Structural changes of the brain in rheumatoid arthritis

Authors


Abstract

Objective

To investigate whether structural changes are present in the cortical and subcortical gray matter of the brains of patients with rheumatoid arthritis (RA).

Methods

We used two surface-based style morphometry analysis programs and a voxel-based style analysis program to compare high-resolution structural magnetic resonance imaging data obtained for 31 RA patients and 25 age- and sex-matched healthy control subjects.

Results

We observed an increase in gray matter content in the basal ganglia of RA patients, mainly in the nucleus accumbens and caudate nucleus. There were no differences in the cortical gray matter. Moreover, patients had a smaller intracranial volume.

Conclusion

Our results suggest that RA is associated with changes in the subcortical gray matter rather than with cortical gray matter atrophy. Since the basal ganglia play an important role in motor control as well as in pain processing and in modulating behavior in response to aversive stimuli, we suggest that these changes may result from altered motor control or prolonged pain processing. The differences in brain volume may reflect either generalized atrophy or differences in brain development.

Previous morphometric studies in patients with chronic pain suggest that pain may be associated with structural changes in the brain (1–5). Apkarian and colleagues (6) have demonstrated that chronic low back pain is associated with a decrease in gray matter in the dorsolateral prefrontal cortex and the thalamus. They suggest that these changes reflect neurodegeneration rather than neuronal reorganization, since these results are consistent with those of other studies showing a decrease in the concentration of N-acetylaspartate, a marker of neuronal well-being, in this region (7). A reduction of gray matter has been described not only in patients with low back pain (1, 2) but also in patients with other chronic pain conditions, such as migraine (8), chronic tension headache (9), chronic fatigue syndrome (10), posttraumatic stress disorder (11), irritable bowel syndrome (3), fibromyalgia (4, 5), and osteoarthritis (12, 13). In contrast, some other studies have shown an increase in gray matter in patients with these conditions (2, 12, 14).

The observed differences may therefore reflect adaptive or maladaptive neuronal plasticity, rather than atrophy, as some structural changes seem to be reversible when the pain is alleviated (12, 13). It remains to be determined for all of these studies whether the structural brain changes are due to the chronic pain itself, the medication used to treat the pain, lifestyle changes related to immobility, or a combination of these factors.

Relatively little is known about structural brain changes in rheumatoid arthritis (RA). Bekkelund and colleagues (15), using manual tracing methods, found brain atrophy only in RA patients with longstanding disease. In contrast, a pronounced reduction in white and gray matter has been described in patients with systemic lupus erythematosus (16), which, like RA, is a systemic, inflammatory, autoimmune disease affecting the connective tissues.

In the present study, we investigated differences in brain structure between RA patients and healthy age- and sex-matched controls. In particular, we examined these patients for evidence of structural brain changes similar to those described in other chronic pain conditions or in systemic lupus erythematosus. Due to the cross-cross-sectional nature of this study, we could not examine the precise temporal or causal nature of any changes related to plasticity.

PATIENTS AND METHODS

Patients and controls.

Patients with active RA who were under consideration for therapy with anti–tumor necrosis factor were consecutively recruited from the rheumatology outpatient department at the Nuffield Orthopaedic Centre in Oxford. All patients had seropositive, erosive arthritis that fulfilled the American College of Rheumatology criteria for RA (17). Healthy volunteers between the ages of 25 and 80 years who had no evidence of joint disease were recruited through poster advertisements.

Patients were excluded for the following reasons: 1) the presence of any neurologic, psychiatric, or medical condition that could affect the results of the study other than depression in the patient group, since depression and anxiety-related disorders show an overlap with chronic pain syndromes (18); 2) use of any medication that acts on the central nervous system, either as a treatment (other than low-dose antidepressants, such as amitriptyline ≤25 mg/day) or for recreational use; and 3) the presence of any medical or other contraindications for magnetic resonance imaging (MRI).

Of the 37 patients initially recruited to the study, data for 6 of them were excluded because the quality of structural scans was not sufficiently good for tissue segmentation. Consequently, data for 31 patients (22 of whom were female) were used for the analyses. The median age of the study patients was 57 years (interquartile range [IQR] 20; range 38–70 years), and the median disease duration was 15 years (IQR 12; range 1.5–40 years). The mean ± SD Disease Activity Score in 28 joints using the erythrocyte sedimentation rate (DAS28-ESR) was 5.9 ± 0.9, the median ESR was 26.0 mm/hour (IQR 28.0), and the median C-reactive protein value was 20.5 mg/liter (IQR 26.0). The median tender joint count in 28 joints was 10 (IQR 7), and the median swollen joint count in 28 joints was 11 (IQR 8). All patients had at least a moderate level of daily pain intensity, with a median pain score of 6.5 (IQR 2.5) on the Numerical Rating Scale (highest possible score 10). The median score on the Beck Depression Inventory (BDI) was 12.0 (IQR 11).

Analgesics were being taken regularly by 28 of the patients, and 27 patients were receiving various combinations of disease-modifying antirheumatic drugs, including methotrexate (n = 11), leflunomide (n = 10), hydroxychloroquine (n = 9), low-dose corticosteroids (≤7.5 mg/day; n = 8), sulfasalazine (n = 3), cyclosporine (n = 2), and azathioprine (n = 1). Four patients were taking low-dose amitriptyline (25 mg/day).

We recruited 28 healthy age- and sex-matched volunteers to serve as control subjects. Two of these subjects were claustrophobic, and the high-resolution structural scans in one of them were of poor quality; thus, data for 25 healthy controls (17 of whom were female) were used for the analyses. The median age in this group was 59.0 years (IQR 16; range 38–71 years). The groups were well matched with respect to age (2-tailed P = 0.9 by Mann-Whitney U test) and sex (Pearson's χ2 = 0.058; P = 0.5).

The study was approved by the Oxfordshire Research Ethics Committee B and was conducted in accordance with the principles of the Declaration of Helsinki. Written informed consent was obtained from all participants at the beginning of the study.

MRI data acquisition.

All scans were acquired using a single-channel transmit–receive head coil and a 3T Tim Trio Siemens MR scanner, which is located at the Oxford Centre for Clinical Magnetic Resonance Imaging Research in the John Radcliffe Hospital. High-resolution T1-weighted structural scans were acquired using a 3-dimensional inversion-recovery magnetization-prepared rapid gradient-echo sequence with the following parameters: repetition time 2,040 msec, echo time 5.56 msec, inversion time 1,100 msec, flip angle 8°, field of view 192 × 192 × 160 mm, and voxel size 1 × 1 × 1 mm.

Analysis of the structural MRI.

The structural data were analyzed using the tools FMRIB Software Library–voxel-based morphometry (FSL-VBM) (19) and FreeSurfer, which is a surface-based morphometric technique (20–22). The subcortical gray matter was also analyzed using FMRIB's Integrated Registration and Segmentation Tool (FIRST), a surfacewise technique (23).

FreeSurfer analysis.

For analysis with the FreeSurfer software, each subject's structural scans were linearly registered to the Montreal Neurological Institute structural template 305 (MNI305) using FMRIB's Linear Image Registration Tool (FLIRT), an affine registration tool (24), and then bias-corrected, skull-stripped using the Brain Extraction Tool (BET) (25), and segmented into white matter, gray matter, and cerebrospinal fluid (CSF). The subcortical gray matter structures were defined according to an atlas-based segmentation method using both intensity and spatial location models for each structure (22).

The segmented white matter was used to create a tessellated surface representing the gray–white matter boundary. The surface was then expanded to model the pial–gray matter boundary. The distance between the gray–white matter boundary and the pial–gray matter boundary was used to estimate the cortical thickness. The gray–white surface mesh was inflated to minimize the distortions and warped to match curvature features across subjects (20, 21).

A study-specific template was generated from the curvature-based registration of all subjects' images. After registration, subject-specific cortical parcellations were created (26). Subsequently, the thickness, volume, and area were calculated for each labeled region and optimized for each subject's specific curvature (27). Each processing step was visually examined and verified.

A cross-subject generalized linear model (GLM) was fitted at each vertex to test groupwise differences in cortical thickness between patients and healthy controls. Age and sex were controlled for, since it has been demonstrated that both affect brain morphometry (28). Participants were classified into 4 groups: male patients, female patients, male controls, and female controls. Age was used as an additional continuous variable.

The age-related change in each group was separately modeled, allowing patients and controls to have different cortical thicknesses, and the thickness could change at different rates.

Each subject's cortical thickness measurements were smoothed using a full-width half-maximum (FWHM) kernel of 15 mm. The group difference t-stat maps were corrected for multiple comparisons across vertices using statistical procedures that controlled for the false discovery rate (FDR) at P = 0.05 (29).

Global brain measures.

The total volume of segmented white and gray matter was also computed from the segmentation files. The total CSF volume was calculated as the sum of the volumes of all ventricles and the extraventricular CSF. Global brain measures were calculated for each subject, since the brain size affects volume and cortical thickness calculations (30). The intracranial volume (ICV) generated by FreeSurfer was used to correct for differences in brain size (31). To verify the ICV results, the atlas scaling factor was also calculated from the linear transformation matrix produced by FLIRT during the FLS-VBM analysis for scaling a subject's skull-stripped brain to the standard template.

Assessment of subcortical structures.

FreeSurfer allows for a vertexwise analysis only of cortical structures; therefore, in order to compare the subcortical structures, their volumes were estimated from the segmented subcortical structures defined at processing stages of FreeSurfer. To correct for differences in brain size, the volumes were normalized by the individual ICV.

Statistical analysis.

The data from global brain measurements and the volumes of subcortical structures were evaluated for between-group differences using a univariate test, with sex and group as fixed factors and age as a covariate. A full factorial model was used, with group difference as a contrast.

Analysis using FSL-VBM.

Whole-brain differences in the topographic distribution of gray matter between RA patients and controls were analyzed using FSL-VBM software (19, 32, 33). First, structural images were brain-extracted using BET. Next, tissue-type segmentation was carried out using FMRIB's Automated Segmentation Tool 4 (FAST4) (34). The resulting gray matter partial volume images were registered to the MNI152 standard space using FLIRT, followed by nonlinear registration using FMRIB's Non-linear Image Registration Tool (FNIRT) (35). The resulting images were averaged to create a study-specific template, to which the native gray matter images were nonlinearly re-registered. The template was created from gray matter images from a random subgroup of 25 patients and all 25 controls. The registered partial volume images were corrected for local expansion or contraction and then smoothed with an isotropic Gaussian kernel with a sigma value of 3 mm (∼7 mm FWHM).

Assessment of group differences.

Differences in the distribution of gray matter between patients and controls were examined using voxelwise permutation-based nonparametric inference testing within the framework of the GLM. To test for between-group differences controlling for the effect of age and sex, the first two regressors in the model represented the group means, and the remaining regressors represented demeaned age and sex. A voxelwise statistical map was created that identified gray matter differences between the groups. Results were considered significant at P < 0.05 (5,000 permutations, initial cluster-forming thresholding at uncorrected P = 0.05), fully corrected for multiple comparisons across space.

Effect of disease duration.

To test for the effects of disease duration, a separate analysis was performed within the patient group. In the GLM, 4 regressors were used, representing the group mean, demeaned age, demeaned sex, and demeaned disease duration. Results were considered significant at P < 0.05 (5,000 permutations, initial cluster-forming thresholding at uncorrected P = 0.05), fully corrected for multiple comparisons.

Analysis using FIRST.

FIRST is a surface-based segmentation tool used to analyze the shape differences of the subcortical gray matter structures (23). First, the structural images were brain-extracted using BET, then registered to the MNI152 standard space using a global affine registration. This registration was then refined using a subcortical mask to improve the joint alignment of the subcortical structures. Following this, a deformable model was fitted to the images in their native space, based on a Bayesian joint shape and appearance model. The model was derived using a 336-image training set and was transferred from standard space using the previously described 2-stage affine registration. To test for local shape differences between groups, a vertexwise analysis was performed on the vertex locations in standard space to normalize for brain size.

The statistics were calculated using a multivariate GLM with Pillai's trace, and corrected for multiple comparisons using FDR. For this analysis, the GLM contained 3 regressors, representing group membership, age, and sex. In the patient group, the effect of disease duration was also investigated using demeaned age, sex, and disease duration as regressors in the GLM model.

RESULTS

Findings of the FreeSurfer analysis.

Surface-based morphometry.

There were no between-group differences in cortical thickness or age-related thickness changes anywhere in the brain after FDR correction.

Global brain measures.

The patients had a significantly smaller ICV than the controls (Figure 1) (F[1,51] = 12.35, P = 0.001, η2 = 0.195, by univariate analysis, for the between-group difference, controlling for effects of age and sex) and a significantly larger atlas scaling factor than the controls (F[1,51] = 5.25, P = 0.026, η2 = 0.093, by univariate analysis, for the between-group difference, controlling for age and sex). The ICV and atlas scaling factor values were highly correlated, both in patients (r = −0.87, P < 0.0005) and in controls (r = −0.89, P < 0.0005). In the control group, 2 male subjects had large ICVs (outliers in Figure 1), but the difference between patients and controls remained significant even when these subjects were excluded.

Figure 1.

Intracranial volume (ICV) in rheumatoid arthritis patients and healthy control subjects. Data are shown as box plots. Each box represents the 25th and 75th percentiles (interquartile range [IQR]). Lines inside the boxes represent the median. Whiskers represent nonoutlier maxima and minima. Circles indicate outliers >3 IQR above the 75th percentile.

Volumes of subcortical structures.

We used FreeSurfer software to estimate volumes of the following subcortical structures: the thalamus, the caudate nucleus, the putamen, the pallidum, the hippocampus, the amygdala, and the nucleus accumbens. After normalizing for the ICV and controlling for the effects of age and sex, patients had a significantly larger volume of the caudate nucleus bilaterally (for the right, F[1,51] = 8.5, P = 0.005; for the left, F[1,51] = 5.9, P = 0.02) and a larger volume of the left putamen (F[1,51] = 4.6, P = 0.04), with a trend toward a larger volume for the right putamen. Patients also had a larger volume of the right nucleus accumbens (F[1,51] = 5.5, P = 0.02). No other significant group differences were found. The results are summarized in Table 1.

Table 1. Between-group differences in normalized volumes of subcortical structures, by univariate analysis*
StructureF[1,51]P
  • *

    FreeSurfer software was used to estimate volumes of subcortical structures in rheumatoid arthritis patients versus healthy control subjects. The differences were determined using univariate analysis, controlling for age and sex, and normalizing for the intracranial volume. P values are uncorrected.

Caudate  
 Right8.50.005
 Left5.90.02
Putamen  
 Right2.70.10
 Left4.60.04
Accumbens  
 Right5.50.02
 Left1.40.25
Thalamus  
 Right1.80.18
 Left0.30.59
Hippocampus  
 Right0.70.40
 Left0.01.00
Amygdala  
 Right0.01.00
 Left0.10.71
White matter  
 Right0.60.46
 Left0.80.36
Gray matter  
 Right2.60.11
 Left2.60.11
All cerebrospinal fluid1.40.24

When the volumes were not normalized by the ICV, there were no significant differences in the volumes of the caudate, the putamen, or the nucleus accumbens. However, patients had significantly smaller non-normalized volumes of the thalamus (for the right, F[1,51] = 13.1, P = 0.001; for the left, F[1,51] = 9.7, P = 0.003), the hippocampus (for the right, F[1,51] F = 11.6, P = 0.001; for the left, F[1,51] = 7.4, P = 0.009), and the right amygdala (F[1,51] = 6.4, P = 0.02) as well as total gray matter (for the right, F[1,51] = 5.2, P = 0.03; for the left, F[1,51] = 5.1, P = 0.03) and total white matter (for the right, F[1,51] = 5.1, P = 0.03; for the left, F[1,51] = 4.6, P = 0.04) (Table 2).

Table 2. Between-group differences in non-normalized volumes of subcortical structures, by univariate analysis*
StructureF[1,51]P
  • *

    FreeSurfer software was used to estimate volumes of subcortical structures in rheumatoid arthritis patients versus healthy control subjects. The differences were determined using univariate analysis, controlling for age and sex. P values are uncorrected.

Caudate  
 Right0.00.99
 Left0.020.89
Putamen  
 Right2.60.11
 Left0.60.46
Accumbens  
 Right0.020.89
 Left1.40.25
Thalamus  
 Right13.10.001
 Left9.70.003
Hippocampus  
 Right11.60.001
 Left7.40.009
Amygdala  
 Right6.40.02
 Left4.70.34
White matter  
 Right5.10.03
 Left4.60.04
Gray matter  
 Right5.20.03
 Left5.10.03
All cerebrospinal fluid0.10.71

Findings of FSL-VBM assessment.

Group differences.

The results of the whole-brain voxelwise analyses showed a larger nucleus accumbens bilaterally in the patient group (Figure 2A). No region was significantly larger in the control group.

Figure 2.

Whole-brain voxelwise morphometric analyses, examining between-group differences in cortical gray matter. FSL-VBM software was used for these analyses (see Patients and Methods for details). A, Voxelwise differences in cortical gray matter between rheumatoid arthritis (RA) patients and healthy control subjects, controlling for the effects of age and sex. Regions with significantly more gray matter in the patient group were identified in the nucleus accumbens bilaterally. B, Voxelwise differences in cortical gray matter between female RA patients and female healthy control subjects, controlling for the effect of age. Regions with significantly more gray matter in female patients were identified in the nucleus accumbens and the caudate nucleus bilaterally, as well as in the right putamen. Between-group differences are represented as statistical maps color-coded on a red–yellow scale, with brighter (more yellow) regions corresponding to more significant differences. Images are presented according to neurologic convention, with right hemisphere structures shown on the right.

To remove the effect of sex, we compared only female patients and controls. In this analysis, we observed an increase in gray matter in the nucleus accumbens and caudate nucleus bilaterally as well as in the right putamen (Figure 2B).

Effect of disease duration.

To investigate whether the structural changes were secondary to disease, we investigated correlations between the gray matter topographic distribution and disease duration. There was a negative correlation between disease duration and gray matter density in the right thalamus when disease duration was used as a single regressor in the GLM. However, this effect was not present when additional regressors, representing age and sex, were used in the model.

Findings of FIRST assessment.

The results of FIRST, the vertexwise analysis of the subcortical gray matter, were consistent with the FreeSurfer results. FIRST models the shape of the following subcortical structures: the thalamus, the caudate nucleus, the putamen, the pallidum, the hippocampus, the amygdala, and the nucleus accumbens. There were significant differences in the shape of the right caudate nucleus between the groups (Figure 3). No other significant between-group differences were found.

Figure 3.

Whole-brain vertexwise morphometric analyses, examining between-group differences in subcortical gray matter using FIRST (see Patients and Methods for details). The regions in blue correspond to the part of the caudate that is larger in rheumatoid arthritis patients than in healthy control subjects. Changes in both the dorsal (top) and ventral (bottom) sides of the caudate nucleus are shown. The color bar indicates the individual vertexwise F statistic values, corrected for multiple comparisons using statistical procedures that controlled for the false discovery rate. Images are presented according to neurologic convention, with the right caudate nucleus shown on the right.

There was no correlation between the change in shape of any of the structures and the disease duration in the patient group, when age and sex were controlled for. There were significant age-related changes in the shape of the thalamus.

DISCUSSION

Several morphometric studies have demonstrated a pronounced decrease in cortical and subcortical gray matter in patients with chronic pain compared to healthy controls (4–6, 12). However, there have been no fully automated morphometric studies of brain structure in RA patients. Bekkelund and colleagues (15) assessed brain changes in RA patients by use of crude measures of atrophy, such as the area of the corpus callosum, the cerebrum, and the cerebellum, as well as the ventricle-to-brain ratio, the bifrontal ratio, and the bicaudate ratio.

In our study, there were no significant cortical differences between the groups, either in the FreeSurfer or in the FSL-VBM analyses. We did not observe extensive (6) or even localized cortical differences between patients with chronic pain and healthy controls, as reported by other authors (1, 2). This suggests that despite chronic pain and active inflammation, there were no changes in the cortical gray matter in the RA patient group.

The global brain measures of the ICV and atlas scaling factor were significantly different between the groups, suggesting that the brain size was smaller in the RA patients. The differences in brain size described by Bekkelund and colleagues (15) were restricted to patients with disease duration in excess of 15 years, who had significantly larger ventricle-to-brain ratios than those in the control group.

The ICV corresponds to the size of the brain at its maximal size, before age-related or disease-related atrophy (36, 37). The ICV used by FreeSurfer is estimated using the automated method that depends on an affine transformation using an atlas scaling factor (31). This method is not completely unaffected by brain atrophy, and its precision declines with brain loss (31, 38). Therefore, some of the ICV differences observed in our study might be related to accelerated atrophy due to pain or inflammation. A smaller ICV, however, may reflect intrinsic differences between RA patients and healthy controls, which may be related to developmental differences or to a particular genetic profile of patients with RA. A further study of patients with early RA would help to clarify this interesting observation.

With all 3 methods used, we consistently observed changes in the basal ganglia of RA patients. For the FSL-VBM results, group differences were present mainly in the nucleus accumbens, the caudate nucleus, and the putamen. The differences were not always bilateral, but it is known from the literature that subcortical nuclei, such as the caudate and putamen, are not symmetric (28), and the asymmetry increases with age (39).

Subcortical changes were confirmed by FIRST and FreeSurfer analyses. FIRST is specifically concerned with assessing shape differences in the subcortical gray matter and can be used to visualize which part of the structure is affected. In our study, there were significant differences in the right caudate nucleus. There were also differences in other structures, but these did not remain after correcting for multiple comparisons.

For the FreeSurfer analysis, patients had a larger ICV-normalized volume of the caudate nucleus bilaterally, as well as the left putamen and right nucleus accumbens, with a trend toward bilateral differences in the putamen. When the ICV was not controlled for, patients had a smaller volume of the thalamus and the hippocampus, as well as the total volume of gray and white matter. Changes in the thalamus and the hippocampus have been previously observed in chronic pain studies (6, 12, 40). However, as we did not find any cortical gray matter changes with the FSL-VBM analysis and no thalamic changes with the FIRST analysis, we suggest that the differences in the non-normalized data are driven mainly by the between-group differences in brain size, and the normalized results are more reflective of true differences between patients and controls. Normalization by brain size is an important part of morphometric analysis, as it reduces the age- and sex-related differences and increases discrimination between the groups (31, 41).

In our study, the subcortical differences in the normalized data were observed mainly in the basal ganglia. The term basal ganglia refers to subcortical gray matter structures, including the striatum (i.e., the caudate nucleus and putamen), the globus pallidus, the nucleus accumbens, the subthalamic nucleus, and the substantia nigra (42). The basal ganglia are involved in the sensory-discriminative, affective, and cognitive dimensions of pain, as well as in modulation of the nociceptive information and gating of sensory information to higher motor areas (43). It has been suggested that changes in basal ganglia connectivity may play an important role in the maintenance of chronic pain (44). An increase of gray matter in the basal ganglia has been previously reported in fibromyalgia (4), chronic lower back pain (2), in vulvar pain (14), as well as in posttraumatic stress disorder (45). Schmidt-Wilcke and colleagues (4) have suggested that structural changes in the striatum may be a result of increased nociceptive input; however, in their earlier analysis (2), these investigators proposed that structural changes might be a cause of pain rather than an effect, as the changes did not correlate with the disease duration but with the reported pain intensity and unpleasantness.

The nucleus accumbens, together with ventral parts of the caudate nucleus and putamen, form the ventral striatum. This region receives input from the insular cortex, the hippocampus, the prefrontal cortex, and the amygdala and projects to the globus pallidus (46). The ventral striatum is involved in anticipation of aversive stimuli as well as in anticipation of placebo-induced pain relief (44). In RA, pain relief is beyond the patient's control and cannot be predicted. Pain relief may be regarded as a form of reward (44). The uncertainty of reward causes tonic activation of dopaminergic neurons in the ventral striatum (47), and a continuous activation of reward circuits may lead to maladaptive changes in the central nervous system (48).

The basal ganglia are typically involved in the integration of sensory (including nociceptive) information for both automatic and voluntary movement (43, 49). The caudate nucleus and putamen are involved in active planning of nonroutine, self-initiated actions and adjustment after set shifting (49). The caudate nucleus plays an important role in evaluation of the agreement between the action and the outcome, as well as planning and performing tasks required to achieve complicated goals (50). In RA, these mechanisms may be disrupted by joint and muscle stiffness, altered posture, joint deformity, and consequent disability. Therefore, the intended movements cannot be normally executed, and there is consequent incongruence between the expected movement-related sensory feedback and the sensory feedback a patient actually receives (51). Some of the symptoms observed in RA and fibromyalgia, such as stiffness, fatigue, altered posture, decreased mobility, and pain, may be at least partly related to changes in the basal ganglia (2).

The nucleus accumbens may be involved in the anticipation of pain and pain relief, the caudate consistently re-plans movements altered by pain and deformity, and the putamen executes those movements, which because of the pain, are not smooth and stereotypical. Altered function of the basal ganglia in RA and compensatory responses to pain and to altered movement control may cause the observed structural changes.

It has been suggested that the increased gray matter within the striatum may be caused by a lack of dopamine or by a reduction of the extracellular dopamine concentration due to an increase in the number of glial cells (52). Dopamine is involved in pain perception, affective pain processing, and pain modulation in healthy people (53) as well as in RA patients (54, 55). It reduces pain by tonically inhibiting nociception in the mesolimbic and mesocortical circuits (56). This mechanism is believed to be dysfunctional in RA, and interestingly, treatment with dopamine agonists improves pain, fatigue, the number of tender and swollen joints, as well as function in RA patients (54, 55). An increase in gray matter might be related to task-specific neuronal activation (57). The observed changes may also be a result of medication or changes in daily activity, or they might be an epiphenomenon.

Disease duration did not affect the gray matter density in the FSL-VBM analysis or the shape of the subcortical gray matter in the FIRST analysis. Disease duration and gray matter volume have been correlated in one study of systemic lupus erythematosus conducted by Appenzeller and colleagues (16). However, no relationships between disease duration and structural changes have been observed in other studies of chronic pain (1, 2), perhaps reflecting the fact that such patients often experience pain months or even years before a clinical diagnosis is made.

This is the first morphometric study comparing brain structure in RA patients and healthy controls. We used unbiased, fully automated morphometric analyses, without prespecifying regions of interest. We observed consistent results across all 3 independent analysis methods when the data were corrected for the ICV. The study groups were well matched with respect to age and sex. In addition, the values were corrected for age, sex, and the ICV, as these variables have a strong effect on volumetric measurements.

It should be noted that the sample size was relatively large for a clinical pain study (for a summary, see ref.14) but was not large enough for more complex statistical analyses. We made a significant effort to account for the effect of age and sex, but because this was a cross-sectional study, temporal changes could not be investigated. Finally, because of the relatively late age at onset of their RA, the study population was rather old, thereby increasing interindividual variations in the cerebral topography and potentially reducing the sensitivity to detect true between-group differences (58). We suggest that future work should perhaps focus on assessing longitudinal changes, especially in early RA in younger patients.

In summary, this study did not identify localized cortical atrophy in RA, despite the fact that a pronounced reduction in gray matter has been previously reported in chronic pain and in systemic inflammatory conditions. Moreover, patients had a smaller intracranial volume, which may reflect a generalized atrophy or differences in brain development. Finally, there was an increase in the subcortical gray matter in the basal ganglia of the patient group, mainly in the right caudate nucleus and in the nucleus accumbens. This probably reflects an effect of pain processing, dysfunctional dopaminergic transmission, or alterations in mobility. Further studies in a larger group of patients are necessary to fully investigate the temporal nature of these morphologic differences. Studies in patients with early RA might be potentially helpful in distinguishing developmental changes from changes secondary to the disease process itself.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Wartolowska had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Wartolowska, Wordsworth, Tracey.

Acquisition of data. Wartolowska, Wordsworth.

Analysis and interpretation of data. Wartolowska, Hough, Jenkinson, Andersson, Wordsworth, Tracey.

ROLE OF THE STUDY SPONSOR

GlaxoSmithKline had no role in the study design or in the collection, analysis, or interpretation of the data, the writing of the manuscript, or the decision to submit the manuscript for publication. Publication of this article was not contingent upon approval by GlaxoSmithKline.

Ancillary