Differences in Brain Structure and Function in Older Adults with Self-Reported Disabling and Non-Disabling Chronic Low Back Pain

Abstract

Objective

The primary aim of this pilot study was to identify structural and functional brain differences in older adults with self-reported disabling chronic low back pain (CLBP) compared with those who reported non-disabling CLBP.

Design

Cross-sectional.

Participants

Sixteen cognitively intact older adults, eight with disabling CLBP and eight with non-disabling. Exclusions were psychiatric or neurological disorders, substance abuse, opioid use, or diabetes mellitus.

Methods

Participants underwent: structural and functional brain MRI; neuropsychological assessment using the Repeatable Battery for the Assessment of Neuropsychological Status, Trail Making Tests A and B; and physical performance assessment using the Short Physical Performance Battery.

Results

In the disabled group there was significantly lower white matter (WM) integrity (P < 0.05) of the splenium of the corpus callosum. This group also demonstrated activation of the right medial prefrontal cortex at rest whereas the non-disabled demonstrated activation of the left lateral prefrontal cortex. Combined groups analysis revealed a strong positive correlation (r_s = 0.80, P < 0.0002) between WM integrity of the left centrum semiovale with gait-speed. Secondary analysis revealed a strong negative correlation between total months of CLBP and WM integrity of the SCC (r_s = −0.59, P < 0.02).

Conclusions

Brain structure and function is different in older adults with disabling CLBP compared to those with non-disabling CLBP. Deficits in brain morphology combining groups are associated with pain duration and poor physical function. Our findings suggest brain structure and function may play a key role in chronic-pain-related-disability and may be important treatment targets.

Introduction

Essential to successful aging is the maintenance of functional independence. Chronic low back pain (CLBP) poses a threat to successful aging as it is one of the most commonly reported chronic pain disorders in older adults and is associated with self-reported difficulty in performing functional tasks [1–3]. However, many older adults remain high functioning despite CLBP [3] and the underlying mechanism linking CLBP and disability is unknown.

Although low back pain is associated with self-reported disability in community-dwellingolder adults [4–9, 3], a number of studies have failed to demonstrate a relationship between low back pain and observed physical performance [3–4, 9]. Further, treatments that target the lumbar spine [10] including decompressive laminectomy [11] do not appear to improve physical performance, which is troubling given the robust relationship between physical performance and future disability [12–16]. Recent findings from our laboratory exploring the relationships among CLBP, brain structure, physical performance (PP), and neuropsychological performance (NP) in older adults suggest a new paradigm for understanding chronic-pain-related-disability that point toward a need for newly targeted treatments.

In our first study [17] we demonstrated that PP and NP were significantly impaired in older adults with CLBP compared to pain-free individuals and that there was a significant correlation between pain and physical performance in CLBP participants. However, the relationship between pain and physical performance was no longer significant when NP was taken into account. These findings suggest that either NP deficits may be a potential mediator of the relationship between pain and physical performance or that NP and PP share a common pathway within the brain. In our second study [18] comparing older adults with CLBP to pain-free individuals, we found CLBP participants performed significantly worse on digit span forward and had smaller regional brain volumes - left parietal lobe grey matter (GM) and middle cingulate white matter (WM). Taken together these findings suggest that there is a relationship between brain structure and function (as assessed by NP) with CLBP, raising the possibility that brain structure and function may play an important role in chronic-pain-related disability. Although altered brain structure and function has been associated with a number of chronic pain syndromes in younger adults [for a review see:19–20], our laboratory is the only one that has looked at brain structure in older adults. It is important to understand the mechanism by which CLBP causes functional limitation so that appropriately targeted treatments can be developed.

The primary aim of this pilot study is to identify structural and functional brain differences in older adults with self-reported disabling CLBP compared with those who report non-disabling CLBP. The secondary aim is to examine relationships among brain structure and function with neuropsychological function (NP) and physical performance (PP).

Methods

This study was approved by the University of Pittsburgh Internal Review Board and written informed consent was obtained at the time of enrollment.

Screening and Enrollment of Participants (See flow diagram: Fig. 1 Screening and Enrollment Process)

Fig. 1.

Screening and Enrollment Process

Sixteen community-dwelling older adults (age ≥65) completed this study. Participants were recruited from the Pittsburgh Claude D. Pepper Older Americans Independence Center registry. A two-part screening process (telephone interview followed by on-site evaluation) was conducted to select candidates who reported CLBP every day or almost every day for ≥3 months of at least moderate intensity using the pain thermometer (a vertical verbal descriptor scale that has been validated in older adults) [21]. Moderate or greater intensity CLBP was required for study inclusion based on evidence from a large epidemiological study that demonstrated older adults with moderate or greater low back pain are significantly more likely to report functional impairment than older adults with less severe pain [3]. We required this level of pain for both disabled and non-disabled participants to avoid pain severity independently influencing brain structure and function. Participants were categorized into two groups: CLBP self-reported as (1) disabling (pain that necessitated cutting back on daily activities or resulted in being bed bound during some days of ≥6 weeks in the past 6 months) or (2) non-disabling (pain that had limited function for < 6 weeks over the past 6 months). We have shown that older adults can accurately recall the degree to which their low back pain has been disabling during the prior 6 months [22]. We operationally defined disabling versus non-disabling CLBP drawing upon the findings of Reid and colleagues [23]. In their prospective 18 month study it was found that older adults who reported having to cut back on daily activities or spend time in bed because of back pain during at least 4 of the 18 months experienced significantly greater decline in gait-speed than other participants.

In order to more specifically exam the relationship of CLBP with brain structure and function, individuals were excluded who experienced pain outside of the lower back of the same or greater frequency or of the same or greater intensity. Individuals were also excluded if they had disorders with known effects on brain structure or function (diabetes mellitus [24], depression [25], anxiety [26–27], post-traumatic stress disorder [28], multiple sclerosis, cerebral tumor, or a cerebrovascular accident), or were taking or had taken in the past 3 months prescription as well as over-the-counter pychotropics (such as St. John’s Wort), including those used for pain management. If an individual scored <24 on the Folstein Mini Mental State Examination (MMSE) [29–30] at the time of the onsite evaluation or if they had a disorder that could impact neuropsychological testing they were excluded (serious hearing or visual impairment, substance abuse, opioid use, history or evidence of traumatic brain injury with loss of consciousness, seizure disorder, or a diagnosis of dementia). Those who were mobility impaired were also excluded so as not to confound PP assessment (Parkinson’s disease or unable to ambulate without a walker or care). Finally, participants were excluded if they were experiencing an acute illness or had conditions that would render MRI performance unsafe (claustrophobia, metal objects in body including a pacemaker, or weight >250lbs). Though hypertension (HTN) is known to affect brain structure [31] this was not an exclusion for the current study. Because of its prevalence in older adults, having HTN as an exclusion would significantly impact our ability to generalize study findings.

Common reasons for exclusion were ineligible pain status (i.e. CLBP < everyday or almost everyday, CLBP < moderate intensity, or pain outside of the back > CLBP) and diabetes mellitus. Of the 168 older adults screened by phone, 36 were deemed eligible for onsite evaluation and 17 enrolled. One enrolled participant who completed the assessments was excluded and the data not analyzed due to chronic headaches that came under neurosurgical evaluation during the study.

Assessments

All participants underwent brain structural and functional MRI within two weeks of the NP and PP assessments. Before undergoing brain MRI, NP and PP testing, a history and physical exam was conducted in order to assure the participant was medically stable and met inclusion criteria. In addition the following data were collected: (1) demographic characteristics, including age, ethnicity, sex, economic, and educational status; pain intensity, using the McGill Pain Questionnaire Short Form [32]; co-morbidity, using the Cumulative Illness Rating Scale (CIRS) [33–34]; and, documentation of all regularly scheduled and as needed prescription medications as well as over the counter (OTC) medications including vitamins, supplements, herbals, and herbal teas.

Brain Imaging and Acquisition: Structural and Function

All images were acquired at the University of Pittsburgh’s Magnetic Resonance Research Center using a Siemens 3T Total Imaging Matrix (TIM) TRIO scanner. Each subject received 4 types of brain MRI scans: Localizer, Diffusion Tensor Imaging (DTI), high resolution anatomical imaging using Magnetization-Prepared Rapid Acquisition Gradient Echo (MPRAGE) acquisition, and resting state fMRI. The 3-plane T1 localizer was used for slice prescription and to ensure that the head was not tilted. DTI was collected using a single shot, spin echo, echo planar sequence with 12 directions with b values of 0 and 1000 (matrix = 128 × 128, TR = 5300ms, TE = 88ms, FOV = 256mm, axial slices = 28, thickness = 3mm, skip = 0 with four repetitions). For drawing grey and white matter Volumes of Interest (VOI), a high resolution 3D T1-weighted (MPRAGE) anatomical scan was collected (matrix = 256 × 256, TR = 2300ms, TE = 3.43ms, FOV = 25, axial slices = 160, thickness = 1.0 mm, and flip = 9). For the resting-state fMRI data, subjects were instructed to keep their eyes closed and rest during 5-minute acquisition block while echo planar imaging (EPI) was collected (matrix = 128 × 128, TR = 2000ms, TE = 34ms, 3 mm oblique axial slices, 28 slices, 2 mm × 2mm in-plane voxel dimensions).

Structural Neuroimaging Processing

Structural analysis included: (1) volumetric analysis of corpus callosum WM (genu, middle, and splenium) and superior parietal GM (these regions were selected based on our previous findings) [18]; and (2) Diffusion Tensor Imaging to assess WM microstructure of the corpus callosum (genu and splenium). The genu of the CC was included as previous studies have implicated prefrontal structural differences in younger patients with CLBP [35]. The splenium was chosen as we found a trend (p = 0.07) in poorer WM integrity of the SCC in CLBP subjects compared to pain free individuals (unpublished data) using a subset of participants from our previous study [18]. Using DTI we also examined (in both hemispheres) the posterior limb of the internal capsule and parietal peri-callosal WM near the sensory cortex (as these regions are associated with motor-sensory processing), and the centrum semiovale (as this is a major anterior to posterior WM tract). All structural MRI processing and analyses were conducted by an assessor masked to participant characteristics including group assignment (disabled vs. non-disabled). In order to ensure reliability of DTI analysis and volume measurements, a random 25% of the sample was measured by a second masked rater and a minimum inter-rater correlation coefficient of 0.85 was considered acceptable.

Structural Neuroimaging: Volumetric Analysis

The cerebellum and pons were masked and excluded so that all measures were based on supratentorial volume. After each brain was stripped and segmented into grey matter, white matter, and CSF using Statistical Parametric Mapping 2 (SPM2, http://www.fil.ion.ucl.ac.uk), the Volumes of Interest (VOI) were defined - WM volumes of the corpus callosum (genu, middle, and splenium) and superior parietal GM volumes. VOI analysis was conducted using a variation of manual tracing [36] using in-house software [37] implemented in Matlab (Mathworks, Sherborn, Mass., USA).

The CC WM VOI was defined by the methods described in our previous work [18, 37]. The CC VOI was defined in the sagittal plane. The mid-point was identified and then 20 slices on either side were included. We included only the middle 10mm as the edges can be difficult to determine. Superior parietal GM VOI were defined by first finding the most posterior portion of splenium of the corpus callosum (CC) at the mid-sagittal point, this being the most anterior boundary of the superior parietal lobe of each hemisphere as seen in the sagittal plane. The 20 slices moving posterior of this anterior boundary of the superior parietal lobe were selected for analysis.

To control for head size, total supratentorial brain volume (SBV) in cm³ was estimated using the methods of Benedetti et al. [39], but only including grey and white matter volumes as CSF segmented volumes were of poor quality. Each between groups volume comparison was based on the mean percentage of the VOI relative to the total SBV. For each VOI, group mean values were calculated and both Student t-test and Statterwaite unequal variance t-test was used for between group comparisons.

Structural Neuroimaging: Diffusion Tensor Imaging (DTI)

In a completely random state bulk water uniformly diffuses (isotropic). In the brain, WM restricts water molecule diffusion (anisotropic). DTI capitalizes on the effects of WM on water diffusion. The diffusion tensor estimates the orientation of the dominant direction of diffusion within a tissue voxel [41–42]. Fractional anisotropy (FA) is a scalar quantity of diffusion calculated for each voxel. The FA value describes the degree of anisotropy, from 0 (isotropic) to 1 (anisotropic) [40–42]. Lower FA values indicate less tissue structure and potential damage [43–46].

All DTI processing was completed using the diffusion tools in the FSL software package (FMRIB’s software library, http://www.fmrib.ox.ac.uk/fsl/). The data were corrected for head motion and eddy current distortion, and then FA maps were calculated. Volumes of interest were drawn on the b=0 DTI images to avoid using the dependent measure to define the dependent measure [47]. The DTI VOIs were defined using modified procedures of Sullivan [48] and Salat [49]. Drawn in the axial plane on two slices using MRICro (www.sph.sc.edu/comd/rorden/mricro.html), the DTI VOI’s were transferred to the FA map and values extracted. The following VOIs were examined: genu and splenium of the corpus callosum, right and left centrum semiovale, left and right posterior limb of the internal capsule, and left and right parietal peri-callosal WM near the sensory cortex. The splenium VOI’s were drawn on the two most superior slices to capture white matter connections of the parietal cortex, not the occipital cortex. For each VOI, group mean FA values were calculated and both Student t-test and Statterwaite unequal variance t-test used for between group comparisons.

Functional Neuroimaging: Resting State Functional MRI and the Default Mode Network

There is a growing body of evidence using fMRI that there is a spatiotemporally correlated resting-state functional network of the healthy brain - a default-mode network (DMN) – that is active during the resting state and attenuated during task performance [50–52, for a review see 53]. Functional connectivity is defined as the temporal correlation of activity between spatially disconnected areas. Functional connectivity is often investigated by correlating the time-course of a seed region with voxel-time-courses across the brain. The posterior cingulate cortex (PCC) has been shown to have consistently greater activity during resting state than during cognitive tasks and is the most often reported region of the DMN identified in healthy controls [52, 54]. Thus, the PCC is often used as a seed region to identify the DMN, as was done for this study.

All functional MRI processing and analyses were conducted by an assessor masked to participant characteristics using SPM2 (http://www.fil.ion.ucl.ac.uk) with standard parameters for motion correction, smoothing (8 mm FWHM Gaussian kernel), and normalization (co-registration of images to T1 ICBM template). A band-pass temporal filter with the cutoff frequencies of (.01 .1) Hertz was used to restrict signal to the DMN frequency band, and remove signal drift and high-frequency noise [55]. The left and right PCC from the Anatomical Automatic Labeling AAL atlas [56] was used as a seed region to generate a resting-state functional connectivity map for each individual, i.e., a map of correlation coefficients (r values) representing the strength of the correlation of that voxel with the PCC. Results were thresholded at 20 voxels (voxel-wise p < 0.0005). These individual resting-state connectivity maps were then compared between groups (disabled and non-disabled) in SPM using a 2-sample T-test. Montreal Neurologic Institute (MNI) coordinates were converted to Talairach space using Gingerale free-ware (http://brainmap.org, [57]) and Talairach Client freeware (www.talairach.org, [58]) was used to specify the grey matter center within and the Brodmann’s Areas (BA) associated with the Talairach space.

Neuropsychological Testing

In order not to fatigue participants the NP battery was limited to 45–60 minutes. In addition to the screening Folstein Mini-Mental State Exam (MMSE) [29,30] NP measures of enrolled participants included the National Adult Reading Test (NART) [59], the Repeatable Battery for the Assessment of Neuropsychological Status (RBANs) [60], Trail making test A and B, and Letter-Number Sequencing (LNS) from the Wechsler Adult Intelligence Scale-III. RBANS assesses a wide range of cognitive function (immediate and delayed memory, visuospatial ability, language, and attention) with norms for older adults [60]. Trail Making Test A is a measure of motor speed [61] and Trail Making Test B a measure of executive function [61] Letter-Number Sequencing also assesses executive function in addition to working memory [62].

Physical Performance

The Short Physical Performance Battery is reliable and well validated in older adults demonstrating strong predictive validity for physical disability and mortality [14, 63]. It consists of tests of balance (standing side-by-side, tandem stand, semi-tandem stand), gait-speed, and a timed standing chair rise. A summary score that is on a continuous scale is calculated for the SPPB. Gait-speed, a component of SPPB, is a reliable and valid measure in and of itself for predicting physical decline [13, 64] and has also been validated as a strong predictor of cognitive disability in older adults [65–66].

Statistical Analysis

Fisher’s Exact test was used for between groups comparisons of sex, ethnicity, economic and educational status. Student’s t-test was used for between groups comparison of age. Depending on whether the data were categorical or continuous and the distributional properties of the continuous data, independent samples t-test or Wilcoxon rank sum tests were used for between groups comparisons of other measures. To strengthen power, groups were combined for association analysis performed with Spearman’s correlation coefficient. We examined the association of the FA values (nine measures total) with the neuropsychological tests (the RBANS final score, Trails A time, and Trails B time), the Short Physical Performance Battery (single score), and gait-speed (single value). We also examined the association of the regional volumes (five measures total) and resting state subtraction map intensities (sixteen values) with the same NP and PP measures.

Results

Patient Characteristics

Participant characteristics and comparisons are displayed in Table 1. There is no statistical difference in age between the disabled (n=8, mean age 74.1) and non-disabled group (n=8, mean age 75.1) nor a statistical difference in ethnicity, sex, income, or education. In addition there is no statistical difference between groups in prevalence of HTN, total months of CLBP or months of everyday CLBP, co-morbidities (CIRS), MMSE, NART, RBANS, Trails (A and B), SPPB, or gait-speed. LNS was not analyzed due to missing data. Overall, participants were healthy and took few medications with the exception of multivitamins and over the counter (OTC) supplements. Given the small sample size, heterogeneity of the OTC’s, and the fact that prescription pharmaceutical use was rare, medication data were not further analyzed.

Table 1.

Participant characteristics.

Variable	Disabling (n=8)	Non-Disabling (n=8)	p value
Age (mean, sd)	74.1 (6.4)	75.1 (7.3)	0.77†
Ethnicity (white/black/Hispanic)	7/1/0	5/3/0	0.60¶
Sex (F/M)	4/4	2/6	0.61¶
Income (<$50K/>$50K)	5/3	5/3	1.00¶
Education			0.46¶
<12 years	n=0	n=2
12 years	n=1	n=0
>12 years	n=7	n=6
Hypertension	n=2	n=1	1.00¶
Total months CLBP (median, IQR)	228 (7–804)	240 (24–480)	0.91§
Total months everyday CLBP (median, IQR)	84 (7–420)	102 (4–180)	0.95§
CIRS (mean, sd)	22 (2.1)	22 (4.1)	0.60†
MMSE (mean, sd)	29 (0.92)	28 (1.5)	0.13†
NAART (mean, sd)	112 (5.64)	107 (10.9)	0.30†
RBANS % (mean, sd)	59.5 (30.4)	57.3 (32.3)	0.88†
Trails A (mean, sd)	34.7(10.1)	42.3(8.17)	0.12†
Trails B (mean, sd)	81.7(27.8)	90.7(25.0)	0.50†
SPPB (median, IQR)	10 (8–12)	10 (7–12)	1.00§
Gait speed m/s (median, IQR)	1.0 (0.88–1.3)	1.8 (0.57–1.3)	1.00§

Statistics:

^†Student t-test (equal variances),

^¶Fisher’s exact test, and

^§Wilcoxon rank.

Key: CIRS (Cumulative Illness Rating Scale); MMSE(Mini-Mental State Exam); NAART (North American Adult Reading Test); RBANS (Repeatable Battery for the Assessment of Neuropsychological Status); SPPB (Short Physical Performance Battery).

Primary aim

Though there are no differences in regional brain volumes between groups, there is significantly lower WM integrity (P < 0.05) of the splenium of the corpus callosum (SCC) in the disabled CLBP group (n=8) compared to non-disabled (n=8). (Fig. 2 and Table 2) The within-group resting-state fMRI showed the expected midline DMN pattern in both groups. The between group contrasts of resting-state fMRI demonstrated some significant differences (Fig. 3). The right medial prefrontal cortex showed higher functional connectivity in the disabled relative to the nondisabled group (degrees of freedom = 7, p < 0.01: MNI −6, 58, 22; Tal −6.69, 50.38, 28.76; Brodmann area 9, Fig 3); and the left lateral prefrontal cortex showed higher functional connectivity in the nondisabled relative to the disabled group (Fig. 3) (degrees of freedom = 7, p < 0.01: MNI 50, 0, 22; Tal 44.99, −3.92, 24.49; Brodmann area 6).

Fig 2.

The disabled CLBP group demonstrates significantly lower white matter integrity of the splenium of the corpus callosum (p < 0.02) shown in black. (This is a single subject image in native space with eddy current and geometric distortions corrected.) Combined groups analysis demonstrates a strong negative correlation (r_s = −0.59, P < 0.02) between the white matter integrity of the splenium of the corpus callosum and total months of CLBP.

Table 2.

Fractional Anisotropy (FA) between groups comparisons.

Variable	Disabled	Non-disabled	T value	Pr > |t|
	Mean (sd)	mean (sd)
GCCFA	0.742(0.073)	0.763(0.032)	−0.88	0.4077
MCCFA	0.659(0.037)	0.654(0.037	−1.01	0.3449
SCCFA	0.729(0.044)	0.774(0.038)	−3.08	0.0179
LPLICF	0.619(0.088)	0.625(0.036)	1.46	0.1874
RPLICFA	0.628(0.027)	0.634(0.041)	−0.05	0.9631
LCSVFA	0.338(0.023)	0.348(0.024)	−1.11	0.3029
RCSVFA	0.343(0.024)	0.333(0.025)	0.12	0.9112
LPPCFA	0.328(0.057)	0.345(0.057)	−1.37	0.2120
RPPCF	0.369(0.053)	0.354(0.041)	−1.44	0.1937

XCCFA: GCCFA (genu), MCCFA (middle), and SCCFA (splenium) of the corpus callosum.

LPLICFA (left posterior limb internal capsule); RPLICFA (right posterior limb internal capsule)

LCSVFA (left centrum semiovale); RCSVFA (right centrum semiovale)

LPPCFA (left pericallosal parietal): RPPCFA (right pericallosal parietal)

Fig. 3.

The resting state fMRI subtraction map for the disabled CLBP group–nondisabled (left sagittal, coronal, and axial image set) demonstrates activation of the right medial prefronal cortex (p < 0.01: MNI −6, 58, 22: Tal −6.69, 50.38, 28.76: Brodmann area 9) whereas the nondisabled-disabled subtraction map (right sagittal, coronal, and axial image set) demonstrates activation of left lateral prefrontal cortex at rest (p < 0.01: MNI 50, 0, 22: Tal 44.99, −3.92, 24.49: Brodmann area 6).

Secondary aims

Combined groups analysis reveals a strong positive correlation (r_s 0.80, P < 0.0002) between higher WM integrity of the left centrum semiovale (LCS) and faster gait (Fig. 4) and a positive correlation between Trail Making Test A and left lateral prefrontal cortical activation at rest (r_s = −0.5916, P < 0.02).

Fig. 4.

Gait-speed and white matter integrity of the left centrum semiovale (LCS) demonstrates a strong positive correlation (r_s = 0.80, P < 0.0002). The LCS is shown to the left as a single subject image in native space with eddy current and geometric distortions corrected. Gait-speed is a sensitive marker and predictor of physical and cognitive disability as well as mortality in older adults (see text).

Post-hoc secondary analysis demonstrates a strong negative correlation (r_s = −0.59, P < 0.02) between total months of CLBP and WM integrity of the splenium of the corpus callosum (Fig. 2) and a strong positive correlation between total months of CLBP as well as total months of everyday CLBP with the right centrum semiovale (r_s = 0.51, P < 0.04 and r_s = 0.64, P < 0.007 respectively).

Discussion

This is the first study in older adults with CLBP to examine differences in brain structure and function in those with self-reported disabling CLBP compared to those with non-disabling CLBP. We found deficits in WM integrity in posterior regions of the brain in those with disabling CLBP and between groups differences in the resting state DMN. Combined groups analyses demonstrated an association of WM structural deficits with poor physical function and pain duration.

Studies of younger adults found an association between GM volume deficits and pain duration [35, 67–68] supporting the hypothesis that brain-structural changes may be secondary to persistent pain. This study found poor WM integrity of the SCC in disabled CLBP participants and a strong negative correlation between total months of CLBP and SCC WM integrity. We also found a positive association between pain duration (total months of CLBP and total months of everyday CLBP) with WM integrity of the right centrum semiovale (r_s = 0.51, P < 0.04 and r_s = 0.64, P < 0.007 respectively).

Only one investigation has followed the longitudinal impact of chronic pain on brain structure using experimental pain [69]. The investigators found significantly increased GM in the midcingulate, somatosensory cortex, and parietal lobe in young males (mean age 26) who received 20 minutes of pain on 8 consecutive days. As evidenced by fMRI [70], these same subjects habituated to the pain with decreased activity in pain transmitting structures along with increased anterior cingulate cortex activity. If the response to repeated pain includes both decreased and increased activity, it is not surprising we found both decreased (SCC) and increased (right centrum semiovale) WM in association with pain duration. Whether disuse or overuse of neural resources for pain processing (or related emotional processing) over time leads to structural changes and subsequent disability cannot determined.

Our first study found parietal GM volume deficits comparing CLBP to pain-free participants [18]. In this study we found WM connectivity differences of the SCC (which projects to the parietal lobes) comparing disabled CLBP to non-disabled. The parietal lobe is important in pain processing [71–77] and parietal structural deficits have been found in other chronic pain syndromes [78–81]. The potential relationship between GM volume and WM connectivity was investigated in one chronic pain study. Using probabilistic tractography the investigators assessed WM connectivity to atrophic GM regions in complex regional pain syndrome (CRPS) participants compared to healthy controls [82]. They found WM tracts ipsilateral to the atrophic GM regions had reduced connections in long distance tracts and total number of connections.. The authors suggested that GM atrophy, perhaps due to the impact of chronic pain, could lead to WM reorganization. In addition, pain intensity and duration were associated with GM volume deficits.

The first study to associate CLBP with structural deficits found reduced GM in the DLPFC (and thalamus) in young adults with CLBP (mean age = 42.9) compared to pain-free individuals [35]. However, hypertension was not excluded and is associated with GM reductions, notably in the DLPFC [83]. Given the sensitivity of frontal regions to vascular disease [83] and the prevalence of HTN in older adults, it is an important confounder to consider in exploring relationships among chronic pain, disability, and brain structure.

Although our participants were not clinically depressed, our SCC findings should be considered in light of data that demonstrate an association between SCC atrophy and late-onset depression. Comparing healthy controls to older adults with early and late onset depression Ballmaier et al. [84] found depressed patients exhibited significant callosal thinning of the genu and splenium of the CC. Thinning of the genu was restricted to early onset depression, but both genu and splenium thinning were associated with late-onset depression. Given our findings of poorer WM integrity of the SCC in those with disabling CLBP, important questions arise: Is disabling chronic pain a somatic manifestation of late onset-depression or, if chronic pain is changing brain structure and function, is this predisposing older adults to late-onset depression and disability? These questions are important because there is a significant overlap in symptoms reported by older adults with chronic pain and depression [85] and because depressive symptoms are highly associated with CLBP that limits physical activity [9, 86–88]. Although we attempted to exclude current central nervous system disorders that could independently impact brain structure or function, we did not ask patients about past disorders such as depression or other types of psychological trauma. Clearly a longitudinal study is needed to more precisely determine the relationships among pain, changes in brain function and morphology, and disability.

This study found right MPFC activation (BA9) in disabled CLBP participants at rest and left LFC (BA6) activation in the non-disabled. Right frontal lateralization of pain processing has been proposed by a number of investigators [89–91] and observed in neuroimaging studies of acute and chronic pain [77, 92–94] as well as studies of rumination and retrieval of negative memories [95–96]. The prefrontal cortex is important in top-down inhibition of stimulus-provoked emotion [97–98], stimulus-independent emotional processing [99] and expectation [100]. It has been found in primates that parieto-frontal connections with BA6 [101–103] play an important role in voluntary movements. BA6 includes the supplementary motor area (SMA proper and pre-SMA [104]) and premotor area. The SMA influences planning, initiation, and coordination of complex movements [105–106]. A recent study of professional athletes found autobiographical negative emotions activated the MPFC while concomitantly suppressing premotor (BA6) and motor regions [107]. The authors suggested that the reciprocal relationship between the MPFC and premotor/motor regions is distinctive to self-referenced negative emotions in the setting of motor-planning. Studies have found that persons with chronic pain display recall bias for unpleasant and pain-related autobiographical memories [108–109]. It is possible the disabled group in this study devotes more of their neural resources to suppression of the emotional experience, memory, or expectation of CLBP, thus affecting motor-planning and response leading to disability, while the non-disabled group does not devote neural resources to negative emotion suppression and is better able to motor-plan. Other studies have found that suppression of negative emotions is particularly demanding and interferes with task performance [99, 110].

Although there were no differences in PP between disabled and non-disabled participants, combined groups analysis revealed a strong positive correlation (r_s = 0.80, P < 0.0002) between WM integrity of the left centrum semiovale (LCS) and gait-speed (See Fig 4). A number of studies have found gait-speed deficits associated with brain structural abnormalities, principally increased WM hyperintensities [111–116] and a strong association between gait-speed [7, 117] and cognitive function deficits [16]. However none of these studies controlled for pain status. It is important to understand the contribution of chronic pain to mobility and cognitive impairment because in older adults, impaired mobility and cognitive function predict future disability [12, 16, 118–119], hospitalization [120], nursing home admission [63, 121], and mortality [63]. As our cohort was relatively healthy, our findings and the literature suggest chronic pain may contribute to brain structural changes leading to mobility disability, although the mechanism is unknown.

Limitations

The main limitation of our study is the small sample size. Although we did not have a pain-free control group, the finding of poorer WM integrity of the SCC in the disabled group is an interesting complement to our previous study, which found decreased volumes of the parietal lobe in CLBP participants compared to pain-free. However, the cross-sectional nature of this investigation prevents us from determining whether the brain differences in disabled compared to non-disabled participants are the result of chronic pain, a predisposition to the development of chronic pain or disability, or some other unidentified factor. Due to the number of assessments, the possibility of spurious results cannot be ruled out. Corrected P values in this exploratory study were not utilized as it was deemed the cost of missing a finding outweighed the value of using corrected P values.

Conclusion: Targeting the Brain in the Treatment of Pain

A fundamental principle of gerontology is that presentation of a new disease depends on the organ system made most vulnerable by previous changes [122]. Our research suggests the brain may be the vulnerable system leading to disability in older adults with CLBP. In support of this hypothesis are a number of studies that have demonstrated success using cognitive behavioral therapies to treat chronic pain in older adults [123–129]. Meditation not only modulates pain processing [129] but in older adults with CLBP decreases self-reported functional limitation [130]. In the study of professional athletes described above, cognitive behavioral intervention focused on performance success resulted in less activation of the MPFC, and restored activation of premotor and motor regions [98]. Taken together, these findings suggest that further investigation of brain structure and function in those with chronic disabling pain and the degree to which the changes observed are modifiable is imperative if we are to effectively treat chronic-pain-related disability in older adults.