Robot Assisted Training for the Upper Limb after Stroke (RATULS): study protocol for a randomised controlled trial

This study is a three-arm, pragmatic, observer-blind, multicentre RCT with embedded economic analysis and a process evaluation. Participants are randomised to receive either: robot-assisted training (in addition to usual NHS care); an enhanced upper limb therapy programme (in addition to usual NHS care); or usual NHS care in accordance with local clinical practice. Figure 1 summarises the study methods. The study is presented according to the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) [10] (SPIRIT Checklist, Additional file 1). Figure 2 shows the SPIRIT schedule of enrolment, interventions and assessments.

The study is being conducted in NHS stroke units in the UK. There are four RATULS study centres (Glasgow, North Tyneside, Northwick Park, and Romford) each consisting of a hub site with an InMotion robotic gym system and spoke sites which are stroke services in adjacent trusts that refer patients to take part in the study and provide usual NHS care.

Adults with a first-ever stroke who fulfil the following criteria are eligible to participate in the trial:

Study participants are recruited from both incident and prevalent stroke populations. Participants can be sought from a number of settings in both primary and secondary care including: stroke units; outpatient clinics; day hospitals; community rehabilitation services; and general practices. The study aims to recruit similar numbers of participants within: 0–3 months of stroke; >3–12 months after stroke; and >12 months to 5 years after stroke.

Individuals who are interested and potentially eligible to take part in the study are given an appointment for further discussion and consent. This may be conducted by a local study coordinator or NIHR LCRN staff. Written informed consent is obtained if the patient wishes to take part in RATULS.

A screening log is kept at each study centre. This records details of all inpatients, outpatients and primary care patients considered for the study and subsequently included or excluded.

Once written informed consent is obtained, a screening assessment is performed by the local study centre coordinator or NIHR LCRN staff. The following data are collected: demography; stroke details; comorbidity; and upper limb function (ARAT score [11]). If the patient fulfils the study inclusion and exclusion criteria, the local study coordinator/NIHR LCRN staff proceeds to the baseline assessment. If it is not possible to complete the baseline assessment on the same day as the screening assessment, eligibility for the study is re-confirmed on the day of the baseline assessment.

The following baseline data are collected: stroke severity (National Institute for Health Stroke Scale [12]); cognitive function (Montreal Cognitive Assessment [13]); language skills (Sheffield Aphasia Screening Test [14]); upper limb impairment (Fugl-Meyer Test (motor and sensory arm sections) [15]); activities of daily living (Barthel ADL Index [16, 17]; quality of life (EQ-5D-5 L [18]); upper limb pain (numerical rating scale [19]) and current upper limb rehabilitation treatments. In addition, patients are given a self-completion questionnaire containing pre-study resource utilisation questions (adaption of the Client Services Receipt Inventory [20‐22]) which they are asked to complete at the end of the assessment.

Randomisation is conducted by the local study coordinator/NIHR LCRN staff following completion of the baseline assessment. A central independent web-based service hosted by Newcastle University Clinical Trials Unit is used. Participants are stratified according to study centre, time since stroke and severity of upper limb function (ARAT score [11]), and randomised to either robot-assisted training, enhanced upper limb therapy, or usual care groups using permuted block sequences.

Outcomes are assessed at 3 months (±7 days) and 6 months (±7 days) following randomisation.

Assessments are undertaken in two stages:

Stage 1 is a self-completion postal questionnaire consisting of the Stroke Impact Scale (SIS) [33] (3 and 6 months) and the adapted Client Services Receipt Inventory resource utilisation questions (6 months only) [20‐22].

Stage 2 is a face-to-face assessment with a researcher blinded to randomisation group. The following data are collected: Barthel ADL Index [16, 17], EQ-5D-5L [18], ARAT [11], Fugl-Meyer Test (motor and sensory arm sections [15]), and adverse events. At the end of the 6-month stage-2 assessment, participants are given a further self-completion questionnaire and are asked to return this by post. This questionnaire contains time and travel resource use questions [34, 35].

Due to the nature of the interventions, it is not possible to blind participants or treating therapists to treatment allocation. It is intended that stage-2 outcome assessments are conducted by a researcher blinded to treatment allocation. After each outcome assessment the researcher is asked to record whether they have unintentionally become aware of treatment allocation due to conversation with the participant. Success of outcome assessment blinding will be reported.

No specific withdrawal criteria have been pre-set. Participants may withdraw from the study at any time for any reason. Data collected prior to withdrawal will be used in the study analysis unless consent for this is specifically withdrawn. Should a decision to withdraw from the study be made, a reason for withdrawal is sought but participants can chose to withdraw without providing an explanation.

Investigators, GPs, stroke physicians and therapists may also withdraw participants from the study at any time if they feel it is no longer in their interest to continue; for example, because of intercurrent illness or adverse events.

The safety of robot-assisted training, enhanced upper limb therapy and usual care is being evaluated by examining the occurrence of all adverse events and serious adverse events in accordance with National Research Ethics Committee (NRES) guidance for non-CTIMP trials [36].

The economic analysis will include a detailed micro-costing analysis, economic evaluation and a longer-term economic model. This will be based upon both a ‘within trial’ analysis and a modelling exercise to explore costs and effects over the longer term. Analyses will be carried out from the perspective of the NHS and personal and social services, but we will also take a societal perspective by including costs borne by the participants and their informal carers. All relevant costs associated with providing the interventions will be measured, this will include the cost of using the InMotion robotic gym system, costed on a per patient basis. All costs will be derived using routine data sources [38] and study-specific estimates. Where appropriate, discounting will be applied to costs and outcomes [39]. Costs in the follow-up period will also be taken into account; this includes secondary care resource, e.g. inpatient stays and outpatient visits; primary care resource use, e.g. general practice, therapy visits and prescription costs. These data will be collected using a health service utilisation questionnaire (adaption of the Client Services Receipt Inventory [20‐22]) administered at baseline and 6 months post randomisation. Patient costs will also be collected via a time-and-travel questionnaire based upon one successfully used in a number of previous NIHR HTA-funded trials [34, 35]. This will include questions relating to travel time, time away from employment (if appropriate) and time spent providing care. The within-trial analysis will also compare changes in health-related quality of life, based on responses to the EQ-5D-5L at baseline, 3 and 6 months post randomisation and scored using population tariffs [40]. These data will be combined with study participants’ mortality to estimate quality-adjusted life years (QALYs). This measure provides a profile of quality of life over time. The results of the analyses will be presented as point estimates of mean incremental costs and QALYs. Techniques, such as bootstrapping, will be used alongside deterministic sensitivity analyses to address uncertainty [41]. In addition, a within trial cost-utility analysis will be performed where both costs and QALY data will be combined into an incremental cost per QALY. The cost-utility analysis will include deterministic and stochastic sensitivity analysis, presented as point estimates and cost-effectiveness acceptability curves (CEACs).

An economic model will also be developed to assess the cost and health consequences measured in terms of QALYs of stroke recovery beyond the 6-month timeframe of the trial. The data from the trial will be the main source of data for this model but further data with which to model outcomes beyond a 6-month follow-up will be systematically derived from the academic literature and other existing data sources following guidance for best practice [42]. These data will include information on factors, such as the incidence of hospitalisation and the need for residential/nursing home care, beyond the trial follow-up period. Sensitivity analysis will be applied to the model using probabilistic and deterministic sensitivity analyses to address parameter and other forms of uncertainty. The data on both costs and QALYs for both trial- and model-based analyses will be reported separately.

Alongside the RCT, a two-stage process evaluation is being conducted to understand both (1) participants‘ and health service professionals’ experiences of robot-assisted training; enhanced upper limb therapy and usual care and (2) factors affecting the implementation of the trial within and across study sites. The process evaluation will capture data concerning feasibility and accumulating experience of the therapies being provided. In stage 1 data collection is by semi-structured interview using a pre-developed and piloted interview schedule. Data collection in stage 2 is primarily by interview; however, analysis also draws upon trial data including baseline, therapy and outcome (3 and 6 month) assessments. Interviews are primarily being conducted face to face; however, due to the geographical spread of the study sites, some follow-up interviews are being conducted by telephone for efficiency (these are particularly appropriate for health service professionals). Data collection and analysis relating to study of implementation factors will be informed by Normalization Process Theory (NPT) [43].

Personal data are regarded as strictly confidential. Original paper Case Record Forms containing study data are stored in the investigator site file at each research site. All study files are securely stored and access restricted to staff involved in the study. Research staff at sites enter data from paper forms onto a secure web-based electronic database run and maintained by Newcastle University. Data are entered using participant-unique study numbers only. Access to this database is password-protected and limited to staff at research sites or Newcastle University who are involved in the study.

The InMotion robotic gym computers store data from each participant session. Data are stored by unique study number only. Periodically, these data are copied from the robot computer system into an electronic database maintained by Newcastle University.

The study complies with the Data Protection Act 1998, and Caldicott Guardian approval for use of patient identifiable data.

The chief investigator has overall responsibility for study conduct. The principal investigators are responsible for the day-to-day study conduct at their individual sites.

The trial is managed by a coordinating centre based at Newcastle University which provides day-to-day support for the sites and provides training through investigator meetings, site initiation visits and routine monitoring visits. A Trial Management Group (TMG) has been convened and the TMG meets regularly during the study.

Quality control is maintained through adherence to Newcastle Biomedicine Clinical Research Platform SOPs, the study protocol and research governance regulations. General monitoring of study conduct and data collected is being performed by a combination of central review and site-monitoring visits. The main areas of focus include consent, serious adverse events and essential documents in study files. All monitoring findings are reported and followed up with the appropriate persons in a timely manner.

A Trial Steering Committee (TSC) has been convened. This comprises an independent chair, three other independent members, a patient and/or a carer representative and the chief investigator. The TSC has agreed a charter of operation and meet at least annually.

An independent Data Monitoring and Ethics Committee (DMEC) has been convened to undertake independent review. This comprises five independent members including expert health care professionals and a statistician. Only the DMEC has access to unblinded outcome data before the trial ends. The DMEC has agreed a charter of operation and meets at least annually.

The data are the property of the chief investigator and co-investigator(s). Publication will be the responsibility of the chief investigator.

The study will be presented at national and international conferences, and reported in peer-reviewed journals and a NIHR HTA monograph. Reports will be written for the study sponsor and regulatory bodies. A summary of the results will be sent to study participants and be available on the study website: https://​research.​ncl.​ac.​uk/​ratuls/​.

Anonymised data will be provided to research databases as requested (e.g. the Cochrane Collaboration, the Virtual International Stroke Trials Archive (VISTA)) to enable future meta-analyses. Anonymised robot kinematic and kinetic data will be provided to co-investigators for exploratory analyses.

RATULS commenced recruitment in April 2014. Four NHS study centres (Glasgow, North Tyneside, Northwick Park, and Romford) are participating. The RATULS trial has recruited 468 patients at the time of submission of this manuscript. Protocol version 3, dated 2 August 2016, was used to prepare the manuscript.