By analyzing clinical and neuroimaging data, including DTI and fMRI results, we provided novel information on clinical and anatomic factors that contribute to an antalgic effect of rTMS on poststroke central pain and identified functional neural correlates associated with rTMS responses. The results of this study demonstrated that high-frequency rTMS applied over the M1 for 5 consecutive days can produce a partial antalgic effect on poststroke central pain. This effect can persist for 2 weeks after the completion of the rTMS intervention. The antalgic effect was more prominent in patients with less depression and preserved superior TCT in the targeted ipsilesional hemisphere.
Reports describing the influence of sensory function on the antalgic effects of rTMS have been inconsistent. Drouot et al20
hypothesized that a preserved sensory neural network is critical for effective modulation of the pain neural network, using surgically implanted devices for epidural cortical stimulation. In contrast, Lefaucheur et al6
found that the degree of sensory loss in the painful area did not influence the antalgic effects of rTMS. However, they also found that pain relief correlated with post-rTMS improvement of warm sensory thresholds in the painful zone of patients with chronic neuropathic pain.21
We did not test for primary sensory appreciation, however.
Studies examining the site of stimulation found that facilitation of the M1 induces an antinociceptive effect on various types of neurogenic pain.9,11
Hirayama et al10
reported an antalgic effect only with M1 stimulation—stimulation over the primary sensory cortex, premotor cortex, and supplementary motor area did not have any effect. rTMS over M1 may spread to other areas of the brain via corticosubcortical and corticocortical connections.22
Although the mechanism of pain relief by rTMS is still unclear, such transsynaptic effects of rTMS along the motor cortex, corticospinal tract, and corticothalamic tracts are believed to play an important role.13,23
The neural network related to poststroke central pain appears to be similar to that of the pain network across several functional imaging studies24,25
and involves the cingulate cortex, insula, dorsolateral prefrontal cortex, SII, inferior parietal lobe, cerebellum, and lateral thalamus.26
Kramer et al27
also reported that the allodynia-associated network included the insula, anterior cingulate gyrus, orbitofrontal cortex, and putamen. In this study, all patients had both allodynia and hyperalgesia, and their baseline fMRI demonstrated a similar activation of the pain network in both the responder and nonresponder groups. However, in the responder group, fMRI activity in the SI and SII, insula, prefrontal cortex, putamen, and cerebellum before rTMS decreased after rTMS. In contrast, neural activity did not change in the nonresponder group. Therefore, we suggest that the antalgic effects of rTMS are mediated by modulation of neural activity in a distributed pain network that includes not only the targeted sensorimotor cortex, but the SII, insula, anterior cingulate gyrus, pre-frontal cortex, and cerebellum.
Such distributed modulation of activity across a neural network by stimulation of a given cortical area may require preserved anatomic connectivity as revealed by our previous study in healthy participants.28
Indeed, careful anatomic analysis in the cat shows that the transsynaptic impact of TMS is correlated with the strength of anatomic connections between the directly targeted brain region and distant, affected areas.29
Two previous studies have pointed to the superior TCT as having a crucial role in poststroke central pain.13,30
Goto et al13
analyzed the corticospinal tract and TCT in patients with poststroke central pain and reported that the effects of rTMS on the pain modulation effect were greater in patients with higher delineation of the corticospinal tract and TCT. Our study confirms and extends those results, in that the antalgic effect of rTMS is greater when the superior TCT is more preserved, presumably enabling greater thalamic impact mediated by modulation of the stimulated sensorimotor cortex.
Looking at the DTI data, there are 2 issues. The first is that a larger stroke lesion would result in additional dysfunction in the pain-related neural network; therefore, pain would be more refractory to rTMS. Stroke lesion volume analysis in the future would clarify this point. The second is a matter of stroke lesion location. Lefaucheur et al6
found that the rTMS effect was least in patients with brainstem stroke and explained this finding as an injury of descending modulation within the brainstem, triggered by the motor corticothalamic output. In this study, we classified stroke lesions according to whether they affected the thalamus, brain stem, and other areas and found that stroke lesion location did not influence the rTMS effect. This may be a result of the small number of participants or lack of evaluation of the cortico-thalamo-spinal pathway.
This study had some limitations. We did not include a sham control group. Also, this study was not aimed at efficacy. We aimed to find novel information about changes in the pain network that may affect responsiveness to rTMS. The effect of rTMS on pain, however, was modest compared with previous studies.6,8
In clinical settings, patients are usually described as “improved” when their VAS score falls 1 point or more relative to their preintervention status. There was a statistically significant difference in the rTMS antalgic effects between the responder and nonresponder groups, so it is reasonable to evaluate the neural correlates of the antalgic effects by comparing these 2 groups. Unfortunately, only 10 patients agreed to DTI and pre-rTMS fMRI studies, so our results may not be generalized to patients with chronic poststroke central pain.
The clinical-anatomic-functional relationship found in this study suggests that depression may affect the pain modulation effect of rTMS; antalgic effects of rTMS occur in patients with preserved superior TCT, and pain relief by rTMS is associated with decreased activation of the pain network. By better defining the patients who are most likely to respond to forms of rTMS, more robust improvements in pain control may become feasible.31