2.1. Demographic and Clinical Findings
There were no significant differences in age between healthy and LBP groups (mean ± SD, 44.6 ± 14.7 years vs. 46.1 ± 11.3 years, p = 0.8). In the LBP group, the mean duration of pain was 8.8 ± 7.2 years, with an average pain intensity of 4.5 ± 1.9 (highest pain intensity of 7.8 ± 1.5, and lowest pain intensity of 2.3 ± 2.3 on Visual Analog Scale, VAS). Nine participants reported constant pain, eight reported intermittent pain, and for two we had no data, but they were referred from a pain management clinic. Pain referral patterns included nine subjects with radiating pain in buttock and/or leg (four left leg) and eight with localized LBP.
Eight participants took regular pain medications (opiate analgesics, e.g., hydrocodone, oxycodone; anticonvulsant, e.g., Neurontin) and four others took nonsteroidal anti-inflammatory drugs on an as needed basis. Fifteen participants were carrying fulltime regular work responsibilities; three were on disability due to back pain and one was retired. The clinical scores of pain, disability, and depression are shown in . Overall, our subjects suffered long duration of pain symptoms with moderate intensity and perceived fear of movement. Despite the general severity of these symptoms, depression and disability were only mild. Pain features (Short Form McGill Pain Questionnare, SF-MPQ and VASave) were positively correlated with perceived fear of movement (r = 0.46, p = 0.05 and r = 0.49, p = 0.04, respectively), disability (r = 0.84, p < 0.001; r = 0.54, p = 0.02), and depression (r = 0.68, p = 0.002; r = 0.51, p = 0.03) in this cohort of LBP.
Clinical scores (mean ± SD) in low back pain (LBP) group.
2.2. 1H-MRS Spectra Quality
There were no group differences in signal-to-noise ratio (SNR, right M1, p = 0.2; left M1, p = 0.2) or total brain tissue volume (right M1, p = 0.3; left M1, p = 0.9) in spectroscopic voxels between groups. Due to poor SNR, we excluded data from one subject’s right hemisphere and two other subjects’ left hemisphere from the LBP group.
2.3. Neurochemical Concentrations
Normal distributions for individual neurochemicals were verified with frequency and Q-Q plots (Kolmogorov-Smirnov test). Subjects with chronic LBP displayed lower NAA concentrations in right M1 (9.0 ± 0.9 vs. 10.2 ± 1.2 mM, p = 0.008) compared to controls. Although lower NAA was found in left M1 it was not significantly different from controls (9.7 ± 0.9 vs. 10.3 ± 1.4 mM, p = 0.2; and A). There were no significant changes in mI in either right or left M1 (4.7 ± 1.0 vs. 5.0 ± 1.4 mM, p = 0.6; 5.0 ± 0.9 vs. 4.9 ± 1.1; p = 0.6 respectively; B). Subgroup analysis of medicated vs. un-medicated LBP subjects showed no statistical differences in NAA or mI concentrations.
Figure 1 (A) Magnetic resonance spectroscopy acquisition: white squares represent the spectroscopic voxels selected in the trunk representation in each primary motor cortex; (B) LCModel output showing N-acetylaspartate (NAA) and myo-inositol (mI) peaks from right (more ...)
Mean (+SD) concentrations of N-acetylaspartate (A) and myo-inositol (B) in control (grey bars) and low back pain (LBP, black bars) groups in right and left M1. Significantly lower NAA has been observed in right M1 in LBP; * p < 0.05.
2.4. Neurochemical Correlations
The strength of correlations was normally distributed in both groups (Kolmogorov-Smirnov test, healthy controls, p = 0.2; chronic LBP, p = 0.2). In the control group, NAA and mI were strongly correlated within each M1 (). All inter-M1 correlations were also strong and statistically significant in controls. In contrast, in the LBP group, most correlations were lower and non-significant, although the correlation for mI between hemispheres reached statistical significance (). Between-group comparison showed lower intra-M1 and inter-M1 mean correlation coefficient in LBP compared to controls (t = 2.31, p = 0.008).
Correlation between N-acetylaspartate (NAA) and myo-inositol (mI) within (intra-M1) and between (inter-M1) in control and low back pain (LBP) groups.
2.5. Correlations between Neurochemical Concentrations and Clinical Scores
Although no correlations between NAA or mI concentrations and clinical measures were found to be significant, we did detect some moderately strong trends between left M1 mI and pain duration, pain intensity, and sensory aspects of the SF-MPQ (r = 0.52, p = 0.05; r = 0.52, p = 0.06; and r = 0.52, p = 0.06, respectively). Left NAA and depression scores were also correlated at moderate strength (r = −0.46, p = 0.08).
Our results demonstrated lower NAA in the trunk representation areas in right M1, and significantly lower correlations between NAA and mI across M1s in participants with chronic LBP compared to controls. There was some evidence that neurochemicals in the left M1 may be correlated with clinical characteristics of pain and depression.
The magnitudes of NAA alterations described here (12% lower in right M1 and 6% lower in left M1) is similar to spectroscopic findings in other brain regions involved in pain processing, e.g., dorsolateral prefrontal cortex (6.5% lower) [10
], anterior insula (4.4% lower) [28
], and SSC (right, 6%; left, 9% lower) [6
]. The significance of altered NAA in M1 is unclear. Although neurodegeneration is associated with lower NAA in some conditions [29
], it is unlikely to explain the current findings since we found no significant differences in brain tissue volume (in spectroscopic voxels) between our groups. Alternatively, several studies have highlighted the dynamic nature of NAA. For example, NAA initially falls immediately after traumatic brain injury but then recovers as cognitive function returns [30
]. NAA is also correlated with brain glucose consumption [31
], suggesting a metabolic role of NAA. Therefore, lower NAA in M1 might suggest altered neuronal mitochondrial metabolism [29
], which could result from altered peripheral input from either lower back [33
] or pain pathway [4
]. Indeed, alterations in motor behavior such as delayed postural control [33
], delayed deep back muscle activation [34
], and altered gait [35
] are reported in these patients. These changes persist beyond resolution of pain symptoms [33
] and may be the result of altered motor output from M1, as suggested from evidence that TMS of M1 reduced pain intensity [19
]. Conversely, changes in the corticospinal M1 excitability are related to pain and disability in chronic LBP [25
]. Taken together, pain studies clearly suggest reorganization not only in the somatosensory cortex but also in the motor cortices [4
]. The interdependence of these two systems is also supported by our findings, e.g., lower NAA in both SSC [6
] and M1 in chronic LBP. Further, lower NAA described here may underlie the functional M1 changes. Future studies could evaluate this relationship.
We did not find significant correlations between NAA levels in M1 and clinical characteristics of pain. It is possible that a more direct measure of the trunk muscle function, such as changes in muscle strength [36
], muscle activation [4
], or muscle volume [38
], could provide a better relationship with neuronal integrity in M1 and should be also considered in future studies. The possible correlation between left NAA and depression scores reported here should be considered with caution and repeated with a large sample size to confirm the reliability of this trend.
Although mI can be an indicator of glial involvement [7
] in chronic pain, previous studies reported conflicting results. Higher but not statistically significant mI in orbitofrontal cortex [10
] and thalamus [11
] has been reported in chronic pain. We found that mI concentrations were not significantly different in M1. mI is a glial cell marker and an osmolyte [39
]. Since glutamate and glutamine, other major brain osmolytes [40
] were not significantly increased (data not presented here), the changes in mI are unlikely to be driven by changes in osmolarity. mI levels in left M1 were moderately correlated with pain characteristics, although these correlations did not reach statistical significance. Since the astrocytes release trophic factors promoting neuronal survival, synaptogenesis, and neurogenesis after nervous system damage [11
] and participate in long-term synaptic plasticity [42
], we can speculate that mI provides information about a potential role of glial cells in left M1 reorganization in LBP.
Previous studies have shown strong correlations between neurochemical concentrations in functionally-related regions in healthy brain [44
]. Following injury, this correlational structure was disrupted. For example, Cirstea et al
. have shown that NAA and mI are highly correlated in the motor cortical network under normal conditions and after subcortical stroke this correlation was diminished [46
]. Similarly, correlations are disturbed in chronic LBP [6
]. For instance, altered neurochemical correlations within DLPFC, anterior cingulate, thalamus [27
], and SSC [6
] suggest disrupted in neurochemical coupling [27
] in brain regions involved in both affective-emotional and sensory-discrimination aspects of pain. In agreement with these findings, we noted significantly lower intra-M1 and inter-M1 neurochemical correlations in our patients compared to controls. Although the significance of lower correlations is not clearly understood, we suggest that the effects of pain are not limited on individual neurochemicals but also on the interactions or “communication” between them.
Although no previous studies have exclusively examined neurochemicals in M1, Grachev et al
] reported no such changes in sensorimotor area in chronic LBP. This contrary finding can be explained by methodological differences between studies. We specifically analyzed individual voxels in the trunk representation area in M1 using multi-voxel MR spectroscopy imaging whereas Grachev et al
] used uni-voxel MRS to examine both M1 and SSC. In addition, we investigated absolute neurochemical concentrations rather than ratios to creatine, which might not be a reliable reference in chronic pain [6
Our study has some limitations that might affect its generalizability. We did not measure the outcomes related to lower trunk area such as trunk muscle strength, activation or volume changes of deep back and abdominal muscles. Such measures may provide better understanding of M1 neural changes and more direct correlations between neurochemical levels and clinical outcomes. Second, our sub-group analysis of pain medication effects on neurochemicals was limited to very small numbers per group. Accordingly, our observation of no significant differences in NAA or mI between medicated and un-medicated groups should be interpreted with caution. Third, this study was restricted to examination of neurochemicals in primary motor cortex and did not include other brain regions involved in pain processing, such as DLPFC, anterior cingulate, and insular cortex. A more widespread study of all areas involved in pain processing might provide a better understanding of their relative contribution in addition to M1 to overall brain plastic changes observed in chronic pain. Finally, due to the point-spread function of 1H-MRS acquisition, the effective voxel size is larger than the selected voxel size. Thus, we cannot exclude the possibility that our measurements include more than trunk representation in each M1.