Clinical effectiveness of subsensory sacral neuromodulation in adults with faecal incontinence: the SUBSoNIC crossover RCT and mechanistic study

Trial design Randomised, multicentre, double-blind crossover trial (with 2 × 16-week periods) of active neurostimulation versus sham stimulation with subsequent open-label follow-up to 58 weeks. Embedded mechanistic sub-study using magnetoencephalography to study bidirectional functional connectivity between brain and anorectum. Methods Participants: adults aged 18–80 years, with chronic symptoms of faecal incontinence refractory to first-line treatments (and meeting national criteria for sacral neuromodulation). Interventions: active: chronic, subsensory (low amplitude) stimulation of a mixed sacral nerve (usually S3) using a commercially available surgically implanted pulse generator; sham: identical implant but turned off (or to 0.05 V). Patient-chosen sub- or supra-sensory open-label stimulation from week 32 to week 58. Primary objectives: (1) to determine whether sub-sensory sacral neuromodulation led to a reduction in total faecal incontinence episodes per week compared to sham stimulation; (2) to identify whether clinical responses to sub-sensory sacral neuromodulation were biologically related to changes in evoked and induced activity between the brain and anorectum. Primary outcome: total faecal incontinence episodes per week based on paper bowel diary performed in the final 4 weeks of each crossover period (allowing 12-week washout). Randomised allocation (1 : 1) to arm 1 (sacral neuromodulation/sham) or arm 2 (sham/sacral neuromodulation) at time of surgery was stratified by sex and centre. Blinding: participants, surgeons and assessors; tamper-proof tape masked stimulation settings. Statistical methods: Poisson regression models failed to converge for the count outcomes, hence paired t-tests were used, and treatment effects summarised by mean differences [with 95% confidence intervals (CIs)]. Sample size: a total of 90 patients (45 per group) were required to detect a 30% reduction in episodes, allowing for 10% loss to follow-up (alpha = 0.05; power 90%). Results Recruitment: a total of 39 patients of 220 screened and 65 pre-enrolled (arm 1: N = 17; arm 2: N = 22) were recruited to the crossover trial at nine sites from the United Kingdom and one site from Ireland between February 2018 and July 2022, of whom only 16 (arm 1: N = 9; arm 2: N = 7) had complete primary outcome data. Nineteen completed follow-up to 58 weeks. Trial delivery was severely affected and terminated early due to COVID-19. Main barriers were the inability to continue face-to-face patient visits, redeployment of research staff to COVID-19 facing clinical roles and cancelling of sacral neuromodulation procedures due to lack of priority for non-urgent surgery. A total of 25 patients underwent magnetoencephalography studies compared to 20 healthy volunteers. Primary outcome (N = 16): sacral neuromodulation conferred a non-significant reduction in mean faecal incontinence episodes per week compared to sham (−0.7, 95% CI −1.5 to 0.0; p = 0.06). Secondary outcomes: in participants who also used the e-event recorder to record the number of faecal incontinence episodes in both periods (n = 7), estimate of effect size was greater but less precise (−1.5, −3.5 to +0.5; p = 0.12). Data suggested successful allocation concealment. Improvements were observed in faecal incontinence symptoms in the follow-up cohort (at 58 weeks) compared to baseline (approx. 3 fewer faecal incontinence episodes per week). A small number of expected adverse events all resolved. Magnetoencephalography studies demonstrated bidirectional afferent evoked cortical and efferent induced anal activity that did not vary greatly from control subjects (n = 20) and appeared unchanged by sacral neuromodulation. Conclusions Due to under-recruitment it is important to interpret the findings on the experimental efficacy of sacral neuromodulation as exploratory. Effects on symptoms observed during double-blinded crossover point to some efficacy over sham, though not large in comparison with placebo responses. The magnitude of effect was highly dependent on method and interpretation of event recording. Study registration Current Controlled Trials ISRCTN98760715. Funding This award was funded by the National Institute for Health and Care Research (NIHR) Efficacy and Mechanism Evaluation (EME) programme (NIHR award ref: 14/144/08) and is published in full in Efficacy and Mechanism Evaluation; Vol. 11, No. 19. See the NIHR Funding and Awards website for further award information. Plain language summary A treatment called sacral neuromodulation is commonly offered to adults experiencing bowel (faecal) incontinence. A battery powered unit is implanted into the lower back in the region of the sacrum (tailbone). This is connected to a specially developed lead with electrodes that rest on the nerves of the lower spine. This stimulator then continuously sends electrical impulses to the nerves and muscles that control the lower bowel (rectum and anus). The aim is to improve bowel control. Previous studies have reported a great benefit of sacral neuromodulation in some patients, but others have little or no response. The SUBsensory Sacral Neuromodulation for InContinence trial recruited 39 patients (of 90 intended) who met the current national criteria for sacral neuromodulation. It compared the effect on numbers of weekly faecal incontinence episodes with the device either on (active) or off (sham) using a special study design called a randomised crossover trial. All participants had the device on and off for 16 weeks in random order (crossing over in the middle). Using stimulation below the level that can be felt (subsensory), both the patients and the research team were unaware of whether the stimulator was on or off (called double blinding). Due to COVID-19, only 16 patients had complete data for analysis, which was much less than the intended number of 90. The results showed that patients experienced reductions in faecal incontinence episodes during both on and off periods (i.e. there was a strong placebo effect). However, slightly greater effects were seen during the on period suggesting a possible genuine biological effect of sacral neuromodulation. The study also showed that the way we record symptoms during research trials for example with paper bowel diaries needs improvement, as the bowel diaries were not fully completed by some participants. Although this is the first double-blind trial of its kind for sacral neuromodulation, all conclusions must bear in mind the poor recruitment and retention of patients. Scientific summary Background Faecal incontinence (FI), defined as the recurrent involuntary loss of faecal material leading to a social or hygienic problem, is a common and debilitating condition with profound effects on quality-of-life and high societal costs. Initial treatments including pharmacological and behavioural therapies (e.g. biofeedback) have variable outcomes and are poorly evidenced. Traditional surgical approaches focusing on anal sphincter reconstruction or augmentation are invasive, irreversible, and risk significant morbidity. A stoma is the final option. Chronic low-amplitude stimulation of the mixed sacral spinal nerves using an implanted electrode and pulse generator – sacral neuromodulation (SNM) is a less invasive alternative, now considered the first-line surgical treatment option for adults with FI in whom non-operative therapies have failed to alleviate symptoms. Current evidence for SNM is based on extensive observational data and few randomised trials that are heterogeneous in design and outcomes. Despite having widespread regulatory approval, SNM remains an expensive intervention with need for greater confidence in efficacy. A further concern regarding SNM therapy is the lack of evidence and understanding of the mechanism of any effect. Objectives Our primary aim was to determine the clinical efficacy of sub-sensory chronic low voltage electrical SNM using a commercially-available implantable device in adults with FI in whom conservative treatment has failed. We sought to determine whether SNM, compared to sham, led to a clinically important reduction in weekly FI episodes. The study also included mechanistic studies to examine whether clinical responses to sub-sensory SNM were biologically related to changes in the central pathway between the brain and anorectum. Methods Trial design SUBsensory Sacral Neuromodulation for InContinence (SUBSoNIC) was a multicentre, randomised double-blind crossover trial at nine UK sites and one site in Ireland in which SNM was compared to sham stimulation. We aimed to randomise 90 eligible participants (adults aged 18–80 years, where non-surgical approaches to National Institute for Health and Care Excellence (NICE) standard have failed and meeting minimum FI severity criterion) to two study arms after SNM implantation. Both arms had two intervention periods (ON-OFF or OFF-ON) of 16-week duration (T0–T16 and T16–T32). Efficacy outcomes were derived from assessments in the final 4 weeks of each cross-over period (T12–T16 and T28–T32) thus allowing for almost 3 months intervention before outcome assessments (and adequate washout for participants in the ON–OFF sequence). Mechanistic studies were performed in the final 2 weeks of the 4-week assessment periods in a subgroup of consecutively consenting participants from both arms until data saturation. After completing the crossover phase of the study, participants were followed up for a further 26 weeks. During this time, participants had either sub- or supra-sensory ‘open label’ stimulation based on preference as would have been normal for routine clinical practice. Further efficacy outcomes were recorded at T54–T58 to provide an indication of the short-term effectiveness of SNM within the rigor of a clinical trial unit (CTU)-monitored prospective study. Interventions Chronic low voltage stimulation of the third or fourth sacral root was achieved by surgical implantation of a commercially available Conformité Européenne-marked active implantable (class III) medical device [Medtronic InterStimTM (Medtronic, Minneapolis, MN, USA)] used in accord with the manufacturer’s instructions and local practice. For the active intervention (ON), the clinical team programmed the device using standard settings of a 14-Hz frequency and 210-µs pulse width. Optimal electrode configuration was determined by cumulatively increasing the amplitude of stimulation by 0.1 V from zero for each electrode until the sensory threshold was reached. The amplitude and site of stimulation were recorded for each electrode with the electrode configuration that achieved sensation in the anus or perineum at lowest amplitude being chosen for chronic stimulation. Sub-sensory chronic stimulation was initiated by reducing the amplitude to a level just below the habituated sensory threshold (for blinding). For the sham intervention (OFF), sensory thresholds were recorded identically; however, the level was then adjusted to zero volts or 0.05 V (the latter was required in some participants due to the new device handset limitations). Mechanistic studies were undertaken at the Institute of Health and Neurodevelopment (IHN) at Aston University in a subgroup of patients identified in the Midlands region (compared to 20 healthy volunteers without FI). A protocol including spatial registration (magnetic resonance imaging head) and a series of magnetoencephalographic (MEG) acquisitions measured induced and evoked cortical activity relevant to determining functional connectivity between the anus and brain (using anal electrical stimulation) and brain and anorectum (using volitional anal squeeze). Control paradigms (tibial nerve stimulation and fist clench) were used respectively. Outcomes The primary clinical outcome was reduction in FI events per week (recorded on paper bowel diaries over a 4-week period) in SNM versus sham phase of crossover (16 and 32 weeks). Secondary clinical outcomes including other bowel diary measures, e-event recording and a panel of summative questionnaires were recorded at 16, 32 and 58 weeks. Mechanistic outcomes included spatial localisation, relative cortical source signal strength and latencies of evoked and induced responses. Allocation and blinding Randomised allocation (1 : 1) to group 1 (SNM/sham) or group 2 (sham/SNM) was performed at the time of surgery using an online randomisation system managed by the Pragmatic Clinical Trials Unit at QMUL, with a randomisation list generated by an independent statistician to ensure allocation concealment. Randomisation was stratified by sex and centre with block sizes of four. Members of the research team, statisticians, surgeons who performed the surgical procedure, and participants were blinded to intervention status (SNM or sham). Participants were informed of the allocation ratio of 1 : 1 and that blinding prevented them from knowing in which group they were participating. Tamper-proof tape was used to mask stimulation settings. Sample size and statistics The study was designed to detect a mean 30% reduction between SNM and sham stimulation in FI event rate (ratio 0.7). At 90% power and 5% significance level with a cross-over design this required 90 participants (45 per group), allowing for 10% loss to follow-up. The pre-specified analysis for the primary outcome involved a mixed Poisson regression applied to the counts of FI events, with fixed effects of cross-over period and stratification factors, a random effect of individual, and a random effect of period within individual (the latter to allow for an over-dispersed Poisson distribution). When it came to the analysis, owing in part to the small numbers, the Poisson regression models did not converge for the count outcomes. Instead we applied a paired t-test to the FI rates in order to estimate the difference between SNM and sham with a 95% confidence interval and p-value. Results Clinical results The COVID-19 pandemic had a major effect on trial recruitment and patient retention. The trial was terminated on 24 July 2022 with just 39 patients randomised. Trial delivery was severely affected and terminated early due to COVID-19. Main barriers were the inability to continue face-to-face patient visits, redeployment of research staff to COVID-19 facing clinical roles and cancelling of SNM procedures due to lack of priority for non-urgent surgery. In total, 220 patients were screened for eligibility at nine sites from the UK and one site from Ireland between February 2018 and July 2022. Of these, 155 patients declined study participation or were ineligible due to study specific exclusion criteria. A total of 65 patients were pre-enrolled and consented to the study, of whom 26 did not meet the baseline minimum frequency criteria of FI episodes per week or did not receive an implant. The remaining 39 patients were randomised (arm 1: N = 17; arm 2: N = 22); however, only 16 completed the primary outcome during both cross-over periods (arm 1: N = 9; arm 2: N = 7). The remaining 23 participants withdrew from the study (N = 12), were excluded on the basis of problems of eligibility (N = 5) or did not complete the primary outcome data (N = 6: still included in the cohort follow-up phase). A total of 22 participants started the cohort follow-up phase, although 3 of these participants did not complete the final follow-up visit, leaving 19 participants for the 1-year effectiveness assessment. There were no major differences at baseline between allocated groups. As predicted, about 90% participants were female with mean age about 57 years. Almost all participants reported symptoms of urgency, combined with varying combinations of passive and urge FI. All participants reported previous conservative management for their FI symptoms (as per NICE guidance). Numbers of FI events at baseline were concordant with design assumptions (based on approx. seven events in a 1-week period). Median St Mark’s incontinence score was 19 in both groups, indicating severe symptoms (max score 24). E-event recordings were only undertaken by a minority (14/39) of participants. Test stimulation was performed using a tined lead in 68.6% participants. General anaesthesia was used in 70.6% of procedures and median operating time was 36 minutes (range 30–55 minutes). The lead was positioned in foramina S3 in most participants (91.4%) with some variations in fidelity of siting based on individual electrode responses (only 50% lead placements achieved the ideal published standard of motor or sensory responses for three electrodes