A single site parallel group, double blinded, sham randomised controlled trial (RCT) of external non-invasive peripheral electrical nerve stimulation (ENPENS), designed to assess the efficacy of ENPENS versus sham in patients with chronic pain following peripheral nerve injury. Patients were randomised to receive either active or control treatment and continued treatment for a period of 3 months (main treatment phase). As a further retention and recruitment aid, following completion of the main treatment phase patients were offered an optional cross over or treatment extension (3 months).
Screened subjects who met the inclusion/exclusion criteria either before or after completion of the extension/swap period were invited to participate in a further open label mechanistic study assessing psychophysical parameters pre and post low frequency stimulation to validate LTD as a working mechanism within a clinical sample.
Suitable patients appearing to fulfil inclusion/exclusion criteria identified from the centre’s pain clinics received a pre-screening telephone assessment. Patients then attended a screening appointment where written informed consent was obtained by the study PI. Separate written consent was obtained for the mechanistic study.
Patients were eligible if they were aged 18 or older and had definite or probable pain post nerve injury of ≥ 12 months duration . They experienced moderate to severe pain intensity (defined as an average of ≥ 5/10 on an 11-point (0–10) numerical rating scale (NRS) recorded daily over 1 week, but not dropping below 4 on any given day), in a localised area (distribution of one to two peripheral nerves to facilitate easily replicable independent stimulation), had brush stroke allodynia in that area (≥ 3/10 NRS) (prioritised by patients as an important clinical outcome) and had trialled first line pharmacotherapy (to ensure patients care was not disadvantaged via inclusion).
Patients were excluded if they had absolute numbness in the affected area, had known treatment contraindications, implantable devices for the same condition, unstable pain intensity of pain medications in the 6 weeks prior to the trial, had diagnosed psychiatric or mental health disorder or other health conditions/pain which in the opinion of the investigators would make the trial unsuitable, or were unable to comply with the study protocol. Additionally, patients were required to stop any medications that numb affected areas prior to the study to enable stimulation of the peripheral nerves. Lidocaine patches 2 weeks prior, Capsaicin treatments (both low-, and high concentration) 4 months prior to stimulation to allow nerve endings to grow back. Patients were requested not to commence any new medications/ treatments that may confuse evaluation of treatment efficacy. All prescribed and non-prescribed treatments were recorded throughout.
The criteria were the same for both studies with the exception that brushstroke allodynia (≥ 3/10 NRS) was not required for the mechanistic study. This study sought to demonstrate a reduction in mechanical hypersensitivity and although brushstroke allodynia is often a feature of this (exhibited by 15/19 patients) significant mechanical hypersensitivity can also exist without the presence of documented brushstroke allodynia .
Once consented patients were randomised to either active or sham treatment. The study trial manager or PI randomised patients using an independent online randomisation service that employed a concealed 1:1 allocation schedule and varying block sizes of 2 and 4. Trial nurses and patients were blinded to treatment allocation and were informed that that the purpose of the trial was to compare two types of stimulation ‘Pen’ and ‘Flat’. They were further informed that efficacy was not related to strength of stimulus but rather determined by the electrical field (see active and sham device). Following randomisation assignment, trial nurses were provided with the appropriate stimulation device to issue to the patient. Randomisation was not employed for the mechanistic study.
Setting of study
The study was conducted at a supra-regional UK national health services (NHS) neurology and neurosurgery hospital. The study was registered on the ISRCTN registry, registry number: ISRCTN53432663. The full protocol was published before initiation of the trial . The trial was conducted in accordance with the original protocol. Data collected ended when all patients had completed the optional treatment extension/swap and was predetermined before commencement of the trial.
The RCT primary objective was to establish whether ENPENS treatment versus sham treatment was effective in reducing pain for people with long-standing neuropathic pain following peripheral nerve injury, as measured by change in pain intensity following 3 months of treatment. Secondary objectives were to assess the benefits associated with treatment to other commonly affected areas such as quality of life, function, mood, self-efficacy (confidence to perform abilities in the presence of pain), reduction of allodynia in the areas of pain, and symptom report. The primary objective for the mechanistic open label study was to establish whether LFS was associated in reduction of enhanced pain-responsiveness in a clinical population, as measured by change in measured sensory features of clinically enhanced pain responsiveness. We hypothesised that within the RCT a significant difference (p = < 0.05) in terms pain reduction, would be observed between groups in favour of the active treatment. Whilst for the open label mechanistic study a significant reduction (p = < 0.05) in mechanical pain sensitivity and dynamic mechanical allodynia would be observed following low-frequency stimulation.
Sample size calculations were conservatively based on detection of a between-group difference of 1.5 following previous observational study data that had shown a mean treatment associated pain reduction of 2.8 units. The standard deviation of the outcome was assumed to be as per this data, 1.9 units . A correlation of 0∙5 between the baseline and outcome pain scores was assumed (0.64 in observational study data). Therefore, based on a 5% significance level, 90% power, and assumed 30% attrition rate, it was calculated that 38 participants per group were required to show a difference of 1.5 units in the primary outcome between groups, further details are provided within the published protocol .
Active and sham interventions
Low frequency nerve stimulation through a transcutaneous-applied, small spherical electrode that induces high current density is a long-established method to localize peripheral nerves for nerve blocks . Active treatment was referred to as the ‘Pen device’ and utilised a transcutaneous peripheral nerve stimulation device with a pen shaped electrode (Xavant stimpod nms460, Pretoria, South Africa) with pre-set parameters of 2 Hz (Hz) and 1.0 ms (ms), and an adjustable stimulation strength of ≤ 30 milliamps (mA). The sham device was referred to as ‘the flat device’ and looked identical but used a flat 5 cm2 square adhesive electrode and parameters of 2 Hz, 0.1 ms, ≤ 6 mA (although appeared to allow 30 mA). The electrode and parameter combination created a perceivable, low current density not eliciting LTD . Electrodes stimulated affected nerves, proximal to the focal area of pain and just outside of the identified area of allodynia. LTD requires delivery of ≥ 1200 pulses, requiring 10 min of treatment [5, 10]. To avoid unblinding due to felt or observed differences related to current density or LTD effect within training, the amplitude and stimulation time was limited (< 5ma and < 5 min) for both the active and sham devices. An independent physiotherapist with experience of stimulation began training by determining the point of stimulation and beginning stimulation for all patients but had no further patient contact- this was safeguard accurate nerve identification for stimulation and to maintain blinding of trial nurses.
Once able to demonstrate independent use of the device, patients were loaned a stimulator for 3 months, stimulating for a minimum of 10 min daily, at a mildly painful but not intolerable amplitude. Patients determined the amplitude, frequency, and timings of stimulation. Weekly telephone calls during the main treatment phase recorded treatment compliance and health care utilization.
Mechanistic study experimental paradigm
At baseline, the area of mechanical hyperalgesia was mapped using a pinprick stimulus of 256 mN along eight equally spaced tracks originating from the epicentre of pain. Then, a circular intervention array of 10 punctate electrodes (each 250 µm diameter), designed to preferentially activate small diameter epidermal nerve fibres (5), was placed over the epicentre of the pain. This electrode delivers electrical pulses using a constant current stimulator (Digitimer DS7A, UK). Baseline QST was then performed on areas directly adjacent the intervention array-electrode to obtain a comprehensive somatosensory profile, using the standardised German Research Network quantitative sensory testing (DFNS) protocol . As part of this protocol, mechanical pain sensitivity (MPS) was assessed as the mean of the pain ratings in response to a geometric series of 8–512 mN calibrated pinpricks at factor 2 progression, and dynamic mechanical allodynia (DMA) as pain to gentle stroking touch (cotton wisp, QT-tip, soft brush); these MPS/DMA series were scheduled at the end of the QST procedure and were spaced by approximately 5 min between each. Then, electrical detection threshold was determined with single 2 ms duration electrical pulses of increasing strength through the array electrode using method of limits (felt/not felt). The subsequent LFS intervention consisted of a train of 2000 electrical, 2 ms duration, stimuli delivered through the array electrode at 1 Hz. Stimulation strength was initially 10 × EDT but, due to poor tolerability was reduced to 5 × EDT for the final 6 study patients. Directly after LFS stimulation, the hyperalgesic area was measured again and then MPS and DMA were determined as before.
RCT outcomes were completed by patients and collected by trial nurses. Outcome questionnaires were scored, and all data entered onto a computer by a technical assistant independent to the trial. Data analysis was conducted by the unblinded statistician after data lock.
The RCT primary study endpoint was the average 24 h pain intensity recorded daily on an 11-point (0–10, 0 = no pain & 10 = worst pain imaginable) numerical rating scale (NRS), averaged over the last 7 days of the three-month treatment phase. At least 1 daily score was required.
Mechanistic study outcomes
The co-primary outcomes were change in intensity of mechanical pain sensitivity and dynamic mechanical allodynia following LFS, normalised to baseline. The secondary outcome was the change in mechanical hyperalgesia area following LFS. Spontaneous post-test pain was not measured as this parameter was considered confounded by the lengthy examination protocol.
RCT secondary outcomes
Text within brackets indicates what each measure was intended to capture.
Brief pain inventory interference subscale (BPI-I) (functional interference) .
EuroQol EQ-5D-5L generic measure of health status. The EQ-5D-5L has two components, a summary index (utility), and the EQ visual analogue scale (EQ-VAS), (health related quality of life) .
RCT exploratory outcomes
Hospital anxiety and depression scale (emotional function) .
Pain self-efficacy questionnaire (perceived confidence to function despite pain) .
Worst pain using BPI (range of pain intensity) .
Dynamic mechanical allodynia determined by manual mapping (change in surface area of allodynia) .
Neuropathic pain symptom inventory (NPSI) (quality of pain) .
Outcomes were recorded at baseline, treatment completion and on completion of the optional treatment extension/swap. Secondary and exploratory endpoints were scores following end of 3-month treatment phase. Treatment phase patient diaries captured daily pain intensities and treatment frequencies.
On study completion
Patient perceived global impression of change (PGIC) .
Telephone interview—qualitative exploration of ‘active’ treatment experience in a proportion of patients.
Perception regarding treatment allocation. Patients were asked if they felt they had been assigned to a more or less effective stimulation.
At every patient contact, safety adverse events (SAEs) and serious adverse events (SARs) were recorded.
RCT statistical analysis
The primary study analysis was intention to treat (ITT) based on all randomized, eligible patients with outcome measures available at the end of the study. For the primary outcome, the primary endpoint was compared between groups by Analysis of Covariance (ANCOVA) using baseline scores as the covariate. In a secondary analysis of the primary outcome (responder analysis) the proportion of patients in each arm that achieved distinct outcomes (≥ 2 points NRS, ≥ 30% and ≥ 50%) were computed. The minimally clinically important difference (MCID) was defined as ≥ 2 points . No stratification variables were included in the primary analyses.
Equivalent methods were used for those secondary and exploratory outcomes measured on continuous scales. For analysis of ordinal outcomes, the Mann–Whitney test was used. MCID was calculated for all further outcomes as follows based on the available literature: BPI-I 2-point reduction, EQ-5D-5L VAS 11 points increase, utility 0.145 increase, HADS 4 points reduction plus movement between severity categories, PSEQ increase of 7 + points plus movement between severity categories, BPI worst pain 3-point decrease, 20% reduction in surface area of allodynia [25, 29,30,31,32]. NPSI was not included due to an absence of appropriate literature. Sensitivity analyses were performed for both the primary outcome, and for the secondary outcomes; multiple imputations (MI) were used to address the missing values, utilising approaches based on the multivariate normal distribution method . Fisher’s exact test was additionally reported as part of secondary analysis to illustrate any association between groups in relation to outcome measures. For all outcomes statistical tests were two-sided, with p-values of p < 0.05 considered statistically significant. All analysis except for post-hoc analysis was prespecified and included as part of the published protocol prior to initiation of the trial. All data was analysed using Stata version 15.1 (Statacorp.2017) statistical software. Adverse events were summarised descriptively.
Health economic analysis
Details of the healthcare resource use and cost analysis are described in the published protocol . A pre-planned health economic analysis to estimate NHS perspective cost-effectiveness, will be reported separately.
Mechanistic study statistical analysis
QST values (excluding paradoxical heat sensations and dynamic mechanical allodynia) were z-transformed using the mean and standard deviation (SD) of the healthy control normative data for age, gender, and body location . Z-scores above zero indicate gain of function and below zero indicate loss of function. QST profiles were compared to normative healthy subject German Research Network quantitative sensory testing (DFNS) protocol data with mean = 0 and SD = 1, using non-paired t-tests . Pain rating data and mechanical hyperalgesia area were Log10-transformed to obtain secondary normal distribution. For all NRS ratings a constant of 0.1 was added to avoid loss of any zero ratings . Raw data were smoothed by 3-point averaging to reduce irregularity caused by swings of single rating values. Results are shown as mean and SEM of log10-transformed data. Data were analysed using Friedman test. All calculations were performed with SPSS 20 (IBMTM) and Excel 2010 (MicrosoftTM).