Central neuropathic pain (CNP) after spinal cord injury (SCI) is debilitating and immensely impacts the individual. Central neuropathic pain is relatively resistant to treatment administered after it develops, perhaps owing to irreversible pathological processes. Although preemptive treatment may overcome this shortcoming, its administration necessitates screening patients with clinically relevant biomarkers that could predict CNP early post-SCI. The aim was to search for such biomarkers by measuring pronociceptive and for the first time, antinociceptive indices early post-SCI. Participants were 47 patients with acute SCI and 20 healthy controls. Pain adaptation, conditioned pain modulation (CPM), pain temporal summation, wind-up pain, and allodynia were measured above, at, and below the injury level, at 1.5 months after SCI. Healthy control were tested at corresponding regions. Spinal cord injury patients were monitored for CNP emergence and characteristics at 3 to 4, 6 to 7, and 24 months post-SCI. Central neuropathic pain prevalence was 57.4%. Central neuropathic pain severity, quality, and aggravating factors but not location somewhat changed over 24 months. Spinal cord injury patients who eventually developed CNP exhibited early, reduced at-level pain adaptation and CPM magnitudes than those who did not. The best predictor for CNP emergence at 3 to 4 and 7 to 8 months was at-level pain adaptation with odds ratios of 3.17 and 2.83, respectively (∼77% probability) and a cutoff value with 90% sensitivity. Allodynia and at-level CPM predicted CNP severity at 3 to 4 and 24 months, respectively. Reduced pain inhibition capacity precedes, and may lead to CNP. At-level pain adaptation is an early CNP biomarker with which individuals at risk can be identified to initiate preemptive treatment.
Pain adaptation test predicts central neuropathic pain after spinal cord injury with high probability and sensitivity and can be used to identify individuals at risk.
Central neuropathic pain (CNP) after spinal cord injury (SCI) has an average prevalence rate of 50% and is one of the most debilitating and difficult conditions to manage. 5,6,21,41,53 Unfortunately, CNP is considered refractory to most of the currently available interventions, 2,20,35,42,46 perhaps because patients receive treatment after CNP has already developed, namely, after substantial and perhaps irreversible neuronal changes have already occurred, dampening the therapeutic potential.
Preemptive treatment may overcome this challenge. Findings from SCI animal models showed that signs of CNP can be prevented or mitigated by early administration of various interventions including neuroprotective agents, 8,25 cell cycle and gap-junction modulators, 39,55 immune-suppressants, 44 and stem cell transplantation. 28,52 Although the aforementioned treatments have not been tested in humans, preemptive administration of medications currently administered for CNP such as pregabalin may improve CNP management. However, due to possible adverse effects and high costs, such treatments should be administered only to individuals who are at risk to develop CNP. Therefore, identifying predictive biomarkers for CNP is imperative. 47
Only a few prospective studies have investigated possible biomarkers for predicting CNP. We have previously found that below-level hypoesthesia and hyperexcitability combined, measured one month after SCI, predicted 89% of the risk for CNP at 18 months. 59 Finnerup et al. 19 further reported that below-level hypersensitivity or dysesthesia at 1 month after SCI predicted 67% of the risk for CNP at 12 months. Despite the rather high prediction levels in both of these studies, these biomarkers were relevant only among individuals with incomplete SCI and could predict only below-level CNP. In addition, cutoff values could not be calculated. In 2 more recent studies, clinical pinprick and light touch scores or ratio tested at the lesion dermatomes at admission predicted CNP at 12 or 18 months with moderate probability. 32,50 However, here again, cutoff values could not be calculated. Furthermore, the prediction was far less effective for individuals with complete SCI. Recently, electroencephalographic recordings revealed a reduced alpha peak frequency during the first 6 months after SCI among individuals who developed CNP later on. 48 However, this reduced frequency also appeared among individuals who did not develop CNP and therefore may not be a specific marker. Furthermore, its recording several months after SCI may not leave a sufficient time window for early, preventive interventions.
Therefore, identifying a single biomarker that can predict both below- and at-level CNP, among all individuals with SCI, as early as possible after the injury, and that enables calculating a cutoff value for clinical decision making is of utmost importance. To fulfill this need, we performed a battery of quantitative sensory tests evaluating, for the first time, both pronociceptive and antinociceptive components of the pain system as early as possible after SCI and followed up on CNP characteristics for 2 years. By comparing the early findings between individuals who eventually developed CNP and those who did not, we could identify a clinically relevant biomarker for the presence of CNP and its severity and infer on its mechanism.
Sixty-seven individuals participated in the study: 47 with acute SCI (average age 46.9 ± 15.5 years, 33 males and 14 females) and 20 healthy controls (HCs) (48.2 ± 16.6 years, 15 males and 5 females). Patients with acute SCI (no longer than 1 month) were recruited from the Department of Neurological Rehabilitation at the Sheba Medical Center, Tel Hashomer, on a voluntary basis. Healthy controls were recruited among the employees of Tel-Aviv University and Sheba Medical Center.
Patients with acute SCI were admitted consecutively to the department between the years 2013 and 2017. Inclusion criteria were as follows: (1) a neurological level of spinal lesion above T10 (to avoid lesions to the conus medullaris and cauda equina) and below C6 for complete lesions or below C4 for incomplete lesions (to ensure intact sensibility in the tested regions), (2) an SCI period of no longer than 3 weeks, and (3) age between 18 and 70 years. Exclusion criteria were as follows: (1) chronic pain before SCI, (2) acute pain other than that involved in postoperative procedures related to SCI, (3) previous or present wounds or skin lesions in the tested regions, (4) known or clinical signs of concomitant cerebral damage, (5) a history of neurological disorders other than SCI (eg, multiple sclerosis, cerebral palsy, traumatic brain injury), (6) concurrent severe medical problems, (7) diseases causing potential neural damage (eg, diabetes mellitus), (8) pregnancy, and (9) any psychiatric or cognitive status that may interfere with the trial. Healthy controls had to be at the same age range of the patients, free of any pain, and they were also subject to aforementioned exclusion criteria no. 3-9.
The study was approved by the human ethics committee of Sheba Medical Center and of Tel Aviv University. Written informed consent was obtained from all subjects, according to the Declaration of Helsinki guidelines after they received a full explanation of the study protocol and goals.
2.2.1. Thermal stimulator
Heat stimuli were delivered using a computerized thermal stimulator (TSA II; Medoc Ltd, Ramat-Ishay, Israel), with a 3 × 3-cm contact probe. Current passing through the Peltier element produces temperature changes at rates determined by an active feedback system. As soon as the target temperature is attained, the probe reverts to a preset adaptation temperature by passing an inverse current. The adaptation (baseline) temperature was set to 32°C.
2.2.2. Water bath
Heat stimuli were also delivered using a 10-L water bath (Chillsafe, ScanVac; Ballerup, Denmark). This circulator bath allows fixed temperatures ranging from 30 to 100°C to be set and maintained (maximum variance ±0.5°C). The water temperature was kept constant at 46°C.
2.2.3. Semmes–Weinstein monofilaments
Mechanical stimuli were delivered with Semmes–Weinstein monofilaments (Touch-Test Sensory Evaluator; North Coast Medical, Inc, Morgan Hill, CA). The kit includes 20 monofilaments that are attached to a plastic holder, ranging between 1.65 and 6.65 calibrated units. Vertical pressure applied with the handle induces a force ranging between 0.008 and 300g.
2.3. Sensory testing
2.3.1. Stimulus-response function for suprathreshold stimuli
A stimulus-response function was created for each subject to extract the stimulation temperatures for subsequent measurements. Subjects received a series of thermal stimuli delivered with the thermal stimulator in an ascending order and were asked to rate their perceived pain using a numerical rating scale (NRS) by choosing a number between the end points 0 = “no pain sensation” and 10 = “the most intense pain sensation imaginable.” The stimuli rose from a baseline temperature of 32°C (the rate of rise was 3°C/second, with an interstimulus interval of 25 seconds) to a destination temperature ranging from 41 to 51°C, or to an intensity eliciting 7 on NRS, where it was maintained for 5 seconds and then returned to baseline. The stimulator probe was moved after each stimulus. From the individual stimulus-response functions, temperatures eliciting values of 3 to 4 and 5 to 6 on NRS were extracted and used in the subsequent testing. 24
2.3.2. Antinociceptive components of the pain system
22.214.171.124. Pain adaptation
Pain adaptation refers to a gradual decrease in pain after repeated/constant, mildly noxious stimuli of fixed intensity, and it reflects antinociceptive mechanisms. 3 Pain adaptation was measured because it has been found to be reduced among SCI individuals with chronic CNP 24,30 but it has not been evaluated so far in the acute phase. To test for pain adaptation, subjects received a noxious heat stimulus with the thermal stimulator at an intensity equivalent of 3 to 4 on NRS (individually adjusted), for 75 seconds. The subjects were asked to rate the amount of perceived pain (using NRS) every 15 seconds (at times 0, 15, 30, 45, 60, and 75 seconds). The stimulus was applied both at and above the injury level (described in section 5). The subjects were not informed of the time that had elapsed from the beginning of stimulation. The magnitude of pain adaptation was calculated by subtracting the first NRS rating from the last. 24
126.96.36.199. Conditioned pain modulation
Conditioned pain modulation (CPM) refers to the diffuse noxious inhibitory control loop 57 wherein pain in one body region is inhibited by pain in another remote region and reflects antinociceptive mechanisms. 31 Conditioned pain modulation was measured because it was found to be reduced among SCI individuals with chronic CNP 24 or who are in the subacute phase, 1 but it has not been evaluated so far in the acute phase. Conditioned pain modulation was measured by administering a noxious stimulus (the test stimulus [TS]) and evaluating its perceived intensity alone and in the presence of a noxious stimulus applied to the contralateral side (conditioning stimulus [CS]). The TS was administered with the thermal stimulator at an intensity equivalent to 5 to 6 on NRS (individually adjusted), for 5 seconds. It was applied both at and above the injury level (described in section 5). Conditioning stimulus was administered to the contralateral hand by immersing it in a hot-water bath (46°C) for 30 seconds. The subjects rated the TS alone, then the contralateral hand was immersed in the hot water for 25 seconds, after which the TS was applied a second time while the hand was still immersed. The magnitude of CPM was calculated by subtracting the NRS rating of the TS in the presence of the CS from the NRS rating of TS alone. 22
2.3.3. Pronociceptive components of the pain system
188.8.131.52. Temporal summation of heat-pain
Temporal summation of heat-pain refers to a phenomenon in which perceived pain gradually increases in response to tonic or phasic noxious stimulation and reflects pronociceptive mechanisms. 43 Temporal summation of heat-pain during tonic heat stimulation was measured because it has been found to increase among SCI individuals with chronic CNP 24 but has not been evaluated so far in the acute phase. To measure TSP, subjects were instructed, in a separate test, to immerse their hand in a hot water bath (46°C) for 30 seconds. During this time, they were asked to rate the amount of perceived pain (using NRS) 3 times: immediately after immersion, after 15 seconds, and 30 seconds after immersion. The subjects were not informed of the time that had elapsed after immersion. The TSP value was calculated by subtracting the first NRS rating from the last. 22,23
184.108.40.206. Mechanical wind-up pain
Temporal summation of heat-pain was also measured with repetitive stimulation at a fast rate, a phenomenon termed wind-up pain, which reflects the frequency-dependent increase in the excitability of spinal cord neurons. 12 Mechanical wind-up (WU) was measured because it has been found to be increased among SCI individuals with chronic CNP 11,18 as well as in the acute phase, before CNP emergence. 59 The examiner applied a Semmes–Weinstein monofilaments no. 6.65 four consecutive times every 3 seconds (0.3 Hz). The subjects were asked to rate the intensity of pain after the first stimulus and fourth stimulus on NRS. The first stimulus of the series produced no pain or minimal pain sensation. The magnitude of wind-up was calculated by subtracting the first NRS rating from the fourth. 59
Allodynia is pain evoked by a nonnoxious stimulus, 37 a phenomenon that reflects the hyperexcitability of the pain system due to enhanced pronociceptive mechanisms. 13,29 Allodynia was measured because it has been frequently recorded among individuals with chronic SCI and CNP 11,16,24 and has been found to precede CNP. 19,59 Mechanical allodynia was examined at and below the injury level, by gently dragging a Semmes–Weinstein monofilaments no. 4.74 along the subjects' skin at 3 cm/second. The subjects were asked to report the quality of sensation evoked by the stimulus, and if pain sensation was evoked during stimulation, allodynia was defined as being present. 59 The presence of other types of allodynia, eg, due to temperature or clothing, was recorded based on interviews with the patients.
2.4. Additional data collection
All the SCI patients were questioned about acute or chronic pain, general health status, and background variables including questions about the injury (eg, age at injury, cause of injury, and motor/sensory level). The information provided was corroborated with the patients' medical records. Patients who developed CNP during the time course of the study were questioned about pain onset, duration, quality, location in the body (on a body chart), dynamic characteristics, use of medication, and alleviating and aggravating factors. To calculate the number of painful body regions, the body was divided into 13 regions: foot, shin, thigh, buttocks, hand, forearm, arm, shoulder, abdomen, chest, lower back, upper back, and neck. Then, the number of areas with CNP from the 13 regions was counted for each patient. The patients were also asked to rate the average severity of their CNP (at or below injury level) in the last 2 weeks on a visual analogue scale (VAS) and they also completed the McGill Pain Questionnaire (MPQ). The MPQ enables quantitative evaluation of the patients' pain. 36 The quantitative indices were (1) the pain rating index (PRI): the total values assigned to the words chosen from a list of 64 pain descriptors; and (2) the number of words chosen (NWC) from that list.
2.5. Experimental protocol
The sample size for this study was estimated based on several considerations: the admission rate of patients with acute SCI to the ward, the ability to fulfill the inclusion and exclusion criteria, the average prevalence rate of CNP, which is about 50%, 15,19,41,59 the sample sizes in our previous longitudinal studies, 32,59 and the expected difference in the mean values of the main outcome measures (CPM and TSP) between the groups. The mean values and standard deviations of CPM and TSP found in our 2 group-design studies on CNP (eg, Refs. 22 and 24 ) were entered into the calculation. For a sample size of 18 and for α = 0.05, the statistical power is 94.6% and for α = 0.01, it is 89.8%.
Figure 1A describes the study's time course for individuals with SCI. The sensory evaluation occurred as soon as the patients with SCI were admitted to the department, provided that they were in a condition that enabled sensory testing (in terms of mental state and adjustment to the department) and provided that they fulfilled the inclusion and exclusion criteria. The average time of the sensory evaluation was 1.5 ± 0.8 months after SCI. Figure 1B describes the sites where quantitative sensory testing (QST) was performed. Because we looked for biomarkers that can predict CNP among both SCI patients with complete lesions and those with incomplete lesions, it was important to perform the testing at the injury level rather than below the injury level, because body regions below the injury level among SCI patients with complete lesions lack sensibility. In addition, given that antinociceptive and pronociceptive tests are intended to examine traits of the pain system, which are not restricted to painful sites 58 and to compare these traits between SCI patients and HCs, we also performed sensory testing above the injury level (intact regions for the SCI patients). Among HCs and among individuals with thoracic SCI (60% of patients), the stimulus-response functions as well as testing of pain adaptation, CPM, TSP, and WU were measured in the volar aspect of the forearm, which was an intact, above the lesion site. Among individuals with cervical SCI (40%), these measurements were conducted on the upper part of the arm or the neck region, depending on the exact injury segments, which were intact, above lesion sites. Among SCI patients, testing at the lesion level included stimulus-response functions, pain adaptation, CPM, and allodynia (which was the only test that was also performed below the injury). The lesion level was determined according to Standards for Neurological Classification of Spinal Cord Injury, which is part of the neurological examination performed at admission, and as corroboration, it is also based on the examiner's crude touch evaluation of the dermatomes of the lesion level. This region was the abdomen or chest area among 81% of the SCI subjects and the neck region among the rest.
The experimental protocol. (A) Time course of evaluations among patients with SCI. Sensory testing was conducted at 1.5 months after SCI, which was followed by 3 additional evaluation time points for determining the presence and characteristics of CNP: 3 to 4 months after SCI (short term), 6 to 7 months after SCI (long term), and 24 months after SCI (follow-up). (B) Sensory tests performed above the level of injury were pain adaptation, conditioned pain modulation, temporal summation of heat-pain, and mechanical wind-up. The 2 first tests were also performed at the level of injury. The presence of allodynia was examined at and below the level of injury. CNP, central neuropathic pain; SCI, spinal cord injury.
Testing took place in a quiet room. Participants sat in their wheelchairs or on a comfortable chair, or lied in their bed. Each testing session lasted about 2 hours. Before sensory testing, the subjects underwent a brief clinical examination to verify the lesion level and the sensory intactness of the testing areas above the lesion. All the participants underwent a training session in which they became familiar with the stimuli and practiced magnitude estimations with NRS. On completion of the training session, testing commenced. First, we determined the stimulus-response function for heat-pain in the intact body region, to extract the individual temperatures to be used for subsequent testing in these regions. We then commenced with the remaining tests above the injury level, which were followed by tests at the injury level. Except for the stimulus-response function, the order of the tests was randomized within each testing site and subjects were given several-minutes rest between tests. The stimulator probe was moved between stimulation trials.
After completing the sensory evaluation, we monitored pain complaints among the patients during their hospitalization to detect the emergence of CNP. If CNP was diagnosed, the patients completed the MPQ and were interviewed about the pain characteristics. Central neuropathic pain was diagnosed according to the criteria of the International Spinal Cord Injury Pain classification 4 : (1) below level pain: spontaneous or evoked burning, stabbing, shooting pain diffusely located in body regions at least 2 to 3 dermatomes below the spinal lesion level, and (2) at the pain level: spontaneous or evoked burning, stabbing, sharp pain located in dermatomes corresponding to the spinal injury level. For this purpose, any relevant data including the results of clinical examinations and diagnostic tests were examined. Because this definition is of exclusion, care was taken to exclude other pathologies that may underlie the pain such as pressure sores, urinary lithiasis, infections, and peripheral neuropathic pain.
After their discharge from the ward, the SCI patients were followed up for 24 months to identify new cases of CNP and to study whether those who developed CNP during hospitalization were still suffering from it after a long time. In addition, the follow-up periods were used to determine whether CNP characteristics change over time. Overall, patients who developed CNP completed the MPQ at 3 time points: at about 3 to 4 months after injury, at about 6 to 7 months after injury, and at 24 months after injury. Because most of the SCI patients arrive at the outpatients' clinics periodically after discharge for continuous rehabilitation or other needs, they could also be interviewed about the pain while completing the MPQ. If patients did not arrive at the hospital, interviews and completion of the MPQ was done through phone.
2.6. Statistical analysis
Data were processed with IBM SPSS statistics software (version 25). First, normal distribution was evaluated with the Kolmogorov–Smirnov (K-S) test. Parametric and nonparametric models were used to compare the demographics and sensory testing between patients who eventually developed CNP (the CNP group), patients who did not develop CNP (the non-CNP group), and HCs. The demographic independent variables were age, sex, marital status/living conditions, and employment. The sensory independent variables were magnitude (delta) and time course of pain adaptation, magnitude (delta) of CPM, magnitude (delta), and the time course of heat-TSP, the magnitude of wind-up pain, and the presence of allodynia. The 2 SCI groups were also compared for the variables related to SCI. These included the time since the injury, the extent of the injury, the cause of the injury, and the completeness of the injury (American Spinal Injury Association Impairment Scale [AIS]). In addition, we tested within each group the existence of pain adaptation, CPM, TSP, and wind-up pain by comparing the first pain ratings with the last pain rating and by comparing the TS alone with the TS in the presence of the CS. The models included the main effects, interactions, and pairwise comparisons (t-tests for the continuous variables, Mann–Whitney for ordinal variables, and χ2 tests for the dichotomy variables). Within the CNP group, repeated parametric and nonparametric models were used to assess changes over time (3 follow-up time points) in CNP characteristics (VAS, PRI, NWC, CNP quality, pain location, the number of painful body regions as well as aggravating and alleviating factors). The models included the main effects and pairwise comparisons (t-tests for the continuous variables, Wilcoxon tests for ordinal variables, and McNemar tests for the dichotomy variables).
To assess the ability of the independent variables to predict the risk for CNP, stepwise logistic regressions were applied with the dependent variable being CNP (yes/no) and the predicting variables being all the sensory indices obtained in body regions above, at, and below the injury level. Spinal cord injury completeness was a covariate in the regression. Based on the results of the logistic regression, we calculated the probability to predict CNP of those variables who were significant predictors according to the regression, using the reliability equation. We then calculated the cutoff value for the prediction that best classifies patients into 2 SCI subgroups, ie, identifies patients at risk for CNP using the receiver operating characteristic (ROC) curve (sensitivity vs 1 − specificity plot). The search for the cutoff value was based on the fact that the closer a result from a contingency table is to the upper left corner of the ROC curve (representing 100% sensitivity and 100% specificity), the better it predicts CNP. To assess the ability of the independent variables to predict CNP severity, linear regressions were applied with the dependent variable being the PRI of the MPQ.
Multiple testing problems were addressed with the false discovery rate procedure, 38 which controls the expected proportion of falsely rejected hypotheses, which is the desirable control, against errors originating from multiplicity. P-values