Probable chronic pain, brain structure, and Alzheimer’s plasma biomarkers in older men

Abstract

Chronic pain leads to tau accumulation and hippocampal atrophy in mice. In this study, we provide one of the first assessments in humans, examining the associations of probable chronic pain with hippocampal volume, integrity of the locus coeruleus (LC)—an upstream site of tau deposition—and Alzheimer’s Disease (AD)-related plasma biomarkers. Participants were mostly cognitively unimpaired men. Probable chronic pain was defined as moderate-to-severe pain in 2+ study waves at average ages 56, 62, and 68. At age 68, 424 participants underwent structural MRI of hippocampal volume and LC-sensitive MRI providing an index of LC integrity (LC contrast-to-noise ratio [LC_CNR]). Analyses adjusted for confounders including major health conditions, depressive symptoms, and opioid use. Models showed that men with probable chronic pain had smaller hippocampal volume and lower rostral-middle—but not caudal—LC_CNR compared to men without probable chronic pain. Men with probable chronic pain also had higher levels of plasma total tau, beta-amyloid-42, and beta-amyloid-40 compared to men without probable chronic pain. These findings suggest that probable chronic pain is associated with tau accumulation and reduced structural brain integrity in regions affected early in the development of AD.

1. INTRODUCTION

Approximately 27%-33% of older adults experience chronic pain, i.e., persistent or recurrent pain lasting 3 or more months.^{20, 63} Chronic pain is related to negative affective symptoms, sleep disruption, reduced physical function, cognitive impairment, and dementia.^{17, 34, 60} This has led researchers to postulate an association between chronic pain and Alzheimer’s disease (AD). So far most human studies have focused on global brain measures, showing evidence of reduced gray matter volume and white matter integrity in adults with chronic pain.^{51, 76} However, chronic pain in older adults is related to smaller hippocampal volume, suggesting the importance of assessing regions particularly affected by AD.²⁶

The association between chronic pain and AD-related neurodegeneration is unclear in humans but has been shown in animals. Guerreiro et al.³² found that induction of chronic pain in wildtype rats led to atrophy and increased levels of tau in the hippocampus compared to unaffected rats. Furthermore, knocking out tau-regulating genes removed the effect of chronic pain on hippocampal atrophy, confirming the neurodegenerative role of tau in this pain model. Additional human studies are needed to examine how chronic pain is associated with measures of AD-related neurodegeneration. If chronic pain is associated with AD-related neurodegeneration then associations should exist beyond the hippocampus, including other AD-affected brain regions and AD-related biomarkers.

One AD-related region of particular interest is the locus coeruleus (LC). Braak et al.¹⁴ have shown that the LC may be the earliest site of abnormal tau deposition. Studies show that tau accumulation typically begins in the LC and later spreads to medial temporal lobe regions, including the hippocampus.^{10, 11, 18, 67} The rostral-middle LC region is particularly vulnerable to phosphorylated tau compared to caudal LC regions.⁶⁷ If chronic pain leads to AD-related neurodegeneration in the hippocampus, then we would expect a relationship with structural LC integrity as well as AD-related biomarkers such as tau. The LC is a small nucleus (30-60mm³) in the brainstem that serves as the central projection site for norepinephrine (NE), and is important for modulating arousal and cognitive function.⁶² Moreover, the LC is also vitally linked to pain processing. Caudal LC projections modulate the transmission of nociceptive signals from the dorsal horn and rostral-middle LC projections modulate sensory and affective pain processing in the cortex.^{44, 47}

Clarifying the relationship of chronic pain with brain regional integrity, including the LC and the hippocampus, and key AD-related biomarkers may help us understand why chronic pain is associated with cognitive decline and increased risk for AD in older adults.^{25, 72} In the present study, we examined the association of chronic pain with hippocampal volume and LC structural integrity in older men. LC integrity was based on MRI acquisition sensitive to signal in the LC region. LC signal been shown to wane in the early stages of AD³⁶ and relates to post-mortem LC neuronal count.³⁸ We examined the rostral-middle and caudal LC regions separately. We hypothesized that men with probable chronic pain would show lower hippocampal volume and lower LC rostral-middle integrity compared to men without probable chronic pain. We also examined group differences in AD-related plasma biomarkers and hypothesized that AD-related plasma biomarkers would be associated with lower hippocampal volume and lower LC integrity.

METHODS

2.1. PARTICIPANTS

Participants were in the Vietnam Era Twin Study of Aging (VETSA), a longitudinal study of men assessed at average ages 56, 62, and 68 – an approximately 12-year time span. All participants served in the military at some time during the Vietnam era (1965-1975); about 80% reported no combat exposure. This sample is representative of general community-dwelling men in their age cohort on health, education, and lifestyle factors based on Centers for Disease Control and Prevention data.⁵⁴ Previous studies have described the methodology and available data in the VETSA project.^{40, 41} Most data are publicly accessible to researchers through data use authorization (https://psychiatry.ucsd.edu/research/programs-centers/vetsa/researchers.html). VETSA procedures were approved by the Institutional Review Boards at the University of California, San Diego (UCSD), and Boston University. Protocols were identical at each site. Participants provided informed consent prior to assessment. MRIs at wave 3 when average age was 68 (2016-2019) were performed at the UCSD site only.

Of eligible MRI participants, 485 (92.4%) met standard MRI inclusion criteria (e.g., no metal in the body) and completed imaging of the LC. Of these, we excluded men with MRI-based cerebral abnormalities (encephalomalacia, meningioma, large infarct, etc.; n = 9) and those who had poor LC imaging quality due to excessive head motion in the scanner (n = 4) as assessed by visual inspection. We also excluded men who did not have at least 2 timepoints of data on pain (n = 48). This resulted in an analytic sample size of 424 men for analyses of LC integrity with an average age of 67.06 years (SD = 2.61, range = 61.00-71.00), of whom 421 had available AD-related plasma biomarkers. Demographic characteristics can be seen in Table 1. For data on hippocampal volume, 47 additional men were excluded who had poor imaging quality (n = 47) resulting in a sample size of 378 with an average age of 67.53 (SD = 2.55, range = 62.37 to 71.00). Of men with data on hippocampal volume, 344 had data available on AD-related plasma biomarkers.

Table 1.

Demographic characteristics of sample (n = 424).

	%	n	M	SD	Range
Age			67.06	2.61	61.00 to 71.00
Probable Chronic Pain	14.2%	60
Racea
Non-Hispanic White	87.7%	372
Hispanic White	2.8%	12
Non-Hispanic Black or African-American	7.3%	31
Hispanic Black or African American	0.7%	3
American Indian or Alaskan Native	0.2%	1
Hispanic –American Indian or Alaskan Native	0.2%	1
More than one race	0.9%	4
Education (years)			14.05	2.13	8 to 20
Physical Morbidities
0	17.7%	75
1	32.3%	137
2+	50.0%	212
Depressive Symptoms			6.65	6.90	0 to 47.00
Opioid Medications			1.36	0.88	0.14 to 9.00
0	96.0%	407
1+	4.0%	17

Notes. Probable chronic pain was defined as pain on the SF-36 Bodily Pain Intensity item (Question 22) greater than mild (>3/6) in two or more study waves. Physical morbidities included major health conditions from the Charlson Index. Depressive symptoms were measured from the Center for Epidemiological Studies Depression scale. Opioid medications were coded as the number of opioid medications recorded during the medical history interview.

^aFor statistical purposes, race was coded as non-Hispanic White versus other race for analyses.

2.2. Measures

2.2.1. Probable Chronic Pain

Pain was assessed using the SF-36 Quality of Life (Version 1.0) Bodily Pain Scale⁷⁰ which was given at each assessment wave. We used the pain severity item, “How much pain severity have you had during the past 4 weeks”. Severity was rated on a 6-point Likert-type scale: “None” (1),“Very Mild” (2),“Mild” (3),“Moderate” (4),“Severe” (5),“Very Severe” (6). Probable chronic pain was defined as moderate to very severe pain (values 4-6 on the scale) at wave 3 and at least 1 of the 2 previous waves. Pain reported over 2 or more waves (approximately 6-12 years) fits the International Association for the Study of Pain definition of chronic pain as pain that lasts or recurs for 3 or more months.⁶³ Furthermore, despite not knowing about pain status between study waves, pain reporting is typically stable in the population making such an approach reasonable.⁴³ Men reporting no pain or pain during only one wave were classified as not having probable chronic pain.

2.2.2. MRI Acquisition and Processing

2.2.2.1. Hippocampal volume.

We previously described our MR imaging procedures.^{23, 24} In brief, participants completed scans in GE 3T Discovery 750x scanners (GE Healthcare, Waukesha, WI, USA) with an 8-channel phased-array head coil. Used for determining regional volumes/thickness, we acquired sagittal T₁-weighted 3D fast spoiled gradient echo (FSPGR) sequences (TE = 3.164 msec, TR = 8.084 msec, TI = 600 msec, flip angle = 8°, matrix = 256 × 192, in-plane resolution = 1 × 1 mm, slice thickness = 1.2 mm, slices = 172). Magnetic resonance images were processed using methods previously described but utilizing the latest software.⁴² When assessing the number of topological defects, as a measure of overall image quality provided from FreeSurfer version 6.0,⁵² there were no statistically meaningful differences between men with and without probable chronic pain (t(413) = −.24, p = .811). In brief, hippocampal volume was derived from atlas-based volumetric segmentation using FreeSurfer version 6.0 (http://surfer.nmr.mgh.harvard.edu)^{.27, 28} Quality of subcortical segmentations was checked using visual review, and participants with inaccurate segmentations (defined by obvious overestimation or underestimation on the segmentation atlas via a detailed lab protocol) were removed from analysis (n = 47). Hippocampal volume was pre-adjusted for an individual’s estimated intracranial volume using residuals from the following linear regression: Y_i(hippocampal volume) ~ β0_i + β1(intracranial volume)_ij + ε_ij, i = individual observation, j = nested within twin pairs.

2.2.2.2. LC integrity.

Detailed descriptions of LC imaging in our sample has been previously published.^{23, 24} Oblique axial FSE-T1-weighted images were obtained (TR = 600 ms; TE = 14 ms; flip angle = 90°; matrix = 512 × 320; FOV = 220 mm; pixel size 0.42 × 0.68 mm; 10 slices; slice thickness = 2.5 mm; interslice gap = 1 mm). Briefly, signal intensities were derived from manually marked regions of interest (ROIs) on 3 axially-oriented slices corresponding to rostral, middle, and caudal LC (see Figure 1 for an illustration). Each image was manually marked by 2 of 4 experienced raters who were blind to any characteristic of the participants, including probable chronic pain status. On each slice, a 3mm²-voxel cross hair-shaped ROI was placed over left and right LC and a 10 mm² square reference ROI was placed in the pontine tegmentum (PT). Mean signal was then extracted from each ROI. The values from left and right LC were then averaged for each slice. LC contrast-to-noise ratio (LC_CNR) values were calculated for each slice as LC_CNR=(LC_intensity − PT_intensity)/PT_intensity. There was 95% interrater reliability across 4 raters (calculated from the results of a mixed model; Wald’s Z = 15.14, p <.001). Higher LC_CNR values are thought to reflect better LC structural integrity. LC_CNR values were averaged from the rostral and middle slices to derive a final rostral-middle LC_CNR value. This was done as the rostral and middle LC regions show prominent changes due to aging and AD pathology compared to the caudal region, and thus allowed us to assess regional specificity of effects.^{11, 30}

Figure 1.

Summary of the manual marking method of the LC. Note. The middle slice is selected 7 mm below the inferior colliculus. Left and right portions of the locus coeruleus (LC) are marked on the rostral, middle, and caudal slices with a 3mm2 cross. Signal intensity is averaged from left and right regions to calculate rostral, middle, and caudal LC intensity. As a reference region, we placed a 10 mm² square placed over the pontine tegmentum (PT). The same marking rules were used to calculate signal intensity for the rostral, middle, and caudal slice of the PT. A contrast to noise (CNR) is created for each region using LC signal intensity subtracted by PT signal intensity and divided by PT signal intensity for each region. For this study, we averaged rostral and middle LC CNR as both regions show similar age and disease-related effects. The caudal LC CNR was used as an exploratory aim and comparison region.

2.2.3. AD-related Plasma Biomarkers

AD-related biomarkers included plasma levels of t-tau, Aβ40, Aβ42, and neurofilament light chain (NfL) collected at Wave 3.³¹ Plasma samples were collected after fasting beginning by 9:00 PM the night before the study appointment. Specimens were collected between 8:00 AM and 8:30 AM the next morning. Levels of t-tau, Aβ40, and Aβ42 were assayed using the high throughput Simoa Human Neurology 3-plex A (N3PA) Quanterix HDX platform. Sample handling and assays were performed following clinical trial standard operating procedures and manufacturer instructions. This approach has been developed by Winston et al and used in multiple publications from the laboratory of Dr. Robert Rissman.^{49, 58, 73, 74} Level of NfL was assayed using single-plex plate on the ultra-sensitive Simoa technology platform HD-1 (Simoa NfL Advantage Kit; Quanterix Corporation) proided by the Alzheimer’s Therapeutic Research Institute (ATRI) Biomarker Core at University of Southern California (PI: Dr. Robert Rissman).⁴⁹ Values were excluded if there was presence of hemolysis and/or a coefficient of variance for plasma concentrations >.20. All biomarkers were log-transformed due to positive skew. All biomarker measures were pre-adjusted for site and storage time using residuals from the following linear regression: Y_i(biomarker) ~ β0_i + β1(site)_ij + β2(storage time)_ij + ε_ij, i = individuals, j = nested within twin pairs.

2.2.4. Covariates

Variables that could confound associations with probable chronic pain and other measures of interest were included as covariates from Wave 3. These included age, race/ethnicity (non-Hispanic white versus other), lifetime education (years), depressive symptoms indexed using the 20-item Center of Epidemiological Studies scale (CES-D⁵⁰), physical morbidities (total number of reported major health conditions from a modified Charlson Index^{15, 16}), and the use of opioid medication based on reports of medications taken in the medical history interview (yes/no). Such variables were selected as important covariates used in prior research.⁶

2.4. Statistical Analysis

We compared key covariates between men with and without probable chronic pain for descriptive purposes. Linear mixed models were conducted using the function GENLINMIXED in SPSS (version 28.01.01, IBM Corp., 2023) to examine the differences in brain measures (rostral-middle LC-CNR; caudal LC-CNR; hippocampal volume) and plasma biomarkers (t-tau, Aβ40, Aβ42, NfL) between men with probable chronic pain and men without probable chronic pain while adjusting for covariates and nesting of data within twin pairs. The basic equation is as follows: Yij(outcome)~ β0ij + β1(probable chronic pain status)_ij + β2(age)_ij + β3(race)_ij + β4(lifetime education)_ij + β5(depressive symptoms)_ij + β6(physical morbidities)_ij + β7(number of opioid medications reported)_ij + ε_ij; i = individual observations across participants, j = twin pairs. Correlation analyses were used to examine how levels of AD-related biomarkers related to hippocampal volume and LC-CNR using polycholoric correlations in OpenMx, which accounts for twin relatedness. All statistical code is provided online (https://github.com/trbellucsd/pain_ADbrainstructure_biomarkers). Statistical significance was set to an α of .05, and 95% confidence intervals were calculated.

3. RESULTS

3.1. Descriptive Statistics

Descriptive statistics are shown in Table 1. For participants completing LC_CNR measurement, 14.2% (n = 60) reported probable chronic pain. Compared with men who did not report probable chronic pain, men reporting probable chronic pain had more depressive symptoms (t(97.44) = 4.30, p <.001, d = .87), more physical morbidities (t(98.26) = 5.16, p = .013, d = 1.04), and were more likely to be using opioid medications (χ²(1) = 25.67, p <.001, η² = .23; 16.3% versus 2.8%), but did not differ in age (p = .069).

3.2. Hippocampal volume

As shown in Table 2 and Figure 2, after controlling for covariates, men with probable chronic pain had significantly smaller hippocampal volumes than men without probable chronic pain (β = −0.24, 95%CI: −0.47 to −0.01, p = .039). Greater depressive symptoms were associated with lower hippocampal volume (β = −0.11, 95%CI: −0.19 to −0.02, p = .011).

Table 2.

Main results of linear mixed models examining association with hippocampal volume.

Variable	β	95%CI	p
Probable Chronic Pain	−.24	−.46 to −.01	.039
Age	−.07	−.15 to .02	.112
non-Hispanic White	.21	−.06 to .47	.122
Education (years)	.03	−.04 to .11	.391
Physical Morbidities	.02	−.06 to .10	.591
Depressive Symptoms	−.11	−.19 to −.02	.011
Opioid Medication Use	−.08	−.18 to .03	.138

Notes. Probable chronic pain was defined as pain on the SF-36 Bodily Pain Intensity item (Question 22) greater than mild (>3/6) in two or more study waves. Physical morbidities included major health conditions from the Charlson Index. Depressive symptoms were measured from the Center for Epidemiological Studies Depression scale. Opioid medications were coded as the number of opioid medications recorded during the medical history interview.

Figure 2.

Differences in hippocampal volume between men with and without probable chronic pain. Hippocampal volume was assessed during wave 3 (average age 68) and was adjusted for effects of intracranial volume and scanner type. Values on the y-axis represent z-scores relative to the samples mean and standard deviation. Hippocampal volume was assessed during wave 3 (average age 68).

3.3. Rostral-middle LC_CNR

As shown in Table 3 and Figure 3, after controlling for covariates, men with probable chronic pain had significantly lower rostral-middle LC_CNR (β = −0.38, 95% CI: −.64 to −.12, p = .004) than men without probable chronic pain. Opioid use was associated with higher rostral-middle LC_CNR (β = 0.13, 95% CI: 0.02 to 0.25, p = .026). Age, race/ethnicity, education, depressive symptoms, and physical morbidities were not significantly associated with rostral-middle LC_CNR (ps > .05).

Table 3.

Main results of linear mixed models examining association with rostral LC_CNR and PT.

Variable	β	95%CI	p
Rostral LCCNR (z-score)
Probable Chronic Pain	−.38	−.64 to −.12	.004
Age	−.18	−.40 to .05	.125
non-Hispanic White	.14	−.21 to .50	.424
Education (years)	.03	−.01 to .08	.150
Physical Morbidities	.06	−.04 to .16	.237
Depressive Symptoms	−.06	−.17 to .04	.243
Opioid Medication Use	.13	.02 to .25	.026
Rostral PT Signal
Probable Chronic Pain	−.06	−.25 to .13	.557
Age	−.03	−.19 to .13	.747
non-Hispanic White	−.21	−.44 to .03	.092
Education (years)	−.04	−.11 to .03	.246
Physical Morbidities	.13	.05 to .20	.001
Depressive Symptoms	−.02	−.10 to .06	.689
Opioid Medication Use	.03	−.05 to .12	.439

Notes. CNR = contrast to noise; LC = locus coeruleus. The rostral LC_CNR value was calculated as the average CNR value from the rostral and middle LC regions. Probable chronic pain was defined as pain on the SF-36 Bodily Pain Intensity item (Question 22) greater than mild (>3/6) in two or more study waves. Physical morbidities included major health conditions from the Charlson Index. Depressive symptoms were measured from the Center for Epidemiological Studies Depression scale. Opioid medications were coded as the number of opioid medications recorded during the medical history interview.

Figure 3.

Differences in the rostral middle and caudal locus coeruleus integrity in older adults with and without probable chronic pain. Locus coeruleus (LC) integrity was assessed using a contrast to noise ratio (CNR). Data for men with or without probable chronic pain are displayed for rostral-middle LC-CNR (A) and caudal LC-CNR (B). LC integrity was assessed during wave 3 (average age 68).

3.4. Caudal LC_CNR

As shown in Table 4 and Figure 3, after controlling for covariates, men with probable chronic pain did not differ from participants without probable chronic pain on caudal LC_CNR (p = .750). No covariate was significantly associated with caudal LC_CNR (ps > .05).

Table 4.

Main results of linear mixed models examining association with caudal LC_CNR and PT.

Variable	β	95%CI	p
Caudal LCCNR (z-score)
Probable Chronic Pain	−.05	−.33 to .24	.750
Age	−.12	−.35 to .11	.295
non-Hispanic White	−.15	−.50 to .20	.408
Education (years)	.05	−.05 to .15	.339
Physical Morbidities	.06	−.05 to .17	.302
Depressive Symptoms	−.01	−.13 to .11	.903
Opioid Medication Use	.05	−.08 to .19	.410
Caudal PT Signal (z-score)
Probable Chronic Pain	−.05	−.24 to .14	.613
Age	−.05	−.20 to .10	.541
non-Hispanic White	−.21	−.44 to .02	.078
Education (years)	−.05	−.01 to .11	.124
Physical Morbidities	.10	.03 to .18	.005
Depressive Symptoms	−.05	−.13 to .03	.203
Opioid Medication Use	.04	−.05 to .12	.372

Notes. CNR = contrast to noise; LC = locus coeruleus. Probable chronic pain was defined as pain on the SF-36 Bodily Pain Intensity item (Question 22) greater than mild (>3/6) in two or more study waves. Physical morbidities included major health conditions from the Charlson Index. Depressive symptoms were measured from the Center for Epidemiological Studies Depression scale. Opioid medications were coded as the number of opioid medications recorded during the medical history interview.

3.5. AD-related plasma biomarkers

As illustrated in Figure 4, men with probable chronic pain had higher levels of t-tau than men without probable chronic pain (β = 0.32, 95% CI: 0.01 to 0.63, p = .047). Men with probable chronic pain also had higher levels of Aβ42 (β = 0.32, 95% CI: 0.03 to 0.62, p = .033) and Aβ40 (β = 0.31, 95% CI: 0.03 to 0.60, p = .032) compared to men without probable chronic pain (see Figure 5). Men with probable chronic pain did not significantly differ on level of NfL compared to men without probable chronic pain (p = .313). Regarding MRI measures, higher levels of Aβ42 (r = −0.19) and Aβ40 (r = −0.15) were associated with smaller hippocampal volume (ps < .01). In addition, higher levels of Aβ40 were associated with lower rostral-middle LC integrity (r = −0.11, p = .034). Associations of tau and NfL with MRI measures did not reach statistical significance (ps > .05).

Figure 4.

Differences in plasma t-tau between men with and without probable chronic pain. Plasma t-tau was measured during wave 3 (average age 68) and was adjusted for the effects of site and storage time. Values on the y-axis represent z-scores relative to the samples mean and standard deviation.

Figure 5.

Differences in plasma Aβ between men with and without probable chronic pain. Comparisons of Aβ42 (A) and Aβ40 (B) are shown for men with or without probable chronic pain. Plasma Aβ was measured during wave 3 (average age 68) and was adjusted for the effects of site and storage time. Values on the y-axis represent z-scores relative to the samples mean and standard deviation.

4. DISCUSSION

Our analyses showed that men with probable chronic pain had smaller hippocampal volumes and lower rostral-middle LC structural integrity compared to men without probable chronic pain at average age 68. Probable chronic pain was associated with higher levels of plasma t-tau, Aβ42, and Aβ40. Higher levels of Aβ42 and 40 were also associated with smaller hippocampal volumes and Aβ40 with lower rostral-middle LC integrity. By examining AD-related brain structure and plasma biomarkers, our work expands research that has largely only focused on the risk of AD dementia in people with chronic pain.^{25, 64, 72} Below we discuss how our findings may indicate AD pathology, but also important caveats and alternative explanations to consider.

Our findings build on an animal model showing that pain appears to play a causal role.³² First, consistent with that model, we found that men with probable chronic pain had smaller hippocampal volume than men without probable chronic pain, a region often typically affected by AD-related neurodegeneration.⁴⁸ Animal studies have shown that chronic pain results in hippocampal atrophy in rats with spinal nerve injury.³² Our finding extends the one prior study of older adults of which we are aware.²⁶ They found that chronic pain was related to smaller right hippocampal volume in a sample with a mean age of 79 years. We found an association for bilateral hippocampal volume in the decade younger VETSA sample (mean age = 67 years). Regarding clinical samples focused on pain (not restricted to older adults), our work is consistent with one study but not another. Lutz et al.⁴⁵ found smaller hippocampal volume in middle-aged women with widespread pain compared to women without pain (mean age = 54). Schweinhardt et al.⁵⁵ found greater hippocampal volume in young women with chronic vulvar pain compared to women without pain (mean age = 25.7). The latter finding may be due to the different type of pain assessed or to the very young age of the sample when AD-related neurodegeneration would not be expected. It is possible that mechanisms are dependent on age and therefore help explain inconsistent findings. Studies in children and young adults with chronic pain often show structural brain differences, thought to mostly reflect cortical re-organization rather than atrophy.¹² Meanwhile, structural brain differences in middle-aged and older adults is thought to reflect impacts of atrophy.² Although this will require further work, our study suggests some involvement of AD-related processes such as tau deposition and amyloidosis that can occur decades before dementia, as early as age 40.^{69, 75} Thus, the impact of pain on brain structure may be influenced by the context in which it occurs and the relative prominence of different processes (e.g., developmental, disease-related, etc.). Inconsistent findings may also reflect sex-related differences. For example, women may show more prominent brain differences than men in the presence of pain.³³ On the other hand, similar pain-hippocampal volume associations have been observed in women⁴⁵ and both older men and women with chronic pain show accelerated brain aging.¹⁹ Further research is needed to fully understand age- and sex-dependent relationships and their implications for AD risk in older adults with chronic pain.

Based on the progression of AD pathology across the brain,¹⁴ we hypothesized that if chronic pain caused hippocampal atrophy, pain should already be associated with LC integrity. Men with probable chronic pain had lower integrity of the rostral-middle LC but not the caudal LC compared to men without probable chronic pain. An association specific to the rostral-middle LC is consistent with AD-related neurodegeneration as the rostral-middle LC has been more strongly related to aging and AD.^{11, 23, 30, 67} This finding advances previous work by establishing a relationship between chronic pain and rostral-middle LC structural integrity in humans for the first time. Animal experiments have examined the causal role of the LC in pain modulation. In rats, stimulation of the LC suppressed pain while lesions to the LC increased pain sensitivity.⁴⁴ Animal models of chronic pain and drug therapy show that targeting the LC-NE system through gabapentinoids³⁵ and serotonin-norepinephrine reuptake inhibitors markedly inhibited neuropathic pain in rats with ablated or constricted nerve injury.¹ Early evidence consistent with the LC-NE system’s role in human pain comes from successful treatment with similar medications.^{5, 57}

Caudal LC-NE projections have been shown to descend into the dorsal horn of the spine where nociceptive signals are gated (allowed to ascend or not).⁴⁴ Rostral-middle LC-NE projections have been shown to ascend into cortical regions involved in sensory and affective pain processing.⁴⁷ The stronger association of rostral-middle LC integrity with both chronic pain and AD suggests that in humans the impact of chronic pain operates primarily through the sensory-affective experience of pain.

Our study showed that men with probable chronic pain had higher levels of t-tau than men without probable chronic pain, which is consistent with the animal model showing that chronic pain increased tau levels.³² Tau deposition is mostly found in the rostral-middle LC region rather than the caudal LC region.^{11, 67} Taken together with evidence that the LC is perhaps the earliest site of tau deposition in the brain,¹⁴ the association of chronic pain with rostral-middle LC integrity and hippocampal volume is consistent with animal model data indicating the causal role of chronic pain in promoting early AD-related pathology. On the other hand, although t-tau was associated with chronic pain, it was not associated with hippocampal volume or LC integrity in our sample. There is, however, evidence that reduction of LC integrity occurs long before notable tau accumulation.³⁶ We do not know if the effects of chronic pain on brain structure and tau accumulation occur simultaneously, or if our assay of tau is not sufficiently sensitive to detect early increases. It is thus possible that associations with tau will emerge in the next wave of the study (currently underway).

Men with probable chronic pain had higher levels of plasma Aβ42 than men without probable chronic pain, which may be elevated in the earliest pre-clinical stages of AD.^{4, 13, 21, 46} Although lower levels of cerebrospinal fluid or plasma Aβ42 are typically associated with greater pathology, a non-linear or biphasic pattern has been demonstrated in at least 3 independent samples.⁵⁹ In those samples, cerebrospinal fluid Aβ42 was positively associated with tau in adults aged 70 and under but negatively associated in those over 70.⁵⁹ Given the age of the VETSA sample, the positive correlation between chronic pain and Aβ42 might become negative in later waves of the study. Alternatively, higher plasma Aβ may reflect contributions of other health-related factors such as vascular disease.⁶⁶

As a caveat, alternative factors may contribute to the observed relationships. For one, the association of probable chronic pain with smaller hippocampal volume and lower LC integrity could be partly explained by the role of the brain structure and function in chronic pain etiology. For example, work by Apkarian et al.³ have shown that abnormal neurogenesis in the hippocampus can contribute to chronic pain. Specifically, downregulating neurogenesis in mice using pharmacological infusion, ablation, and transgenics, diminished or blocked the occurrence of chronic pain, while methods of upregulation led to prolonged chronic pain and negative mood. Furthermore, morphological differences in the hippocampus and the LC may be part of larger networks involved in chronic pain initiation and duration, such as the sensorimotor, salience detection, and the default mode networks.¹² Indeed, studies show an important role of the limbic system and the LC in interacting with such networks.^{65, 68} As such, it is possible that smaller hippocampal volume and lower rostral-middle LC integrity in probable chronic pain is related to its etiology, suggesting bidirectionality. Associations with the LC and AD-related biomarkers suggest some role of AD pathology in later life.

Regardless of mechanisms, differences in brain structure may be related to symptoms in older men with chronic pain; however direction of causality cannot be determined in these analyses. Lower hippocampal volume has been linked to problems with mood and cognitive performance^{26, 32} which are more common in older adults with chronic pain than those without chronic pain.^{8, 9} Previously, we found that poorer rostral-middle LC integrity in this sample was related to poorer cognitive performance, increased risk of amnestic mild cognitive impairment, and lower cortical microstructural integrity.²⁴ Our prior work in this sample also indicates a potential role of rostral-middle LC in other symptoms such as daytime sleep-related dysfunction,²³ a symptom frequently reported in people with chronic pain.²² We have also shown that men with lower rostral-middle LC integrity had greater subjective cognitive complaints⁷ which are more common in older adults with chronic pain compared to those without chronic pain.⁷¹ In older adults, chronic pain might be linked to these clinical symptoms partly due to AD-related neurodegeneration of the LC.

Our study has both limitations and strengths. We cannot directly address causality due to the cross-sectional observational nature of our study. However, previous work shows that chronic pain leads to neurodegeneration in animal and human studies, and that this involves AD-related pathology.^{32, 51} It is still possible that neurodegeneration in the LC also leads to chronic pain due to increased pain sensitivity, suggesting bidirectionality.⁴⁴ It is also possible that our biomarkers are not AD-specific but reflect other peripheral conditions affecting plasma levels such as poorer physical health.²⁹ Our sample is all male, making generalizability to women uncertain. Our sample was also primarily non-Hispanic white, limiting generalization to racial/ethnic minorities. Regarding operationalization of chronic pain, our definition was based on pain severity at waves of observational data rather than clinical diagnosis. However, this method has produced results similar to those using diagnostic codes.^{39, 72} A small number of individuals were classified with probable chronic pain (14% of sample), which produced very unbalanced sample sizes across groups. However, we employed linear mixed models which are robust to issues arising from unbalanced sample sizes across groups, such as heteroscedasticity.^{37, 53} Furthermore, we were focused on cases of probable chronic pain but more work may be done to understand the basic relationship between general pain reporting and pain processing with AD-related biomarkers and brain structure, though beyond the scope of our study.

Regarding measurement limitations, it is possible that the caudal LC is more difficult to image than the rostral region which could explain a lack of significant findings for the caudal LC region. However, histological studies do show the caudal regions to be relatively spared from neurodegeneration compared to the rostral LC, and rostral LC is particularly vulnerable to tau-related neurodegeneration.^{14, 67} Regarding tau, at the time of this writing, we only had t-tau assays. Plasma p-tau may be more sensitive and specific to AD-specific processes, and thus more strongly associated with AD-related brain atrophy than t-tau. Also differences across the t-tau and Aβ plasma biomarkers were of small effect sizes. This is common when examining differences in AD-related plasma biomarkers across health conditions that are considered important AD risk factors among individuals without dementia.⁶¹ There was also considerable within-group variability in the biomarkers, but this is common even among comparisons of people with and without AD dementia.⁵⁶

Regarding strengths, we used a novel MRI approach to provide a measure of structural LC integrity, unobservable in typical imaging approaches. We included an assessment of AD-related plasma biomarkers in addition to MRI measures to provide additional indicators of AD-related pathology, and showed an association between higher biomarker levels and less brain integrity in hippocampus and LC. Another strength is that we measured chronic pain longitudinally rather than making a judgment based on intensity at a single timepoint or retrospective reporting by participants.

4.1. Conclusion

Chronic pain is a costly condition in later life associated with increased dementia risk ^{25, 64, 72}. Identifying the biological mechanisms behind these associations and whether these involve AD-related processes is thus a worthwhile goal and could identify potential treatment targets. Our study provides some insight, showing that probable chronic pain is associated with lower hippocampal volume, worse integrity of the rostral-middle LC, and higher plasma levels of t-tau and Aβ. Finally, although the focus of treatment for chronic pain is typically on physical health and physical function, our results suggest that it may be beneficial to pay more attention to the potential association of chronic pain with AD-related brain and biomarker indicators.

PERSPECTIVE:

Probable chronic pain was associated with plasma biomarkers and brain regions that are affected early in Alzheimer’s disease (AD). Reducing pain in midlife and elucidating biological mechanisms may help to reduce the risk of AD in older adults.

AVAILABILITY OF DATA AND MATERIALS:

Data are available for research access (http://www.vetsatwins.org/for-researchers/).