Are cardiometabolic markers of allostatic load associated with pronociceptive processes in Native Americans?: A structural equation modeling analysis from the Oklahoma Study of Native American Pain Risk

Abstract

Native Americans (NAs) experience higher rates of chronic pain than the general U.S. population, but the risk factors for this pain disparity are unknown. NAs also experience high rates of stressors and cardiovascular and metabolic health disparities (eg, diabetes, cardiovascular disease) consistent with allostatic load (stress-related wear-and-tear on homeostatic systems). Given that allostatic load is associated with chronic pain, then allostatic load may contribute to their pain disparity. Data from 302 healthy, pain-free men and women (153 NAs, 149 non-Hispanic Whites [NHW]) were analyzed using structural equation modeling to determine whether cardiometabolic allostatic load (body mass index, blood pressure, heart rate variability) mediated the relationship between NA ethnicity and experimental measures of pronociceptive processes: temporal summation of pain (TS-pain) and the nociceptive flexion reflex (TS-NFR), conditioned pain modulation of pain (CPM-pain) and NFR (CPM-NFR), and pain tolerance. Results indicated that NAs experienced greater cardiometabolic allostatic load that was related to enhanced TS-NFR and impaired CPM-NFR. Cardiometabolic allostatic load was unrelated to measures of pain perception (CPM-pain, TS-pain, pain sensitivity). This suggests cardiometabolic allostatic load may promote spinal sensitization in healthy NAs, that is not concomitant with pain sensitization, perhaps representing a unique pain risk phenotype in NAs.

Introduction

Back pain^8,37, arthritis^46,83, headaches¹⁰⁸, dental pain⁷⁷, and other chronic pain states are more prevalent in Native Americans (NAs) than the general U.S. population^8,67,146. Moreover, NAs have endured generations of extreme stressors, including genocide, forced relocation, and boarding school placement⁴³, and studies suggest they continue to experience high levels of psychosocial stressors (eg, trauma, adversity)^15,17,38,150. Thus, stressors may contribute to their pain disparity^72,135, as it does with other health disparities¹⁴¹.

Allostasis is the physiologic balance maintained in the face of change⁸⁶; but, when constantly pushed out of balance by stressors, this causes wear-and-tear on systems that maintain balance (ie, allostatic load)¹²¹ and can ultimately lead to disease⁸⁸. Allostatic load can be measured by neuroendocrine (eg, cortisol), cardiovascular (eg, blood pressure), metabolic (eg, body mass index), immune (eg, C-reactive protein), and parasympathetic (eg, heart rate variability) outcomes¹²¹ and are often combined to create a cumulative score^70,121,126. Surprisingly few studies have directly studied allostatic load in NAs^140,141, but numerous studies have shown that NAs experience high rates of allostatic load corollaries. For example, NAs are 1.5× more likely to be obese, 1.3× more likely to have high blood pressure, and 1.2× more likely to be physically inactive, compared to non-Hispanic Whites (NHWs)^24,26. They also have the highest age-adjusted rates of diabetes in persons ≥20 years old²⁵. Moreover, one study found that young NA adults showed blunted cortisol and cardiovascular reactivity to acute stress⁶⁸, a marker of allostatic load associated with early life adversity^23,70.

Chronic pain is often comorbid with secondary and tertiary effects of allostatic load on the cardiovascular and metabolic systems (eg, cardiovascular disease, diabetes)^{45,54,81,82,138,142}; but, there have been relatively few attempts to assess the relationship between allostatic load and pain in disease-free persons. One study of U.S. adults found that allostatic load was associated with a higher prevalence of severe headaches, pain lasting ≥24 hours, and widespread body pain¹³³ and other studies found that it was associated with chronic pain severity in Norwegian adults¹³¹ and older English adults¹²⁹. However, these studies cannot rule out that these relationships are due to chronic pain promoting allostatic load, given that chronic pain is a stressor¹³⁰. Alternatively, one could study pain-free participants to determine if there is a relationship between allostatic load and pronociceptive mechanisms known to increase chronic pain risk (eg, enhanced pain facilitation, impaired descending inhibition, hyperalgesia)¹⁵³. To our knowledge, no study has addressed this issue.

The current study addressed this issue via an ancillary analysis (153 NAs, 149 NHWs) of data from the Oklahoma Study of Native American Pain Risk (OK-SNAP)^114,115. Healthy participants were studied to ensure that any observed relationships were not due to disparities in chronic pain severity and/or pain treatment^28,85,143. Neuroendocrine and immune markers were not assessed in OK-SNAP; thus, a “cardiometabolic allostatic load” latent variable was modeled from cardiovascular (mean arterial blood pressure, MAP), metabolic (body mass index, BMI), and parasympathetic (heart rate variability, HRV) indicators. Although we found few ethnic differences in pronociceptive processes in OK-SNAP, indirect pain promoting pathways can be identified using mediation analysis^102,103. Indeed, ethnic difference in pronociceptive processes may only emerge in the presence of other factors (eg, pain-related anxiety, pain catastrophizing), as we have demonstrated previously^113,144. Thus, current analyses examined whether indirect (mediated) paths linked NA ethnicity to pronociceptive processes via cardiometabolic allostatic load. Analyses were conducted with and without controlling for variables that could influence this relationship. We predicted that NAs would experience higher cardiometabolic allostatic load which would be associated with impaired descending inhibition (ie, conditioned pain modulation), enhanced nociceptive facilitation (ie, temporal summation), and reduced pain tolerance (latent variable assessed from heat, cold, ischemia, and electric pain tolerances).

Materials and Methods

Participants

Healthy, pain-free participants were recruited in OK-SNAP. Recruitment efforts included tribal and non-tribal newspaper ads, fliers, personal communications with NA groups, email announcements, online platforms (eg, Facebook), and word of mouth. Those deemed eligible following an initial phone screen were invited to attend a laboratory testing day, which began with a thorough screening for inclusion/exclusion criteria. Data collection occurred between March 2014 and October 2018.

Participants were excluded for: 1) <18 years old, 2) history of self-reported cardiovascular, neuroendocrine, musculoskeletal, neurological disorders, 3) self-reported chronic pain or current acute pain, 4) BMI≥35 (due to difficulties recording electromyogram for NFR), 5) current/recent use of anti-depressants, anxiolytic, analgesic, stimulant, or anti-hypertensive medication, 6) current psychotic symptoms (assessed by Psychosis Screening Questionnaire¹²) or substance use problems, and/or 7) an inability to read and speak English. NA status was verified from Certificate of Degree of Indian Blood or tribal membership cards. NA participants represented tribal nations predominately from the southern plains and eastern Oklahoma tribes. To respect tribal confidentiality, tribal affiliations are not reported. Before they provided verbal and written informed consent, all participants were given an overview of procedures and told they could withdraw at any time. Participants received a $100 honorarium for the completion of each testing day. The study was approved by IRBs of The University of Tulsa, Cherokee Nation, and the Indian Health Service Oklahoma City Area Office.

Power analyses for primary aims of the parent (OK-SNAP) study were conducted with G*Power (version 3.1.9.2). These analyses were based on effect sizes of pain outcomes derived from our pilot study of pain processing in NAs (Cohen’s d’s ranged from 0.10 to 0.89)⁹⁷, as well as effect sizes for race/ethnic differences in CPM from Campbell et al (d=0.53 to 0.72)²². These analyses suggested that 120 per group (N=240 total) would result in power of 0.80 for most of the outcomes associated with the primary aims of the parent study (power=.80 for effect sizes d≥.36 at alpha=.05). Thus, that was the targeted sample size for completion of all study tasks. Racial/ethnic group differences associated with the main study findings have been reported elsewhere^114,115.

Of the 329 found eligible, 2 participants’ data were lost due to a computer malfunction, 22 participants were non-NA minorities and thus were excluded from analyses, and 3 were excluded for having Type 1 or Type 2 diabetes. Thus, a total of 302 participants (153 NA and 149 NHW) were available for analysis in the current study. 227 (75%) of the 302 participants completed both testing days, 46 (15%) completed one testing day, and 29 (10%) completed only partial tasks during one testing day. When participants quit during testing, 29 quit during pain modulation day (4 during NFR threshold, 4 during 3-stimulation threshold, 2 during single suprathreshold electric stimulations, 9 during TS-NFR, 7 during CPM, 3 during ECON) and 3 quit during the pain tolerance day (2 suprathreshold heat, 1 due to inclement weather) – Note: these numbers do not sum to 29 because some participants quit during a task on the second day. Table 1 presents means and SDs of study variables by ethnic group, as well as reports the samples sizes for each variable.

Table 1.

Racial/ethnic group differences in study variables (N=302).

	non-Hispanic Whites				Native Americans
Variable	N	M	SD		N	M	SD		t-test	p-value	Cohen’s d
Age (years)	149	28.56	13.53		153	30.60	12.67		−1.84	0.067	0.212
Physical Activity (IPAQ; MET mins/week)	149	7905.90	7000.14		152	9930.62	8631.91		−1.49	0.138	0.171
Subjective Sleep Quality (PSQI 9; 0–3)	117	0.96	0.64		115	1.26	0.85		−3.08	0.002	0.405
Perceived Stress (PSS; 0–40)	146	13.10	6.00		147	14.54	6.04		−2.04	0.042	0.238
Psychological Distress (GSI of SCL-90; 0–1)	146	0.32	0.32		147	0.42	0.40		−2.54	0.012	0.296
General Health Scale (SF-36; 0–100)	146	80.45	13.62		147	78.64	13.52		1.14	0.256	0.133
Mean Arterial Blood Pressure (mmHg)	147	83.16	8.19		147	88.60	10.69		−4.96	<.001	0.579
Body Mass Index (BMI; kg/m2)	148	24.15	3.80		149	25.90	4.48		−3.62	<.001	0.420
Heart Rate Variability (RMSSD; ms)	129	51.74	37.13		122	50.97	40.02		0.46	0.645	0.058
CPM of Pain (Δpain; −100 to +100)	127	−5.69	7.26		119	−7.63	8.48		1.93	0.055	0.246
CPM of NFR (ΔNFR; d-units)	126	−0.05	0.39		117	0.01	0.40		−1.10	0.275	0.141
TS-Pain (Δpain; −100 to +100)	122	11.11	14.37		119	13.16	13.78		−1.13	0.258	0.146
TS-NFR (ΔNFR; d-units)	128	0.49	0.54		128	0.58	0.48		−1.42	0.158	0.177
Heat Tolerance (°C)	129	45.81	2.07		124	45.55	1.59		1.15	0.253	0.143
Cold Tolerance (sec)	130	1.80	0.42		124	1.66	0.31		2.98	0.003	0.371
Ischemia Tolerance (sec)	129	2.20	0.42		125	2.12	0.45		1.50	0.136	0.188
Electric Tolerance (mA)	129	30.51	13.31		125	31.12	12.94		−0.37	0.712	0.046

Note: For ease of interpretation, means/SDs that are reported here are untransformed and prior to multiple imputation. The Cohen’s d effect sizes are calculated from variables that were corrected for non-normality and outliers, but prior to multiple imputation. IPAQ=International Physical Activity Questionnaire; MET= metabolic equivalence of task; PSS=Perceived Stress Scale; GSI=Global Severity Index; SCL=Symptom Checklist; RMSSD=root mean square of successive differences; CPM=conditioned pain modulation; TS-pain=temporal summation of pain; TS-NFR=temporal summation of the nociceptive flexion reflex; mA=milliamperes.

The sample size estimates needed for structural equation modeling are typically based on N-to-parameter ratios, with ratios varying between 10:1¹²⁵ and 20:1⁷⁴. In the present study, the model predicting TS/CPM outcomes had a ratio of 16:1 and the model predicting pain tolerance had a ratio of 14:1. However, when control variables were added to these models, the ratios were lower, so the models with controls are interpreted with caution and primarily used to examine whether results were similar with statistical controls in place.

Brief Overview of Procedures

A full description of OK-SNAP procedures is reported elsewhere¹¹⁵. Pain testing was conducted over a 2-day period (4–6 hours/day). Informed consent and inclusion/exclusion screening were conducted on the first day. On one of the testing days, electric, heat, cold, and ischemia pain tolerance were assessed with breaks in between tasks. On the other testing day, measures of pain modulation were assessed. Specifically, conditioned pain modulation (CPM) was used to assess descending inhibition and temporal summation (TS) was used to assess nociceptive facilitation. Importantly, pain report and the nociceptive flexion reflex (NFR) were assessed as outcomes during CPM and TS. NFR is a polysynaptic, spinal reflex elicited in response to Aδ fiber activation, so it is used as a measure of spinal nociception¹²⁴. Although the reflex arc does not require supraspinal input, the reflex can be modulated by descending input from supraspinal centers¹²⁴. Thus, assessing the NFR during CPM (ie, CPM-NFR) and TS (ie, TS-NFR) allow investigators to examine the effects of these processes on spinal nociception, while at the same time assessing the effect on pain perception (ie, CPM-pain, TS-pain). Fig 1 provides an overview of study procedures, but the reader is referred to the parent study for a full description of all procedures¹¹⁵.

Figure 1.

Study procedures across the two testing days.

Order of testing days was randomized but blocked by ethnicity and sex. Moreover, tests within each day were partly randomized to avoid order effects. Breaks were provided between tasks to minimize carryover.

Testing Environment and Apparatus

During testing, participants were seated in a comfortable reclining chair (Perfect Chair Zero Gravity Recliner, Human Touch, Long Beach, CA) in a sound-attenuated and electrically shielded room. The experimenter sat in an adjacent room and monitored testing using audio and video equipment. Questionnaires were presented by a computer. Custom-built software (LabVIEW; National Instruments, Austin, TX) was used to control most experimental procedures.

Electric pain stimuli were delivered via constant current stimulator (Digitimer DS7A; Hertfordshire, England) and a bipolar electrode (Nicolet; 30 mm inter-electrode distance) filled with a conductive gel (EC60, Grass Technologies) that was attached on the participant’s left ankle. Heat stimuli were delivered using a Medoc (Haifa, Israel) Pathway device with a Contact Heat Evoked Potential Stimulator (CHEPS) thermode. A circulating water bath (Thermo Fisher Scientific, Pittsburgh, PA) was used to assess cold pain tolerance, and as the painful conditioning stimulus in the CPM task. The water level was kept constant (6″ deep) across all participants to keep procedures standardized and to maintain similar cold exposure to participants’ hands/forearms.

For procedures that assessed the nociceptive flexion reflex (NFR), the skin was first cleaned with alcohol and then exfoliated (Nuprep gel; Weaver and Company, Aurora, CO) to achieve impedances <5kΩ. Recording electrodes filled with conductive gel (EC60; Grass Technologies, West Warwick, RI) were placed over the left biceps femoris (hamstring) muscle to assess the reflex from electromyogram (EMG). The signal was filtered (10 Hz to 300 Hz) and amplified (×10,000) using a Grass Technologies Model 15LT amplifier with AC Module 15A54. Sampling and digitization (1000 Hz) of the signal was achieved using a National Instruments (Austin, TX) analog-to-digital converter, before being stored to a computer hard drive.

Pain Modulation: Temporal Summation (TS) and Conditioned Pain Modulation (CPM)

Determination of Electric Stimulus Intensity for TS and CPM.

To determine the electric stimulus intensity (in mA) to use during pain modulation tasks, 3 procedures were conducted to calibrate the stimulus to each participant (for a full description of these procedures see,¹¹⁵):

1. NFR threshold – To assess the intensity of the stimulus to reliability elicit the reflex (ie, NFR threshold), 3 ascending/descending staircases of electric stimuli were presented in which the stimulus intensity was increased until a reflex was obtained, and then the intensity was decreased until the reflex disappeared (this happened 3 times)^50,110,

2. Pain30 – if electric stimuli delivered at NFR threshold did not evoke a 30 out of 100 rating on a visual analog scale (VAS) ranging from “no pain sensation” to “the most intense pain sensation imaginable,” then an ascending series of electric stimulations was administered to determine stimulus needed to evoke Pain30, and

3. 3-stimulation threshold – an ascending staircase of trains of 3 stimuli (at 2 Hz; ie, 0.5-s interstimulus interval [ISI]) were delivered that was used to determine the stimulus intensity needed to evoke an NFR on the 3^rd stimulus in the train.

Stimuli during TS-NFR assessment were set at 1.2× NFR threshold or 1.2× of 3-stimulation threshold (whichever was higher), whereas stimuli during CPM assessment were set at the highest of 1.2× NFR threshold, 1.2× 3-stimulation threshold, or 1× Pain30. Together, these procedures ensure that suprathreshold stimuli evoke reliable NFRs and suprathreshold pain perception.

Temporal Summation of the Nociceptive Flexion Reflex (TS-NFR).

TS-NFR is believed to be a physiological marker of human central (spinal) sensitization, because a series of constant-intensity electric stimulations evokes a summation of the reflex^4,6 that is dependent on NMDA receptors^5,58 (similar to wind-up in animals)^34,90,91. Although NFR is driven by A-δ fiber activation, the electric stimulus series depolarizes both A-δ and C-fibers, and the C-fiber activation thus leads to windup of wide dynamic range (WDR) neurons. Supporting this, animal studies have simultaneously recorded the activity of single motor units (part of the circuitry associated with the motor output of the NFR) and the activity of WDR neurons and found that summation of the single motor units was correlated with the summation of WDR neurons^157,158. Moreover, TS-NFR is enhanced in individuals with chronic pain and central sensitization⁷. When taken together, this evidence indicates that TS-NFR can be used to assess facilitation of spinal nociceptive neurons.

TS-NFR was defined as the degree of reflex summation following a series of 3 suprathreshold stimuli. To assess this, 5 series of 3 suprathreshold stimuli (0.5 s ISI; ie, 2 Hz) were delivered. We have shown that TS-NFR is best assessed from a series of 3 suprathreshold electric stimulations at this frequency¹³⁷ because it maximizes the summation in NFR (responses to longer series results in asymptote of the reflex) and minimizes participant exposure to noxious stimuli. After each series, participants were instructed to rate the pain intensity for each of the 3 stimulations, using a set of 3 computer-presented VASs (described previously). After the participant completed the ratings, there was an inter-series interval of 8–12 s. The baseline EMG in the 60 ms prior to the third stimulus in the series was visually inspected for excessive muscle tension or voluntary movement (ie, mean rectified EMG >5μV). If present, the series was repeated. NFR magnitudes were calculated in d-units by subtracting the 60 ms baseline prior to the first stimulus in each series from the EMG response 70–150 ms after each stimulus in the train (Note: this post-stimulus interval starts earlier than the 90–150 ms interval used for NFRs elsewhere in this study because prior research has indicated TS procedures can decrease the onset latency of NFRs¹³⁷). This difference was then divided by the average of the SDs of the rectified EMG from these two intervals. TS-NFR was defined as the difference in the NFR magnitude in response to the 3^rd stimulus in the series minus the NFR magnitude in response to the 1^st stimulus in the series. This TS-NFR change score was averaged across the 5 series.

Temporal Summation of Pain (TS-Pain).

Temporal summation of pain is believed to be the psychophysical correlate of spinal neuron facilitation^40,104, but TS-NFR and TS-Pain can diverge¹¹⁶ indicating that the nociceptive signal may undergo additional supraspinal modulation after ascending from the spinal level to impact the degree of pain perception facilitation. Thus, for the current study, TS-Pain will be used to assess pain facilitation.

Assessing TS-Pain from the 3-stimulus series noted above is not ideal, because the 2 Hz series is too fast for participants to make pain ratings immediately following each stimulus. Thus, all 3 ratings must be made after the last stimulus, which can be affected by recall bias, especially for the first two stimuli in the series.

Therefore we took a different approach to define TS-Pain; one that has been used by other studies using both electric⁴⁴ and mechanical⁵³ stimuli. 5 single electric stimuli were delivered using the same suprathreshold intensity as that used during TS-NFR. However, these stimuli had an interstimulus interval that would preclude the evocation of temporal summation (i.e., 8–12 s ISI; 0.125 to 0.08 Hz). Participants made pain ratings immediately after each stimulus using the VAS described above. Thus, TS-Pain was defined as the difference between the average rating of these single stimuli (in which no pain summation had occurred) and the average rating of the 3^rd stimulus in the 3-stimulus train delivered during TS-NFR (in which pain summation had occurred).

Conditioned Pain Modulation (CPM).

Our CPM procedures involved assessment of pain and NFR in response to electric test stimuli delivered to the ankle before and during a tonic conditioning stimulus (CS; circulating, 2-min, 10±0.1°C water bath). When applied in this way, prior research has shown that two dissociable modulatory circuits are engaged: a cerebrospinal circuit that modulates NFR and supraspinal inhibitory circuit that modulates pain perception⁹⁹. Thus, CPM-NFR was used to assess descending inhibition of spinal nociception (shortened to “spinal inhibition” in the Tables/Figures), whereas CPM-Pain was used to assess pain inhibition.

Each CPM phase was 2-min and consisted of a 20 s wait period, followed by 5 electric test stimuli (random 8–12 s inter-stimulus interval). There was a 2-min break between CPM phases. Participants provided electric pain ratings during CPM using a numerical rating scale (NRS) that was displayed on a computer screen in front of them. The scale had verbal anchors every 20 points (0 “no pain”, 20 “mild pain”, 40 “moderate pain”, 60 “severe pain”, 80 “very severe pain”, and 100 “worst possible pain”), but participants were encouraged to use the full scale. Participants were asked to say their ratings aloud (due to having their hand in the cold water) and an experimenter recorded the ratings while they observed from the adjacent room. After the CS was administered, participants rated it on the same NRS. As we have reported elsewhere¹¹⁵, the CS elicited moderate to severe pain that did not differ between the groups (NRS mean rating=54).

NFR magnitudes in response to electric test stimuli were used to assess changes in spinal nociception¹¹¹. NFR magnitudes were calculated as a d-score [NFR d = (mean rectified EMG of 90 to 150 ms post-stimulation interval minus mean of rectified EMG from −60 to 0 ms prestimulation interval) divided by the average SD of the rectified EMG from the two intervals]. Trials with NFR baselines whose average EMG was >3.0 μV were excluded (3% of trials were excluded). CPM-pain/CPM-NFR was defined as the difference in the average electric pain/NFR assessed during the CS minus the average electric pain/NFR assessed during the baseline phase. A “sham” CPM task was also employed which was identical to the real CPM task with the exception that a 26°C non-painful water bath was used as the conditioning stimulus (results not presented). The order of CPM and sham CPM was randomized and there was a 15 min break between the two to avoid carry over effects. HRV was recorded during the breaks that followed each CPM task.

Pain Tolerance

Although different stimulus modalities are known to impact different receptor types and nociceptive processes¹⁰⁵, different measures of pain tolerance should assess a common underlying latent construct associated with the ability to tolerate noxious stimuli. Consistent with this, we found that electric, heat, cold, and ischemic pain tolerance were all correlated about r=.40 (see Table 2). Therefore, we used SEM to model a latent pain tolerance variable from these observed variables.

Table 2.

Means, SDs, and Pearson’s correlations among all study variables (N=302)

Variable	M	SD	CPM pain	CPM NFR	TS pain	TS NFR	Heat Tol	Cold Tol	Isch Tol	Elec Tol	Race	Sex	Age	Exer	Sleep Q	Stress	Psych	Gen Hlth	MAP	BMI
CPM-pain	−6.63	7.91	1
CPM-NFR	−.02	.39	−.014	1
TS-pain	12.12	14.09	.021	.012	1
TS-NFR	.53	.52	−.076	.226*	.160*	1
Heat Tol	45.68	1.85	−.079	.023	−.042	.146*	1
Cold Tol (log)	1.73	.38	.091	−.141*	−.047	−.021	.429*	1
Isch Tol (log)	2.16	.44	.010	−.149*	−.067	.098	.442*	.380*	1
Elec Tol	30.81	13.11	.091	−.081	−.035	.091	.401*	.422*	.375*	1
NA Race	—	—	−.123	.070	.073	.088	−.072	−.183*	−.094	.023	1
Female Sex	—	—	.057	−.140*	.039	−.120	−.354*	−.129*	−.159*	−.060	.085	1
Age (log)	1.44	.17	.139*	.152*	.184*	.295*	.056	−.168*	−.029	−.005	.106	.045	1
Exercise (log)	3.76	.45	.117	.116	.188*	.124*	.058	.046	.014	.094	.086	−.117*	.066	1
Sleep Q	1.11	.76	.085	.059	.185*	.047	.043	.054	.012	.131*	.200*	−.055	−.093	.118	1
Stress	13.82	6.05	.031	.081	.074	.035	−.092	.079	−.002	.089	.119*	−.013	−.201*	.063	.337*	1
Psych (log)	.12	.09	.055	.111	.153*	.075	.022	.106	.057	.095	.147*	−.021	−.084	.076	.426*	.705*	1
Gen Hlth	79.54	13.58	.069	−.008	−.004	−.039	.026	.048	−.042	−.064	−.067	.029	.063	.035	−.215*	−.333*	−.341*	1
MAP	85.57	8.98	−.084	.153*	.128*	.224*	.216*	−.024	.044	.036	.279*	−.194*	.433*	.002	.110	−.034	.065	−.086	1
BMI	25.03	4.24	.013	.142*	−.014	.137*	.037	−.095	−.151*	.068	.206*	−.082	.345*	.064	−.003	−.002	.039	−.081	.362*	1
HRV (log)	1.62	.28	−.065	−.151*	.017	−.124	−.021	.103	−.028	−.054	−.029	−.111	−.408*	.122	.072	.015	.013	.053	−.275*	−.141*

Note: Means, SDs and intercorrelations are based on variables that were transformed for normality and for which missing variables were estimated using multiple imputation. CPM=conditioned pain modulation; NFR=nociceptive flexion reflex; TS=temporal summation; Tol=Tolerance; NA Race (0=non-Hispanic Whites, 1=Native American); Female Sex (0=male, 1=female); MAP=mean arterial pression; BMI=body mass index; HRV=heart rate variability; Sleep Q=sleep quality; Gen Hlth=General Health.

^*p<.05

Electric Tolerance.

Electric pain tolerance was assessed using a single ascending staircase of stimulations that started at 0 mA and increased in 2 mA steps. After each stimulus, the participant rated their pain on the computer-presented VAS described earlier. The stimulations increased until the participant rated the stimulus as the maximum tolerable pain and the stimulus intensity (mA) was recorded as pain tolerance¹¹¹.

Heat Tolerance.

Heat pain tolerance was assessed 5 times (1 practice, 4 averaged trials) by attaching the CHEPS thermode to the participants’ left volar forearm²¹. Each trial started from a baseline of 32°C and heated at a rate of 0.5°C/s until the participant terminated the stimulus by pushing a button when the heat became intolerable. The maximum intensity was set to 51°C.

Cold Tolerance.

Cold pressor tolerance was assessed by asking participants to submerge their hand and forearm (fingers spread apart) into a 6±0.1°C circulating water bath^21,89,109. During submersion, participants made continuous ratings on the same VAS described earlier. The time (sec) until a rating of maximum tolerable was defined as pain tolerance (or 5-mins max was reached).

Ischemia Tolerance.

Ischemia tolerance was assessed using a forearm tourniquet test⁴⁷. Participants conducted hand exercises using a dynamometer (Lafayette Instrument Company, IN) at 50% grip strength for 2 min (1×/sec), then raised their arm for 15 s to allow desanguination. A blood pressure cuff was then inflated over the arm’s bicep muscle to 220 mmHg to occlude blood flow. During occlusion, participants continuously rated their pain on the previously described VAS and the time taken (sec) to achieve a rating of maximum tolerable was defined as pain tolerance (or 25-min max was achieved).

Cardiometabolic Allostatic Load

Allostatic load represents wear-and-tear on stress regulation systems (cardiovascular, metabolic, immune, neuroendocrine, parasympathetic). In early studies¹²⁶, allostatic load was measured from 10 indicators that were divided into primary (ie, DHEA, cortisol, epinephrine, norepinephrine) and secondary (ie, systolic/diastolic blood pressure, waist-to-hip ratio, HDL cholesterol, HbA1c) outcomes. This list has been expanded to include several measures of cardiovascular (eg, blood pressure, total cholesterol, HDL cholesterol, triglycerides, heart rate, homocysteine, FEV1/FVC, pulse pressure), metabolic (eg, body mass index, waist-to-hip ratio, HbA1c, albumin, glucose, waist circumference, LDL cholesterol, triglycerides, creatine, homocysteine, HOMA-IR, apolipoprotine A1 and B), immune (eg, C-reactive protein, IL-6, IL-1, IL-10, herpes simplex, fibrinogen, insulin-like growth factor-1, white blood cell count, TNF-α), neuroendocrine (eg, cortisol, epinephrine, norepinephrine, DNEA, dopamine, aldosterone, adnephrin), and parasympathetic (eg, heart rate variability) systems¹²¹. It is rare that all indices are used, but the individual indices are typically combined in some way to create a measure of overall allostatic load¹²¹.

OK-SNAP was designed to assess pronociceptive mechanisms in otherwise healthy, pain-free NAs, but it was not specifically designed to assess allostatic load. For this reason, the present study focused on markers of allostatic load that were collected as part of the general health of participants in the study. The 3 indices that were available primarily assessed cardiovascular and metabolic domains (blood pressure, heart rate variability, body mass index). Prior studies have shown that all 3 are individually associated with pronociceptive processes and/or chronic pain^3,19,62,155.

The overall allostatic load variable is oftentimes created by first calculating quartiles of each marker and then a count is created from the number of markers that fall into the high risk quartile (typically the upper quartile)¹²¹. A limitation of this approach is that it does not consider the full range of risk for each marker, because it assumes that only the upper quartile confers risk. Thus, other approaches have been used to account for the full range of risk. For example, the z-score allostatic load index⁷⁰ first converts each variable to a z-score based on the sample’s distribution and then a linear combination of the z-scores is created by summing them to get a total score. We took a similar approach and created a latent variable to form a linear combination of the 3 markers of cardiovascular (blood pressure), metabolic (body mass index) and parasympathetic (heart rate variability) functioning. This allowed us to create a continuous measure of cardiometabolic allostatic load that accounts for the full continuum of risk. Plus, this approach has the advantage of modeling and removing measurement error from each indicator⁶⁴.

Body Mass Index (BMI).

Weight and height were assessed from a medical scale to calculate BMI. Higher BMI is indicative of greater allostatic load¹²¹.

Blood Pressure.

Blood pressure (mean arterial blood pressure, MAP) was assessed 3 times (3 min inter-test interval) at rest on the first testing day (Dinamap; Tampa, FL) before any testing occurred. Participants sat still in a chair with their arm on the armrest. Higher blood pressure is indicative of greater allostatic load¹²¹.

Heart Rate Variability (HRV) During Pain Recovery.

The parasympathetic nervous system is the branch of the autonomic nervous system responsible for “rest and digest” functions and serves to return homeostasis following “fight or flight” (ie, stress) reactions⁸⁴. Given this, impairments of parasympathetic control can lead to allostatic load¹³⁹. Respiratory sinus arrythmia (RSA) refers to the increase and decrease in heart rate caused by inhalations and exhalations, respectively, of the respiratory cycle. RSA is believed to be mediated exclusively by the parasympathetic nervous system (ie, vagal nerve)² and can be measured by some HRV metrics, such as the root mean square of successive differences (ie, RMSSD) between beat-to-beat intervals in the electrocardiogram (ECG) signal^2,128. To assess the functioning of the parasympathetic nervous system during recovery from pain tasks, HRV was assessed from two 5-min breaks that occurred in between pain tasks on the pain modulation assessment day. During the assessment, participants were seated in a reclining chair and told to get into a comfortable position, relax, and not move while baseline physiology was recorded for 5-mins. The recording began only after they pressed a button to indicate that they were ready.

During each 5-min period, ECG was measured using a Grass Technologies (West Warwick, RI) Model 15LT amplifier (with AC Module 15A54). One active electrode was placed near the right clavicle bone and the other electrode was placed on the lower left abdomen. ECG was sampled at 1000 Hz. Beat-to-beat (ie, RR) intervals were identified offline by an experimenter using custom-built LabVIEW software. In the event of an ectopic beat, the R-spike was estimated to be equally spaced between the adjacent R-spikes. HRV analysis software⁹⁵ was used to trim 30-s off the beginning and end of each record (to account for subjects acclimating to the task), then the trimmed RR-interval records were used to calculate RMSSD. The average RMSSD of these two rest periods was used as the measure of parasympathetic activity. Lower RMSSD is indicative of greater allostatic load¹²¹. As evidence of the validity of our HRV measurement, Table 2 demonstrates that HRV was negatively correlated with age and BMI, as shown in prior research¹. Further, mean (untransformed) RMSSD was 51.36 ms (SD=38.49), which is within the normative range of other studies (ie, M=42 ms, Range=19–75)¹²⁸, and mean heart rate during this period was 66.45 (SD=11.73), suggesting that participants did not continue to experience significant stress during these break periods.

Control Variables

The statistical models were tested with and without control variables that might influence the relationship between cardiometabolic allostatic load and pronociceptive processes. Older age^32,127, biological sex (particularly male sex for cardiovascular indices)⁷⁰, and lower physical activity level⁴⁹ (assessed from the International Physical Activity Questionnaire^31,60) are associated with higher allostatic load; therefore, they were controlled in the pathway linking NA ethnicity to cardiometabolic allostatic load to ensure greater allostatic load in NAs was not due to individual differences in these variables. Moreover, older age⁵⁹, female sex⁹, and lower physical activity^93,94 are also associated with increased pain risk; therefore, these variables were also controlled in the paths between cardiometabolic allostatic load and pain outcomes. In the parent OK-SNAP study¹¹⁵, we found ethnic group differences on general health perception (subscale of the Medical Outcomes Study 36-item Short Form Health Survey, SF-36)¹⁴⁸, self-reported sleep quality (assessed from the subscale of the Pittsburgh Sleep Quality Index, PSQI)²⁰, perceived stress (assessed from the Perceived Stress Scale, PSS)²⁹, and psychological distress (assessed from the Global Severity Index [GSI] of the Symptom Checklist-90-Revised, SCL-90-R)³⁶, so these variables were controlled in the paths linking allostatic load to pain outcomes. And finally, because we individually calibrated the electric stimulus intensity during TS and CPM to each participant’s reflex thresholds rather than their pain threshold/tolerance, this could have impacted modulation of their pain perception (given that the perception of each stimulus varied across participants). As a result, stimulus intensity was added as a control to TS-pain and CPM-pain.

Data Analysis

Structural equation models were conducted with LISREL 8.8⁶⁹. Separate models were used to predict TS/CPM outcomes versus pain tolerance to reduce the number of parameters and complexity of the models and because the models would not converge on a solution when included together. A few participants’ data were excluded from some analyses for failing to follow instructions or poor effort (ie, 1 was excluded from heat tolerance, 1 was excluded from ischemia tolerance). Additionally, some participants were excluded from pain modulation analyses for having baseline ratings at ceiling or floor; thus, impeding any ability to assess modulation. 2 participants were excluded from analyses of CPM-pain, because all ratings during baseline were at ceiling (NRS≥95; none were at floor NRS≤5). 20 participants (8 NHW, 12 NA) were excluded from analysis of TS-Pain because ratings of baseline stimulations were at ceiling (VAS≥95) or floor (VAS≤5), because the intensity of electric stimuli used during TS-pain were calibrated to NFR threshold, rather than pain perception. There were no group differences between individuals excluded for ceiling/floor effects and those used in the analysis on biological sex (χ²=0.009, p=.924), ethnicity (χ²=0.747, p=.387), BMI (t=0.043, p=.966), age (t=1.441, p=.151), physical activity (t=0.330, p=.742), sleep quality (t=.224, p=.823), perceived stress (t=0.247, p=.805), psychological distress (t=0.529, p=.597), mean arterial blood pressure (t=0.849, p=.397), or HRV (t=0.110, p=.913), but general health was slightly better in the group with omitted data (85.5 v. 79.1; t=2.045, p=.042). Cold pressor tolerance, ischemia tolerance, age, physical activity scores (IPAQ), HRV (RMSSD), and psychological distress (GSI) were log10 transformed to reduce positive skew. Outliers were identified using Wilcox’s MAD-median procedure and then winsorized to the next nearest non-outlier value¹⁵¹. All pain outcome variables, except electric tolerance, were winsorized, as were blood pressure (MAP), general health (SF-36), perceived stress (PSS), psychological distress (GSI), physical activity (IPAQ), and HRV (RMSSD). Independent-samples t-tests were conducted in SPSS to examine ethnic differences on study variables for descriptive purposes. Significance was set at p<.05 (2-tailed) for all analyses.

LISREL uses full information maximum likelihood (FIML), which means the models are run with all available data, even if there are missing values (the current state-of-the-art method for handling missingness⁷⁹). Model fit was assessed from the root mean square error of approximation (RMSEA). As noted by Kline⁷⁴, values ≤0.05 are considered “close-fit”, values between .05 and .08 are considered a “reasonable approximate fit,” and values ≥ .10 are considered “poor fit.” Moreover, the “Test of Close Fit” provided by LISREL was also reported. The Test of Close Fit assesses the null hypothesis (H₀) that RMSEA < 0.05, thus if you reject H₀ the model is not close fitting. The chi-square goodness of fit is also reported; however, this metric is highly sensitive to sample size and thus is significant in most large samples. For this reason, the chi-square is rarely used as a measure of fit, although it is almost always reported¹⁴⁷. For both structural models, paths from NA ethnicity to cardiometabolic allostatic load, as well as paths from cardiometabolic allostatic load to pain outcomes were assessed. We have shown elsewhere that ethnic group differences in pain outcomes are not present in this study¹¹⁵, so the direct pathways between NA ethnicity and pain outcomes were not modeled. However, current statistical theories on mediation (indirect tests) do not assume a direct effect between the IV (ethnicity) and DV (eg, pain outcome) is necessary for there to be significant mediation that forms a potential causal pathway between the IV and DV^102,103. Specifically, there may be an indirect/mediated relationship between NA ethnicity and pronociceptive mechanisms that is mediated by allostatic load (NA→allostatic load→pro-nociception), even when NA ethnicity is not directly related to pain outcomes to generate observable group differences in pain processing (ie, even if there is no “total effect”). If there is an indirect effect in the absence of a total effect, this implies that ethnic differences in pro-nociceptive processes exist only in the presence of cardiometabolic allostatic load. For this reason, the LISREL tests of indirect effects that assess for mediation are also reported.

Results

Final Sample

Of the 302 participants, 25% of participants had partial missing data (13% were missing TS/CPM procedures; 15% were missing pain tolerance procedures). Thus, analyses were conducted to determine whether there were differences in those with and without completed data on available background characteristics. There were no differences in biological sex (χ²=3.04, p=.081), ethnicity (χ²=1.78, p=.183), BMI (t=0.273, p=.785), age (t=1.056, p=.292), physical activity (t=1.244, p=.214), sleep quality (t=1.131, p=.259), perceived stress (t=0.954, p=.341), psychological distress (t=1.410, p=.160), general health (t=0.047, p=.962), mean arterial blood pressure (t=0.241, p=.810), or HRV (t=0.246, p=.806). LISREL’s FIML was used to address any missingness on variables for both SEM analyses. For the model predicting CPM/TS outcomes, 3.38% of values were missing. For the model predicting pain tolerance, 1.95% of values were missing.

Background Characteristics

In the full sample of 302 participants, 153 were NA (58% female) and 149 were NHW (50% female). Table 1 presents means and standard deviations (SD) for study variables, separated by ethnic status. As shown, NAs reported worse sleep quality, higher stress, more psychological symptoms, higher blood pressure and BMI, and reduced cold tolerance. There were no group differences on biological sex, χ²=2.20, p=.138. Means, SDs, and Pearson’s inter-correlations for study variables after they were corrected for non-normality are presented in Table 2.

Cardiometabolic Allostatic Load and Pain Modulation

The model without control variables (Table 3, Figure 2) demonstrated a close fit, RMSEA=0.0343 (90% CI: 0.00, 0.0669; test of close fit p=0.753), and a non-significant chi-square (χ²=26.22, df=20, p=0.16). As shown, NAs experienced greater cardiometabolic allostatic load. In turn, greater cardiometabolic allostatic load was associated with greater spinal facilitation (greater TS-NFR), and reduced inhibition of spinal nociception (less CPM-NFR). To visualize this latter effect, a scatterplot was generated that graphed latent scores on cardiometabolic allostatic load against the CPM-NFR change score (Figure 3). As can be seen, not only was CPM-NFR impaired, but higher cardiometabolic allostatic load was associated with a transition to NFR facilitation during CPM. Cardiometabolic allostatic load was not significantly associated with pain facilitation or pain inhibition in this model. Moreover, there were significant indirect effects. These indirect effects provide evidence that NA ethnicity is associated with pronociceptive outcomes as a result of higher cardiometabolic allostatic load in the NA group. Specifically, due to higher cardiometabolic allostatic load, NAs have increased spinal facilitation (TS-NFR; indirect effect=0.116, SE=0.037, p=0.040) and impaired inhibition of spinal nociception (CPM-NFR; indirect effect= 0.069, SE= 0.026, p=0.007), but not pain facilitation (TS-Pain; indirect effect= 1.463, SE= 0.820, p= 0.074) or pain inhibition (CPM-Pain; indirect effect= −0.499, SE= 0.434, p=0.251). The significant indirect effect for TS-NFR indicates that the TS-NFR magnitude is 0.116 d-units higher (greater spinal sensitization) in NAs than NHWs as a result of higher cardiometabolic allostatic load in the NA group. Similarly, the indirect effect for CPM-NFR indicates that NFR inhibition is 0.069 d-units less (impaired spinal inhibition) in NAs than NHWs as a result of higher cardiometabolic allostatic load in the NA group.

Table 3.

Unstandardized coefficients for the model predicting temporal summation and conditioned pain modulation outcomes (N=264)

	Model without Control Variables				Model with Control Variables
Measurement Model	Estimate	SE	Z-test	R2	Estimate	SE	Z-test	R2
Native American Status --> Race (NHW=0, NA=1)	1.000	—	—	—	1.000	—	—	—
Cardiometabolic Allostatic Load --> MAP	1.000	—	—	0.578	1.000	—	—	0.434
Cardiometabolic Allostatic Load --> BMI	0.285	0.064	4.489	0.205	0.328	0.055	5.968	0.203
Cardiometabolic Allostatic Load --> HRV	−0.013	0.004	−3.661	0.106	− 0.019	0.004	−5.361	0.164
Pain Facilitation --> TS-pain	1.000	—	—	—	1.000	—	—	—
Spinal Facilitation --> TS-NFR	1.000	—	—	—	1.000	—	—	—
Pain Inhibition --> CPM of Pain	1.000	—	—	—	1.000	—	—	—
Spinal Inhibition --> CPM of NFR	1.000	—	—	—	1.000	—	—	—
Error in Race
Error in MAP	33.890	9.265	3.658		45.473	5.664	8.028
Error in BMI	14.663	1.534	9.561		14.694	1.407	10.440
Error in HRV	0.067	0.006	10.465		0.063	0.006	10.505
Error in TS-pain	—	—	—		—	—	—
Error in TS-NFR	—	—	—		—	—	—
Error in CPM of Pain	—	—	—		—	—	—
Error in CPM of NFR	—	—	—		—	—	—
Structural Model	Estimate	SE	Z-test	R2	Estimate	SE	Z-test	R2
Predictors of Allostatic Load				0.117				0.653
Race --> Cardiometabolic Allostatic Load	4.673	1.030	4.534		3.862	0.798	4.839
Age --> Cardiometabolic Allostatic Load	—	—	—		25.701	2.831	9.078
Sex --> Cardiometabolic Allostatic Load	—	—	—		− 3.025	0.821	−3.686
Physical Activity --> Cardiometabolic Allostatic Load	—	—	—		− 1.760	0.927	−1.898
Predictors of Pain Facilitation (TS-pain)				0.023				0.106
Cardiometabolic Allostatic Load --> Pain Facilitation	0.313	0.167	1.871		0.062	0.310	0.201
Age --> Pain Facilitation	—	—	—		14.338	9.960	1.440
Sex --> Pain Facilitation	—	—	—		1.924	1.935	0.994
Physical Activity --> Pain Facilitation	—	—	—		5.196	2.072	2.507
Sleep Quality --> Pain Facilitation	—	—	—		2.282	1.359	1.679
Perceived Stress --> Pain Facilitation	—	—	—		− 0.077	0.211	−0.365
Psych Distress --> Pain Facilitation	—	—	—		21.865	14.938	1.464
General Health --> Pain Facilitation	—	—	—		0.052	0.069	0.754
Stimulus Intensity--> Pain Facilitation	—	—	—		− 0.023	0.081	−0.282
Predictors of Spinal Facilitation (TS-NFR)				0.109				0.158
Cardiometabolic Allostatic Load --> Spinal Facilitation	0.025	0.007	3.717		0.025	0.012	2.153
Age --> Spinal Facilitation	—	—	—		0.321	0.365	0.878
Sex --> Spinal Facilitation	—	—	—		− 0.054	0.067	−0.768
Physical Activity --> Spinal Facilitation	—	—	—		0.128	0.074	1.732
Sleep Quality --> Spinal Facilitation	—	—	—		− 0.011	0.047	−0.225
Perceived Stress --> Spinal Facilitation	—	—	—		0.003	0.007	0.454
Psych Distress --> Spinal Facilitation	—	—	—		0.263	0.520	0.507
General Health --> Spinal Facilitation	—	—	—		− 0.000	0.002	−0.123
Predictors of Pain Inhibition (CPM-pain)				0.009				0.161
Cardiometabolic Allostatic Load --> Pain Inhibition	−0.107	0.091	−1.170		− 0.573	0.199	−2.881
Age --> Pain Inhibition	—	—	—		21.112	6.126	3.447
Sex --> Pain Inhibition	—	—	—		− 0.555	1.148	−0.484
Physical Activity --> Pain Inhibition	—	—	—		0.397	1.212	0.328
Sleep Quality --> Pain Inhibition	—	—	—		1.273	0.732	1.738
Perceived Stress --> Pain Inhibition	—	—	—		0.033	0.117	0.281
Psych Distress --> Pain Inhibition	—	—	—		4.479	8.235	0.544
General Health --> Pain Inhibition	—	—	—		0.040	0.038	1.050
Stimulus Intensity --> Pain Inhibition	—	—	—		0.118	0.0389	3.048
Predictors of Spinal Inhibition (CPM-NFR)				0.066				0.100
Cardiometabolic Allostatic Load --> Spinal Inhibition	0.015	0.005	3.004		0.019	0.009	2.025
Age --> Spinal Inhibition	—	—	—		− 0.095	0.293	−0.322
Sex --> Spinal Inhibition	—	—	—		− 0.051	0.056	−0.914
Physical Activity --> Spinal Inhibition	—	—	—		0.092	0.060	1.548
Sleep Quality --> Spinal Inhibition	—	—	—		− 0.000	0.037	−0.003
Perceived Stress --> Spinal Inhibition	—	—	—		0.003	0.006	0.484
Psych Distress --> Spinal Inhibition	—	—	—		0.341	0.421	0.812
General Health --> Spinal Inhibition	—	—	—		0.001	0.002	0.752
Residuals
Residual for Cardiometabolic Allostatic Load	41.049	10.133	4.051		12.106	4.128	2.933
Residual for Pain Facilitation	193.204	17.844	10.827		176.848	16.178	10.932
Residual for Spinal Facilitation	0.236	0.022	10.545		0.223	0.021	10.790
Residual for Pain Inhibition	61.855	5.612	11.023		52.333	5.259	9.950
Residual for Spinal Inhibition	0.144	0.014	10.603		0.139	0.013	10.564

Note: z-tests ≥ 1.96 or ≤ −1.96 are statistically significant at p<.05. Bolded estimates are statistically significant. The R-squared values that are reported in the structural model that includes control variables is the total variance explained in the latent variable. NHW=non-Hispanic White; NA=Native American; MAP=mean arterial blood pressure; BMI=body mass index; HRV=heart rate variability; TS=temporal summation; NFR=nociceptive flexion reflex; CPM=conditioned pain modulation.

Figure 2.

Structural equation model linking Native American ethnicity (coded 0=non-Hispanic White [NHW], 1=Native American [NA]) to cardiometabolic allostatic load and markers of pain modulation. Standardized coefficients and R-squared values without parentheses are from the model without control variables, whereas those in parentheses are from the model with control variables. MAP=mean arterial blood pressure, BMI=body mass index, HRV=heart rate variability, TS-pain=temporal summation of pain, TS-NFR=temporal summation of the nociceptive flexion reflex, CPM-pain=conditioned pain modulation of pain ratings, CPM-NFR=conditioned pain modulation of the nociceptive flexion reflex, RMSEA=root mean square error of approximation. *p<.05.

Figure 3.

Scatterplot depicting the relationship between the cardiometabolic allostatic load latent variable (standardized into z-scores) and conditioned pain modulation of the nociceptive flexion reflex (CPM-NFR). Higher cardiometabolic allostatic load was associated with a transition from spinal inhibition (negative CPM-NFR scores) to spinal facilitation (positive CPM-NFR scores).

The model with control variables (Table 3, results in parentheses of Figure 2) had a slightly worse fit than the model without the controls, but the RMSEA=0.055 (90% CI: 0.0365, 0.0733; test of close fit p=0.30) still suggested a close fit to these data, although there was a significant chi-square (χ²=90.31, df=50, p=0.0004). Even after controlling for age, sex, and physical activity, NAs still experienced greater cardiometabolic allostatic load. Moreover, after controlling for age, sex, physical activity, sleep quality, stress, distress, general health, and stimulus intensity, cardiometabolic allostatic load was still associated with greater spinal facilitation (TS-NFR) and less inhibition of spinal nociception (CPM-NFR). In this model, cardiometabolic allostatic load was also significantly associated with greater pain inhibition (CPM-Pain). This is likely due to controlling for age, because age was significantly related to CPM-pain. Similar to the model without controls, this model also found that there were significant indirect effects indicating that, in the presence of cardiometabolic allostatic load, NAs have increased spinal facilitation (TS-NFR; indirect effect= 0.096, SE=0.045, p=.032) and impaired inhibition of spinal nociception (CPM-NFR; indirect effect= 0.073, SE= 0.036, p=.044), but not pain facilitation (TS-PAIN; indirect effect= 0.240, SE= 1.196, p=.840). The significant indirect effect for TS-NFR indicates that the TS-NFR magnitude is 0.096 d-units higher (greater spinal sensitization) in NAs than NHWs as a result of higher cardiometabolic allostatic load in the NA group. Similarly, the indirect effect for CPM-NFR indicates that NFR inhibition is 0.073 d-units less (impaired spinal inhibition) in NAs than NHWs. Moreover, in this model there was also a significant indirect effect indicating that, in the presence of cardiometabolic allostatic load, NAs have improved pain inhibition when age was controlled (CPM-Pain; indirect effect= −2.212, SE= 0.774, p= .004). This indirect effect indicates that pain was inhibited by an additional 2.212 VAS units (improved pain inhibition) in NAs than NHWs as a result of higher cardiometabolic allostatic load in the NA group.

In terms of the effects of the control variables (Table 3), being male and being older were associated with greater cardiometabolic allostatic load, higher levels of physical activity were associated with greater pain facilitation (TS-pain) and being older was associated with less pain inhibition (CPM-Pain). Higher stimulus intensity was associated with reduced pain inhibition (CPM-Pain).

Cardiometabolic Allostatic Load and Pain Tolerance

Results for these models are presented in Table 4 and Figure 4. Initial attempts to model these tolerance data resulted in a negative variance (Heywood case) until the errors for blood pressure and heat tolerance were allowed to covary (Figure 4). Thus, both models described below included this modification.

Table 4.

Unstandardized coefficients for the model predicting pain tolerance (N=256)

	Model without Control Variables				Model with Control Variables
Measurement Model	Estimate	SE	Z-test	R2	Estimate	SE	Z-test	R2
Native American Status --> Race (NHW=0, NA=1)	1.000	—	—	—	1.000	—	—	—
Cardiometabolic Allostatic Load --> MAP	1.000	—	—	0.849	1.000	—	—	0.458
Cardiometabolic Allostatic Load --> BMI	0.189	0.069	2.723	0.140	0.331	0.055	6.062	0.229
Cardiometabolic Allostatic Load --> HRV	−0.009	0.004	−2.482	0.080	−0.017	0.003	−4.745	0.144
Pain Tolerance --> Heat Tolerance	1.000	—	—	0.437	1.000	—	—	0.500
Pain Tolerance --> Cold Tolerance	0.198	0.028	7.194	0.407	0.181	0.024	7.452	0.387
Pain Tolerance --> Ischemia Tolerance	0.220	0.031	7.045	0.376	0.207	0.028	7.397	0.377
Pain Tolerance --> Electric Tolerance	6.777	0.950	7.137	0.394	5.960	0.828	7.195	0.346
Error in Race
Error in MAP	12.865	24.146	0.533		45.357	6.541	6.935
Error in BMI	15.772	1.652	9.548		14.120	1.409	10.022
Error in HRV	0.067	0.007	10.130		0.063	0.006	10.044
Error in Heat Tolerance	1.886	0.240	7.848		1.667	0.233	7.164
Error in Cold Tolerance	0.083	0.010	8.297		0.086	0.010	8.710
Error in Ischemia Tolerance	0.118	0.014	8.684		0.118	0.013	8.828
Error in Electric Tolerance	103.686	12.239	8.472		111.902	12.244	9.139
Structural Model	Estimate	SE	Z-test	R2	Estimate	SE	Z-test	R2
Prediction of Cardiometabolic Allostatic Load				0.094				0.629
Race --> Cardiometabolic Allostatic Load	5.199	1.086	4.786		4.212	0.858	4.911
Age --> Cardiometabolic Allostatic Load	—	—	—		25.133	2.823	8.903
Sex --> Cardiometabolic Allostatic Load	—	—	—		−3.141	0.848	−3.702
Physical Activity --> Cardiometabolic Allostatic Load	—	—	—		−1.829	0.975	−1.877
Prediction of Pain Tolerance				0.000				0.125
Cardiometabolic Allostatic Load --> Pain Tolerance	0.003	0.012	0.255		−0.030	0.032	−0.930
Age --> Pain Tolerance	—	—	—		0.491	1.017	0.482
Sex --> Pain Tolerance	—	—	—		−0.843	0.211	−3.993
Physical Activity --> Pain Tolerance	—	—	—		0.111	0.221	0.504
Sleep Quality --> Pain Tolerance	—	—	—		0.104	0.140	0.741
Perceived Stress --> Pain Tolerance	—	—	—		−0.028	0.022	−1.231
Psych Distress --> Pain Tolerance	—	—	—		2.420	1.642	1.473
General Health --> Pain Tolerance	—	—	—		0.004	0.009	0.495
Residuals
Residual for Cardiometabolic Allostatic Load	65.301	24.807	2.632		14.205	5.226	2.718
Residual for Pain Tolerance	1.466	0.297	4.929		1.461	0.282	5.188

Note: z-tests ≥ 1.96 or ≤ −1.96 are statistically significant at p<.05. Bolded estimates are statistically significant. The R-squared values that are reported in the structural model that includes control variables is the total variance explained in the latent variable. NHW=non-Hispanic White; NA=Native American; MAP=mean arterial blood pressure; BMI=body mass index; HRV=heart rate variability; TS=temporal summation; NFR=nociceptive flexion reflex; CPM=conditioned pain modulation.

Figure 4.

Structural equation model linking Native American ethnicity (coded 0=non-Hispanic White [NHW], 1=Native American [NA]) to cardiometabolic allostatic load and pain tolerance. Standardized coefficients and R-squared values depicted without parentheses are from the model that did not include control variables. Standardized coefficients and R-squared values without parentheses are from the model without control variables, whereas those in parentheses are from the model with control variables. MAP=mean arterial blood pressure, BMI=body mass index, HRV=heart rate variability, Heat Tol=heat tolerance, Cold Tol=cold pressor tolerance, Isch Tol=ischemic tourniquet test tolerance, Elect Tol=electric tolerance, RMSEA=root mean square error of approximation. *p<.05.

The model without control variables had a RMSEA= 0.0711 (90% CI: 0.0425, 0.0999), suggesting a reasonable fit, but the test of close fit was non-significant (p=0.10) suggesting a close fit. The chi-square test was significant (χ²=41.30, df=18, p=0.0014). As shown in Fig 4, NAs experienced greater cardiometabolic allostatic load, but cardiometabolic allostatic load was not associated with pain tolerance. Not surprisingly, there was not a significant indirect effect linking NAs and pain tolerance in the presence of cardiometabolic allostatic load (indirect effect= 0.016, SE= 0.062, p=.799).

Results from the model with the control variables were similar. That model had a RMSEA= 0.0653 (90% CI: 0.0488, 0.0818) suggesting a reasonable fit, but the test of close fit was non-significant (p=0.063) suggesting close fit. The chi-square test was significant (χ²=119.28, df=57, p<.001). As shown by the coefficients within parentheses of Fig 4, being NA was associated with greater cardiometabolic allostatic load, but cardiometabolic allostatic load was still not associated with pain tolerance. Moreover, there was not a significant indirect effect linking NAs and pain tolerance in the presence of cardiometabolic allostatic load (indirect effect= −0.125, SE= 0.133, p=.3498).

In terms of the effects on the control variables (Table 4), being male and being older was associated with greater cardiometabolic allostatic load, whereas being female was associated with lower pain tolerance.

Discussion

This study found that NAs experienced greater cardiometabolic allostatic load^24–26,141 that was in turn associated with enhanced TS-NFR and reduced CPM-NFR, but not CPM-pain, TS-pain, or pain tolerance. These results were generally replicated even after controlling for age, sex, physical activity, sleep quality, stress, distress, general health perceptions, and stimulus intensity, with one exception. Cardiometabolic allostatic load was associated with improved CPM-pain when controlling for age. Implications are discussed below.

Does Cardiometabolic Allostatic Load Confer Pain Risk in Native Americans by Promoting Latent Sensitization?

Interestingly, we found that cardiometabolic allostatic load was not associated with TS-pain or CPM-pain; rather, it was only associated with enhanced TS-NFR and impaired CPM-NFR. The NFR is a spinal nociceptive reflex¹²⁴ whose arc does not require supraspinal input¹²⁰, even though NFR can be modulated by descending controls from supraspinal centers^117,152. Thus, it is used as an indirect marker of spinal nociception^124,132. Given this, cardiometabolic allostatic load appears to amplify spinal nociception (TS-NFR) perhaps by impairing descending inhibition of spinal nociception (CPM-NFR).

Typically, it is assumed that facilitation of spinal nociception will correspond to a similar pain facilitation; however, evidence suggests that dissociable modulatory circuits can be triggered during CPM. Piche et al⁹⁹ employed a nearly identical CPM paradigm as ours and found that NFR modulation was associated with a cerebrospinal circuit (prefrontal cortex, cingulate cortices, parahippocampal gyrus, supplementary motor area, paracentral lobule, primary somatosensory cortex, thalamus, periaqueductal gray, pons, rostral ventromedial medulla), whereas electric pain modulation was associated with a supraspinal circuit (orbitofrontal cortex, cingulate cortices, medial prefrontal cortex, anterior insula, parahippocampal gyrus, and amygdala). Thus, these separate circuits provide a means by which NFR and pain could diverge. Specifically, cardiometabolic allostatic load may facilitate spinal nociception by impairing cerebrospinal circuitry, but if the supraspinal circuit is not impaired (lack of association between allostatic load and CPM-pain), or even augmented (allostatic load associated with improved CPM-pain in the model with controls) then the supraspinal inhibitory circuit may offset the amplified ascending signal (Fig 5). This could explain why cardiometabolic allostatic load (in healthy, pain free participants) is not associated with enhanced TS-pain or reduced pain tolerance. The present findings are consistent with other OK-SNAP findings indicating that exposure to major life stressors (eg, trauma) produce sensitization of spinal nociception without concomitant pain sensitization^61,71,72,135.

Figure 5.

Putative mechanisms linking cardiometabolic allostatic load to pronociceptive processes in healthy, pain-free individuals. Cardiometabolic allostatic load was associated with impaired cerebrospinal inhibitory mechanisms and sensitization of spinal nociception. However, intact supraspinal inhibitory mechanisms may have inhibited the amplified ascending signal at supraspinal levels thus leading to a lack of pain facilitation or hyperalgesia. This could cause the dissociation between spinal and perceptual measures of pain in the current study.

Sensitization of spinal nociception without pain sensitization could reflect a phenomenon referred to as latent sensitization in rodents (Fig 6)^27,76,136. In rodents, latent sensitization is triggered by a physical insult or environmental stressor and produces sensitization of spinal nociception and (initially) pain hyper-responsivity¹³⁶. However, the pain hyper-responsivity is short-lived and remits after a few days. Interestingly, the spinal sensitization persists for months following the triggering event but is masked by compensatory inhibitory mechanisms³⁰, thus causing it to be “latent” because it is no longer associated with pain hyper-responsivity. Nonetheless, it persists as a vulnerability for chronic pain that can be triggered by future events (eg, injury, stressors) that impair pain inhibitory mechanisms and unmask the spinal sensitization to cause hyperalgesia and persistent pain.⁷⁶

Figure 6.

Hypothetical model linking allostatic load to latent sensitization and pain risk in Native Americans. In rodent models¹³⁶, a triggering event (eg, environmental stressor, physical insult) can cause spinal sensitization and hyperalgesic behavior, but after a few days endogenous inhibitory mechanisms are engaged that eliminate the hyperalgesia. Nonetheless, the spinal sensitization persists and becomes “latent” because it no longer produces pain hypersensitivity. This latent sensitization can persist for long periods of time and serve as a vulnerability for future chronic pain following events (eg, environmental stressors) that disrupt endogenous inhibitory mechanisms that keep the spinal sensitization in check. Findings from OK-SNAP suggest that allostatic load and/or major life stressors might trigger spinal sensitization (eg, enhanced TS-NFR, impaired CPM-NFR) that is offset by compensatory pain inhibitory mechanism (eg, CPM-pain), thus eliminating hyperalgesia (ie, no lowered pain tolerance). However, this pain phenotype could represent a unique risk factor for chronic pain risk if future events (eg, greater allostatic load, additional stressors, injury, advanced age) cause compensatory inhibitory mechanisms to fail and unveil the spinal sensitization to produce enhanced pain. Given that rodent studies suggest the initial hyperalgesia is short-lived relative to spinal sensitization¹³⁶, hyperalgesia would not be observed unless pain testing occurred immediately following a triggering event. Moreover, recruiting only pain-free individuals would further reduce the chance that hyperalgesia would be detected in participants in the current study.

Stressors and the resulting allostatic load could trigger spinal sensitization and compensatory endogenous pain inhibition that keeps the sensitization from producing hyperalgesia. However, over time, latent sensitization could itself promote wear-and-tear of pain inhibitory mechanisms (e.g., the supraspinal inhibitory circuit) which could ultimately lead to chronic pain. Several factors could promote this, such as advancing age (older age = impaired CPM-pain and enhanced TS-pain⁵⁹), physical trauma (increasing nociceptive input and central sensitization⁷⁵), additional adversity (stress unmasks hyperalgesia in rodent models of latent sensitization)^27,76, or chronic illness (eg, diabetes impairs pain inhibition^55,56). The present study found that older age was associated with impaired CPM-pain, so when age was controlled, cardiometabolic allostatic load was associated with improved CPM-pain. This might represent augmented supraspinal pain inhibition processes that counteract spinal sensitization to make it latent.

An additional finding from OK-SNAP that supports this interpretation comes from preliminary analyses of our follow-up data. We assess participants every 6-months to determine if they develop pain lasting ≥3-months that does not remit at subsequent follow-ups. 73% of participants have responded, and 16% of those have developed persistent pain¹²². Remarkably, NAs are transitioning at ^~3-fold the rate of NHWs¹²², and impaired CPM-NFR is associated with this transition⁶⁵. So, when taken together with other OK-SNAP results, adversity and allostatic load may increase pain risk in NAs.

Allostatic Load as a Potential Mediator of Sociocultural Determinants of Pain Risk in Native Americans

NAs have experienced significant historical adversity^{18,43,98,134,141}, including forced relocation and assimilation, forced separation of children from families to attend boarding schools, political oppression, and genocide. Such experiences can significantly impair stress regulation mechanisms (eg, HPA axis) and promote allostatic load^14,87. Via epigenetics, these effects can even be transmitted across generations^18,35,51,154. Moreover, many NAs continue to experience higher levels of adversity than the general U.S. population. For example, NAs experience more adverse childhood experiences (ACEs) and adulthood trauma10,11,73,78,80,100,118,119, food insecurity⁹⁶, and discrimination^48,57. NAs are also more likely to live in poverty¹⁴⁵. Together, these historical and ongoing stressors likely contribute to a “3-hit process of vulnerability”³³ in which genetic/epigenetic predispositions (hit-1), early life adversity (hit-2), and later life environment (hit-3) accumulate to increase allostatic load^{18,43,98,134,141}. The contribution of these factors to pain risk in NAs should be explored in future studies.

Implications: Potential Interventions for Reducing Pain Risk in Native Americans

Given the high levels of adversity, stressors, and allostatic load for NAs, there are multiple levels at which to intervene⁴³. At the societal level, changes in policies and funding are needed to improve environmental and socioeconomic factors^43,141. At the community level, work is being done to maintain NA cultures and languages, and improve access to resources, such as healthy food environments, medical care, and education^13,66. At the individual level, efforts are being made for early detection of allostatic load conditions^63,149,159. Moreover, culturally-responsive (eg, group interpersonal psychotherapy) and empirically-validated (eg, prolonged exposure) treatments are available for addressing the negative psychological sequalae of trauma and adversity^{16,92,101,106,107}.

However, it is important to note that reducing stress may not be enough to eliminate pain disparities related to allostatic load. Our results suggest allostatic load is pronociceptive even after controlling for psychological stress/distress. Therefore, interventions may need to directly target spinal sensitization. To this end, there are a few emerging therapies that may help. For example, NFR threshold is increased (indicative of greater spinal inhibition) by brief relaxation^41,42 and biofeedback¹¹², whereas secondary hyperalgesia (marker of spinal sensitization) is reduced by written disclosure about stressful experiences¹⁵⁶ and treatments that reduce pain catastrophizing¹²³. As Tiedt and Brown note¹⁴¹, such interventions for NAs may benefit from taking into account cultural beliefs, using tactics that are respectful and holistic, and emphasizing balance and connectedness. Working collaboratively with NA communities to develop them is important⁵².

Strengths and Limitations

To our knowledge, this is the first study to assess the relationship between allostatic load and pain processes in NAs. Moreover, both psychophysical and neurophysiological outcomes were used to assess pain perception and spinal nociception, and structural equation modeling was used to analyze these relationships. Nonetheless, a few limitations should be noted.

First, we only assessed 3 markers of allostatic load, which left the neuroendocrine and immune systems unmeasured. Immunity/inflammation may be particularly relevant to chronic pain risk¹³³ and a combined score assessing all systems may lead to a different result than we obtained. Additionally, we only measured HRV between pain tasks. Measurement at another time point (eg, rest day) or as a change from resting baseline may have produced different results. Second, excluding persons with chronic health conditions, current substance use disorder, and psychiatric medication use could limit the generalizability and/or obscure ethnic differences due to the differential prevalence rates in NAs and NHWs. Moreover, only recruiting healthy, pain-free individuals likely restricted the range of allostatic load markers. However, this choice was necessary so that we could try to identify markers of pain risk before its onset. Future studies could recruit participants with subclinical disease states associated with high allostatic load (eg, prediabetes) to help overcome this limitation, as well as test these relationships in less healthy populations. Third, our sample was large for this type of study, and is the largest study of NA pain mechanisms, but was too small to test more complex models or to ensure the stability of the models with control variables. Nonetheless, the results with and without controls were similar. Fourth, there were a few participants that did not complete all study tasks. Although this could limit generalizability, analyses of completers vs. non-completers have found few notable differences¹¹⁵. And finally, NAs in OK-SNAP were primarily from northeastern Oklahoma; there may be important differences associated with NAs from other regions that could alter allostatic load and/or pain processing (eg, diet, beliefs, acculturation, access to health care)³⁹.

Conclusions

NAs experienced greater cardiometabolic allostatic load that was in turn associated with a pronociceptive pain phenotype indicative of latent spinal sensitization (ie, facilitation of spinal nociception, no hyperalgesia). Thus, cardiometabolic allostatic load could contribute to pain disparities in this population.

Highlights

• Native Americans (NAs) experience high rates of chronic pain

• NAs have suffered generations of sociocultural stressors

• Allostatic load (AL) is stress-related wear-and-tear on homeostatic systems

• Cardiometabolic AL linked NA ethnicity to sensitization of spinal nociception

• Cardiometabolic AL was not linked to pain sensitization

Perspective:

Healthy, pain-free Native Americans experienced greater cardiometabolic allostatic load that was associated with a pronociceptive pain phenotype indicative of latent spinal sensitization (ie, spinal sensitization not associated with hyperalgesia). This latent spinal sensitization could represent a pain risk phenotype for this population.