Subjects
Thirteen SCI-NP subjects and 13 age- and sex-matched HC were recruited for this study (Table 1). SCI subjects were recruited from the Spinal Cord Injury Center at Balgrist University Hospital and the Swiss Spinal Cord Injury Cohort Study database, while HC were recruited via online flyer advertisements. The inclusion criteria for SCI subjects were: i) the presence of NP by the current diagnostic criteria (Finnerup 2016), ii) 18-80 years old, iii) non-cervical SCI, i.e., thoracic or high lumbar level of injury, iv) no contraindications for MRI, and v) no history or presence of other neurological, psychological or medical conditions (e.g., traumatic brain injury, diabetes, cancer). The inclusion criteria for HC were: i) 18-80 years old, ii) no neurological or psychiatric conditions, iii) no history of chronic pain or pain during participation and iv) no intake of psychoactive medication. The 13 age- and sex-matched HC were selected from a primary cohort of 40 HC. The psychophysical and neuroimaging data obtained from this cohort will be reported separately. The matching process of the 13 HC was performed by a secondary investigator (MH) and the primary assessor (VH) was blinded from this selection procedure. Written informed consent was acquired from all subjects prior to the assessments. All procedures described are in accordance with the Declaration of Helsinki and the study has been approved by the local ethics board ‘Kantonale Ethikkommission Zürich, KEK’ (EK-04/2006, PB_2016-02051, clinicaltrial.gov number: NCT02138344).
NP characterization in SCI subjects
Diagnosis of NP in SCI subjects adhered to current recommendations which includes the presence of a neurological lesion of the spinal cord 22. Overall, presence of typical sensory signs and symptoms is required and the area of pain needed to follow a plausible neuroanatomical distribution with respect to the lesion level.
To assess the intensity and spatial extent of NP, each subject completed a pain drawing prior to MRI scanning (Rosner et al., 2021). The drawing consists of a body schematic (front and back) where subjects were asked to draw the current pain location, distribution alongside its intensity indicated on an 11-point numerical rating scale (NRS) (0 = ‘no pain’ to 10 = ‘worst pain imaginable’). A standardized scheme presenting the full body dermatomes was laid on the pain drawing in order to delineate at- and below-level NP, which was distinguished as pain within or below three dermatomes of the lesion, respectively. After identifying the NP, each affected area was highlighted and quantified into an overall percentage of NP extent (%), which was determined by the sum of total pixel count from both front and back of subjects’ digitalized pain drawings divided by the total pixel count of the body schematic. NP extent as characterized by these pain drawings has been shown to have excellent inter-session reliability 23. Averaged NP intensity of each SCI subject was obtained by calculating the sum of NP intensity divided by the amount of particular regions that the subjects rated as painful.
Study design
All subjects completed questionnaires, thermal thresholds and a familiarization procedure for the heat stimuli of the following CPM paradigm. The Pain Catastrophizing Scale (PCS) 24 and Beck Depression Inventory version II (BDI) 25 were filled out in order to provide psychological outcome measures of each subject. To ensure sensory integrity of the tested area, i.e., the volar forearm as an area above the neurological level of injury in our paraplegic subjects, thermal thresholds were examined according to the quantitative sensory testing protocol of the German Research Network on Neuropathic Pain 26. Warm detection and heat pain thresholds (WDT and HPT, respectively) were assessed with the PATHWAY Pain & Sensory Evaluation system using the 3cm x 3cm square ATS thermode (Medoc Ltd, Ramat Yishai, Israel) at the volar forearm. Each trial began with a baseline thermode temperature of 32°C and increased at a rate of 1°C per second. Subjects were required to click the response unit as soon as they perceived a change in temperature (WDT) or at the initial sensation of pain (HPT). Thermal thresholds were determined by averaging three trials of individual stimuli. Safety cut-off temperatures for WDT and HPT were set at 55°C. All subjects were blinded from the operator screen during threshold testing. After obtaining thermal thresholds, subjects were acquainted with the instructions of MRI acquisition and the CPM procedure including the pain rating process. Those who understood the instructions clearly were familiarized with a heat stimulus on the volar forearm on the side of the non-dominant side.
Conditioned Pain Modulation
The CPM paradigm is summarized in Figure 1. Inside the MRI scanner, a parallel CPM paradigm with two conditions was performed with each subject: i) TS with a CS (TS-CPM) and ii) TS with a sham condition (TS-Sham). The two conditions were randomized across subjects in a balanced fashion and lasted 6:10 mins with a five-minute break in between. The TS was applied with a 3cm x 3cm square ATS thermode (Medoc Ltd, Ramat Yishai, Israel) attached to the volar forearm of subjects’ dominant hand. Per condition, eight TS were applied with an inter-stimulus interval of 35 seconds with each TS having a fixed target temperature of 47.5°C and lasted a total of 10 seconds including ramp time (2.5 seconds ramp up, 5-seconds plateau, 2.5 seconds ramp down) (Sprenger et al., 2011). Between each condition, the position of the thermode was slightly shifted to alleviate sensitization effects. Following each TS, subjects had 20 seconds to rate their perceived pain of the TS and CS (10 seconds each) on a numeric rating scale (NRS, 0 being "no pain" and 10 "worst pain imaginable"). The NRS was projected on the NordicNeuroLab 32” screen (NordicNeuroLab, Norway and USA, https://www.nordicneurolab.com) and the subjects rated using a manual response unit placed in their dominant hand. This unit was programmed to either move a box up or down the NRS per click of the allocated button, i.e., if subjects perceived the pain of the TS to be a “five” subjects clicked button 1 five times to move the box up the NRS. Subjects were always prompted to rate the TS followed by rating of the CS. The CS consisted of two ice bags covering the non-dominant hand for the whole duration of the condition. Each ice bag contained ~600g of ice and 250ml of water guaranteeing a stable temperature of 0°C. Pain rating of the CS from pilot data (n=5) indicated it being an appropriate noxious stimulus with an averaged pain perception of NRS 7.0 (range= 6.3-8.0). As recommended by Yarnitsky et al. (2015), an appropriate noxious stimulus is determined at an intensity of 4/10. For the sham condition, two bags with water at skin temperature (~32°C) were used. This CPM paradigm was adapted from a previous study utilizing similar ice bags, noxious stimuli and block timings (Sprenger et al., 2011). In particular, we applied the CS at the hand rather than the whole leg to prevent autonomic dysreflexia in our SCI-NP cohort 28.
MRI data acquisition
All subjects’ MRI images were obtained with the 3.0 Tesla Philips Ingenia system (Philips Medical Systems, Best, the Netherlands) using a 32-channel Philips head coil. 3D T1-weighted structural images were acquired with a Turbo Field Echo sequence with the following parameters: repetition time (TR), 8.1ms; echo time (TE), 3.7ms; flip angle (FA), 8°; number of slices, 160; slice thickness, 1mm; field of view (FOV), 240 x 240 x 160mm3; matrix, 240 x 240 and isotropic voxel 1 x 1 x 1 mm3. Scan time was a total of 4:53 mins. Resting-state functional images were acquired with an echo-planar-imaging sequence with the following parameters: TR, 2000ms; TE, 30ms; FA, 78°; number of slices, 36; FOV, 220 x 136 x 220mm3; matrix, 72 x 74; voxel size, 3.0 x 3.0 x 3.0mm; reconstructed voxel size, 1.72 x 1.72 x 3 mm3 and a scan time of 5:00 mins. During the resting-state acquisition, subjects were instructed to relax and asked to fixate on a motionless cross, projected on a NordicNeuroLab 32” screen (NordicNeuroLab, Norway and USA, https://www.nordicneurolab.com). To minimize motion, cushions were placed around each subject’s head. Structural and resting-state functional MRI data acquisition was performed prior to the CPM paradigm. This sequence and paradigm was used in a previous study 29. Task-related functional MRI data was acquired during the CPM paradigm, but this data will be reported separately.
Data analyses
CPM psychophysics
For each subject, the CPM effect was calculated as the averaged pain ratings of the eight TS during the TS-Sham condition, subtracted from the averaged pain ratings of the eight TS during the TS-CPM condition (CPM effect = averaged pain ratings of TS (TS-CPM) – averaged pain ratings of TS (TS-Sham)). This provided an overall sham-corrected CPM effect score with negative numbers denoting pain inhibition and positive numbers denoting pain facilitation 6. Individual CPM effect scores were used in the neuroimaging analyses to investigate brain correlates of CPM effect.
Statistical analyses were performed with the Statistical Package for the Social Sciences (SPSS, version 24). To test normality of subjects’ characteristics and CPM psychophysical readouts, histograms, Q-Q plots and Shapiro-Wilk tests were applied. For non-normally distributed variables non-parametric tests were used. Within each cohort, overall CPM effects was tested with a dependent two-tailed t-test between the averaged pain ratings of the TS in each condition (TS-CPM vs TS-Sham). Correlation analyses were used to test relationships between subjects’ age, WDT, HPT, BDI and PCS scores and pain ratings of the CS with individual CPM effect. Between-group differences in demographics and pain ratings were tested with independent two-tailed t-tests and chi-square tests. In SCI-NP subjects only, partial correlations were implemented to assess associations between CPM effect and NP characteristics, e.g., NP intensity and extent with age, sex, with PCS and BDI scores as covariates of no interest. Results were deemed significant at p<0.05.
Pre-processing for neuroimaging analysis
Structural T1-weighted images and resting-state fMRI images were pre-processed using Statistical Parametric Mapping (SPM12) software (Wellcome Department of Imaging Neuroscience, London, United Kingdom: (http://www.fil.ion.ucl.ac.uk/spm/) implemented in MATLAB 2017a (The Mathworks, Inc, Natick, MA). Prior to pre-processing, structural and functional images of each subject were realigned and centered to the anterior commissure (Montreal Neurological Institute (MNI) coordinates; MNI = 0, 0, 0) using the SPM12 display function. One SCI-NP subject showed distortion artefacts of their resting-state fMRI data and was excluded from further analysis (n = 12).
Structural scans were segmented into grey matter, white matter and cerebrospinal fluid maps using the New Segment tool 30. Functional images were pre-processed as follows: realignment (head motion correction), centering (to anterior commissure, MNI co-ordinates= 0,0,0), slice-timing correction (ascending), outlier detection and scrubbing (using ARtifact detection Tools) during the denoising step 31,32, MNI normalization and smoothing with a 6mm Gaussian full width at half maximum (FWHM). The pre-processing steps generated interpolated 2 x 2 x 2 mm3 resolution images for the analyses. Head motion during the resting-state scan was assessed with the three translational and rotational dimensions for each scan. Subjects whose mean head motion during the functional scan exceeded +1.5mm for translation and/or 1° for rotation were removed from rsFC analyses. During the denoising step, normalization of voxel-to-voxel connectivity values were performed in addition to linear detrending and subjects that showed normally distributed data after denoising were included for rsFC analyses.
Seed-to-voxel rsFC analyses
Resting-state functional MRI data was analysed with the CONN toolbox (CONN 18b; www.nitrc.org/projects/conn) 33. CONN utilises a component-based noise correction method (CompCor) which increases selectivity, sensitivity and permits a higher degree of inter-scan reliability (Behzadi et al., 2007). A band-pass filter of 0.01-0.1 Hz was applied to remove linear drift artefacts and high-frequency noise. CONN also accounts for outlier data points and movement time courses as nuisance regressors. For each subject, the six motion parameters, activity from segmented white matter and cerebrospinal fluid maps were included as regressors of no interest, thereby reducing noise and signal unlikely to reflect neuronal activity related to functional connectivity.
A general linear model was implemented to test between-group differences in seed-to-voxel rsFC and the relationship between CPM effect and seed-to-voxel rsFC. To this end, a one-way ANCOVA was implemented (Whitfield-Gabrieli and Nieto-Castanon, 2012) to firstly compare differences of seed-to-voxel rsFC without CPM effect. Secondly, differences in regressions of CPM effect and rsFC between HC and SCI-NP subjects were explored. Here, seed rsFC values were the dependent variable, and group-by-CPM interactions were the independent variables. A priori ROIs were used as seed regions for seed-to-voxel analyses, these ROIs were areas involved with descending pain modulation, i.e. ACC, amygdala and vlPAG. The left and right ACC 35 and amygdala 36 were created in the SPM anatomy toolbox 37. ROI maps of the vlPAG (left, right and bilateral) were provided by Ezra et al., 38,39. Age and sex were included as covariates of no interest for all analyses and significant results are reported at p<0.05 Family-Wise Error (FWE) level correction (Whitfield-Gabrieli and Nieto-Castanon, 2012). For visualization purposes, CPM effect was plotted against the rsFC strength (Fisher transformed correlation coefficients) between ROIs showing significant associations. Pearson’s correlation coefficient from partial correlations between CPM effect and rsFC strength were also provided for visualization (age and sex were included as covariates of no interest).