Abstract
Background:
Explicit Sensory Retraining (SR) has been suggested for pain management in several neuropathic pain syndromes.
Objectives:
To study the effectiveness of SR on Central Post Stroke Pain (CPSP) symptoms.
Methods:
A preplanned subgroup of 23 subjects post-stroke reporting central pain were randomized within a larger study, to receive either explicit sensory discrimination retraining (SR) (in subgroup n = 11) or implicit repeated exposure to stimuli (RE) (in subgroup n = 12), applied to the lower limb. Pain intensity measured by VAS, measures of sensory ability; balance and gait activities; and participation were assessed by a blinded assessor at four points in time.
Results:
A group*time interaction effect was demonstrated (p = 0.04) for VAS, as for the SR treatment group VAS decreased from 56/100 to 23/100. Nine out of eleven subjects in the SR group demonstrated positive response to treatment. Pain alleviation was maintained at follow-up. A similar interaction effect was found for tactile threshold (p = 0.03). Significant improvement was noted for all other variables with no group differences.
Conclusion:
This study provides preliminary evidence to the clinically relevant positive effect of explicit sensory retraining for subjects with CPSP. The positive effect on tactile threshold detection could potentially indicate pain alleviation mechanisms.
Keywords
Introduction
Central Post Stroke Pain (CPSP) is a neuropathic pain syndrome, characterized by pain in parts or in the entire contralesional hemibody post stroke, not explainable by other diagnosis or pathology (Finnerup et al., 2016; Jensen & Finnerup, 2014). Central pain is highly correlated with sensory impairment post stroke, such as hypoesthesia and reduced ability to discriminate sensory input (Finnerup et al., 2016; Treede et al., 2008), and therefore is described as paradoxical pain (“I can’t feel my leg, but it hurts”). The pain, which is frequently of high intensity, may appear as numbness, burning, cold, needling or aching (Jensen & Finnerup, 2014). CPSP’s prevalence is 11% of stroke survivors (Liampas et al., 2020) and up to 18–20% of post-stroke population with sensory impairments (Klit et al., 2009; Krause et al., 2016). Suggested treatment includes several pharmacological options and invasive/noninvasive brain stimulation techniques. However, no physical therapy treatment has been proven effective to date (Boldt et al., 2014; Flaster et al., 2013; Harrison & Field, 2015; Jensen & Finnerup, 2014; Klit et al., 2009; Liampas et al., 2020; Mulla et al., 2015; Treister et al., 2017).
Already widely used to treat chronic pain of different etiologies, sensory stimulation treatment through sensory repeated exposure to a variety of stimuli (e.g. manual therapies, electrotherapy, thermotherapy etc.) has been suggested for use in neuropathic pain syndromes, relying on implicit learning and bottom-up induced plasticity (Cakici et al., 2016; Cruccu et al., 2007; David et al., 2015; Gibson et al., 2017; Gok Metin et al., 2017; Vance et al., 2014). Transcutaneous Electrical Nerve Stimulation treatment has been tested specifically for individuals with CPSP and failed, as it alleviated the pain for only 3 out of 15 and increased the pain of 5 out of 15 subjects (Leijon & Boivie, 1989).
An alternative form of sensory focused treatment is explicit sensory discrimination retraining (SR). SR is a form of perceptual learning treatment, explicitly directing the attention and awareness of the individual to the stimulus and its features. Different studies have used various explicit sensory retraining protocols for pain management or for sensory rehabilitation. Examples for trained sensory modalities in these studies include discrimination of textures (Carey, 2012; Schmid et al., 2017), two point discrimination (Moseley et al., 2008), electric current features recognition (Flor et al., 2001) and proprioceptive joint angle perception (Carey, 2012).
Moseley et al. (2008, 2012) suggest that attentive SR, rather than repeated sensory stimulation alone, should be implemented in order to more effectively treat chronic neuropathic pain (Moseley & Flor, 2012; Moseley et al., 2008). This statement is based on studies in which SR was successfully used to reduce post amputation phantom limb pain (Flor et al.) and Complex Regional Pain Syndrome (Moseley et al., 2008; Schmid et al., 2017). The suggested mechanism for the alleviation of pain relies on the assumption that disrupted cortical somatosensory representation is involved in the mechanism of chronic pain and that sensory discrimination training allows for cortical reorganization of the relevant body part somatosensory representation area (Flor et al.; Moseley & Flor, 2012). While this assumption has been recently questioned (Mancini et al., 2018), the evidence as to the effect of discrimination training on pain is quite compelling. Acerra et al. (2007) suggest that since commonality of symptoms and neuro-pathophysiology exist between Complex Regional Pain Syndrome, phantom limb pain and CPSP, a common treatment strategy should be investigated.
Purpose
Within a larger study of sensory retraining (Ofek et al., 2022), the aim of this subgroup analysis was to evaluate and compare immediate and long term (at three month follow-up) effectiveness of explicit learning sensory discrimination retraining therapy (SR) and of implicit learning repeated exposure to sensory stimuli treatment (RE) applied to the lower limb of chronic stroke subjects with sensory impairments and with Central Post Stroke Pain (CPSP). Both groups’ protocols included balance and mobility training, as well as TENS and non-task specific exercise. Effectiveness was measured on pain level, along with other measures of sensory performance, balance, gait, and participation.
Methods
Participants
Twenty three subjects post stroke with contralesional ongoing pain, fitting characteristics of CPSP (Finnerup et al., 2016; Klit et al., 2009), were enrolled as a subgroup within a larger sample (n = 64) of subjects post-stroke with sensory impairments participating in a sensory rehabilitation training study (Ofek et al., 2022). Pain occurrence and intensity were part of the planned protocol in order to create a pre-planned pain subgroup (Finnerup et al., 2016; Klit et al., 2009; Krause et al., 2016; Treede et al., 2008), as the correlations of sensory loss and CPSP are well known. Full trial protocol is available (Ofek, 2018; Ofek et al., 2022). CPSP characteristics for inclusion in the pain subgroup were: (a) contralesional extensive ongoing pain at the lower extremity; (b) non-specific pain location; (c) pain within sensory impaired location as determined by sensory tests, at the screening assessment session; (d) no orthopedic or other diagnosis to explain pain origin. General inclusion criteria for the sensory retraining study were: (e) at least six months post infarct or hemorrhagic stroke; (f) ability to walk at least six meters with or without aid and/or personal assistance of one person; and (g) adequate comprehension of instructions and ability to respond yes/ no verbally or non-verbally. Exclusion criteria were: (a) other central neurologic condition (e.g. Alzheimer’s disease); (b) evidence of peripheral neuropathy unrelated to the stroke (determined by bi-lateral failure to detect a standard 5.07 Semmes Weinstein monofilaments (Plucknette et al., 2012)); (c) use of pacemaker (a contraindication to electrical stimulation); (d) hemispatial neglect as determined by positive screening star cancellation test (Friedman, 1992); and (e) participation in another clinical trial at the same time. The Helsinki Ethics Committee of Clalit Health Services approved the study. All study subjects provided signed informed consent prior to participation in the screening assessment. The study was registered on ClinicalTrials.gov (identifier: NCT01988220).
Experimental design
A parallel design RCT was conducted: eligible subjects were randomly assigned to the Sensory Retraining (SR) intervention group or to the Repeated Exposure (RE) intervention group; with a 1 : 1 allocation ratio using computer generated randomization and concealed envelopes. Randomization and assigning of participants to intervention groups were conducted by a research assistant. The primary investigator, performing the assessments, was blinded to the treatment allocation. The randomization was carried out within the larger study (screening n = 82; randomized n = 64). The assessor performed two pre-treatment assessments: screening (a week ahead, providing a baseline) and pre (right before treatment commencement), and two post-treatment assessments: immediately post treatment and at three month follow-up. The participants were not made aware of the differences between the two treatment groups. Furthermore, pain management was not mentioned as a goal to either the treating therapists or to the patients, as the explicit goal of treatment was sensation, balance, and mobility improvement. All treating therapists were physical therapists specializing in neuro-rehabilitation, who underwent training in performing both intervention protocols.
Using a checklist, referred patients were screened for compliance with the inclusion and exclusion criteria. Screening assessment sessions included an interview for demographic data and sensory experience (hypoesthesia, dysesthesia, aggravating and alleviating factors etc.); assessment of intensity of pain as measured by Visual Analogue Scale (VAS); assessment of tactile and proprioceptive threshold and discrimination; and Timed Up and Go test for mobility assessment. Pre-post treatment and the three month follow-up assessment sessions included repeat of pain intensity evaluation along with other sensory tests. The pre-post and follow-up assessment sessions also included assessment of activities and participation variables, as well as the McGill questionnaire and open question narrative gathering of sensation, pain, and function experiences.
The intervention included ten 45-minute sessions, carried out once or twice a week for up to 12 weeks, by the treating therapists. Assessments and treatment sessions were carried out in outpatient Physical Therapy clinics, belonging to Clalit Health Services, Haifa and Western Galilee region. Figure 1 presents the overall outline for the RCT.
Study allocation flow.
Outcome measures were collected during all assessment sessions in an orderly manner, with sensory assessment first, activities (balance, gait) second and activity and participation questionnaires last.
Pain assessment: For pain assessment in this subgroup, the additional primary outcome measure was pain intensity as measured by a VAS: participants were asked to grade their pain on a line scale (happy face to sad face with a verbal anchor –the happy face stands for “no pain at all” and the sad face for “the strongest pain you can imagine”). The assessor noted the score on a 0–100 numeric scale that was hidden from the subject (Price et al., 1983; Williamson, 2005). The McGill questionnaire was filled out to capture a broader perspective on pain behavior. Unfortunately, only half of the already small subject sample was able to complete the questionnaire, due to comprehension difficulty (aphasia and/or Hebrew not their mother tongue). Therefore, statistical analysis was not conducted on this measure.
Sensory function outcome measures: 1. Tactile detection threshold, as measured by Semmes-Weinstein monofilaments (SWM) (Bell Krotoski & Tomancik, 1987). A mean was calculated for two measurement locations on the dorsal foot: at mid-ankle and above the head of the first metatarsus; 2. Trans-cutaneous Electrical Nerve Stimulation (TENS) detection threshold, as described by Eek et al, 2003 (Eek & Engardt, 2003); 3. Texture discrimination ability at the sole of the foot, as assessed by the Lower Extremity Texture Discrimination Test (LETDT) (Ofek et al., 2017); and 4. Proprioception performance as assessed by the Lower Extremity Position Test (LEPT): a proprioception assessment tool, quantifying combined ankle-knee movement reproduction error (Ofek et al., 2019).
Activities (gait, mobility, balance, ADL) and participation (confidence in mobility and social participation) outcome measures: 1. Mobility was assessed by the two minute walk test (Rossier & Wade, 2001); 2. Gait and balance were assessed by miniBEStest and its Timed up and Go (TUG) subtests (single and dual task) (Franchignoni et al., 2010; Godi et al., 2013); 3. Activities of Daily Living (ADL) independence was measured with the Stroke Impact Scale (SIS) questionnaire ADL section (Lai et al., 2003); 4. Social participation was measured with the SIS participation section (Lai et al., 2003); and 5. Confidence in mobility was measured by the Activities-specific Balance Confidence (ABC) scale (Powell & Myers, 1995).
Treatment intervention
Two distinctive protocols of sensory rehabilitation treatment were conducted. Both the SR and RE treatments included three 15-minute subsections, each addressing a sensory modality group: 1. Superficial sensory input (tactile including texture and shape, temperature, pressure); 2. Deep sensory input (proprioception including limb position and kinesthesia, weight); and 3. TENS. Training of both groups was conducted in different positions (lying, sitting, standing, walking, and stairs). The SR protocol was based on explicit perceptual learning exercises, including tactile discrimination training (for example: “Where am I touching you now?” “Is this the wet towel or the dry one?”), proprioceptive discrimination training (for example: “Tell me when I’m moving your ankle.” “Is this the high step or the low one?”) and discrimination of TENS input (for example: “Is the stimulation now stronger or weaker?” “Do you feel tingling at your foot or knee?”). SR used vision input to the uninvolved lower extremity, auditory cues, etc. to enhance sensory learning through anticipation and calibration (Carey, 2012). The RE protocol included implicit augmented sensory input (for example: standing and walking barefoot on turf, balancing on a wobbling board, receiving usual care TENS treatment).
Statistical analysis
Demographic data was analyzed using chi-square tests or t-tests, as appropriate. Intention to treat pre-post analysis, using imputation for pre and post results, included two-way (2X2) ANOVAs mixed design. A two-way ANOVA (2X3) without imputation was conducted for the pre-post-follow-up analysis (due to additional fall-out). Studentized Maximum Modulus (SMM) post hoc tests were used to indicate significant time and interaction effects. Simple Mean Analysis (SMA) was used to reveal significance of time within group. Level of significance was p≤0.05. Statistical analyses were performed using SAS 9.4 for Windows. In order to evaluate the treatment’s effect on pain, subjects were divided into 4 groups: with negative response to treatment, with no response, with response > 10%, and with high response > 30% (Busse et al., 2015). Absolute Risk Reduction (ARR) and Number Needed to Treat (NNT) were calculated for high response only, as this is a minimal clinically important change.
Results
Twenty-three subjects with CPSP were included in this pain subgroup, out of 64 subjects with sensory deficits post-stroke (36%) in the larger RCT. Imaging data was available for only half the subject sample: ischemic stroke (n = 8); hemorrhagic stroke (n = 4); unknown (n = 11). Stroke location: cortical lesion (n = 2); subcortical lesion (n = 7); mixed lesion (n = 3); unknown (n = 11). Table 1 presents demographic characteristics of participants in both treatment groups.
Demographic characteristics
Demographic characteristics
Pain level was measured by VAS, as the primary outcome measure for the CPSP subgroup. Table 2 demonstrates pain levels as measured by VAS, by intervention group, and across time. There was no difference between groups for pain level at baseline (p = 0.64). Baseline pain level was maintained between screening test pain levels (mean 57.9±SD 23.4) and pre-test pain levels (mean 53.0±SD 21.7), as a two-tailed paired t-test demonstrated no significant change (p = 0.86) for the entire group or for each group independently (SR p = 0.96; RE p = 0.56).
Pain levels as measured by VAS, mean±standard deviation (min-max), by group, across
assessments
Pain levels as measured by VAS, mean±standard deviation (min-max), by group, across assessments
VAS=Visual Analogue Scale for pain of contralesional lower extremity.
As described in the methods section, a two-stage analysis was performed: 1. Pre-post comparison using intention to treat analysis, and 2. Pre-post-follow-up comparison without imputations for missing values. When only pre-post scores are addressed, a mixed design two-way ANOVA found no significant group effect (p = 0.58), a trend for time effect (p = 0.09), and significant group*time interaction (p = 0.04), as demonstrated in Fig. 2. The SMM post hoc test indicated a significant difference for the SR group between times (p = 0.01), as the mean score decreased from (56.2±21.2) pre-treatment to (24.0±33.3) post-treatment for the SR treatment group, with no significant difference for the RE group, as shown in Fig. 2.
Changes in pain level, as measured by VAS, by group, for time *p < 0.05.
When analyzing pre, post, and follow-up assessment points, a mixed design two-way ANOVA found no significant group or time effects. A trend towards group*time interaction was demonstrated (p = 0.1). A SMM post hoc showed improvement for the SR group (p = 0.01 for time pre-post; p = 0.04 for pre-followup), as the mean VAS decreased from (56.2±21.2) pre-treatment to (24.0±33.3) post-treatment with no significant change between post-treatment and follow-up (follow-up mean VAS (27.2±26.8). No significant changes in VAS were demonstrated for the RE group, as shown in Table 3 and in Fig. 2.
Changes of pain level across time (p-value) (pre- n = 23, post- n = 19, follow-up n = 15), by group at pre-post-followup
*Denotes statistical significance p≤0.05. **denotes statistical significance p≤0.01. FU = follow-up, VAS = Visual Analogue Scale for pain of contralesional lower extremity.
Distribution to responding effect groups is demonstrated in Table 4: highly positive effect of more than 30% alleviation of VAS score, positive effect of 10%, no response if no alleviation or augmentation was reported within 10%, and negative response when augmentation of pain was reported by more than 10%. For four subjects of the SR group and for two of the RE group, pain that had endured for months and years was completely resolved. If only highly positive effect is counted (due to minimal clinical importance difference data), then calculated ARR is –0.39 and the number needed to treat with sensory retraining is three.
Distribution to responding effect groups, per treatment group
Table 5 demonstrates outcomes of sensory, activities, and participation performance measures as collected at pre-treatment, post-treatment and at follow-up. No significant changes were found between the two pre-treatment assessments.
Sensation, activities, and participation variables, mean±standard deviation, by
group, across assessments
Sensation, activities, and participation variables, mean±standard deviation, by group, across assessments
Abbreviations: SWM = Semmes Weinstein Monofilaments, TENS = Transcutaneous Electrical Nerve Stimulation, LETDT = Lower Extremity Texture Discrimination Test, LEPT-22 = Lower Extremity Position Test, ABC = Activities-specific Balance confidence Scale, miniBEST = mini Balance Evaluation Systems Test, TUG = Timed Up and Go test, SIS = Stroke Impact Scale, ADL = Activities of Daily Living, part = participation.
A mixed design two-way ANOVA found no significant group effect, a significant time effect (p = 0.001), and a significant group*time interaction effect (p = 0.03) for tactile detection as measured by Semmes Weinstein Monofilaments (SWM), when only pre-post scores were addressed. SMM post hoc test indicated a significant difference for the SR group between times (p = 0.0003), as the mean detection threshold score increased from (4.82±0.53) pre-treatment to (4.34±0.72) post-treatment for the SR treatment group, with no significant difference for the RE group, as shown in Fig. 3. When pre-post-follow-up analysis was made, no such interaction was found.
Changes in tactile detection threshold of the contralesional foot, as measured by SWM, by group, for pre and post treatment assessments *p < 0.05.
ANOVA found no significant group effect, a significant time effect (p = 0.02), and no significant group*time interaction for Lower Extremity Position Test (LEPT) of the involved foot. No changes were found for TENS detection threshold or for texture discrimination.
Similar improvements for both groups were determined (no group effect, a significant time effect, and no interaction) for all activities and participation outcome measures (apart from SIS-participation, where no significant improvement was seen for both groups).
This study demonstrates for the first time that explicit sensory retraining can alleviate the level of pain in CPSP. The pain of the SR group lessened by more than 30 mm on the VAS, which exceeds the minimal clinically important difference or minimal detectable change determined for painful conditions (Busse et al., 2015; Lee et al., 2003; Salaffi et al., 2004). In addition, this therapeutic effect was maintained at the three month follow-up. These findings are particularly important, as current therapeutic options for relief of CPSP are either pharmacological or involve expensive equipment (such as transcranial magnetic stimulation) (Harrison & Field, 2015; Klit et al., 2009), and since our treatment protocol is simple, cheap and applicable at any clinic worldwide.
In keeping with the findings of other studies suggesting the effectiveness of explicit sensory retraining for CRPS and amputees (Acerra et al., 2007; Flor et al.; Moseley & Flor, 2012; Moseley et al., 2008; Pleger et al., 2005; Schmid et al., 2017), the SR explicit training was effective on average, and for most allocated subjects, whereas the RE implicit protocol was not. In other words, attention demanding, sensory discriminative, recognition training generated a change which repeated exposure to sensory stimuli failed to make. Both groups received balance and mobility weight-bearing task training and TENS, so motor learning effect and implicit TENS effect can be excluded as confounding variables here. Our results suggest that in order to alleviate pain of central origin, explicit training, involving elements of attention, anticipation, prior expectation and recognition, needs to be designed in addition to the bottom-up input through sensory stimulus and implicit sensory and sensorimotor experience, that others have suggested (Heeger, 2017; Moseley & Flor, 2012).
Based on the literature addressing other neuropathic pain syndromes (Acerra et al., 2007; Flor et al.; Moseley & Flor, 2012; Moseley et al., 2008; Pleger et al., 2005; Schmid et al., 2017), we hypothesized that the explicit sensory retraining group would demonstrate alleviated pain and better sensory performance (SWM, TENS thresholds, texture discrimination, and proprioception) post treatment, when compared to the implicit treatment group. Our results demonstrate some advantage to sensory performance improvement in the SR group compared with the RE group. Significant interaction effect was found for improvement in tactile detection threshold (SWM) post treatment for the SR group only. There was proprioceptive improvement for both groups, while changes in TENS detection threshold and in tactile discrimination ability were not registered for either. This may be due to the small sample size. Another explanation to consider would be that pre-treatment allodynia of some of the participants was resolved during treatment. For example, although sensation abnormality was lessened, TENS threshold was not lower on average, since for some subjects, allodynia was resolved and they became less sensitive. No major difference between groups was found in activities and participation improvement, as both have improved for those variables. Pain may not be the major burden for this population in terms of mobility, and the alleviation of pain is not sufficient for additional fundamental changes in activities of participation.
In addition, this study raises questions about the prevalence of CPSP in the post-stroke population. The study group –individuals post-stroke reporting pain that matches the characteristics of CPSP –was recruited as a preplanned subgroup, from an initial study population that included participants post-stroke with sensory loss (n = 64). Our review of the literature led us to expect CPSP in up to 18% of the sensory impaired population post-stroke (Klit et al., 2009), yet we found 36% who fit the characteristics of CPSP (n = 23). It is possible that this high percentage of CPSP is due to the very chronic stage of our population group, as CPSP can emerge with months’ latency post-stroke. In addition, possibly, our focused sensory testing, including detection and discrimination components (Kim & Choi-Kwon, 1996), identified a different population group than that used in previous studies. The high percentage of CPSP found here provides further evidence for the known correlation between sensory impairment post stroke and the incidence of CPSP (Jensen & Finnerup, 2014; Klit et al., 2009).
This study is limited by its small sample size, and the fact that the study was not initially and deliberately designed for individuals with CPSP. We were unsuccessful in gathering imaging data (ischemic or hemorrhagic stroke and stroke location) for half the population. In addition, sensory performance outcome measures do not include temperature spinothalamic performance measures, known to be in correlation with central pain. Therefore, further research should include a larger sample and specifically target individuals post-stroke with central pain. We suggest a trial of SR treatment in the early stages post-stroke, as a method for the prevention of CPSP, and the repeat of this study with other central pain populations (incomplete SCI, MS, Parkinson’s, TBI etc.).
Conclusion
This study suggests that explicit sensory retraining can be an effective therapeutic tool in the treatment of CPSP. Explicit sensory discrimination training, and not sensory exposure alone, should be further studied to determine effect in physical therapy for alleviation of pain in this population.
Conflict of interest
The authors report no declarations of interest.