (87, 91)	85 (78, 89)	< 0.0001 b
FIM Cognitive at Follow-Up, Median (IQR)	33 (31, 34)	30 (27, 33)	< 0.0001 b
DRS at Follow-Up, Median (IQR)	1 (0, 1)	4 (2, 5)	< 0.0001 b
GOSE at Follow-Up			< 0.0001 c
Severe Disability, N (%)	41 (9.5%)	158 (35.0%)	*
Moderate Disability, N (%)	171 (39.8%)	236 (52.2%)	*
Good Recovery, N (%)	218 (50.7%)	58 (12.8%)	*
Chronic Pain Characteristics			p-value
Pain Intensity, Mean (SD)	3.7 (1.7)	5.9 (1.8)	< 0.0001 a
Pain Interference, Mean (SD)	2.5 (2.0)	6.4 (2.1)	< 0.0001 a
Pain Location‡, Median (IQR)	2 (1, 5)	5 (3, 7)	< 0.0001 b
Back, N Yes (%)	246 (56.8%)	322 (70.3%)	< 0.0001 c
Legs or Feet, N Yes (%)	242 (55.8%)	302 (66.1%)	< 0.0001 c
Shoulder, N Yes (%)	179 (41.2%)	253 (55.2%)	< 0.0001 c
Head, N Yes (%)	145 (33.5%)	280 (61.1%)	< 0.0001 c
Neck, N Yes (%)	147 (33.9%)	239 (52.2%)	< 0.0001 c
Arms or Hands, N Yes (%)	147 (33.9%)	194 (42.4%)	0.0099 c
Hips, N Yes (%)	125 (28.8%)	185 (40.4%)	0.0003 c
Face or Jaw, N Yes (%)	51 (11.8%)	109 (23.8%)	< 0.0001 c
Pelvic Area or Groin, N Yes (%)	55 (12.7%)	88 (19.2%)	0.0081 c
Buttocks, N Yes (%)	40 (9.3%)	64 (14.0%)	0.0287 c
Widespread Pain or Fibromyalgia, N Yes (%)	30 (6.9%)	74 (16.2%)	< 0.0001 c
Chest, N Yes (%)	24 (5.6%)	71 (15.5%)	< 0.0001 c
Abdomen, N Yes (%)	26 (6.0%)	74 (16.2%)	< 0.0001 c
Frequency of Pain (Most Bothersome Location)			< 0.0001 c
Several Times a Week, N (%)	136 (32.1%)	78 (17.3%)	*
Daily, N (%)	194 (45.7%)	184 (40.8%)	-
Constant, N (%)	94 (22.2%)	189 (41.9%)	*
Neuropathic Pain Type			< 0.0001 c
Clinically Significant Positive, N (%)	28 (6.9%)	120 (28.0%)	*
Clinically Significant Negative, N (%)	333 (81.8%)	199 (46.4%)	*
Unclear, N (%)	46 (11.3%)	110 (25.6%)	*
Headache Type§			< 0.0001 c
Migraine/Probable Migraine, N (%)	100 (23.1%)	245 (54.0%)	*
Other Headache, N (%)	45 (10.4%)	31 (6.8%)	-
No Head Pain, N (%)	288 (66.5%)	178 (39.2%)	*
Treatment History Characteristics
Total Medical Services or Treatments, N (%)	385 (89.7%)	439 (96.5%)	< 0.0001 c
Medication, N (%)	368 (87.0%)	432 (95.2%)	< 0.0001 c
Injection, N (%)	116 (26.9%)	162 (35.6%)	0.0053 c
Surgery, N (%)	73 (16.9%)	77 (16.9%)	0.9955 c
Implanted Device, N (%)	8 (1.9%)	22 (4.8%)	0.0143 c
Non-Implanted Electrical Stimulator, N (%)	120 (27.8%)	110 (24.2%)	0.2134 c
Total Exercise Therapies, N (%)	332 (77.0%)	370 (81.3%)	0.1158 c
Physical Therapy, N (%)	271 (63.0%)	317 (69.7%)	0.0363 c
Pool or Aquatic Therapy, N (%)	113 (26.3%)	119 (26.2%)	0.9662 c
Occupational Therapy, N (%)	136 (31.6%)	200 (44.0%)	0.0002 c
Yoga, N (%)	105 (24.4%)	89 (19.6%)	0.0808 c
Tai Chi, N (%)	20 (4.7%)	24 (5.3%)	0.6697 c
Home Exercise Program, N (%)	258 (59.9%)	263 (57.8%)	0.5338 c
Total Psychological Services, N (%)	150 (34.8%)	232 (51.0%)	< 0.0001 c
Psychotherapy (e.g., CBT, ACT), N (%)	48 (11.1%)	103 (22.6%)	< 0.0001 c
Biofeedback, N (%)	29 (6.7%)	29 (6.4%)	0.8238 c
Deep Breathing, N (%)	123 (28.5%)	184 (40.4%)	0.0002 c
Guided Imagery, N (%)	41 (9.5%)	67 (14.7%)	0.0184 c
Mindfulness Therapy, N (%)	80 (18.6%)	87 (19.2%)	0.8192 c
Hypnotherapy, N (%)	12 (2.8%)	12 (2.6%)	0.8929 c
Total Complementary/Alternative Medicine, N (%)	239 (55.6%)	264 (58.0%)	0.4638 c
Chiropractic, N (%)	130 (30.2%)	134 (29.5%)	0.7994 c
Acupuncture, N (%)	86 (20.0%)	94 (20.7%)	0.8076 c
Massage, N (%)	189 (44.0%)	208 (45.7%)	0.5986 c
Comprehensive Pain Rehabilitation Program, N (%)	45 (10.5%)	75 (16.5%)	0.0090 c

SD = standard deviation; IQR = interquartile range;

^†Participants may self-report more than one race;

^‡Participants may self-report more than one pain location;

^§Headache Type was classified for those who reported head pain;

SCI = spinal cord injury; FIM = Functional Independence Measure; DRS = Disability Rating Scale; GOS-E = Glasgow outcome scale – extended; BPI = Brief Pain Inventory;

^a:two-sample test;

^b:Wilcoxon rank-sum test;

^c:chi-square test;

bold indicates statistically significant differences at α = 0.05;

^*indicates categorical levels with large (>2) cell chi-squares contributing to global differences

The low catastrophizing phenotype was characterized by older age, higher rates of male sex, White race, non-Hispanic ethnicity, married, college education, employed/student, military history, living with a spouse/significant other, moderate alcohol use, higher FIM motor and cognitive scores (more independence), lower DRS scores (less disability), good recovery on the GOS-E, lower pain intensity, less frequent pain, non-neuropathic pain type, and no head pain. The high catastrophizing phenotype, on the other hand, was characterized by younger age, higher rates of female sex, Black race, Hispanic ethnicity, never being married, high school education or lower, being retired or unemployed, living with other family, illicit drug use, abstaining from alcohol use, lower FIM motor and cognitive scores (less independence), higher DRS scores (more disability), severe disability or moderate disability on GOSE, higher pain intensity, constant pain, neuropathic type pain, and migraine type pain. Individuals in the high catastrophizing phenotype also tended to endorse more pain locations and more treatment utilization than those in the low catastrophizing phenotype.

Multivariable Predictive Model for Discriminating Extreme Pain Catastrophizing Phenotypes

The LASSO regression process identified the following set of variables as important unique predictors of extreme phenotypes for pain catastrophizing (in order of relative importance): pain interference, military history, age, pain intensity, FIM cognitive, FIM motor, neuropathic pain type, level of education, and headache type. The estimated odds ratios and 95% confidence intervals discriminating extreme high catastrophizing and low catastrophizing phenotypes based on the multinomial logistic regression model with this set of predictors are summarized in Table 4. Significantly increased odds of being in the extreme high catastrophizing versus extreme low catastrophizing phenotype were associated with younger age (OR = 1.24 for a 10-year decrease), below a college level of education (OR = 1.70–2.53), no military history (OR = 3.25), lower FIM motor (OR = 1.20 for a 5-unit decrease) and FIM cognitive (OR = 1.53 for a 5-unit decrease) scores, higher pain intensity (OR = 1.22 for a 1-unit increase), higher pain interference (OR = 1.93 for a 1-unit increase), neuropathic type pain (OR = 1.82–1.86), and migraine type headache versus no head pain (OR = 1.65). Pairwise AUC for discriminating extreme low catastrophizing and high catastrophizing phenotypes was 0.922 (outstanding). Sensitivity analyses excluding the single VA site found overall results to be largely consistent with the full sample. However, the significance of differences related to education (HS vs College) and military history were not significant when excluding these cases.

Table 4:

Multivariable Logistic Model for Discriminating Extreme Low versus Extreme High Pain Catastrophizing Phenotypes (N = 1556 of 1699)

	Extreme High vs Extreme Low Pain Catastrophizing Phenotype
	Odds Ratio	95% CI	p-value
Age at Follow-Up (10-year decrease)	1.235	(1.093, 1.395)	0.0007	*
Level of Education at Follow-Up
Less than High School vs College	1.703	(0.906, 3.201)	0.0983	-
High School vs College	2.532	(1.470, 4.362)	0.0008	*
Some College vs College	1.862	(1.140, 3.043)	0.0131	-
Less than High School vs Some College	0.914	(0.529, 1.582)	0.7489	-
High School vs Some College	1.360	(0.858, 2.155)	0.1906	-
Less than High School vs High School	0.673	(0.377, 1.199)	0.1784	-
Military History (No vs Yes)	3.254	(2.075, 5.103)	< 0.0001	*
FIM Motor at Follow-Up (5-unit decrease)	1.202	(1.059, 1.365)	0.0045	*
FIM Cognitive at Follow-Up (5-unit decrease)	1.531	(1.148, 2.041)	0.0037	*
Pain Intensity (1-unit increase)	1.224	(1.078, 1.389)	0.0018	*
Pain Interference (1-unit increase)	1.927	(1.731, 2.146)	< 0.0001	*
Neuropathic Pain Type
CS Positive vs CS Negative	1.818	(1.038, 3.186)	0.0367	-
Unclear vs CS Negative	1.885	(1.194, 2.974)	0.0065	*
CS Positive vs Unclear	0.965	(0.509, 1.828)	0.9124	-
Headache Type
Migraine vs No Headache	1.648	(1.095, 2.480)	0.0167	*
Migraine vs Other Headache	1.289	(0.654, 2.540)	0.4636	-
Other Headache vs No Headache	1.279	(0.671, 2.438)	0.4552	-

CI = confidence interval; CS = clinically significant;

^*statistically significant at α = 0.05 or after Bonferroni correction of α = 0.0083 (level of education) or α = 0.0167 (neuropathic pain type and headache type)

DISCUSSION

To our knowledge, this study is the first to define and explore the concept of extreme phenotypes in pain catastrophizing among individuals with mild to severe TBI requiring inpatient rehabilitation and comorbid chronic pain across 30 years of recovery. Generally, the low catastrophizing phenotype had better functional outcomes, lower pain-related characteristics, and fewer pain treatment utilization rates compared to the high catastrophizing phenotype, with the exception of utilization of recreational and complementary therapies. Factors that uniquely discriminated high catastrophizing from low catastrophizing phenotypes included younger age, not having a college education, no military history, lower FIM motor and cognitive scores, higher pain intensity, higher pain interference, and reports of neuropathic and/or migraine type pain.

The study provides evidence that pain catastrophizing is a robust correlate of adverse pain experiences⁸ among individuals with TBI requiring inpatient rehabilitation. Use of pain catastrophizing to define and characterize individuals yielded distinct profiles of individuals whose characteristics can highlight those most likely to endorse maladaptive pain responses. Lower functioning with high catastrophizing matches the mild TBI literature, where greater pain catastrophizing is associated with higher pain intensity or severity, worse cognitive functioning, or persistent cognitive and functional difficulties.^16–18

Our finding that not having a college level of education was associated with higher pain catastrophizing reflects the extant literature on chronic pain,⁴³ and adds to current knowledge of socioeconomic characteristics (e.g., Black race, public insurance³²) associated with greater pain catastrophizing following TBI. Education has been identified as a moderator in the relationship between cognitive and/or affective responses to pain (e.g., pain catastrophizing, depression, kinesiophobia) and functional outcomes.⁴⁴ This suggests certain socioeconomic risk factors for worse pain outcomes may be due in part to maladaptive coping responses. Interventions with strategies in mind to reduce individual disparities related to education on active coping may minimize engagement in pain catastrophizing, and mitigate the impact of chronic pain on outcomes. Our finding that those in the high pain catastrophizing phenotype were younger may reflect general differences in how pain catastrophizing changes over the lifetime and is consistent with literature that pain catastrophizing tends to decrease with age.⁴⁵ In the current study catastrophizing was measured in terms of helplessness to effectively respond and adapt to pain. Our finding may reflect a tendency among younger adults in which pain catastrophizing is preferentially associated with emotional responses, whereas in older adults it is more likely to be preferentially associated with pain intensity ratings.⁴⁶

Individuals with no military experience were more likely to reflect the high pain catastrophizing phenotype. This is surprising considering the insalubrious impact of catastrophizing on pain related outcomes and the extant literature indicating greater chronic pain prevalence as well as more severe and disabling pain among military Veterans.⁴⁷ Veterans and service members also experience post-traumatic stress disorder and other psychiatric conditions associated with increased pain-related catastrophizing at higher rates.⁴⁸ Comparisons on pain catastrophizing between Veterans and nonveterans with TBI is limited. In the current study, participants from VA centers are frequently involved in long-term rehabilitative care programs tailored to their complex needs which differ from civilian care settings and may influence recovery trajectories and ultimately appraisal of pain experiences by the individual. Earlier work comparing civilian versus VA pain care provider perceptions reflects determinants that may differentially support the ability of those with TBI at multiple touchpoints in the process of pain treatment access received within VA beyond rehabilitative care.^49,50 Additionally, VA has implemented systematic provider trainings to ensure that Veterans have access to evidence-based psychotherapeutic approaches, including cognitive behavioral therapy for chronic pain.⁵¹ It is possible that some military and Veteran participants received this intervention, a major treatment benefit of which is reduced pain-related catastrophizing, regardless of where they received their initial TBI care.

It is noteworthy that likely neuropathic and migraine type pain were both associated with high pain catastrophizing phenotype, as pain catastrophizing may be more prevalent in contexts where pain may be experienced as more debilitating, uncontrollable, or unpredictable – reinforcing beliefs of helplessness in terms of pain management.^15,52 The role that pain catastrophizing may play for these pain types following TBI may be a critical factor in preventing the chronification and amplification of pain.²² This is especially important for the TBI population, who have been shown to have higher rates of migraine type headache^37,53 than expected from general population data. This suggests that persons with TBI may be more vulnerable to pain catastrophizing compared to the general population.

Strengths and Limitations

This study addresses pain catastrophizing, known to be an important factor influencing pain adjustment, chronification, and function. Extreme phenotyping, widely utilized in precision medicine to identify individuals with the best and worst outcomes, is a novel approach to learn about the complex problem of chronic pain. Additionally, the recruitment approach utilized a large, well-characterized TBIMS cohort 1– 30 years post-TBI.

The cross-sectional derepresign of the current study precludes causal inference. A wide range of time since injury (1–30 years) in the current sample allows us to examine pain across time, but the lack of repeated intraindividual measures limits our ability to identify changes in catastrophizing across time. Information regarding the context (i.e., military versus civilian) in which the index TBI occurred is not available. Where cause of injury is reported (e.g., falls, MVAs, violence), findings should not be interpreted to exclusively mean military or combat related. Groups were defined using quartiles, similar to other phenotype approaches, to reflect generally accepted “at risk” elevations; however, phenotype groups could be defined and compared using more extreme definitions (e.g., 10^th percentiles) or using statistical classification (e.g., ±1.96 SD). To reduce participant burden, a subscale of the Coping Strategies Questionnaire (i.e., helplessness domain) was used to measure catastrophizing. Including additional domains of this construct (e.g., rumination, magnification) may yield different results. Analyses conducted herein are constrained by data available, and unmeasured confounders may bias results. As with all studies relying on self-report, findings may be subject to recall bias.

While the TBIMS sample has been established to be largely representative of the population of adults admitted for acute, inpatient rehabilitation in the US with a primary diagnosis of TBI, the TBIMS tends to include participants with overall younger age at time of injury, higher employment rates at injury (and lower retirement rates), and longer rehabilitation length of stays.⁵⁴ It is important to note that of the 1762 participants included in this study, 286 were from a single VA polytrauma rehabilitation center and when excluded, phenotypes were no longer distinguished based on military history in multivariate models. Approximately 13% of participants in the Traumatic Brain Injury Model Systems (TBIMS) database are Veterans or service members who received TBI rehabilitative care at civilian sites.⁵⁵ In the current study, these participants represented approximately 16% (n=120) of the sample. Although effect sizes were small, participants who completed the Pain Survey tended to be younger, had higher levels of education, and tended to have better functional outcomes compared to TBIMS participants who did not complete the Pain Survey.² These differences may affect the generalizability of our results to the larger TBI population.

CONCLUSION

Extreme phenotyping was applied to pain catastrophizing in a community-based sample of individuals with chronic pain enrolled in the TBIMS, 1–30 years post-TBI. Younger age, lower education, no military history, reduced motor and cognitive function, greater pain intensity and pain interference, and neuropathic type pain and/or migraine type headache distinguished high from low pain catastrophizing phenotypes. Extreme phenotypes based on pain catastrophizing may be useful for future research prioritizing pain-focused interventions specifically in TBI subpopulations where this cognitive-affective response to pain is likely to impede adaptive reactions to the experience of pain, potentially altering the course of pain chronification in at risk polytrauma samples. A clinical focus on adapting evidence-based interventions for this population that specifically target pain catastrophizing with these factors in mind is warranted.