Biobehavioral Predictors of Pain Intensity, Pain Interference, and Chronic Pain Episodes: A Prospective Cohort Study of African-American Adults

Abstract

Racial disparities in pain experiences are well-established, with African-American adults reporting higher rates of daily pain, increased pain severity, and greater pain-related interference compared to non-Hispanic Whites. However, the biobehavioral factors that predict transition to chronic pain among African-American adults are not well understood. This prospective cohort study provided a unique opportunity to evaluate predictors of chronic pain onset among 130 African-American adults (81 women), ages 18 to 44, who did not report chronic pain at their baseline assessment and subsequently completed follow-up assessments at 6- and 12-months. Outcome measures included pain intensity, pain-related interference, and chronic pain status. Comprehensive assessments of sociodemographic and biobehavioral factors were used to evaluate demographics, socioeconomic status (SES), stress exposure, psychosocial factors, prolonged hypothalamic-pituitary-adrenal secretion, and quantitative sensory testing (QST) responses. At baseline, 30 adults (23.1%) reported a history of prior chronic pain. Over the 12-month follow-up period, 13 adults (10.0%) developed a new chronic pain episode and 18 adults (13.8%) developed a recurrent chronic pain episode. Whereas SES measures (i.e., annual income, education) predicted changes in pain intensity over the follow-up period, QST measures (i.e., pain threshold, temporal summation of pain) predicted changes in pain interference. History of chronic pain and higher depressive symptoms at baseline independently predicted onset of a new chronic pain episode. The present findings highlight distinct subsets of biobehavioral factors that are differentially associated with trajectories of pain intensity, pain-related interference, and onset of chronic pain episodes in African-American adults.

Racial differences in the experience, impact, assessment, and treatment of pain are pervasive in the United States [3; 51]. African-American (AA) adults are more likely than their non-Hispanic White (NHW) counterparts to exhibit heightened pain sensitivity, pain intensity, and pain-related interference [3; 57; 97]. Following a traumatic injury, AA adults are more likely than NHW adults to transition from acute to chronic pain [74–76] and to suffer from long-term pain-related life interference [36]. Following surgical procedures, AA adults are also more likely than NHW adults to develop chronic postsurgical pain (CPSP) [90]. Compounding these disparities among those who develop pain, AA adults are less likely than NHW adults to receive services across the continuum of pain care [3; 57].

The disproportionate impact of pain on minoritized racial groups is likely to be explained by a variety of biobehavioral factors [41; 45; 77]. To date, evidence of biobehavioral risk factors for pain comes from predominantly non-minoritized samples followed in epidemiological, post-surgical, and posttraumatic injury designs. These studies evaluate predictors of continuous (i.e., pain intensity/interference) and dichotomous (i.e., chronic pain onset) outcomes; the former provide insight into predictors of changes (i.e., slopes) in pain worsening or resolution within individuals, which are critical for understanding pain chronification [2; 49; 65]. Prospective studies of large community samples evaluating first-onset of chronic pain conditions [1; 30; 46; 59; 66; 72; 80; 88; 104; 109; 112] have identified the following predictors: female gender, older age, low socioeconomic status [SES]), stress exposure, greater depressive and somatic symptoms, quantitative sensory testing (QST) responses, and sleep difficulties. One study of healthy individuals recently hired into an occupation deemed high risk for neck pain found that QST responses (i.e., conditioned pain modulation [CPM]) and depressive symptoms predicted chronic pain onset [101]. Prospective studies of individuals undergoing planned surgical interventions [8; 50; 61; 95; 96; 100; 116] have identified several predictors of CPSP, including female gender, lower education level, lack of employment, weaker CPM, higher anxiety, pain cognitions, and lower social support. Prospective studies following individuals recently exposed to traumatic injuries/stressors [5; 26; 55; 64; 113] indicate that childhood trauma, elevated stress responses, and greater pain catastrophizing and depressive symptoms predict worse pain outcomes. Few prospective studies, however, have evaluated racial differences in biobehavioral determinants of pain [9; 10] and none, to our knowledge, have focused on predictors of chronic pain specifically in healthy AA individuals.

The present study addressed several critical gaps in our understanding of longitudinal changes in pain intensity and interference (i.e., pain trajectories) and the onset of chronic pain episodes. First, we focused on healthy AA adults without current chronic pain. This allowed us to evaluate biobehavioral predictors of pain outcomes without needing to consider variables important to other prospective study designs (e.g., injury characteristics, surgical procedures). Understanding how and why pain worsens among minoritized racial groups is particularly challenging following injury or planned surgery given the potential influence of racial health care inequities [3]. Second, we considered the unique contributions of a broad array of theory-driven and empirically-supported biobehavioral factors implicated in risk for negative pain outcomes, including demographic, socioeconomic, stress exposure, stress response, psychosocial, and QST responses [18]. A comprehensive assessment of biobehavioral factors is essential given evidence for the multifactorial nature of pain [41] and involvement of a variety of biobehavioral factors, each associated with small effects but collectively contributing to more accurate prediction of pain [64]. To date, however, broad assessments of biobehavioral factors within the same individuals are relatively uncommon [77]. Focusing on only a subset of biobehavioral factors is likely to limit understanding of pain risk, because “pain is sculpted by a mosaic of factors that is completely unique to each individual at a given point in time, and this mosaic must be considered in order to provide optimal pain treatment” [41]. Third, we examined similarities and differences in relations between biobehavioral factors and distinct pain outcomes: within-person changes (trajectories) of pain intensity and pain-related interference and first onset/recurrence of chronic pain episodes.

METHODS

Participants

This prospective study included 130 healthy adults (i.e., no current chronic pain condition) who self-identified as AA (81 women [62%]) and were between the ages of 18 to 44 (inclusive). Recruitment was conducted in a mid-size metropolitan area through online research participant registries, community and university webpages advertising research studies, flyers distributed at local Historically-Black Colleges and Universities, and flyers placed in waiting rooms of local clinics serving the AA community. Participants were excluded if they had a current chronic pain condition at baseline assessment (defined below in ‘Pain measures’ section). This exclusion criterion was due to our particular interest in identifying early markers of risk for daily pain among healthy individuals that could inform the development of programs designed to prevent chronic pain episodes. Additional exclusion criteria were presence of sickle cell disease, medical condition (e.g., Cushing’s disease, hyperthyroidism, pregnancy) or taking daily medications (e.g., opioid analgesics, corticosteroids) that could affect pain or stress (i.e., hypothalamic-pituitary-adrenal axis) responses, or meeting criteria for a substance use disorder in the three months prior to baseline assessment (see Supplemental Figure 1 for details). This study was approved by the institutional review boards at Meharry Medical College and the University of Mississippi Medical Center.

Participants included in this study were those seen at baseline and at least one of two scheduled follow-up assessments at 6- and 12-months after baseline. Each of these three assessments included the same protocol for hair sampling, QST, and self-report questionnaires (administered via research electronic data capture system [REDCap]). The data inclusion rule was developed due to a substantial number of missed follow-up assessments for enrolled participants during cessation of non-essential research activities at the institution caused by the COVID-19 pandemic. Of the 130 AA adults who completed at least one follow-up assessment, 17 and 50 missed their 6- and 12-month follow-ups, respectively.

For analyses examining the development of a chronic pain episode at follow-up (new onset or recurrence of past chronic pain), data analysis was restricted to participants who had either (a) confirmed absence of chronic pain across all three assessments or (b) confirmation of an incident chronic pain episode (defined below in ‘Pain measures’ section) for at least one follow-up assessment. This resulted in 78 participants (45 women) contributing to analyses of chronic pain prediction and trajectories of biobehavioral factors distinguishing chronic pain from pain-free AA adults. Participants contributing to analyses of chronic pain onset did not differ significantly from those included in analyses of pain intensity and interference trajectories on any baseline demographic characteristics, biobehavioral factors, or pain outcomes (p’s>.05) with the exception of CPM (t=2.02, p=.045). AA adults who contributed data to the chronic pain episode onset analyses exhibited stronger pain inhibition (mean CPM=9.19, SD=19.24) than those who contributed data only to the trajectory analyses (mean CPM=2.35, SD=18.13).

Pain measures

The key outcomes in this study targeted changes in pain status over the follow-up period. These were assessed in two ways: 1) changes in continuous measures of pain intensity and pain interference (to capitalize on both the larger sample size available and the greater statistical power of continuous measures) and 2) first occurrence or recurrence of a chronic pain episode (dichotomous). For the latter analyses, we combined individuals who were pain-free at baseline and developed either a new or recurrent chronic pain episode over follow-up due to our interest in identifying potential targets for both primary prevention (among those with no prior chronic pain history) and secondary prevention (among those with a history of chronic pain but currently not in an episode). The McGill Pain Questionnaire-Short Form (MPQ-SF [79]) was used to assess the intensity of participants’ daily/ongoing pain. Participants indicated on a four-point scale (0 = “None” to 3 = “Severe”) the extent to which they had experienced a variety of sensory and affective pain characteristics. The MPQ-SF total score has good to excellent psychometric properties [60]. The MPQ-SF Pain Rating Index (MPQ-SF-PRI) was computed as the sum of ratings on all 15 items, with higher scores reflecting greater pain intensity (score range from 0 to 45). Cronbach’s alphas for the present sample indicated excellent internal consistency at all assessments (α’s ranged from .94–.96).

The PROMIS Pain Interference – Short Form (PROMIS-INT [87]) is an 8-item measure that was used to assess the degree to which pain interfered with daily life activities in the past seven days. Item scores ranged from 1 (not at all) to 5 (very much), with total scores ranging from 8 to 40 (higher scores reflect greater pain-related interference). Cronbach’s alphas for the present sample indicated excellent internal consistency at all assessments (α’s ranged from .92–.97).

Dichotomized chronic pain status was determined using a modified version of the Persistent Pain Questionnaire (PPQ[23]) with presence of recent chronic pain defined as experiencing pain for the past 3 or more months, occurring daily or almost every day, with a typical pain rating of ≥ 30 out of 100, in at least one of 8 body regions (i.e., head; neck; shoulder, arm, or hand; chest; abdomen; pelvic area; upper or lower back; leg, foot, knees, ankles). This approach has been used in prior work [107; 114].

Sociodemographic measures

A brief demographics form was used to determine participant age, gender, education level (years completed), and total annual family income (1=$0–10,000; 2=$10,001–20,000; 3=$20,001–30,000; 4=$30,001–40,000; 5=$40,001–50,000; 6=$50,001–60,000; 7=$60,001–70,000; 8=$70,001–80,000; 9=$80,001–90,000; 10=$90,001–100,000; 11=$100,000+; “Estimate not available”).

Stress exposure measures

The Childhood Trauma Questionnaire (CTQ[15]) is a 28-item, self-report measure used to assess the frequency of different types of abuse experienced as a child and adolescent. CTQ item scores ranged from 1 (never true) to 5 (very often true), with total scores ranging from 28 to 140 (higher scores reflect more childhood trauma exposure). Reliability for the CTQ at baseline was good (α=0.82).

A brief 10-item version of the Family Adversity Questionnaire (FAQ) was used to assess the number of non-sexual, adverse early life experiences, including parental incarceration, illness, disability, death and severe poverty [62]. Items were dichotomous (0 = No; 1 = Yes), with total scores ranging from 0 to 10 (higher scores reflect more adverse early life experiences). Reliability for the FAQ at baseline was good (α=0.76).

The Chronic Burden Scale (CBS) is a 21-item scale used to measure the degree to which a variety of chronic stressors (e.g., economic, employment, crime, and legal problems) had been a problem for participants in the past 6 months [53]. CBS item scores ranged from 1 (not a problem for me) to 4 (a major problem for me), with total scores ranging from 15 to 84 (higher scores reflect greater burden from chronic stressors). Cronbach’s alphas for the present sample indicated good internal consistency at each assessment (α’s ranged from .83–.87).

The Perceived Stress Scale-10 (PSS-10 [29]) is a 10-item measure used to assess the degree to which individuals perceive their lives as stressful in the past month. PSS-10 item scores ranged from 0 (never) to 4 (very often), with total scores ranging from 0 to 40 (higher scores reflect higher levels of perceived stress). Cronbach’s alphas for the present sample indicated good internal consistency at each assessment (α’s ranged from .84–.86).

The 17-item Brief Perceived Ethnic Discrimination Questionnaire – Community Version (BPEDQ-CV [21]) was used to assess lifetime experiences of discrimination because of one’s race and ethnicity. Participants indicated the frequency of experiences with racial discrimination on a scale of 1 (never) to 5 (very often), with total scores ranging from 17 to 75 (higher scores reflect more experiences with racial discrimination). Cronbach’s alphas for the present sample indicated good-to-excellent internal consistency at each assessment (α’s ranged from .87–.91).

Psychosocial measures

Current depression severity was assessed with the Beck Depression Inventory second edition (BDI-II [12]). The BDI-II is a 21-item scale with scores ranging from 0 to 63, with higher scores indicating more severe depressive symptoms. Cronbach’s alphas for the present sample indicated excellent internal consistency at each assessment (α’s ranged from .91–.96).

The Pain Catastrophizing Scale (PCS [108]), a 13-item self-report measure, was used to measure dispositional tendencies to engage in catastrophic thought while experiencing pain. Participants indicated the degree to which they experience catastrophic thoughts during a painful experience on a scale of 1 (not at all) to 4 (all the time). Scores range from 0 to 52, with higher scores indicating a greater tendency to engage in catastrophic thoughts. Cronbach’s alphas for the present sample indicated excellent internal consistency at each assessment (α’s ranged from .92–.93).

The Pain Resilience Scale (PRS [105]) is a 14-item self-report measure that was used to assess cognitive/affective positivity and behavioral perseverance during a painful experience. Participants were asked to report how frequently they experienced cognitive, emotional, or behavioral responses to pain on a scale of 0 (not at all) to 4 (all the time). Scores range from 0 to 56, with higher scores indicating higher levels of pain-specific resilience. Cronbach’s alphas for the present sample indicated excellent internal consistency at each assessment (α’s ranged from .95–.96).

Prolonged Hypothalamic-Pituitary-Adrenal (HPA) secretion

Prolonged HPA secretion – a physiological proxy of response to ongoing chronic stress distinct from self-report measures of perceived stress [106] - was determined by hair cortisol concentrations (HCC), as previously described [17; 44; 52; 111]. Three cm hair segments (hair closest to the posterior vertex of the scalp) were collected and stored at room temperature in dry and dark conditions. Each segment provided an index of cumulative cortisol secretion integrated over the previous three months of hair growth. Hair was washed twice with isopropanol at room temperature for 30 seconds to remove external contamination. Hair was dried, weighed using a high-precision analytical balance, and milled using 7 mm stainless steel balls. Cortisol was extracted with methanol overnight, acetone for 5 minutes, and then with methanol overnight once more. Pooled solvent fractions were removed under a compressed air stream. Samples (along with a blank) were dissolved in the assay diluent, distributed to avoid a batch effect, and analyzed in duplicate using Arbor Assays cortisol enzyme-linked immunosorbent assay. The R² value for testing linear fit for cortisol amount versus the amount of milled homogenized sample in the 5 to 25 mg range was 0.99 (linearity experiment). For 6 homogenized individually extracted replicates, the cortisol extraction variability was 8.5% (replication experiment). The serial dilutions of ELISA standards versus the experimental cortisol extract (parallelism experiment) produced results as expected. For validation, we satisfactorily completed linearity, parallelism, and replication experiments. The median HCC was 6.5 pg/mg and the limit of detection was 1.6 pg/mg. Intra-plate variability (1.7%) and inter-plate variability (3.3%) were both excellent. If relative standard deviation for a sample was over 7%, the measurement was repeated (5% of all samples were randomly repeated). All high-concentration samples were diluted and remeasured. HCC measurements were conducted at Stress Bioanalytics LLC.

Quantitative sensory testing protocol

The QST protocol in this study has been used by this group with a variety of populations to assess static (i.e., heat pain threshold and tolerance) and dynamic (i.e., TSP and CPM) evoked pain responses [81; 85]. Participants were instructed not to take any pain medications within four hours of their visit. Upon arrival, they completed orientation and training for pain testing procedures and stimuli. The QST protocol was administered through a computerized Medoc TSA-II NeuroSensory Analyzer (Medoc US, Minneapolis, MN), using commercially available software (TPS-CoVAS version 3.19; Medoc Inc, Ramat Yishay, Israel), as well as a Boekel General Purpose Water Bath (Boekel Scientific, Feasterville, PA).

Heat pain threshold and tolerance were both determined using a thermode attached to the ventral forearm of each participant’s nondominant arm. For heat pain threshold, participants were instructed to terminate the stimulus by clicking on a computer mouse when they first perceived it as “painful”. For heat pain tolerance, participants were instructed to terminate the stimulus by clicking on a computer mouse “when you can’t stand the heat pain any longer.” Four trials each for the threshold and tolerance tasks started at an adaptation temperature of 32°C (threshold) or 40°C (tolerance), followed by temperature increase at a ramp rate of 0.5°C per second until either threshold or maximum tolerance were reached. The thermode was moved upwards on the forearm to a new, nonoverlapping location during each 25-second interstimulus interval. The maximum temperature limit was 51°C. Heat pain threshold and tolerance were each computed as the mean of the temperatures for the last three trials.

TSP was assessed with a standardized oscillating thermal stimulation protocol used previously by this group [27; 33; 84] and others [42]. A sequence of 10 heat pulses with a 48°C target stimulus intensity was applied to the ventral forearm. Each pulse was 0.5 seconds in duration and started at a temperature of 40°C, with sequential pulses administered at a frequency of 0.4 Hz. During the TSP protocol, participants rated the intensity of pain sensation shortly after the peak of each heat pulse using a 0 (no pain) to 100 (worst pain possible) scale. TSP was computed for each individual as the change from their first pain rating to their highest pain rating (larger positive values indicate more pronounced TSP).

For the CPM protocol, the test stimulus was the thermode applied to the ventral nondominant forearm and the conditioning stimulus was the hot water bath maintained at a steady temperature of 46.5°C in accordance with established guidelines [116]. First, a “P-60” temperature was determined for each participant as the thermode temperature eliciting pain ratings between 50 and 70. The thermode was applied to the nondominant ventral forearm in sequences of 15-second pulses at 45°C, 46°C, 47°C, and additional higher or lower temperatures as warranted until P-60 was identified. Second, three preconditioning pain ratings – at 10-second intervals - were obtained for the P-60 temperature applied to a nonoverlapping location on the nondominant ventral forearm for a 30-second period. Third, participants took a 5-minute break from the QST protocol. Fourth, participants immersed their dominant hand in the hot water bath (conditioning stimulus) for 30 seconds and then provided ratings of pain intensity for the test stimulus, set to deliver the P-60 temperature, at 40, 50, and 60 seconds. CPM was computed for each individual as the difference between the mean conditioning and pre-conditioning pain ratings, with positive values indicating stronger pain inhibition.

Data analysis

The distributions of predictors and outcomes were examined and two variables with significant positive skew were identified (i.e., HCC, pain interference). HCC outliers (i.e., values 1.5 times the interquartile range [IQR] larger than the 3^rd quartile or 1.5*IQR smaller than the 1^st quartile) were removed due to concerns regarding unreported drug or topical cortisol use. Pain interference outliers were winsorized to the nearest non-outlier (determined in the same manner as for HCC) neighbor value prior to analysis.

Primary analyses.

Multilevel models (MLMs) were conducted in Hierarchical Linear Models (HLM) v.8 [98] to evaluate relations between baseline biobehavioral factors (level 2) and within-person changes (i.e., slopes) in continuous pain outcomes (i.e., daily pain intensity, pain-related interference) across all three assessments (level 1). For pain intensity and interference outcomes, data analyses included baseline, 6-, and 12-month follow-up assessments. MLMs included 130 AA adults who completed at least one follow-up assessment: 17 and 50 missed their 6- and 12-month follow-ups, respectively. Missing data at level 1 were handled in MLMs using maximum-likelihood estimation, which assumes missingness is random. In the present study, missing follow-up assessments were predominantly caused by institutional (i.e., restrictions on research activities during the pandemic) rather than individual factors. Of primary interest were the cross-level interactions between behavioral measures and linear change in pain intensity and interference.

Biobehavioral factors were tested for each pain outcome in the following domains: socioeconomic status (SES; i.e., education level, annual family income); stress exposure (i.e., childhood trauma, family adversity, chronic burden, perceived stress, lifetime discrimination); psychosocial (i.e., depressive symptoms, pain catastrophizing, pain resilience); prolonged HPA secretion (i.e., HCC levels); thermal QST responses (i.e., pain threshold, pain tolerance, TSP, CPM). Preliminary analyses examining predictors of pain intensity/interference within each of these domains revealed no problems with multicollinearity (VIF’s < 1.8, tolerances > 0.55) [28].¹ Significant interactions were probed and simple slopes were calculated using an online calculator [93]. Given established sex differences in pain processing [99] as well as known associations between age, sex, and chronic pain [31], all models controlled for age and gender. Power analysis for MLMs based on prior estimates for correlations between biobehavioral factors and continuous pain outcomes (i.e., r = 3 between QST indices and pain severity [116]), assuming a two-sided test (alpha = 0.05) and power of at least 0.80, indicates a necessary sample size of 82 [40]. Hence, the present sample size (n = 130) was sufficient to detect anticipated effect sizes for primary analyses of pain intensity and pain interference.

Secondary analyses.

For chronic pain episodes, analyses included data from baseline, 6-, and 12-month follow-up assessments. Logistic regression and MLMs included a subset of 78 AA adults who met the selection criteria described above in the ‘Participants’ section (i.e., confirmed presence or absence of a chronic pain episode during the study). Secondary analyses involved a series of multivariate logistic regressions conducted in SPSS v.28 (IBM, Corp.) to evaluate relations between baseline biobehavioral factors and occurrence of a chronic pain episode (i.e., 0 = no chronic pain, 1 = first onset or recurrence of chronic pain). Biobehavioral factors were first evaluated within domains (i.e., SES, stress exposure, psychosocial, prolonged HPA activation, thermal QST response), controlling for age and gender. All baseline predictors with significant odds ratios within domains were then entered into a final multivariate logistic regression model. To account for multiple testing, we used the Benjamini-Hochberg false discovery rate correction to control for the rate of type I errors by adjusting the p-value based on the number of significant results in a family of tests [13] – in this case biobehavioral domains. Power analysis for logistic regression models based on prior estimates for biobehavioral predictors of chronic pain onset (i.e., OR = 3 for depressive symptoms [101]), assuming a two-sided test (alpha = 0.05) and power of at least 0.80, indicates a necessary sample size of 75 [40]. Hence, the present sample size (n = 78) was sufficient to detect anticipated effect sizes for secondary analyses of chronic pain onset.

Exploratory analyses.

Finally, exploratory analyses (MLMs) evaluated chronic pain status (i.e., 0 = no chronic pain, 1 = first onset or recurrence of chronic pain) as a potential moderator of changes in biobehavioral factors over time, controlling for age and gender. The goal of these MLMs was to identify biobehavioral trajectories that distinguished individuals who developed chronic pain episodes from those who did not, which could be evaluated in future studies as potential mechanisms of risk or resilience. These analyses were limited to biobehavioral factors assessed across follow-ups (i.e., chronic burden, perceived stress, discrimination, depressive symptoms, pain catastrophizing, pain resilience, HCC levels, QST responses) and did not include factors assessed only once (education level, family income, childhood trauma, or family adversity).

RESULTS

Baseline descriptive statistics and correlations for pain, demographic, SES, stress exposure, psychosocial, prolonged HPA secretion, and thermal QST measures are presented in Table 1. Participants were mostly young (mean age=25.7, SD=6.3) and well-educated (mean years of education=15.9, SD=3.6) adults with a modal annual family income of between $40,000 and $50,000. Although no participants were experiencing chronic pain (as defined above) at baseline due to study entry criteria, 31 participants (23.8%) developed a new or recurrent chronic pain episode over the follow-up period. Eighteen participants (13.8%) reporting a history of chronic pain prior to baseline evaluation developed a recurrent episode of chronic pain during the study period as did 13 participants (10.0%) reporting no prior chronic pain history. Individuals from higher family income categories reported lower levels of childhood trauma exposure, family adversity, and pain catastrophizing. Higher levels of pain catastrophizing were associated with greater exposure to most types of adversity (i.e., childhood trauma, chronic burden, perceived stress, racial discrimination), higher depressive symptoms, lower pain resilience, higher pain intensity, and pain-related interference. Notably, mean scores for depressive symptoms and pain catastrophizing at baseline (i.e., no current chronic pain) were in the minimal range [11] and below clinical cutoffs [108], respectively. In addition, higher pain intensity was associated with higher perceived stress, greater lifetime racial discrimination, higher depressive symptoms, lower pain resilience, and higher pain threshold. Higher pain-related interference was associated with greater childhood trauma exposure and higher perceived stress. Four participants did not complete the full TSP protocol due to pain ratings of 100 (a pre-determined stopping rule): three of these participants (all female) did not contribute TSP data and were excluded from analyses; the remaining participant contributed 6 ratings and was included in analyses. One participant (male) did not complete the full CPM protocol because they could not keep their hand immersed in the hot water bath for at least 30 seconds.

Table 1.

Baseline descriptive Statistics Summary for African-American Adults.

Variables	M	SD	1	2	3	4	5	6	7	8	9
1. Age	25.7	6.3	-
2. Education	15.9	3.6	.32 ***	-
3. Income	6.5	3.3	−.19 *	.06	-
4. Childhood trauma	40.0	14.0	.15	−.03	−.19 *	-
5. Family adversity	2.1	2.2	.19 *	.08	−.20 *	.56 ***	-
6. Chronic burden	28.3	7.8	.15	−.03	−.16	.45 ***	.51 ***	-
7. Perceived stress	15.1	6.8	−.09	−.07	−.08	.23 *	.18 *	.38 ***	-
8. Discrimination	31.3	9.5	.13	−.03	−.03	.33 ***	.28 **	.46 ***	.36 ***	-
9. Depressive symptoms	6.8	7.4	−.07	−.24 **	−.12	.36 ***	.26 **	.54 ***	.57 ***	.38 ***	-
10. Pain catastrophizing	8.9	9.3	−.11	−.12	−.19 *	.28 **	.18	.28 **	.32 ***	.19 *	.25 **
11. Pain resilience	34.7	12.7	.11	.04	−.01	−.07	−.07	−.20 **	−.32 ***	−.12	−.24 *
12. HCC levels	8.2	8.7	.04	−.10	−.11	−.05	−.10	.06	−.02	−.04	−.05
13. Pain threshold	42.8	3.9	.20 *	.19 *	.07	.05	.14	.09	−.04	.12	−.14
14. Pain tolerance	47.1	2.0	.05	.10	.06	.04	.06	.06	−.10	.20 *	−.07
15. TSP	7.4	9.8	−.12	.02	−.10	−.09	.10	.08	.07	.03	.10
16. CPM	6.4	19.0	−.12	−.01	.03	.04	.07	.06	.11	−.05	.09
17. Pain intensity	7.2	9.1	−.08	.01	−.03	.11	.17	.07	.22 *	.19 *	.19 *
18. Pain interference	10.0	4.2	.06	.03	.02	.19 *	.07	.15	.19 *	.10	.04

^***<p.001

^**p<.01

^*p<.05.

Note. Education = years completed; income (categories ranging from 1 [$0–10,000] to 11 [$100,000+]; Discrimination = lifetime racial discrimination; HCC = hair cortisol concentrations (pg/mg); TSP = temporal summation of pain; CPM = conditioned pain modulation.

Primary Analysis: Predicting Pain Intensity Trajectories

MLMs revealed that both SES variables (i.e., family income, education level) independently predicted changes in pain intensity over follow-up, controlling for age and gender. The family income by time interaction was significant (b=−.241, SE=.098, p=.015); simple slope analysis revealed that whereas individuals reporting higher annual income (+1 SD) did not exhibit significant changes in pain intensity over the follow-up period (b=−.046, SE=.142, p=.746), those reporting lower annual income (−1 SD) exhibited increases in pain intensity (b=.436, SE=.138, p=.002) over follow-up (Figure 1). The education level by time interaction was also significant (b=.255, SE=.101, p=.012); simple slope analysis revealed that whereas individuals reporting lower levels of education (−1 SD) did not exhibit significant changes in pain intensity (b=.032, SE=.270, p=.906), those with higher levels of education (+1 SD) did exhibit increases in pain intensity (b=.570, SE=.268, p=.035) over follow-up (Figure 2). None of the interactions between time and stress exposure (p’s>.11), psychosocial (p’s>.45), prolonged HPA (p=.29), or thermal QST response (p’s>.43) variables predicted changes in pain intensity over follow-up. Results of MLMs predicting pain intensity are presented in Supplemental Table 1.

Figure 1.

Multilevel model testing annual family income as a moderator of changes in pain intensity among African-American adults. *p<.01

Figure 2.

Multilevel model testing years of education completed as a moderator of changes in pain intensity among African-American adults. *p<.05

Primary Analysis: Predicting Pain Interference Trajectories

MLMs revealed that two thermal QST response variables independently predicted changes in pain interference over follow-up, controlling for age and gender. The heat pain threshold by time interaction was significant (b=−.018 SE=.007, p=.014); simple slope analysis revealed that while individuals with lower (−1 SD) pain threshold at baseline (i.e., more pain sensitive) exhibited increases in pain interference over follow-up (b=.201, SE=.083, p=.017), those with higher (+1 SD) pain threshold did not exhibit significant changes in pain interference (b=−.017, SE=.075, p=.821) (Figure 3). The TSP by time interaction was also significant (b=.016 SE=.006, p=.009); simple slope analysis revealed that whereas individuals with higher (+1 SD) TSP at baseline (i.e., greater central sensitization) exhibited increases in pain interference over follow-up (b=.232, SE=.070, p=.001), those with lower (−1 SD) TSP did not exhibit significant changes (b=−.048, SE=.070, p=.492) in pain interference (Figure 4). None of the interactions between time and SES (p’s>.08), stress exposure (p’s>.17), psychosocial (p’s>.13), prolonged HPA (p=.10), or other QST response (i.e., pain tolerance, CPM) variables (p’s>.11), predicted changes in pain interference over follow-up. Results of MLMs predicting pain interference are presented in Supplemental Table 2.

Figure 3.

Multilevel model testing heat pain threshold as a moderator of changes in pain-related interference among African-American adults. *p<.05

Figure 4.

Multilevel model testing temporal summation of pain (TSP) as a moderator of changes in pain-related interference among African-American adults. *p<.01

Secondary Analysis: Predicting Occurrence of a Chronic Pain Episode

Preliminary logistic regressions conducted within biobehavioral domains revealed that prior history of chronic pain, years of education, depressive symptoms, and pain resilience were associated with likelihood of developing chronic pain over the follow-up period. A multivariate logistic regression model including these biobehavioral factors and controlling for age and gender revealed that only prior history of chronic pain (OR=8.088, 95%CI=1.845 to 35.456, p=.006) and higher depressive symptoms at baseline (OR=1.151, 95%CI=1.028 to 1.288, p=.015) were independently associated with risk for occurrence of a chronic pain episode. Results of the multivariate logistic regression model predicting chronic pain episodes are presented in Supplemental Table 3.

Exploratory Analysis: Biobehavioral Trajectories Distinguishing Chronic Pain Groups

Exploratory MLMs evaluated chronic pain status (i.e., 0 = no chronic pain, 1 = first onset or recurrence of chronic pain) as a moderator of changes in biobehavioral factors over the follow-up period. Results revealed that changes in biobehavioral factors across follow-up did not differ between chronic pain groups (p’s>.07).

DISCUSSION

Efforts to understand the transition from acute to chronic pain typically anchor follow-up assessments to an index event, such as a traumatic injury or planned surgery. However, chronic pain may also emerge without a clear event [115]; in such cases, the transition is from wellness to illness. Under these circumstances, the biobehavioral factors that drive individual differences in the emergence of pain can be evaluated without the need to consider the complex and often confounding effects of individual differences in treatment. This conflation of pain and health care inequities is common among minoritized racial groups, who are disproportionately impacted not only by the burden of pain but also by disparities in the treatment they receive. Given the absence of prior work on factors driving pain intensity, interference, and chronicity - specifically in AA populations - the present study sought to advance understanding of the biobehavioral factors that predict worse pain outcomes in AA adults. Results indicated that lower pain threshold and higher TSP predicted increases in pain interference, history of chronic pain and depressive symptoms predicted development of a chronic pain episode, and SES factors (i.e., income, education) predicted changes in pain intensity.

In the present study, healthy AA adults with higher pain sensitivity (i.e., lower pain threshold, higher TSP) exhibited increases in pain-related interference over the 12-month follow-up period; however, pain inhibition (i.e., CPM) and pain tolerance were not predictive of pain interference. These findings are somewhat consistent with prior prospective studies indicating the utility of evoked pain responses as predictors of transition from acute to chronic pain following surgery, predictors of pain persistence among those with chronic pain, and, to a lesser degree, the emergence of chronic pain among healthy, pain-free individuals [101]. Meta-analyses support QST indices as predictors of pain interference among adults with musculoskeletal pain conditions [48]. Moreover, elevated TSP prospectively predicted poorer physical function among individuals with chronic low back pain [89] and moderated the relation between psychosocial factors and worse pain function after total knee arthroplasty [38]. QST indices also predict trajectories of pain interference among youth with chronic pain, over and above psychosocial predictive factors [81]. Reasons why CPM and pain tolerance did not predict changes in pain outcomes in the current study despite prior support in multiracial samples [101] remain to be explored, but may relate to well-documented differences in evoked pain responses between AA and other racial/ethnic groups [63; 97]. Present findings extend prior work by showing for the first time that heat pain threshold and TSP – established predictors of pain interference among injured patients and non-minoritized groups – are also important predictors of the extent to which pain hinders engagement in important activities for AA adults without chronic pain.

Biobehavioral factors prospectively predict chronic widespread pain [73], fibromyalgia [30], TMD [43], musculoskeletal pain [86], low back pain [67; 68; 86], and the transition from acute to chronic pain [56]. These factors include depressive symptoms, pain-related cognitions (e.g., fear-avoidance beliefs, pain catastrophizing, expectations about recovery), obesity, and smoking [4]. Depressive symptoms, in particular, have been consistently identified as a prospective predictor of pain outcomes [91]. Depressive symptoms predicted back pain in a large sample of middle-aged women [19], new onset of knee pain [59], and new onset TMD in healthy young females [103] and adolescents [37; 66]. The present study demonstrated for the first time in a prospective and initially pain-free sample of AA adults that higher depressive symptoms (and prior history of chronic pain) were the only factors that independently predicted occurrence of a chronic pain episode. The odds of developing a new chronic pain episode over follow-up increased by 15% for each unit increase in baseline depressive symptom severity scores and were more than 8 times higher among AA adults with a history of chronic pain compared to those without such history. Our findings are also similar to those of an epidemiological study of older adults that found these same two factors were the only independent predictors of new back pain onset [35]. Several features are notable about the present findings when considered in the context of the extant literature on chronic pain prediction. First, the African Americans in this study were mostly young adults, which contrasts with prospective studies focused on much older age groups [35]. Second, the 1-year follow-up period was also much shorter compared to other studies (e.g., 3 to 15 years) examining new onsets of chronic pain episodes [19; 43; 59]. Finally, incidence of chronic pain in this study was comparable to some prospective studies of widespread pain [72], but 2–3 times higher than rates reported in larger epidemiological studies [39]. Taken together, these results suggest that young AA adults are at elevated risk for onset of a chronic pain episode, with this risk driven by psychosocial factors similar to those identified in other racial groups and older segments of the population.

Risk for negative pain outcomes is associated with the SES of individuals and their environments [71; 92; 94]. Lower individual-level SES has been linked to greater pain severity and pain interference [20], worse post-injury pain [110], higher pain prevalence [14], and, in one large epidemiological study, predicted increased risk for chronic musculoskeletal pain onset over a 6-year period [47]. Lower neighborhood SES has been linked to alterations in evoked pain responses (i.e., impaired CPM) in youth with chronic pain [82]. The present study adds to this growing literature by demonstrating for the first time in a prospective study of initially pain-free AA adults that higher household income served as a buffer against pain, such that only individuals who reported lower annual income exhibited increases in pain intensity over the 12-month follow-up period. In contrast to findings for annual income, higher educational attainment predicted increases in pain intensity among AA adults. Though counterintuitive, this contributes to a growing literature showing that whereas higher social status is associated with greater health benefits (e.g., lower pain intensity) among White adults, Black adults exhibit worse pain outcomes as they climb the social ladder [6]. Upward social mobility differentially impacts health outcomes for Black as compared to White adults, which may be attributed to factors such as greater exposure to interpersonal and institutional discrimination [54]. Stressors linked to the pursuit of higher education likely require prolonged and effortful coping, which for some AA adults may contribute to worse pain outcomes via the effects of “Jonn Henryism” (i.e., a strategy for coping with persistent social stressors that involves high levels of effort expenditure likely to incur cumulative physiological costs [e.g., allostatic load]) [58]. Together, these findings complement prior work showing that SES measures are not interchangeable [102]. Notably, the observed associations in the present study were unique to pain intensity: neither income nor education predicted changes in pain interference or occurrence of a chronic pain episode. Future studies will be needed to determine the extent to which the association between income and pain intensity is attributable to individual-level (e.g., higher levels of health-promoting behaviors) versus structural (e.g., increased access to health care) factors.

There were several notable differences between the present findings and prior work on pain prediction conducted by our own group and others. First, despite strong support for pain catastrophizing as a predictor of pain intensity and interference [113], and evidence that AA are more likely than NHW adults to endorse it [78], pain catastrophizing did not predict any pain outcomes among AA adults without chronic pain. Pain catastrophizing levels at baseline were low, on average, in this sample without chronic pain, but may be more strongly predictive of pain outcomes when scores surpass clinical cutoffs [108]. Second, despite evidence that adverse exposures across the lifespan, including racial discrimination, are associated with higher daily pain, pain sensitivity, and chronic pain status [22; 24; 25; 34; 83], stress exposures did not predict trajectories of pain intensity or interference or the development of new chronic pain episodes. Whether these differences can be explained by sample characteristics (e.g., the cumulative impact of stress exposures on pain outcomes are more pronounced for older as compared to younger African Americans) or study design (e.g., pain catastrophizing correlated with pain intensity and interference at baseline but did not prospectively predict pain trajectories) indicates an important avenue of inquiry.

The major limitation of the present study relates to missing follow-up data. Institutional restrictions on research implemented during the COVID-19 pandemic led to forced cancellation of follow-up assessments for a subset of enrolled participants. Hence, missingness was attributed to a random external factor rather than participants’ chronic pain status or their pain-related behaviors. Although MLMs can handle missing data for continuous pain outcomes (i.e., intensity, interference) through restricted maximum likelihood estimation, our sample size was reduced substantially for models predicting the dichotomous chronic pain status outcome. Another limitation is that this sample included some individuals with a prior history of chronic pain even though none were experiencing chronic pain at the time of study entry based on inclusion/exclusion criteria. Therefore, factors predicting occurrence of a chronic pain episode in the current work cannot necessarily be generalized to populations with no prior history of chronic pain. Given the modest sample size, particularly for the first onset group, we were not able to compare the characteristics or predictors of first onset versus recurrent chronic pain episodes. Future studies should attempt to disentangle these groups, which may exhibit different risk/resilience factors and require different interventions.

Another limitation is that the majority of this sample resided in one greater metropolitan area. This precluded examination of place-based indicators (e.g., rurality/urbanicity, area deprivation) that have been associated with self-reported pain experiences and could inform the interpretation of associations between individual-level SES indicators and pain outcomes. Findings from this relatively young AA adult sample, selected to minimize potentially confounding influences of age on lifetime exposure to chronic pain risk factors (e.g., cumulative stress, lifetime racial discrimination), may not generalize to older samples. For example, new stressors related to parenting and caregiving and new resilience factors such as wealth accumulation may amplify or dampen risk for chronic pain among middle- and older-age AA adults. Although the present statistical models controlled for age and gender, future work in larger samples with wider age ranges should evaluate these two variables as potential moderators of relations between biopsychosocial factors and subsequent chronic pain risk. Finally, this study included one measure of cognitive/affective positivity and behavioral perseverance during a painful experience. Future prospective studies should adopt a multisystem resilience framework [7] to evaluate profiles of psychosocial, sociological, and health-related resilience in pain-free individuals that predict first and/or recurrent chronic pain episodes.

Conclusion and Implications

Although AA adults experience greater pain intensity and interference and are more likely to transition from acute to chronic pain than their NHW counterparts [57; 74], prospective studies examining the biobehavioral determinants of pain among healthy AA adults are lacking. Pain intensity and pain-related interference are moderately correlated dimensions of pain experience - particularly among individuals with chronic pain. However, there are pronounced individual differences in how individuals fare when confronted with pain. Consistent with this multifactorial nature of pain [41], the present findings demonstrate unique sets of biobehavioral factors that prospectively predict distinct dimensions of pain experience – intensity (i.e., education and income), interference (i.e., heat pain threshold and TSP), and occurrence of a chronic pain episode (i.e., depression and chronic pain history). It follows that the prevention of negative pain outcomes among AA adults requires a multimodal approach that is tailored to specific pain dimensions. The components of such a prevention program could focus on the following buffers: financial counseling/coaching to reduce pain intensity among those with lower income; pain neuroscience education to reduce pain interference among those with greater pain sensitivity [69]; preventive psychological and pharmacological interventions to reduce risk for chronic pain onset among those with elevated depressive symptoms [16]. Although these prevention components have previously been evaluated in non-minoritized populations, cultural adaptations of treatment materials may be necessary to address pain risk and resilience factors that operate differently among AA adults, such as educational attainment and broader dimensions of social status. Early detection of biobehavioral risk for chronic pain in AA adults is critical for the development and refinement of prevention programs to mitigate racial disparities in pain [51]. Once chronic, pain can become ‘stuck’ and resistant to first-line pharmacological and psychotherapeutic interventions [18; 32; 70].

Data availability:

The datasets generated and analyzed for the current study are not publicly available due to concerns regarding the privacy of research participants. However, de-identified datasets are available from the corresponding author on reasonable request.