Functional brain networks
that predict transition from sub acute to chronic pain.
JAH.
Introduction:
Chronic low
back is highly prevalent in the general population. Neuroimaging studies have
demonstrated that brain structure and function is significantly altered in
chronic back pain. Whether pre existing differences in brain
networks predispose subjects towards developing chronic back pain has not been
studied. Here we investigated the neural
mechanisms underlying the transition of sub-acute back pain to chronic back pain
in a longitudinal brain imaging study.
Methods:
Patients suffering from a single episode of sub-acute
back pain (SBP) were recruited. Pain intensity measurements and brain imaging
data were obtained at five specific time periods spanning over a period of one
year. Structural T1 images were acquired followed by functional magnetic
resonance imaging scans of subjects while they continuously rated spontaneous
back pain intensity or while they rated the changes in the length of a visual
bar as a control. On study completion, pain decreased >20% in nearly half of
the subjects (n=17) and persisted in the remaining 19 patients. The latter represents
the group that transitioned to chronic back pain and the two groups were
designated to a SBP recovering and SBP persisting group. Data analyses were
geared towards investigating mechanisms associated with predisposition towards
development of chronic back pain. Towards this goal, brain activity was
contrasted between the SBP persisting and SBP recovering groups at baseline to
delineate brain mechanisms that differ between the groups early on. The data
were preprocessed and analysed using FSL and MATLAB.
Results:
First, pain ratings activated the thalamus, brain
stem, amygdala and hippocampus in all subjects. An unpaired t-test between
groups showed that the SBP persisting group had significantly greater activity
in the left insula than the SBP recovering group. In contrast, the SBP persisting
group (corrected at p<0.05, z = 2.3) showed more response than the SBP
recovering group in the dorsal lateral prefrontal (DlPFC, corrected at
p<0.05, z = 2.0). To confirm this finding, the BOLD time series were
extracted from these regions and epochs spanning the time point where the
spontaneous pain showed increase were averaged per subject and group. These, trigger
evoked time series showed that the BOLD signal in the insula increased in
association with increase in pain, but this response was significantly more
exaggerated in the SBP persisting group. In the DLPFC, the pain trigger related
BOLD response increased in the SBP recovering group but in the SBP persisting
group, it showed the opposite pattern by decreasing when pain intensity
increased. The control task did not show visual task related increase in activity
in the insula or DLPFC and group differences were not significant. Network
analysis using ICA with dual regression showed that the fronto parietal attention
network increased in activity in the SBP recovering group and decreased in
activity in the SBP persisting group during phases when the subjects reported
an increase in pain. Moreover, the left insula was significantly more connected
to the left fronto parietal attention network and salience network in the SBP
recovering group (corrected at p<0.05, z = 2.3). Finally, the DLPFC and insula were significantly
more synchronized with each other in the SBP recovering group but in the SBP
persisting group, these two regions had little or no connectivity.
Conclusion:
Over all, these findings demonstrate that sub acute
back pain patients that later develop chronic back pain have increased activity
in the insula early on in the disease. In addition, sub acute pain patients
that later recover from pain have greater activity in the DLPFC and fronto
parietal attention networks. Moreover, a
enhanced synchrony between DLPFC and insula observed in the SBP recovering
group suggests a direct top down regulatory association between the two regions
that hinders the transition of pain from sub acute to chronic back pain.
GLM analysis, using binarised spontaneous pain vectors.
1. Contrasts between CBP and SBP. All contrasts corrected at 0.05 zthresh 2.3 using OLS. The only contrast not significant was the mean CBP. Unocrrected, it shows the mpfc. In the corrected contrast CBP > SBP , CBP shows the usual network.
CBP n=32
SBP n=36
SBP recovering versus persisting GLM contrasts using OLS:
SBP rec=17
SBP per =19
Standard visual controls
GLM and roi analaysis showed regional differences, but are these regions a part of specific networks? does each region contribute to differences in more than one network?
For this dual regression was used and voxel wise contrast showd the following spatial connectivity differences that were significant between SBP recovering and persisiting:
The above results showed that regions similar to those observed in the glm (e.g dlpfc and insula and dmpfc) were differently connected between the two groups. Suggests that the insula (the region that was more prominently activated by pain in sbp persisting in the glm analysis) is connected more to networks containing a prominent dlpfc element e.g attention, salience (?) in the recovering group. A shown below, three networks (attention, insula, dlpfc and dmpfc or indd network 1, and one other network that has many pain related regions, not shown below) show presence of left anterior insula only in recovering group. Suggests that these networks may be regulating insular activity that leads to resolved pain in the recovering group.
The over lappin regions (dlpfc, dmpfc and insula) were selected as roi's and the time series was extracted to calculate trigger evokes averages to see whether above differences are pain specific or not. These regions are distinct from the glm analysis, but show similar results where the insula seems to have more activity in persisting group and the prefrontal regions are activated more in the recovering group.
One question is, that are the prefrontal region connected with insula more in the recovering group. A difference in synchrony between these two groups (insula and prefrontal regions:dlpfc and dmpfc) would indirectly suggest that the correlation between insula and prefrontal regions are bound up in the recovering group and hence the recovering group shows less of a response to pain. The dlpfc, dmpfc and insula may represent a network that is more synchronised in the recovering group than in persisting group. The correlations between time series was calculated for all scans and showed that indeed the insula, dlpfc and dmpfc were more correlated to each other in the recovering group than in persisting predominantly during earlier scans.
Supplementary
Conclusion:
The GLM resulsts show
1. that back pain activates the MPFC in CBP patients and the acute pain networks such as thalamus, insula, amygdala and hippocampus in SBP patients.
2. Back pain is represented more strongly in the insula in SBP persisting patients. The SBP recovering showed more dlpfc activation than SBP persisting in the contrast, but it was below threshold after correction, so can't conclude anything with this result alone. However, complemented by the ica and pain triggered events results, the SBP recovering do show more dlpfc activity in response to pain.
3. The pain triggered responses show that insula and amygdala show more pain associated activity in the persisting group. The recovering group shows more pain triggered activity in prefrontal regions including dlpfc, dmpfc and mpfc.
4. ICA results show that the insula is bound up in cognittive networks such as the fronto parietal attention network and salience networks (insula, dlpfc, dmpfc containing networks) in the recovering group.
5. The regions and the networks ?? show temporal changes time locked with pain onset. The recovering and persisting groups show different activation patterns in response to pain especially in the attention network and default mode network.
