DiPALS: Diaphragm Pacing in patients with Amyotrophic Lateral Sclerosis a randomised controlled trial

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 09/55/33. The contractual start date was in July 2011. The draft report began editorial review in July 2015 and was accepted for publication in January 2016. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Andrew Bentley reports honoraria and travel subsistence from BioMarin to attend clinical advisory group in Berlin (April 2015) on respiratory management of mucoploysaccharoidoses. Anita K Simonds reports being on the steering committee of trial of adaptive servo ventilation in heart failure patients with predominant obstructive sleep apnoea (SERVE-HF). Chin Maguire reports grants from Motor Neuron Disease Association and non-financial support from Synapse Biomedical Inc. during the conduct of the study. Christopher J McDermott reports grants from the National Institute for Health Research Health Technology Assessment programme, grants from the Motor Neurone Disease Association and non-financial support from Synapse Biomedical Inc. during the conduct of the study and separate to the Diaphragm Pacing in patients with Amyotrophic Lateral Sclerosis (DiPALS) trial. Lyn Taylor reports personal fees from PAREXEL International Corporation not associated with the DiPALS trial. Roger Leek reports membership of the Motor Neurone Disease Association. John Stradling reports personal fees from ResMed UK not associated with the DiPALS trial. Martin Turner reports grants from Medical Research Council, personal fees from Biogen Idec, personal fees from GlaxoSmithKline, personal fees from Neuraltus Pharmaceuticals, personal fees from Ontario Brain Institute, personal fees from GLG Consulting and personal fees from Vertex Pharmaceuticals not associated with the DiPALS trial. Richard W Orrell reports grants from the Motor Neurone Disease Association of England, Wales and Northern Ireland during the conduct of the study. Wisia Wedzicha reports grants from Johnson and Johnson, grants from Vifor Pharma, grants from Takeda, grants from GlaxoSmithKline, personal fees from Novartis, personal fees from GlaxoSmithKline, personal fees from Astra Zeneca, personal fees from Boehringer, personal fees from Takeda and personal fees from Vifor Pharma outside the submitted work.

Copyright statement

© Queen’s Printer and Controller of HMSO 2016. This work was produced by McDermott et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Management of amyotrophic lateral sclerosis

Management is largely aimed at easing symptoms and supporting patients to maximise their function through a multidisciplinary approach. 4 One treatment, riluzole, can marginally slow down disease progression, prolonging survival, usually by around 3 months. 5,6 The exact mechanism by which riluzole affects the disease course is unclear, although it is known to modulate glutamate release, among other effects. However, the greatest impact on the disease course comes from the use of non-invasive ventilation (NIV). A randomised controlled trial (RCT) demonstrated an improvement in quality of life (QoL) and a median survival benefit of approximately 7 months (p = 0.006), in ALS patients using NIV who had good bulbar function. 7 NIV, however, is not without its problems, as some individuals are unable to tolerate it because of problems with claustrophobia and mask interface issues. Furthermore, there comes a point in the course of the disease when NIV is no longer effective.

Phrenic nerve stimulation, in which the diaphragm is stimulated into contracting, is a potential alternative or complementary method of providing respiratory support. The approach originated in patients with spinal cord injury, and historically it required direct phrenic nerve stimulation. The challenges with this approach have been the significant risk of iatrogenic phrenic nerve injury and, until recently, the need to undertake a thoracotomy. 8 More recently, phrenic nerve stimulation has been achieved using less invasive techniques. The NeuRX^® RA/4 diaphragm pacing system (DPS)™ (Synapse Biomedical, Oberlin, OH, USA) is a four-channel percutaneous neuromuscular stimulation system. DPS has an advantage over the earlier direct approach, in that the stimulating electrodes can be inserted into the under-surface of the diaphragm using a minimally invasive laparoscopic technique. 9 The leads are then tunnelled to an exit site on the abdomen and an external stimulator delivers the stimulus pulses and provides respiratory timing. Initial experience with the NeuRX RA/4 DPS in the spinal cord injury population10 suggested diaphragm pacing (DP) could reduce time spent on mechanical ventilation. 11 The NeuRX RA/4 DPS is now licensed for use in spinal cord injury in many countries, including the USA, and within the European Union.

Evidence for diaphragm pacing in amyotrophic lateral sclerosis

To date, the evidence for DPS in the ALS population is limited to case series and one uncontrolled multicentre cohort study. Their findings are consistent with those from the spinal cord patient population, and highlighted the apparent simplicity and operative safety of the NeuRX RA/4 DPS. 7,10 The US Food and Drug Administration (FDA) approved the NeuRX RA/4 DPS as a humanitarian-use device (HUD) in ALS following the submission of a humanitarian device exemption (HDE) application. The FDA summary of safety and probable benefit document (SSPB) summarises the evidence on which the HDE approval is based, which is largely on data from the aforementioned cohort study. 11 The main inclusion criteria for the study were patients with ALS with evidence of residual bilateral phrenic nerve function and a forced vital capacity (FVC) of less than 85% at screening and above 45% at DPS implantation. The full data from this study, including baseline characteristics, Consolidated Standards of Reporting Trials (CONSORT) diagram, survival, etc., on the 144 patients enrolled have not yet been published. Of the 144, 106 were implanted with the NeuRX RA/4 DPS and survival data on 84 are presented in the SSPB. 11

Median survival data are reported as 39 months from ALS diagnosis and 19 months post implantation in the SSPB on 84 implanted patients. As the data are from a cohort study there is no randomised control population to compare survival with. Instead, a subgroup of the HUD patients (n = 43) was compared with a historical control group of NIV users. The DP HUD group patients demonstrated survival of 37.5 months from diagnosis, compared with 21.4 months from diagnosis for the historical NIV control group (p < 0.001).

Safety data have been published in the SSPB for 86 implanted patients from the cohort study. In these patients, there were no reported deaths and none who needed a tracheostomy with permanent ventilation within the 30-day postoperative period. Four serious adverse events (SAEs) were reported relating to implantation [capnothorax (n = 2), respiratory failure due to complications of surgery (n = 1) and serious anaesthetic reaction (n = 1)]. Therefore, the implantation procedure itself appeared to be relatively safe. Important adverse events (AEs) over the 12-month protocol reported were capnothoraces in 16 patients (19%), percutaneous site infection in eight patients (9%) and pacing-related discomfort, which was described as mild in 20 patients (23%) and moderate in two patients (2%). In the 86 patients implanted, there were no SAEs involving malfunctioning device components. There were 18 reports of anode malfunction in 18 patients (21%) and 44 reports of electrode malfunction in 28 patients (33%). Although there was relatively frequent electrode breakage, there was no need to reimplant any electrodes, as all occurred external to the body at the connector holder.

Following the HUD approval of the NeuRX RA/4 DPS for ALS, there has been increasing use of this therapeutic option worldwide. 12 The promising survival data, lack of apparent harm and absence of alternatives have made this an appealing option, especially among patients who are unable to tolerate NIV, who may account for up to 50% of patients with ALS. 2,3,13,14 The evidence base for the NeuRX RA/4 DPS for ALS is limited. Moreover, pacing is expensive and it is not known whether or not DPS would meet the National Institute for Health and Care Excellence’s threshold for cost-effective interventions for end-of-life care. At the time of conducting the Diaphragm Pacing in patients with Amyotrophic Lateral Sclerosis (DiPALS) trial, the cost of DPS and implantation procedures (the excess treatment costs) was approximately £16,500 per participant. Therefore, although the preliminary data are promising, DPS is unlikely to be widely introduced without robust, randomised evidence together with a formal analysis of cost-effectiveness. This was our motivation for undertaking the DiPALS trial.

Research objectives

The aim of this study was to perform a definitive RCT on the efficacy and safety of the NeuRX RA/4 DPS when used in addition to NIV, compared with standard care of NIV alone, in patients with ALS. Specifically, we wished to assess whether or not the use of DP in addition to NIV prolongs survival, and also to quantify its impact on:

• QoL of the participant

• QoL of the main carer

• safety (AEs) and tolerability (withdrawal from treatment)

• quality-adjusted life-years

• views and perceptions of patients and family carers regarding acceptability and impact on life.

Trial design

The DiPALS trial was a multicentre, parallel-group, open-label RCT incorporating health economic analyses and a qualitative longitudinal substudy. The primary objective was to assess survival over the study duration of patients with ALS with respiratory failure, allocated to either standard care (NIV alone) or standard care (NIV) plus DPS using the NeuRX/4 DPS. Participants from seven UK hospitals were allocated to treatment using minimisation methods (see Appendices 2 and 3).

The trial was non-commercial, with input from Synapse Biomedical for quality assurance purposes, specifically for the training of surgeons and trial staff on implantation and technical aspects of the NeuRX/4 DPS.

Recruitment took place over 24 months (from December 2011 to December 2013).

Participants and eligibility criteria

Patients aged ≥ 18 years with a confirmed diagnosis of ALS were identified and screened at participating sites or patient referral sites against the following eligibility criteria.

Participant identification

Potentially eligible patients were identified from NHS hospital clinics at lead research sites or participant identification centres by neurology or respiratory clinicians or a study research nurse (see Appendix 4).

All patients aged ≥ 18 years with a confirmed diagnosis of ALS underwent a ‘prescreen’ whereby their last routine respiratory measure was checked to determine whether or not:

1. patients were ineligible because of poor respiratory criteria or other reason

2. patients were potentially eligible

3. patients may be eligible in the future as their respiratory measure was currently too good.

All those potentially eligible were approached at clinic to discuss the trial in detail. If they were identified from the clinic list prior to attendance the study information was sent to them by post.

Recruitment

Patients were approached by a member of the research team at sites when they attended routine clinic appointments. The study patient and carer information sheets formed a basis for a discussion with potential participants in clinics. Patients were given as long as they required to consider taking part in the trial, with further discussion by telephone or at another clinic offered. Informed consent was taken only if the researcher was satisfied that the patient fully understood the study procedures, was willing to undergo screening and participate in the trial, and was capable of giving full informed consent. Consent was obtained using participant and carer consent forms approved by the Research Ethics Committee (REC), as either full written consent, verbal consent given or consent given via the use of a communication aid (for REC-approved patient information sheets see Appendices 5–8 and for REC-approved consent forms see Appendices 9–12). When non-written consent was obtained an independent witness was asked to sign the consent form to verify that consent was taken.

If participants were willing to provide reasons for declining participation in the trial, these were entered onto the screening log. Anonymised basic details (age, sex and reason for exclusion) were collected on all eligible patients to allow completion of the CONSORT flow chart.

Participants were free to withdraw from the trial at any point without giving a reason, but data collected up to the point of study withdrawal were maintained and used in study analyses. Patients who asked to withdraw from treatment were encouraged to complete follow-up for QoL and safety to reduce attrition bias. All participants were followed up for survival until 12 months after the last participant was recruited, unless they specifically requested otherwise (Figure 1).

FIGURE 1.

Participant flow diagram. a, At 9 months, only ED-5D-3L, DPS/NIV use, AEs and resource use collected. CBI, Caregiver Burden Inventory; EQ-5D-3L, European Quality of Life-5 Dimensions, three levels; SAQLI, Sleep Apnoea Quality of Life Index questionnaire; SF-36, Short Form questionnaire-36 items.

Screening

Once patient consent had been confirmed, a member of the site study team initiated the screening process. Patient eligibility was assessed against both non-clinical and clinical criteria, and baseline assessments. Eligibility for the study was based on all the inclusion and none of the exclusion criteria being met. To confirm eligibility, a combination of techniques was used. Researchers checked medical notes/data to assess past medical history, current prescriptions and any other known information relating to eligibility criteria. At the screening visit, participants were asked to perform a respiratory test to determine their FVC or other measure. An assessment was made of the participant’s phrenic nerve function – a measure intended to ensure that, if the patient were randomised to DPS, it was likely that it would be possible to sufficiently stimulate the participant’s diaphragm once the system was fitted. All criteria were checked and signed off by the recruiting clinician. Once eligibility was established, participants were asked to complete baseline assessments and could proceed to randomisation and study procedures (see Figure 1).

Randomisation and blinding

Once eligibility had been confirmed and consent acquired, the participants were randomly allocated to a treatment group. The recruiting clinician accessed a centralised web-based randomisation system provided by the Sheffield Clinical Trials Research Unit (CTRU) in partnership with a University of Sheffield wholly owned subsidiary software development company, epiGenesys. Sites were able to log on to the system using a site-specific username and password. Researchers were prompted to enter patient details (identification number, date of birth and the minimisation factors) and to confirm that consent and eligibility was complete. Following this, the randomisation system notified the user and the study manager of the treatment allocation. Patients were informed of their treatment allocation within 1 week of randomisation by telephone or letter.

The study was open label; it was not considered feasible to blind participants, carers or site staff as the study intervention involves a surgical procedure and ongoing use of an implanted medical device. A sham surgical procedure was considered but discounted as being an unnecessary burden to participants, given the objectivity of the primary outcome measure.

Patients were allocated their treatment (NIV alone or NIV plus DPS) by method of minimisation, using baseline bulbar function, baseline FVC, age and sex as the minimisation factors. Factors were categorised as follows: bulbar function (mild, moderate or severe); FVC (50–59%, 60–69% or ≥ 70%); age (≤ 39 years, 40–79 years or ≥ 80 years); and sex (male or female). The allocation incorporated a burn-in period of 10 participants (i.e. the first 10 participants were allocated using simple randomisation). Thereafter, allocation was performed using minimisation incorporating a random probability element of 80% into the allocation algorithm. That is each participant was allocated to the arm that reduced treatment imbalance with 80% probability and to the opposite arm with 20% probability.

The minimisation system was prepared by the study statistician; the study team did not have access to the allocation list in order to maintain allocation concealment. The study team and governance committees were blinded during the course of the study with the following exceptions: (1) the trial statistician and data management team had full access to unblinded outcome data; (2) the Data Monitoring and Ethics Committee (DMEC) had access to unblinded summaries of outcome data; (3) site staff were aware of the treatment allocation of their own participants; and (4) the chief investigator, study manager and site monitor were aware of individual participants’ allocations and AEs but did not see unblinded comparative summaries. The trial unblinded statistician was responsible for providing the DMEC with the accumulating unblinded data summaries, but did not attend the closed meetings during which the DMEC reached its decisions.

The original analysis plan was signed off after recruitment start, by which point one participant had died. The updated analysis plan was written by the unblinded trial statistician and endorsed with minimal suggestions by the unblinded DMEC; all other reviewers and approvers did so blinded to outcome data.

Interventions

Participants were randomly allocated to receive either standard care (NIV) or standard care in addition to DPS using the NeuRX/4 DPS.

Outcomes

Data collection and management

Database

Trial data were entered into a validated bespoke web-based database system (Prospect) managed by the Sheffield CTRU in partnership with a University of Sheffield wholly owned subsidiary software development company, epiGenesys. Prospect stores all data in a PostgreSQL 9.1 (open-source software from The PostgreSQL Global Development Group) database on virtual servers hosted by Corporate Information and Computing Services at the University of Sheffield. Prospect uses industry standard techniques to provide security, including password authentication and encryption using Secure Sockets Layer/Transport Layer Security. Access to Prospect was controlled by usernames and encrypted passwords, and a comprehensive privilege management feature was used to ensure that users had access to only the minimum number of data required to complete their tasks. An automated audit trail recorded when (and by which user) records were created, updated or deleted. Prospect provides validation and verification features which were used to monitor study data quality, in line with CTRU’s standard operating procedures (SOPs).

Methods used for treatment allocation, sequence concealment and blinding

Patients were allocated their treatment (NIV alone or NIV plus DPS) by a method of minimisation, using baseline bulbar function, baseline FVC, age and sex as the minimisation factors. Factors were categorised as follows: bulbar function (mild, moderate or severe); FVC (50–59%, 60–69% or ≥ 70%); age (≤ 39 years, 40–79 years or ≥ 80 years); and sex (male or female). The minimisation was non-deterministic and incorporated a burn-in period of 10 participants and a random probability element of 80% into the allocation algorithm. In other words, the first 10 participants were allocated using simple randomisation; thereafter, each participant was allocated to the arm that reduced treatment imbalance with 80% probability and to the opposite with 20% probability. A centralised, web-based randomisation system hosted by the CTRU was used to allocate treatment allocations. Sites were able to log on to the system using a site-specific username and password. Researchers were prompted to enter patient details (identification number, date of birth and the minimisation factors) and to confirm consent and eligibility were complete. Following this, the randomisation system notified the user and the study manager of the treatment allocation.

The study was open label: it was not considered feasible to blind participants, carers or site staff.

Sample size

The sample size calculation was based on log-rank test, using Simpson’s rule19 as implemented in Stata version 11.1 (StataCorp LP, College Station, TX, USA) to allow for the unequal length of follow-up. The study duration comprised an 18-month recruitment period and a 12-month follow-up period, giving a maximum follow-up of 30 months and a minimum of 12 months. Assuming control group survival proportions of 45%, 20% and 10% at the minimum, average and maximum follow-up times, respectively, a hazard ratio (HR) of 0.45 and an additional 10% loss to follow-up, a total of 108 patients (54 per group) were needed to ensure a power of 85% using a two-sided type I error of 5%. The control group figures were conservative estimates based on the sole RCT of NIV,1 which is now considered standard care in the UK. We estimated the sample size on a conservative (but clinically important) 1-year difference in survival of 45% versus 70%, which produced the estimated HR of 0.45. It was expected that complete survival data would be available on all participants recruited, based on previous experience in ALS trials. We did, however, allow for a 10% loss to follow-up in the sample size calculation.

With regard to QoL data, we expected a low level of missing data due to loss to follow-up. We reviewed the patients who were initiated on NIV between July 2008 and June 2009 and had maintained contact with 100% of those patients surviving at 12 months. The appointment of a research nurse at each study site enabled home visits to collect the QoL data when necessary.

Early stopping

No interim analyses or early stopping was foreseen. However, in December 2013 the DMEC recommended that recruitment to DiPALS should cease on safety grounds (discrepancy in survival between the two treatment arms), and a final decision to stop the trial was made in June 2014. The recommendations and actions taken are reported fully in results (see Chapter 3).

Statistical methods

Economic evaluation

Prior to the trial commencing, a modelling exercise was undertaken to assess the feasibility that pacing could be cost-effective at standard willingness-to-pay thresholds for end-of-life care. This modelling confirmed that pacing would be cost-effective if the trial were to demonstrate a treatment effect of similar magnitude as demonstrated by the SSPB cohort in comparison with historical data. A full cost–utility analysis was planned for this trial, but following the early termination of the trial the chief investigator and funder agreed that this was no longer necessary.

Research governance

Patient and public involvement

Patient and public involvement was sought throughout the trial. The Sheffield Motor Neurone Disease Research Advisory Group (SMNDRAG), an independent research advisory group, was approached and agreed to be involved at an early stage. The SMNDRAG was established in 2008 following the principles of INVOLVE. 26 The group comprises members of the public, including several carers of ALS patients, patients and other lay individuals who responded to the call for members. Research training is provided to all members of the public that volunteer for the SMNDRAG as part of their induction and ongoing support. The SMNDRAG has been a valuable part of the DiPALS trial, collaborating as part of the research team at all stages of the research process.

Recruitment suspension

The fourth DMEC meeting was held on 16 December 2013. The ‘open’ part of the meeting discussed trial progress with both internal and external DMEC members. This was followed by a ‘closed’ meeting when the external, independent members reviewed unblinded safety data.

Following the closed meeting, the chairperson of the DMEC contacted the chief investigator on 18 December to recommend that the trial suspend recruitment temporarily based on safety grounds, citing a discrepancy in survival between the two arms. In doing so, the DMEC acknowledged that the sample size was relatively small (74 randomised, 24 deaths at the point of this recommendation), and that their decision would be reviewed as additional data became available.

The DMEC provided the following response:

The DMEC has reviewed the unblinded data on survival in the DiPALS study and the members of the Committee are unanimous in recommending that recruitment should cease as soon as practicable for reasons of safety, that monitoring of safety and survival should continue, and that the DMEC should review the updated survival data at the end of February 2014, and periodically thereafter.

Suspension of recruitment did not constitute an ‘urgent safety measure’.

In summary, the DMEC recommended that:

• recruitment be suspended with immediate effect

• implantation of new pacing devices be suspended

• other aspects of the trial remain unaltered; in particular, patients in the pacing arm should be encouraged to continue using their device.

The CTRU suspended the online randomisation system on the same day to ensure that no further patients were recruited to the trial. The TMG (PIs, co-investigators, research nurses and relevant members of the CTRU) were all notified of this decision immediately (18 December 2013). All planned surgeries for the DPS were cancelled. The PIs and chief investigator discussed the key message to give to any concerned trial participants while still remaining blinded to the data.

The TSC convened on 20 December 2013 to discuss the implications of this decision and their concerns that the data had not been thoroughly scrutinised. These included the consideration of centre effects; the effect of patient withdrawal and non-implantation with the device; chance; and non-compliance with either NIV or DPS. A second unblinded report was produced and circulated to only the external, independent DMEC members that incorporated analyses to ensure all these issues were considered.

A joint TSC and DMEC meeting was held on 13 January 2014 to allow all members to present any remaining concerns before responding to the DMEC. The TSC and study team remained blinded to trial results. The DMEC upheld its initial decision by providing the following response to the additional data analyses:

The DiPALS DMEC members, having reviewed the updated unblinded report and having considered the points raised during discussion with the TSC, remain unanimous in their advice that recruitment into the DiPALS trial should cease, but that follow-up should continue. This advice is based on analysis of the primary end point (patient survival during the course of the trial). The DMEC considers that secondary analyses are unlikely to alter the safety concerns raised by the primary intention-to-treat analysis. The DMEC will be happy to consider further reports and to provide advice as required.

The TSC jointly agreed and communicated its agreement with the DMEC to suspend recruitment to the trial team late on 15 January 2014. The TSC requested that the DMEC continue to review unblinded data every 3 months to capture any further changes that would warrant discontinuation of the use of the DPS for participants in follow-up.

Recruitment and participant flow

In total, 74 participants (37 per arm) were randomised between 5 December 2011 and 18 December 2013, when the DMEC recommended that recruitment cease. Study follow-up concluded in December 2014, by which time 47 patients had died; one patient was last followed up in August 2014, with the remaining 26 known to be alive in December 2014 (Figure 2).

FIGURE 2.

Trial profile (CONSORT diagram). MND, motor neurone disease; NK, not known.

Baseline data

A total of 74 participants were allocated (37 to each arm). The characteristics of the participants are presented in Table 2. Participants were predominantly male with sporadic (usually limb) onset and with mild bulbar impairment (74%). Despite age being included as a minimisation factor, the NIV plus DPS arm was slightly older (average 60 vs. 54 years), a consequence of most participants falling into the middle category of age 40–79 years; this imbalance was addressed in all comparisons, however, as all prespecified analyses included age as a covariate (continuous, rather than the categorisation used for the purpose of minimisation).

Primary outcome (overall survival)

Primary overall survival analyses

The Kaplan–Meier curve for overall survival is presented in Figure 3. The median survival was 11.0 months (95% CI 8.3 to 13.6 months) in the pacing arm and 22.5 months in the control arm. As the upper bound of the Kaplan–Meier survival curve never reached 50%, the upper limit of the 95% CI is unknown: the lower bound is 13.6 months. The HR (adjusting for minimisation covariates) was 2.27 (95% CI 1.22 to 4.25; p = 0.01), indicating the risk of death at any point in time was higher in the NIV plus pacing arm than in the control arm. A permutation test approach, as recommended by Scott et al. ,25 produced a p-value of 0.013 confirming the statistical significance of this difference.

FIGURE 3.

Overall survival.

Sensitivity analyses were undertaken to assess the consistency of findings to different covariates and to different models. First, we allowed for between-site differences via a stratified log-rank test in which the strata comprised the seven hospitals. The findings from doing this were not materially changed. There was, however, evidence of non-proportionality of the hazards of the two groups, as detected by the Grambsch–Therneau test20 for the correlation between residuals and time (ρ = –0.31; p = 0.03). In other words, although overall NIV plus DPS was associated with a twofold increase in hazard, the impact was greater in early months and smaller among longer-term survivors. Two alternative modelling approaches were fitted: first, an accelerated failure time model (in which the time, rather than hazard, is modelled); and, second, a non-parametric restricted mean survival time, in which the difference in mean survival is estimated based on the area between the Kaplan–Meier curves. 22 The findings from the sensitivity analyses analysis are presented in Table 3. Overall, the magnitude of the survival deficit in the NIV plus DPS group was remarkably consistent, with survivorship always significantly lower in the NIV plus DPS group.

TABLE 3 - Sensitivity analyses

Method of analysis	Trial arm		Comparison
	NIV plus DPS	NIV
Cox proportional hazards regression	Deaths per person-year follow-up		HR (95% CI)	p-value
Log-rank (univariate)	0.74	0.37	2.24 (1.30 to 4.19)	0.005
Log-rank (stratified by centre)	–	–	2.02 (1.21 to 3.84)	0.012
Cox regression (univariate)	–	–	2.28 (1.27 to 4.10)	0.006
Cox regression (including minimisation factors as covariates)	–	–	2.27 (1.22 to 4.25)	0.012
Accelerated failure time regression	Median survival time (months)		Time ratio (95% CI)	p-value
Gehan–Wilcoxon test (univariate)	11.0	22.8	–	0.002
Gehan–Wilcoxon test (stratified by centre)	–	–	–	0.029
Log-normal regression (univariate)	–	–	0.52 (0.36 to 0.77)	0.001
Log-normal regression (including minimisation factors as covariates)	–	–	0.52 (0.35 to 0.77)	0.001
Restricted mean failure time	Mean survival time over first 30 months		Difference (95% CI)	p-value
Univariate	13.7	20.6	–6.9 (–11.4 to –2.4)	0.003
Including minimisation factors as covariates	–	–	–6.8 (–11.5 to –2.2)	0.005

Overall survival at Data Monitoring and Ethics Committee intervention

The DMEC did not prespecify rules for stopping because of safety, preferring to use clinical judgement about participant outcomes as a whole. Nevertheless, its recommendations were based primarily on overall survival, and for reference we present overall survival as of the following dates:

• 10 December 2013 (the date that data were seen by the DMEC, when it recommended suspension of recruitment and implantation)

• 10 June 2014 (the date that data were seen by the DMEC, when it recommended participants in the NIV plus DPS arm should cease pacing).

The statistical significance of the overall survival difference varied markedly during the data collection phase, both across time and also according to which of the aforementioned models were used to derive it. Nevertheless, as shown in Table 4 and Figures 4 and 5, overall survival was significantly worse in the NIV plus DPS group at both time points.

TABLE 4 - Statistical significance of the overall survival difference during data collection

Univariate Cox regression.

Thirty-six participants in each arm at this time. A further two participants were randomised between production of the database report and the DMEC meeting (18 December 2013). The number of participants implanted remained at 32.

Data collection time point	Trial arm, number (%) of deaths			Intention-to-treat comparison (univariate)
	NIV plus DPS		NIV
	All patients (n = 37)	Excluding non-implanted (n = 32)	All patients (n = 37)	HR (95% CI)a	p-value
As of 10 December 2013	16 (44b)	14 (44)	8 (22b)	2.45 (1.05 to 5.74)	0.039
As of 10 June 2014	23 (62)	21 (66)	11 (30)	2.73 (1.33 to 5.60)	0.006
At end of study	28 (76)	26 (81)	19 (49)	2.28 (1.27 to 4.10)	0.006

FIGURE 4.

Survival at 10 December 2013.

FIGURE 5.

Survival at 10 June 2014.

Subgroup analyses of overall survival (preplanned)

We prespecified three subgroup analyses: (1) by NIV tolerance, based on estimated average NIV usage; (2) by bulbar function, as derived from the ALSFRS-R questionnaire; and (3) a per-protocol analysis excluding participants who were inappropriately randomised or who did not adhere to NIV and/or pacing (if that was their randomised allocation).

The results of these are given in this section. In all cases the results were not adjusted for other covariates, as the small number of participants in each subgroup means that more complex models are likely to be unstable. Subgroup analyses were by univariate Cox regression, and tests of interactions were derived by adding treatment, subgroup and their interaction to the model. It should also be noted that small subgroup sizes results in wide CIs and low statistical power both for the main effects and interactions.

Overall survival by baseline bulbar function

The subgroup analysis by bulbar function was originally defined to look at survival in severe (functional score of 0–4) and mild/moderate (functional score of 5–12) separately. As all bar five participants were in the latter group, we have chosen to split the mild and moderate subgroups out, but retain severe bulbar function as its own (small) category.

Overall survival was inferior in the NIV plus DPS arm in both mild and moderate subgroups, although the difference was statistically significant only in the former. There were too few participants in the severe subgroup to allow a meaningful interpretation to be made. Overall, there was little evidence of a differential effect of DP according to bulbar function with a non-significant p-value for the interaction between bulbar function and group (likelihood ratio test χ²₍₂₎ = 3.44; p = 0.18) (Figure 6).

FIGURE 6.

Overall survival by bulbar function. (a) Mild bulbar impairment; (b) moderate bulbar impairment; and (c) severe mild bulbar impairment.

Overall survival by non-invasive ventilation tolerance

The subgroup analysis by NIV use was again originally defined to look at survival in two subgroups: NIV tolerant (≥ 4 hours typical daily use) and NIV intolerant (< 4 hours). Post hoc, we have split the latter category further into low NIV use (1–3.9 hours daily, which might confer some benefit) and non-NIV use (< 1 hour daily, which confers virtually no benefit). The rationale is to allow better assessment of the relationship between pacing and outcome. Pacing has been suggested as possibly having particular benefit among patients who do not tolerate NIV, and this analysis allows us to put that hypothesis to the test. Moreover, the difference in survival may (theoretically) be attributed to participants in the NIV plus DPS arm using pacing in place of NIV; this allows us to compare users within each NIV use subgroup.

We categorised average daily NIV use into < 1 hour, 1–3.9 hours or ≥ 4 hours in order to assess whether or not the difference between groups in overall survival was similar. The Kaplan–Meier graphs are shown in Figure 7. In six participants there were inadequate data from which to assess NIV usage, and small numbers hamper interpretation, but survival in the NIV plus DPS was not superior to NIV alone in any subgroup, with the largest difference (in favour of NIV) among the subgroup of participants who did not use NIV. As with bulbar function there was no evidence of a differential effect according to NIV use at conventional statistical levels (likelihood ratio test χ²₍₂₎ = 4.14; p = 0.13).

FIGURE 7.

Overall survival by NIV use. (a) Non-use (< 1 hour); (b) low use (1–3 hours); and (c) NIV tolerant (≥ 4 hours).

Per-protocol analysis of overall survival

Protocol compliance was based on excluding participants who:

• were randomised in error (i.e. in breach of inclusion/exclusion criteria)

• did not tolerate NIV (< 4 hours’ average daily use)

• did not undergo successful DP implantation

• were non-users of DP.

This subgroup was very similar to that shown in the ‘tolerant’ (≥ 4 hours) subset in subgroup analysis. No participants were erroneously randomised and only one non-user of the pacing device was NIV tolerant, meaning that the total number included reduced from 34 to 33. The HR for NIV plus DPS compared with NIV in this subgroup was 1.92 (95% CI 0.81 to 4.56; p = 0.141).

Non-invasive ventilation and pacing usage

Non-invasive ventilation use

Non-invasive ventilation was initiated in 70 out of 74 patients. In total, 57 patients were initiated within 2 weeks of randomisation, a further six within 1 month and the remaining 7 between 3 and 11 months. NIV usage was similar between the two groups (Table 5 and Figure 8).

TABLE 5 - Average daily NIV usage after initiation

Variable	Trial arm
	NIV plus pacing (n = 37)	NIV (n = 37)
Number of patients with data	34	34
Average daily usage (hours)
Mean	5.2	4.8
Median (IQR)	3.2 (0.5–8.2)	4.6 (0.0–7.8)
Number (%) by subgroup
< 1 hour	9 (26%)	11 (32%)
1–3.9 hours	9 (26%)	5 (15%)
≥ 4 hours	16 (47%)	18 (53%)

FIGURE 8.

Average daily NIV usage after initiation.

Overall survival by non-invasive ventilation and pacing

The relationship between overall survival and NIV use (categorised as non-use, low use and tolerant over the study period) is covered further in this section. Furthermore, the use of overall survival and NIV use in relation to pacing is investigated in more detail. The role of post hoc and exploratory analyses is to attempt to understand the unexpectedly poor survival experienced by participants in the NIV plus DPS group.

Non-invasive ventilation use and overall survival

The problems in evaluating the association between therapeutic regimens and patient outcomes in an uncontrolled, observational setting are well documented. Using NIV is generally accepted as beneficial,27 but the extent to which a participant uses it depends, to a degree, on prognosis. A patient who uses NIV at low levels may do so because their respiratory physician considers them well enough, or they consider themselves well enough, to have no need for high levels of use. Conversely, ALS patients often use NIV at high levels (≥ 12 hours per day) towards the end of their life. There are other factors besides this which impact on NIV use, but the aforementioned factors illustrate how the true impact of NIV is confounded by prognosis and can thereby be underestimated. With this in mind, the overall lack of association between NIV use and survival in the remainder of this section should not be taken as evidence against NIV being beneficial. Rather, this section attempts to assess whether or not NIV use gave rise to differential treatment effect and, crucially, whether or not some participants in the NIV plus DPS group used pacing in place of NIV.

We also note the difficulties in quantifying NIV use. In some cases there was a lag between NIV initiation and regular use, and in a handful of cases several months passed between initiation and regular use. It could be argued that the analysis should take NIV use starting from the date of first regular use, but the limitation of doing so is that ‘regular’ use is not unequivocally defined. Some patients started NIV at low levels (1–2 hours per day), whereas others used NIV on trial basis, only to stop (sometimes for several months) before recommencing. Furthermore, the participant diary data were in many cases not recorded in sufficient detail (completeness) to allow a date of ‘regular’ use to be determined.

Non-invasive ventilation use in the 12-month follow-up period

We first looked at NIV from the point of initiation to the end of the 12 months’ follow-up or death, whichever occurred first. Data were available for 68 of the 74 participants over the study duration.

The primary motivation, however, is not to quantify the association between NIV and survival per se, but to investigate whether or not the difference between the randomised NIV plus DPS and NIV arms may be affected by NIV rather than pacing alone.

The problem in visualising survival data is that the survival times are censored for a proportion of the participants. The relationship between NIV use and survival can be displayed graphically only if some assumptions are made as to what the censored times would have been, had they been observed. There is necessarily considerable uncertainty in doing so. For the purpose of displaying the data in graphical form we followed the approach proposed by Royston and Parmar22 to estimate the underlying (but unobserved) survival times based on each individual’s censored survival time, treatment group and prognostic covariates. A Q–Q plot of the data showed the log-normal survival distribution to give a reasonable fit to the data; therefore, a truncated log-normal regression model was fitted. From this, the overall survival for each individual was being estimated based on their predicted mean survival, their censored survival time, and the root-mean-square error of the model. We fitted two models with minimisation covariates also included in both: first, for all patients and then second, within each group separately. The estimated survival was taken as the average of these two predicted times.

It is important to note that we estimate the survival times only for the purpose of visualising the relationship between NIV, and not for the analysis itself. In all cases, the graphical display distinguishes estimated survival times from observed survival times.

Overall, no relationship was evident between typical NIV use and survival, using either the level of NIV or its categorised version (< 1 hour, 1–3.9 hours or 4 hours). These are demonstrated graphically for the NIV use measured in hours (Figure 9), using the estimated/extrapolated survival for censored survival and in a more conventional Kaplan–Meier graph for the categorisation of NIV use (Figure 10).

FIGURE 9.

Survival by NIV use: NIV use in hours. Censored survival times are estimated based on a censored log-normal distribution using treatment and minimisation covariates. The line is a locally weighted scatterplot smoother (lowess) with a bandwidth of 0.6. Circles denote actual survival times; stars are estimated times.

FIGURE 10.

Survival by NIV use: NIV use category. Censored survival times are estimated based on a censored log-normal distribution using treatment and minimisation covariates. The line is a locally weighted scatterplot smoother (lowess) with a bandwidth of 0.6.

In the categorised version of NIV there was no association between average daily usage and overall survival, using either the standard log-rank test (χ²₍₂₎ = 0.64; p = 0.72) or the log-rank test for trend across the categories (χ²₍₁₎ = 0.64; p = 0.46). As shown in Table 9, the largest difference between NIV plus DPS and NIV alone was observed in the low-use subgroup (n = 20 participants), but the interaction term between treatment group and NIV category was not statistically significant at conventional levels in a Cox regression model (likelihood ratio test χ²₍₂₎ = 4.14; p = 0.13).

TABLE 9 - Overall survival by bulbar function and by NIV use

Variable	Trial arm, number of participants (deaths)		Comparison
			HR (95% CI)	p-value
	NIV plus DPS	NIV
Overall survival by bulbar function
Overall test of interaction between treatment and bulbar function				0.179
Mild (9–12)	26 (21)	29 (14)	2.99 (1.49 to 6.01)	0.002
Moderate (5–8)	8 (6)	6 (3)	1.95 (0.48 to 7.92)	0.349
Severe (0–4)	3 (1)	2 (2)	0.62 (0.05 to 7.00)	0.697
Overall survival by NIV use
Overall test of interaction between treatment and NIV use				0.126
Non-use (< 1 hour daily)	9 (8)	11 (5)	3.90 (1.23 to 12.4)	0.021
Low use (1–3.9 hours daily)	9 (5)	5 (3)	1.50 (0.35 to 6.38)	0.584
Tolerant (≥ 4 hours daily)	16 (12)	18 (11)	1.63 (0.70 to 3.78)	0.253

No relationship was found when looking at the average number of hours (i.e. as a continuous measure) and overall survival. The linear association as derived from a univariate Cox regression model showed a small and non-statistically significant increase in survival with increasing hourly use of NIV (HR 0.98, 95% CI 0.92 to 1.04; p = 0.52). The possibility of a non-linear relationship was assessed using fractional polynomials regression, but the best non-linear model still failed to find any association (likelihood ratio test χ²₍₂₎ = 1.26; p = 0.53) or of any interaction between treatment and NIV use (likelihood ratio test χ²₍₁₎ = 0.68; p = 0.41).

Non-invasive ventilation use in the first 3 and 6 months post randomisation

We also looked at whether overall survival is associated with NIV use in (1) the first 3 months post randomisation or (2) the first 6 months post randomisation. Although this does not use the more complete NIV data over the 12 months’ follow-up, focusing on short-/medium-term use might be seen as a surrogate for intention-to-use NIV, as opposed to the level of use necessitated by disease progression. Participants for whom NIV initiation was delayed are included; a participant who was initiated at 4 months would be defined as having zero use within the 0–3 months period.

However, the findings were similar to those reported for 12-month NIV use above. Overall survival was weakly associated with NIV use, but was not statistically significant: the HR for 3-month use was 0.96 (95% CI 0.87 to 1.05; p = 0.33) and for 6-month use was 0.98 (95% CI 0.91 to 1.07; p = 0.68). Non-linear associations derived from fractional polynomial regression also failed to reach statistical significance (likelihood ratio test for 3-month follow-up: χ²₍₂₎ = 2.95; p = 0.23; 6-month follow-up likelihood ratio test χ²₍₁₎ = 0.89; p = 0.64). Again, no interaction was found between treatment group and NIV use over either period.

Pacing use and overall survival

We assessed the relationship between pacing use and overall survival. Unlike NIV, no normative data exist for pacing use; consequently, we did not categorise pacing usage into groups. The caveats that apply when assessing the relationship between NIV and survival also apply here.

Figure 11 shows average pacing use against overall survival, again estimating survival times as described in Non-invasive pacing and overall survival. As with NIV, no discernible relationship was evident. The HR for each additional hour of use was 1.01 (95% CI 0.94 to 1.09; p = 0.707); if non-implanted patients were excluded, the HR was 1.00 (95% CI 0.92 to 1.09; p = 0.931). Non-linear models, again derived using fractional polynomials, did not improve the fit of the model.

FIGURE 11.

Survival by pacing use. Censored survival times are estimated based on a censored log-normal distribution using treatment and minimisation covariates. The line is a locally weighted scatterplot smoother (lowess) with a bandwidth of 0.6. Circles denote actual survival times; stars are estimated times. N, not implanted.

Non-invasive ventilation, pacing and overall survival

The final analyses looked at the association between NIV and pacing use within the 37 participants randomised to receive the intervention. A potential theory was that participants in the NIV plus pacing arm may have used pacing in place of NIV. If so, this could partly explain the between-group differences, as NIV use is known to prolong survival.

Nevertheless, this theory was not borne out by the data. The average use of pacing and NIV is shown in Figure 12; the relationship between the two measures was positive rather than negative, suggesting participants who use NIV more also tended to use pacing more. We also split the NIV plus DPS arm into NIV-tolerant users (≥ 4 hours per day on average) and NIV-non-tolerant users (< 4 hours per day on average), with the Kaplan–Meier curves as shown in Figure 13. Participants in the NIV plus DPS arm who used NIV for an average ≥ 4 hours a day had a better survival than those who did not, but their survival remained lower than the NIV alone group. Neither comparison was statistically significant at the 5% significance level; however, the small numbers in each subgroup mean that the statistical power is low for this comparison.

FIGURE 12.

Relationship between NIV and pacing use. N, not implanted.

FIGURE 13.

Overall survival by pacing and NIV use.

Secondary outcomes

Participant quality of life

European Quality of Life-5 Dimensions, three levels

The participant health utility as measured by EQ-5D-3L is shown in Figures 14 and 15. Among those surviving to 12 months and with complete data, the difference in EQ-5D-3L scores was –0.18 (95% CI –0.44 to 0.08; p = 0.164). However, over the follow-up period as a whole, the EQ-5D-3L score declined quicker in the NIV plus DPS arm than in the NIV alone arm, with the longitudinal analysis providing an average difference across time of –0.13 (95% CI –0.25 to –0.00; p = 0.042). Imputing missing data made no material difference to the conclusions. The preplanned analysis of EQ-5D-3L scores according to NIV tolerance and bulbar function was not undertaken because of small numbers (especially in the NIV plus DPS arm) and the apparent deficit in overall survival.

FIGURE 14.

The EQ-5D-3L health utility among surviving participants over time: EQ-5D-3L health tariff.

FIGURE 15.

The EQ-5D-3L health utility among surviving participants over time: patient EQ-5D-3L health tariff, deaths imputed as zero.

A further analysis of EQ-5D-3L scores included both surviving and non-surviving participants, with a score of 0 assigned to the latter from the point of death onwards. Doing so demonstrated a lower QoL both at 12 months (mean difference –0.12, 95% CI –0.24 to –0.01; p = 0.040) and also longitudinally (mean difference –0.14, 95% CI –0.24 to –0.04; p = 0.001).

The findings of the EQ-5D-3L thermometer scale were consistent with those for the main EQ-5D-3L questionnaire (Table 10). All longitudinal analyses were repeated using an unstructured correlation matrix, as the random intercept model makes the assumption of a common correlation between repeated measures, irrespective of their proximity in time. Although correlation decreased with time, the estimated difference between intervention and control was virtually unaltered.

TABLE 10 - Quality of life

Completeness is number of questionnaires obtained within time windows as a ratio of the number expected (i.e. not including post death). Denominator includes all participants, although not all participants had assigned carers.

Imputation comprised a health utility or thermometer score of 0 from the point of death onwards.

Instrument	End of follow-up analysis				Longitudinal analysis
	Trial arm		Comparison		Comparison
	NIV plus DPS (n = 37)	NIV (n = 37)	Mean difference (95% CI)	p-value	Mean difference (95% CI)	p-value
Number (%) of participants alive at 12 months	16 (43)	27 (73)	–	–	–	–
Patient QoL
EQ-5D-3L health state (% completea)	131/178 (74)	161/209 (77)	–	–	–	–
Complete case, mean (standard error)	0.02 (0.09)	0.19 (0.08)	–0.18 (–0.44 to 0.08)	0.164	–0.13 (–0.25 to 0.00)	0.042
With imputation for surviving participants, mean (standard error)	0.02 (0.09)	0.13 (0.08)	–0.11 (–0.39 to 0.16)	0.410	–0.12 (–0.24 to 0.00)	0.056
With imputations for all participants,b mean (standard error)	0.01 (0.03)	0.11 (0.05)	–0.12 (–0.24 to 0.01)	0.040	–0.14 (–0.24 to 0.04)	0.001
EQ-5D-3L thermometer scale (% completea)	132/178 (74)	160/209 (77)	–	–	–	–
Complete case, mean (standard error)	34.4 (6.3)	42.3 (5.1)	3.4 (–14.5 to 21.2)	0.699	–8.3 (–17.4 to 0.8)	0.074
With imputation for surviving participants, mean (standard error)	36.0 (6.6)	40.0 (5.2)	1.3 (–17.6 to 20.1)	0.893	–5.6 (–14.5 to 3.2)	0.212
With imputation for all participants,b mean (standard error)	14.8 (3.9)	27.4 (4.8)	–13.4 (–25.9 to 0.9)	0.036	–12.0 (–20.8 to 3.1)	0.008
SF-36 aggregate physical health score (% completea)	110/154 (72)	133/174 (76)	–	–	–	–
Complete case, mean (standard error)	25.4 (4.2)	21.8 (1.8)	7.4 (–1.9 to 16.8)	0.110	0.2 (–2.2 to 2.5)	0.900
With imputation, mean (standard error)	23.8 (3.2)	20.6 (2.4)	8.7 (–1.3 to 18.7)	0.089	0.3 (–2.0 to 2.7)	0.780
SF-36 aggregate mental health score (% completea)	110/154 (72)	133/174 (76)	–	–	–	–
Complete case, mean (standard error)	44.1 (3.7)	47.0 (4.2)	–7.8 (–20.5 to 4.9)	0.210	–4.1 (–8.6 to 0.3)	0.066
With imputation, mean (standard error)	42.7 (4.2)	47.4 (3.5)	–8.6 (–22.2 to 5.1)	0.218	–3.5 (–7.9 to 0.8)	0.112
SAQLI (% completea)	110/154 (72)	132/174 (76)	–	–	–	–
Complete case, mean (standard error)	4.1 (0.5)	4.2 (0.3)	–0.2 (–1.3 to 1.0)	0.751	–0.3 (–0.7 to 0.1)	0.163
With imputation, mean (standard error)	3.9 (0.4)	4.5 (0.3)	–0.6 (–1.7 to 0.6)	0.329	–0.3 (–0.7 to 0.1)	0.117
Carer QoL
EQ-5D-3L health state (% completea)	109/178 (61)	148/209 (71)	–	–	–	–
	0.78 (0.11)	0.82 (0.06)	–0.08 (–0.38 to 0.23)	0.600	–0.08 (–0.17 to 0.01)	0.077
EQ-5D-3L thermometer scale (% completea)	110/178 (62)	149/209 (71)	–	–	–	–
	81.3 (7.1)	71.0 (6.4)	12.1 (–13.1 to 37.2)	0.329	–0.2 (–7.4 to 7.1)	0.966
CBI (% completea)	93/154 (60)	121/174 (70)	–	–	–	–
	28.0 (3.2)	29.6 (3.2)	3.1 (–7.5 to 13.8)	0.536	1.2 (–2.7 to 5.0)	0.558

Short Form questionnaire-36 items

As expected, overall physical health (as derived from SF-36) is considerably lower than population norms in both groups, although mental health was comparable. No differences were evident between the two groups over the duration of the study. As with EQ-5D-3L, the difference in aggregate mental health scores between the groups was not statistically significant among surviving participants at 12 months (mean difference –7.8, 95% CI –20.5 to 4.9; p = 0.210), but was of borderline significance over the duration of the study (mean difference –4.1, 95% CI –8.6 to 0.3; p = 0.066). Physical health appeared similar between groups on both scales (Figures 16 and 17). Relaxing the assumption of a common correlation in the longitudinal analyses made virtually no difference to the findings, with the correlation being relatively constant with time.

FIGURE 16.

The SF-36 aggregate physical health component.

FIGURE 17.

The SF-36 aggregate mental health component.

Sleep Apnoea Quality of life Index

Sleep apnoea appeared similar at all time points between the two groups (Figure 18).

FIGURE 18.

Sleep Apnoea Quality of Life Index.

Carer quality of life

The carer QoL is presented in Table 10 and, graphically, in Figures 19 and 20. There was little in the way of difference between the groups.

FIGURE 19.

Carer EQ-5D-3L health tariff.

FIGURE 20.

Caregiver Burden Inventory.

Relationship between quality of life, pacing and baseline symptoms

We have compared baseline scores with 12-month averages in EQ-5D-3L, SF-36 physical component summary score, SF-36 mental component summary score and total SAQLI and see no relationship between symptoms present and QoL scores. Figures 21 and 22 show these for SAQLI; similar patterns were observed on other measures.

FIGURE 21.

Change (12 months – baseline) in SAQLI score in relation to baseline SAQLI score (Q12–Q14). (a) NIV plus DPS; and (b) NIV.

FIGURE 22.

Change (12 months – baseline) in SAQLI score in relation to baseline ALSRFS-R sum (Q11–Q12). (a) NIV plus DPS; and (b) NIV.

Safety

Physiological measurements

In the light of the early stopping, the TSC requested additional respiratory function data to be collected to augment that which was already collected at baseline (and for DP, immediately pre surgery). Specifically, we wished to assess (1) whether or not the groups were comparable at baseline; (2) whether or not decline appeared more pronounced in the post-surgery period among the NIV plus DPS group than in the NIV group; and (3) the trajectory across time, in general, to see if it offered any other clues by which to explain the findings. We were able to obtain data for patients at some of the sites for FVC, arterial carbon dioxide and ALSFRS-R.

The purpose of the analysis was not to assess efficacy but to investigate possible treatment harm and, consequently, we analysed the group based on actual exposure (i.e. implanted with the pacing device) as opposed to the more conventional intention-to-treat analyses. Measurements taken more than 45 days prior to randomisation or > 1 month after the end of scheduled follow-up were excluded. A generalised least squares regression model was fitted in which the treatment received, time since randomisation and their interaction were the covariates. No attempt was made to impute missing data.

The figures show some evidence that FVC did indeed decline faster in participants who were implanted than those who were not, with the average rate of decline per month being 2.1% steeper among implanted participants than in those without implantation (95% CI –3.4% to –0.8%; p = 0.001). It is not easy to distinguish whether or not the decline is more pronounced in the postoperative period than the preoperative period (and, hence, directly attributable to operation), as the latter is a short period of time. (Figures 23 and 24).

FIGURE 23.

Forced vital capacity against time for implanted participants. The triangle denotes date of procedure and the cross denotes date of death.

FIGURE 24.

Forced vital capacity against time for non-implanted participants. The cross denotes date of death.

The trajectory for PaCO₂ is shown in Figures 25 and 26. In contrast to FVC, the change in PaCO₂ is more modest, with no obvious change in levels in either group.

FIGURE 25.

The PaCO₂ against time for implanted participants. The triangle denotes date of procedure, the cross denotes date of death.

FIGURE 26.

The PaCO₂ against time for non-implanted participants. The cross denotes date of death.

Finally the trajectory is shown for ALSFRS-R, a measure of ALS symptomology based on a 12-item questionnaire and ranging from 0 points (most severe) to 48 points (least severe). There was substantial decline in the control group (on average 1.1 point per month), but the rate of decline was greater still among participants who underwent implantation (mean difference 0.4 points per month, 95% CI –0.7 to 0.0 points per month; p = 0.035) (Figures 27 and 28).

FIGURE 27.

The ALSRFS-R score against time for implanted participants. The triangle denotes date of procedure and the cross denotes date of death.

FIGURE 28.

The ALSRFS-R score against time for non-implanted participants. The cross denotes date of death.

Outcomes

The primary end point was the perceptions of patients and carers regarding acceptability and impact of the device.

Qualitative substudy methods

Data analysis methods

Interviews were transcribed verbatim and transcript data were analysed using techniques of thematic analysis, where similar concepts or ideas across the interviews are brought together and assigned a code. Relationships between codes are then examined to develop a network of themes and subthemes recurring across the data. Systematic coding and retrieval of data was supported by the ATLAS.ti software (version 7, ATLAS.ti Scientific Software Development GmbH, Berlin, Germany). Data were shared and discussed at several qualitative team meetings to establish consensus for the coding network.

Participant characteristics

Fourteen patients and eight carers took part in the qualitative substudy. Eight patients were male and six were female; 10 were aged in their sixties, and four were aged in their fifties. On the ALS FRS-R at entry into the study, 11 patients were in the mild impairment range for bulbar function and three were in the moderate bulbar impairment range. In terms of limb function, one patient scored > 9, 12 scored 5–8 (moderate impairment), and one scored < 5. Patients were under the care of five different ALS centres. Nine participants were interviewed at both time points, two patients had died in the interval period and two patients preferred not to be followed up because of disease progression. One further person was not using the system at initial visit, or subsequently, and declined follow-up.

Results

Patient and carer experiences

Participants described the process from initial approach to be involved in the study, to ongoing usage of the pacer and perceived outcomes. The recurring themes and subthemes are presented in Figure 29. Owing to the extensive volume of material we focus on the themes most frequently found and include a number of verbatim quotes to illustrate key points.

FIGURE 29.

The patient journey trialling DP.

Main findings

The main finding was that DP had a deleterious effect on overall survival. In the DP plus NIV arm compared with the NIV alone arm, the observed reduction in overall survival was 11.5 months and the reduction in survival from symptom onset was 17 months.

The impact on QoL was at best minimal, with no statistically significant differences observed between the arms in patient or carer preplanned QoL measures SF-36, EQ-5D, SAQLI and CBI. The lack of benefit on the SAQLI is particularly noteworthy, as the one RCT of NIV demonstrated improved QoL and respiratory symptoms using this inventory. 1 A planned cost–utility analysis did not go ahead following the lack of efficacy. However, we analysed the EQ-5D-3L scores, as this is a valid QoL tool in its own right. The patient health utility (EQ-5D-3L) was slightly lower in the NIV plus pacing group than in the NIV alone group, especially when a score of 0 was imputed to EQ-5D-3L following death. Differences were modest at any individual time point, but longitudinal analysis demonstrated statistically and clinically significant differences on all patient EQ-5D-3L questionnaires, indicating that at least on one measure, the addition of DP, was associated with a reduction the QoL of patients.

There were more AEs in the NIV plus DPS arm, and many of these were related to the surgery, for example postoperative pain. The type and frequency of the AEs are consistent with those published for ALS, and device-related AEs (e.g. discomfort on pacing or wire failure) were less common than observed in the SSPB, as submitted to the FDA. 11 There are three immediate plausible explanations for the latter finding: (1) improvements in using the device with time, (2) patient under-reporting or (3) the fact that AEs were reported over a longer follow-up (in our study, AEs were reported up to death or 1 year post randomisation, whichever came first). Nevertheless, the low rate of pacing-related AEs support our assertion that the implantation procedures and management of the NeuRX RA/4 DPS was conducted in accordance with previous reports and the manufacturer’s recommendations.

More deaths were observed in the NIV plus pacing arm; however, the causes of death were similar across the arms and consistent with expected causes of death in ALS. This observation suggests that the NeuRX RA/4 DPS is not modifying the disease phenotype of ALS leading to death through some unusual mechanism. The causes of death would support the hypothesis that the disease course is accelerated by the intervention. It is perhaps important to emphasise that postoperative mortality was not encountered, with only one participant death within 3 months of implantation (and this was 45 days after the procedure). This rules out the direct effects of the implantation procedure itself as the cause of the observed harm in the NIV plus pacing arm.

A full health economic analysis of the cost-effectiveness of the NeuRX RA/4 DPS was planned. Given the lack of efficacy and apparent harm, this analysis did not go ahead. We have, however, compared the health-care resource use between the two arms. This is helpful in exploring if there were differences in other treatments or resources that could have impacted on the results. We found no difference in the way participants accessed health care. Similarly, access to other respiratory interventions, such as cough assist, breath stacking and suction, was comparable. There was a tendency for patients in the NIV arm to use cough assist and suction more frequently, whereas those in the NIV plus pacing arm used breath stacking more. The impact of these differences cannot be quantified, but they are unlikely to explain the differences in survival we observe.

Survival in relation to non-invasive ventilation and diaphragm pacing use and compliance

One concern was that perhaps patients in the NIV plus pacing arm would reduce NIV use in favour of DP use. If this occurred, then the negative impact of not using NIV, a proven intervention to prolong life, may have caused the survival difference between arms. Somewhat paradoxically, NIV use was similar in both arms, and, when modelled, NIV use did not influence survival, although this is likely to reflect increasing use of ventilation as prognosis deteriorates. We explored whether or not bulbar function at the time of randomisation and whether or not NIV tolerance influenced survival. In the RCT,1 NIV individuals with poor bulbar function did not experience the survival benefits that those with good bulbar function obtained. Bulbar dysfunction is one obstacle to successful NIV use, but there are others, and indeed, some patients with significant bulbar dysfunction can manage NIV. Therefore, looking at NIV tolerance is likely to be the most informative. The impact of bulbar function at the time of randomisation was negligible among the subgroup of patients who were intolerant of NIV and had a trend towards worse survival in the NIV plus pacing arm than in the NIV arm. Although the numbers are small, this would suggest that DP should not be considered an alternative in patients not able to tolerate NIV. This is particularly disappointing, as despite all efforts there is a consistent 20% or so of individuals with ALS who cannot get on with NIV and would benefit from an alternative means of respiratory support. 2,3

Diaphragm pacing was well tolerated, with the majority of patients reaching the set targets for pacing. We explored whether or not the amount of DP use influenced survival and did not find an association. One might have expected, given the overall findings, that the more an individual used DP, the worse the outcome would be. We did not observe such a dosing effect. The study was not designed to look for a dosing effect and one needs to be cautious in interpreting observational data. However, this observation would lend soft support to the hypothesis that the indirect effects of surgery itself may have been the major contributor to the harm.

Qualitative study

The inclusion of qualitative data relating to patient and carer experiences involved in trials is important, as it ensures that research does not focus only on intervention effectiveness, but also considers elements of implementation and acceptability to participants. The use of a qualitative substudy is a significant strength of this trial, as it provides a comprehensive analysis and interpretation of patient perspectives.

The qualitative data provide valuable insights into the views and experiences of patients and carers following their decision to take part in the trial of DP. For most participants there were few factors adversely impacting on their use of the intervention, although some patients in contrast described severe pain or discomfort.

The qualitative data confirm previous reports of pain experienced by some users of DP and highlight the wide variation in experiences: a longer than expected hospital stay; the perception among some that the surgery had negatively impacted on their functioning; the perceived ease of operation; issues of portability; and concern regarding fragility of the device.

In general, the participants found the device to be acceptable in terms of ease of operation and limited impact on QoL. The majority of users described hoped-for rather than perceived benefit. Although the majority of participants were uncertain regarding any benefit obtained, none appeared to regret their decision to trial the device. They described the limited options available to people with ALS and their willingness to try any available interventions.

Although the data offer important insights into patient and carer views and experiences of DP, the qualitative data may be limited by the small sample size and by the fact that these participants had been invited and agreed to participate in a clinical trial. We were successful in achieving a sample relatively balanced by participant sex; however, other differences between our qualitative substudy sample and wider groups of patients with motor neurone disease may need consideration.

Understanding the results

There has been a presumption that DP offers probable benefit; however, the mechanism by which such benefit is conferred is unclear and unproven. One potential mechanism put forward is that DP reverses an effect on the diaphragm muscle fibres whereby a shift from a predominance of efficient type I fibres to inefficient type IIb fibres occurs. 12 This hypothesis is extrapolated from observations in a population receiving closed-system invasive ventilation support and it is not known whether or not NIV has a similar effect on the diaphragm muscle fibres. 28 A further hypothesis is that DP leads to conditioning of the diaphragm muscle (i.e. increase in muscle mass). There are a few case reports of this, although the functional consequence of this is unknown should it occur. 29 Improved compliance of the lungs and prevention of basal collapse; restoration of the co-ordination of breathing, lost as a result of upper motor neurone dysfunction; and increased tidal volumes are other potential as yet unproven mechanisms. Unhelpfully, we can be no more certain as to why DP may be harmful. The possibilities are a direct effect of pacing (i.e. electrical stimulation) on either the muscle or the phrenic motor neurones. The physiological effects of pacing have not been studied in humans. In canines and rodents, it is clear that neuromuscular damage can be induced depending on the parameters of pacing and that the effects observed differ between healthy and disease models. 8,9 A simpler explanation may be that pacing causes excessive muscle fatigue or that asynchrony between pacing induced diaphragm contraction, patient- and/or NIV-triggered breaths is an issue.

Undertaking surgery in patients with ALS and respiratory failure does present some challenges and angst among health-care professionals. Our study confirmed the finding of others that general anaesthesia can be undertaken safely in terms of perioperative mortality, with no deaths within 30 days of the implantation procedure and only one within the first 3 postoperative months. 7 However, there is evidence that the consequences of surgery and general anaesthesia may be less immediate and not apparent initially. A retrospective review of ALS patients undergoing surgery for any reason demonstrated an apparent acceleration of ALS disease progression post surgery, suggesting a potential disease-modifying effect, albeit one that is not well understood. 10 Direct effects of anaesthetic agents and surgical stress, with the release of systemic pro-inflammatory cytokines, are postulated to contribute to the observed effect. It is possible that such indirect effects of surgery are contributing to the survival differences in the DiPALS trial; however, this is far from certain. In an attempt to minimise patient burden and focus on survival and cost-effectiveness, during the grant peer-review process secondary outcome measures detailing progression of ALS (ALSFRS-R score) and respiratory function (FVC) were removed from the protocol. When the negative outcome was identified, the TSC requested that an attempt was made to retrospectively collect any functional data on participants during the follow-up period. Clearly, this is an unplanned and retrospective analysis on incomplete data and conclusions based on this analysis need to be cautious. The baseline data at randomisation demonstrate the two arms to be balanced in terms of respiratory function and ALSFRS-R score. However, it does appear that there was a change in the rate of decline of FVC and ALSFRS-R score following implantation. This finding, however, would be consistent with either an indirect effect of surgery or a direct effect of pacing, as described in the Discussion.

The results of the Diaphragm Pacing in patients with Amyotrophic Lateral Sclerosis trial in context

Our findings are at odds with the FDA summary of SSPB, which reported a survival advantage for DP of 16.1 months from symptom onset and 9 months from the point of NIV initiation, compared with NIV alone, in a historical cohort. 11,30 This compares with our findings of a reduction in survival in the NIV plus DPS group by 17 months from symptom onset and 11.8 months from randomisation. The median survival from symptom onset in DiPALS was 45 months in the NIV arm, 28 months in the NIV plus pacing arm and 56 months in the SSPB pacing study.

This raises several possibilities, which include that there are differences between the populations in the DiPALS trial and in previous cohort studies; the DP intervention has been delivered differently; or other treatments that impact on survival are different between the DiPALS trial and previous studies. We will discuss these three possibilities next.

The Diaphragm Pacing in patients with Amyotrophic Lateral Sclerosis trial population compared with other studies of diaphragm pacing

The population in the DiPALS trial other than having a male preponderance is relatively typical of an ALS population with age, proportion of limb to bulbar onset, rate of decline of ALSFRS-R score, all as would be expected. 31 The complete data set, including baseline study population characteristics, from the uncontrolled multicentre cohort study which led to FDA approval on humanitarian grounds of the NeuRX RA/4 DPS (HDE), has to date not yet been published. Therefore, it is not possible to compare the populations and, therefore, it is challenging to fully understand the differences in the reported outcomes. The broad inclusion criteria for the pilot and pivotal phases of the cohort study were evidence of residual bilateral phrenic nerve function and a FVC of < 85% at screening and > 45% at DPS implantation, but specific details, such as the mean age, FVC and bulbar function of the recruited population, are not publicly available. 11 The FVC inclusion is higher than the 75% used in the DiPALS trial and if a significant number of patients lay in the 75–85% range for FVC, then this would contribute to some of the difference observed. More immediately, however, it is noteworthy that of the 144 patients enrolled only 106 were implanted and 84 included in the analysis of overall survival. The reasons for these exclusions are not provided, but we speculate that these may, again in part, explain the disparity. The cohort study contained a lead-in phase of 3 months before implantation during which time patients were monitored. The details of 38 patients who did not undergo implantation are not reported. It seems likely that some have been excluded because they rapidly deteriorated, but this is not clear and there may have been other criteria used. Of the 106 implanted, the data of only 84 patients contribute to the SSPB report, two patients having been lost to follow-up and 20 not meeting the HUD criteria. This cohort is therefore a highly selected group and may not be generalisable to the wider ALS population. In contrast, our trial used an intention-to-treat approach in which all consenting participants were analysed, including patients who subsequently declined rapidly in either group. Therefore, we suspect baseline difference in the study populations may be a major cause of observed survival differences.

Outside ALS, DP is licensed for patients with spinal cord injury in several territories including the USA and Europe. As with ALS, the marketing licence was granted on humanitarian grounds and on the basis of a small (n = 50) cohort study of patients with spinal cord injury resulting in a poorly controlled diaphragm and requiring continuous mechanical ventilation. 11,30 Forty-eight of the 50 had achieved a successful outcome, defined as a continuous period of at least 4 hours without the assistance of a mechanical ventilator; one participant was unable to achieve diaphragmatic pacing. The average follow-up was 1.4 years, during which four deaths were reported, none of which was considered to be related to the device. 11 Thus, as with the ALS cohort data, there is limited evidence to support or refute the safety of DP.

Delivery of diaphragm pacing in the Diaphragm Pacing in patients with Amyotrophic Lateral Sclerosis trial compared with other studies

The device manufacturer (Synapse Biomedical) provided training for surgeons and site staff in all aspects of the intervention, as is standard practice when it supplies the NeuRX RA/4 DPS within a service. For example, an experienced surgeon from Synapse Biomedical attended the implantation procedure and proctored the local site surgeon until he or she was competent to insert the devices. At all surgeries Synapse Biomedical staff were present to provide standard technical support. It is likely that the implantation procedures, which followed the manufacturer’s recommendations, were comparable to those undertaken during the earlier cohort study. Supporting this is the fact that the number of surgical complications was lower than previously reported and there was no mortality within the 30-day perioperative period. All diaphragms were stimulatable at surgery supporting our combined clinical and ultrasound assessment of residual phrenic nerve function. No surgeries needed to be abandoned, as planned in the protocol, if on intraoperative mapping of the diaphragm stimulation was not achievable. Following implantation the majority of patients became regular users (mean daily use 6.2 hours) and titrated their use over the course of the study, in line with the protocol and the manufacturer’s recommendations. There were six non-users in the NIV plus pacing arm: five of these did not undergo surgery and one stopped pacing after pulling the wires out after 1 month. The overall survival analysis in the DiPALS trial was by intention to treat; however, if these non-users were excluded, then the HR remained significant (HR 2.71, 95% CI 1.39 to 5.27).

Other confounders

Interventions proven to influence survival in ALS are riluzole, multidisciplinary team care and NIV. 1,4–6 Riluzole use was equal across groups, as this was one of the inclusion criteria. NIV use was comparable across both arms. All participating centres deliver multidisciplinary team care and were specialist centres for treating patients with ALS and respiratory failure. The heath-care resource use indicated patients accessed care in a similar manner, regardless of treatment group.

Limitations

Given that patients allocated DP underwent surgical intervention, it was unavoidable that participants were unblinded as to the intervention. The study assessors were also unblinded to the intervention. The trial statistician was unblinded, but withheld accumulating data from the study team. As the primary outcome measure was objective (overall survival), the risk of bias is small, but there is an unavoidable risk of bias in the subjective patient-reported secondary outcome measures. We considered inserting the DP devices in the control group but not connecting them (sham pacing), to reduce the risk of bias and also to be able to offer DP to control subjects at the end of the 12-month follow-up period, but concluded that implanting control subjects appears less reasonable if one considerd a possible negative outcome. The effect of DP on the ongoing use of NIV was a concern and we asked the question of whether or not patients stopped using the NIV system, which has established survival benefit, in favour of the DP system. However, our data show that this was not the case, with similar daily periods of NIV use across both groups.

There are imbalances between the treatment arms in our study, the most notable of which is that the NIV plus pacing arm was slightly older. We have adjusted the HRs for this (and other) covariates, but also propose that it is unlikely that such a small age difference would have a large impact on ALS survival. Older age is negatively associated with survival, but less so within the ALS population, in which prognosis is poor across all demographics; certainly, we are unaware of any previous work that has identified age as having an impact that could explain a difference of this magnitude. Similar numbers of patients across each group are reported to receive additional respiratory interventions. However, there are differences in the frequency of cough assist use and breath stacking use, among those given devices, across the treatment groups. The effects of these differences are unknown but, again, are unlikely to explain the observed poor survival in the NIV plus pacing group.

Developments since the completion of the Diaphragm Pacing in patients with Amyotrophic Lateral Sclerosis trial

Recently, a second randomised trial [Early Stage Amyotrophic Lateral Sclerosis Phrenic Stimulation (RespiStimALS); NCT01583088] has also stopped early citing ‘absence of benefit [and] a statistically significant excess mortality in the group of patients receiving active stimulation’. 32,33 There are two notable differences between this study and the DiPALS trial. First, DP implantation was carried out prior to respiratory failure, with the primary end point being the time to requiring NIV. Second, all participants underwent surgical procedure, but with the control group having a ‘sham’ procedure in which pacing was not instigated until after NIV had started. The fact that an excess mortality was observed in the intervention arm offers compelling evidence that pacing is actively harmful as opposed to surgery. Subsequently, a third randomised study of DPS (NCT01938495) in the USA has suspended enrolment pending further follow-up data from patients already randomised in the study.

Meaning of the study and implications for clinicians or policy-makers

Diaphragmatic pacing should not be used as a routine treatment for all patients with ALS in respiratory failure. We cannot exclude the possibility that it is beneficial in a subgroup of patients; however, this should not be assumed. Our findings demonstrate that insertion of the NeuRX RA/4 DPS for pacing is harmful when instigated at the point at which an individual with ALS develops respiratory failure. These findings are further supported by a second RCT in which implanting was performed earlier in the disease trajectory, which also stopped early because of excess mortality in the pacing arm (NCT01583088). 32

A poor prognosis and the absence of curative therapy understandably encourage a ‘nothing to lose’ approach among patients and some clinicians alike, with an attendant lowering of the standards of evidence required to adopt a new intervention. Our trial demonstrates the potential for harm that can arise from adopting this approach.

Recommendations for future research

The findings of our trial, and also those of Gonzalez-Bermejo,32 are at odds with the promising survival data presented in a previous cohort11 and, anecdotally, a small number of individual cases using diaphragmatic pacing in our own clinics. Before any further study of DP in ALS is undertaken in the patient population, we would encourage a meta-analysis of all current trial data on DPS in ALS. This may help to identify whether or not there are any specific circumstances in which a patient may benefit from DP, such as those with predominant upper motor neuron disease. In addition, the emerging findings within ALS should prompt some consideration of the evidence for DP use among patients with spinal cord injury. If any further studies take place in the future, they should include measures to understand the mechanism by which harm or benefit occurs as a result of DP.

The Diaphragm Pacing in patients with Amyotrophic Lateral Sclerosis Research Team

Contributions of authors

The following produced the first draft of the report: Christopher J McDermott (Consultant Neurologist), Mike J Bradburn (Statistician), Chin Maguire (Trial Manager), Cindy L Cooper (Professor of Health Services Research and Clinical Trials), Wendy O Baird (Professor of Health Services Research), Susan K Baxter (Qualitative Researcher), Judith Cohen (Research Fellow) and Hannah Cantrill (Research Assistant).

The following conceived of or designed the work: Christopher J McDermott (Consultant neurologist), Mike J Bradburn (Statistician), Cindy L Cooper (Professor of Health Services Research and Clinical Trials), Wendy O Baird (Professor of Health Services Research), Simon Dixon (Professor of Health Economics), Roger Ackroyd (Consultant Surgeon), Simon Baudouin (Senior Lecturer in in Critical Care), Andrew Bentley (Consultant in Respiratory and Intensive Care), Stephen Bianchi (Consultant in Respiratory Medicine), Stephen C Bourke (Consultant Physician in Respiratory and General Medicine), John Ealing (Consultant Neurologist), Patrick Fitzgerald (Health Economist), Simon Galloway (General and Upper Gastrointestinal Surgeon), Dayalan Karat (Consultant Surgeon), Nick Maynard (Consultant Surgeon, Upper Gastrointestinal Surgery), John Stradling (Professor of Respiratory Medicine), Kevin Talbot (Professor of Motor Neuron Biology), Tim Williams (Consultant Neurologist) and Pamela J Shaw (Professor of Neurology).

THE following were involved in the acquisition of data for the work: Christopher J McDermott (Consultant Neurologist), Roger Ackroyd (Consultant Surgeon), Simon Baudouin (Senior lecturer in Critical Care), Andrew Bentley (Consultant in Respiratory and Intensive Care), Richard Berrisford (Surgeon), Stephen Bianchi (Consultant in Respiratory Medicine), Stephen C Bourke (Consultant Physician in Respiratory and General Medicine), John Ealing (Consultant Neurologist), Mark Elliott (Consultant Respiratory Physician), Simon Galloway (General and Upper Gastrointestinal Surgeon), Hisham Hamdalla (Consultant Neurologist), C Oliver Hanemann (Consultant in Neurology), Philip Hughes (Consultant in General and Respiratory Medicine), Ibrahim Imam (Consultant Neurologist), Dayalan Karat (Consultant Surgeon), Nick Maynard (Consultant Surgeon, Upper Gastrointestinal Surgery), Richard W Orrell (Senior Lecturer and Consultant Neurologist), Abeezar Sarela (Consultant in Upper Gastrointestinal Surgical Oncology, Bariatric and Metabolic Surgery and Minimally Invasive Surgery), John Stradling (Professor of Respiratory Medicine), Kevin Talbot (Professor of Motor Neurone Biology), Martin Turner (Senior Clinician Scientist), Tim Williams (Consultant Neurologist) and Wisia Wedzicha (Consultant Respiratory Physician).

The following were involved in the analysis of data: Christopher J McDermott (Consultant Neurologist), Mike J Bradburn (Statistician) and Stephen C Bourke (Consultant Physician in Respiratory and General Medicine).

The following were involved in the interpretation of data for the work: Christopher J McDermott (Consultant Neurologist), Chin Maguire (Trial Manager), Judith Cohen (Research Fellow), Roger Ackroyd (Consultant Surgeon), Simon Baudouin (Senior Lecturer in in Critical Care), Andrew Bentley (Consultant in Respiratory and Intensive Care), Richard Berrisford (Surgeon), Stephen Bianchi (Consultant in Respiratory Medicine), Stephen C Bourke (Consultant Physician in Respiratory and General Medicine), Roy Darlison (TSC PPI representative), John Ealing (Consultant Neurologist), Mark Elliott (Consultant Respiratory Physician), Simon Galloway (General and Upper Gastrointestinal Surgeon), Hisham Hamdalla (Consultant Neurologist), C Oliver Hanemann (Consultant in Neurology), Philip Hughes (Consultant in General and Respiratory Medicine), Ibrahim Imam (Consultant Neurologist), Dayalan Karat (Consultant Surgeon), Roger Leek (TSC lay representative), Nick Maynard (Consultant Surgeon, Upper Gastrointestinal Surgery), Richard W Orrell (Senior Lecturer and Consultant Neurologist), Abeezar Sarela (Consultant in Upper Gastrointestinal Surgical Oncology, Bariatric and Metabolic Surgery and Minimally Invasive Surgery), John Stradling (Professor of Respiratory Medicine), Kevin Talbot (Professor of Motor Neuron Biology), Lyn Taylor (Principal Biostatistician), Anita K Simonds (Professor of Respiratory and Sleep Medicine), Tim Williams (Consultant Neurologist), Wisia Wedzicha (Consultant in Respiratory Physician) and Carolyn Young (Consultant Neurologist and Honorary Professor of Neurology).

The following drafted the monograph: Christopher J McDermott (Consultant Neurologist), Mike J Bradburn (Statistician), Chin Maguire (Trial Manager), Wendy O Baird (Professor of Health Services Research), Susan K Baxter (Qualitative Researcher) and Judith Cohen (Research Fellow).

The following revised the work critically for important intellectual content: Christopher J McDermott (Consultant Neurologist), Mike J Bradburn (Statistician), Chin Maguire (Trial Manager), Cindy L Cooper (Professor of Health Services Research and Clinical Trials), Wendy O Baird (Professor of Health Services Research), Susan K Baxter (Qualitative Researcher), Judith Cohen (Research Fellow), Simon Dixon (Professor of Health Economics), Andrew Bentley (Consultant in Respiratory and Intensive Care), Mark Elliott (Consultant Respiratory Physician), Philip Hughes (Consultant, General and Respiratory Medicine), Richard W Orrell (Senior Lecturer and Consultant Neurologist), Kevin Talbot (Professor of Motor Neurone Biology) and Tim Williams (Consultant Neurologist).

The following were involved in the final approval of the version to be published: Christopher J McDermott (Consultant Neurologist), Mike J Bradburn (Statistician), Andrew Bentley (Consultant in Respiratory and Intensive Care), Mark Elliott (Consultant Respiratory Physician), Philip Hughes (Consultant in General and Respiratory Medicine), Richard W Orrell (Senior Lecturer and Consultant Neurologist), Kevin Talbot (Professor of Motor Neurone Biology) and Tim Williams (Consultant Neurologist).

All authors agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Publications

McDermott CJ, Maguire C, Cooper CL, Ackroyd R, Baird WO, Baudouin S, et al. Protocol for diaphragm pacing in patients with respiratory muscle weakness due to motor neurone disease (DiPALS): a randomised controlled trial. BMC Neurol 2012:12:74.

McDermott CJ, Bradburn MJ, Maguire C, Cooper CL, Baird WO, Baxter SK, et al. Safety and efficacy of diaphragm pacing in patients with respiratory insufficiency due to amyotrophic lateral sclerosis (DiPALS): a multicentre, open-label, randomised controlled trial. Lancet Neurol 2015;14:883–92.

Data sharing statement

Please direct any requests to ctru@sheffield.ac.uk.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health.