Brain Responses to Noxious Stimuli in Patients With Chronic Pain

This meta-analysis assesses differential brain responses to noxious stimuli in patients with chronic pain using functional magnetic resonance imaging while adhering to current best practices for neuroimaging meta-analyses.


Data Extraction
Coordinates and information about each experiment were extracted manually by at least one author (AX or VS) and checked independently by another member of the study team (AX or VS). The following information about each paper was extracted for analyses: sample size; and whether the coordinate space was MNI or Talairach. If multiple related contrasts were present in an article, we included all contrasts but treated them as one experiment, thereby using only one set of coordinates in the metaanalysis per article. To provide more detail on experiments included in the meta-analysis, we also extracted more information about patients. This information included diagnosis, diagnosis method, duration of pain, medication status, modality of noxious stimulus (e.g., thermal, mechanical, chemical, or electrical), and location of induction.

©2021 Xu A et al. JAMA Network Open.
We conducted meta-analyses using the coordinate-based meta-analytic method activation likelihood estimation (ALE) using a revised algorithm that allows for random effects inference. [44][45][46][47] Briefly, for each experiment included, ALE treats coordinates for the foci of reported clusters as the center of an uncertainty function modeled by a 3D Gaussian probability distribution. The full width at half-maximum (FWHM) of this 3D Gaussian kernel was determined by empirical data on between-subject and betweentemplate (i.e., MNI or Talairach space coordinates) variance. Specifically, the algorithm takes into account between-subject variance by using a tighter Gaussian distribution for experiments with greater sample sizes to represent that these experiments should provide more reliable results of a true activation effect. 45 It also models uncertainty due to template use and transforms coordinates reported in Talairach coordinates into MNI coordinates. 48 This model then provides probabilities for all activation foci in each experiment, which were combined for each voxel, resulting in an individual modeled activation (MA) map for each experiment. By taking the union across all the MA-maps, we generated voxelwise ALE scores that describe the convergence of results at each particular location. 45 Note that MA-values reflect data for a single experiment while ALE-values integrate data across multiple experiments.
For ALE maps, the p-value was defined as the proportion of values obtained under a null distribution reflecting a random spatial association between experiments. The resulting non-parametric p values were subsequently thresholded using a voxelwise threshold of p < 0.001, reflecting current recommendations for best practices. 49,50 At this voxelwise threshold, the significance of cluster extent was estimated using 10,000 Monte-Carlo simulations; this distribution was calculated specifically for each metaanalysis conducted. 46,47 Clusters were considered significant if they achieved a family-wise corrected significance of p < 0.05. Prior to display, p-values were transformed into z-scores.
ROI-based analyses used an ROI mask constrained to regions activated by pain stimulation in healthy participants that was derived from a previous meta-analysis. 51 ROI-based meta-analyses resulted in an outcome measure of the sum of ALE scores across our mask (i.e., the ALE integral). This ALE integral was then compared with the sum of a null set of ALE scores generated from a random spatial distribution, which was estimated from 10,000 random realizations of the ALE integral in the search mask (thereby defining the null hypothesis). The ALE integral exceeding the sum of the null set would reflect whether the sum of ALE scores within the directed search mask were activated above chance.

Sub-analyses of experiments matching noxious stimuli by perceptual rating
Sub-analyses of experiments matching noxious simulation to patients and controls by similar perceptual ratings of pain (n = 18) did not reveal any significant results (eFigure 1). There were not enough experiments matching noxious stimulation to patients and controls by similar stimuli intensity to conduct robust sub analyses (n = 11).

eFigure 1. Unthresholded Maps of Subanalyses of Experiments Matching Noxious Stimuli by Perceptual Rating
Maps display unthresholded effects of experiments matching noxious stimuli between groups by perceptual rating of pain. No between-group differences were significant at our pre-registered statistical threshold (voxel height P < 0.001, FWE-corrected cluster significance P < 0.05).

Post-hoc regional sub-analysis of pain network show greater responses to pain in patients with chronic pain
Based on examination of unthresholded maps related to sub-analyses of experiments reporting greater activity in patients in response to pain (n = 23), we conducted post-hoc regional analyses focused on a previously reported pain network. 51 This ROI-based analysis revealed overall convergence of greater brain activity in patients in response to chronic pain. Sub-analysis of experiments reporting less activity in patients were not conducted due to an insufficient number of experiments to reliably meta-analyze.

eFigure 2. Post hoc Regional Analysis of Pain Network Suggest Greater Activity in Response to Pain in Patients With Chronic Pain
Histogram shows distribution of summed ALE scores within the pain network under a null hypothesis of random, permuted spatial distribution. Inset displays the pain network masked used based on a prior meta-analysis of brain responses to pain in healthy volunteers. 51 Observed value from non-permuted data suggest a significant, overall convergence of greater responses to pain within the pain network in patient

eTable. Study Information About Articles and Experiments Included in Metaanalyses
Description of sample, contrasts used, medication inclusion criteria, pain duration, stimulus modality and location, baseline used to measure against pain, method of matching noxious stimuli to patients and controls, and coordinates used. Citations of included experiments are listed in eReferences. IBS = irritable bowel syndrome, SE = standard error, SD = standard deviation, IQR = interquartile range  Table 4