The effect of rehabilitation interventions on physical function and immobility-related complications in severe stroke: a systematic review

Objective To evaluate the effectiveness of rehabilitation interventions on physical function and immobility-related complications in severe stroke. Design Systematic review of electronic databases (Medline, Excerpta Medica database, Cumulative Index to Nursing and Allied Health Literature, Allied and Complementary Medicine Database, Physiotherapy Evidence Database, Database of Research in Stroke, Cochrane Central Register of Controlled Trials) searched between January 1987 and November 2018. Methods The Preferred Reporting Items for Systematic Reviews and Meta-Analysis statement guided the review. Randomised controlled trials comparing the effect of one type of rehabilitation intervention to another intervention, usual care or no intervention on physical function and immobility-related complications for patients with severe stroke were included. Studies that recruited participants with all levels of stroke severity were included only if subgroup analysis based on stroke severity was performed. Two reviewers screened search results, selected studies using predefined selection criteria, extracted data and assessed risk of bias for selected studies using piloted proformas. Marked heterogeneity prevented meta-analysis and a descriptive review was performed. The Grading of Recommendations Assessment, Development and Evaluation approach was used to assess evidence strength. Results 28 studies (n=2677, mean age 72.7 years, 49.3% males) were included in the review. 24 studies were rated low or very low quality due to high risk of bias and small sample sizes. There was high-quality evidence that very early mobilisation (ie, mobilisation with 24 hours poststroke) and occupational therapy in care homes were no more effective than usual care. There was moderate quality evidence supporting short-term benefits of wrist and finger neuromuscular electrical stimulation in improving wrist extensor and grip strength, additional upper limb training on improving upper limb function and additional lower limb training on improving upper limb function, independence in activities of daily living, gait speed and gait independence. Conclusions There is a paucity of high-quality evidence to support the use of rehabilitation interventions to improve physical function and reduce immobility-related complications after severe stroke. Future research investigating more commonly used rehabilitation interventions, particularly to reduce poststroke complications, is required. PROSPERO registration number CRD42017077737


STRENGTHS AND LIMITATIONS OF THIS STUDY
 This is the first systematic review to investigate rehabilitation interventions specifically to survivors of severe stroke  The review included outcomes on immobility-related post-stroke complications, which contribute to high levels of caregiver burden  Marked heterogeneity of included studies prevented meta-analysis  Most included studies were rated as low or very low-quality evidence due to unclear or high risk of bias as well as recruitment of very small samples Severe stroke can be understood as a stroke resulting in a significant amount of brain tissue damage and multiple neurological impairments, which leads to a significant loss of function and residual disability.|12| Dependent upon how it is measured, 14 -31% of people who sustain a stroke globally are classified as having a severe stroke,|13-18| a cohort of the stroke population that experiences worse outcomes compared to survivors of less severe stroke.|19-30| In the initial hospitalisation phase post-stroke, they are more likely to develop acute medical complications, which are negatively associated with functional recovery.|19| Three month mortality can be as high as 40%, compared to just under 5% for those patients with mild stroke.|20-22| Survivors of severe stroke pend longer in hospital, resulting in increased hospital costs, and demonstrate slower and less functional recovery, resulting in greater dependency when they are discharged from hospital.|14,15,23,25| For those discharged from hospital, survivors of severe stroke are at least eight times more likely to be discharged to a nursing home.|25,26| Longer-term care costs, which mostly support survivors of severe stroke, represent 49% of total stroke care spending globally.|27| In the first year post severe stroke, mortality can be as high as 60%|20| and survivors of severe stroke also experience very high levels of immobility-related complications, such as falls, contracture, pain, and pressure sores.|28,29| Due to this residual disability, the physical assistance provided by caregivers to look after survivors of This systematic review aims to establish the effectiveness of rehabilitation interventions on physical function and immobility-related complications for survivors of severe stroke and identify areas for future rehabilitation research for these patients.

METHODS
The systematic review has been reported according to the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) statement (see supplementary file 1).|32| The protocol for the systematic review has been published previously.|33|

Study design
The systematic review included randomised controlled trials (RCTs). The systematic review excluded quasi-experimental, correlational and descriptive study designs. Studies were selected according to the PICO (participant, intervention, comparator and outcome) format.
The systematic review protocol provides full details of the PICO components.| 33|

Search strategy
Information sources included studies were hand searched and any potentially relevant study was included for review. Forward citation checks of included studies were also performed. To avoid language or cultural bias, studies in any language or geographical location were included.

Data management and study selection
The results from the literature search were uploaded to a reference management programme (Refworks) and duplicate references were removed. A final list of non- independently performed data extraction for all eligible articles using a data extraction proforma previously piloted.

Risk of bias and quality assessment
Risk of bias was assessed by two review authors independently (MM and JJ) using the Cochrane Collaboration tool for assessing the risk of bias across six main domains (sequence generation, allocation concealment, blinding, incomplete outcome data selective outcome reporting, other bias) .|34| A risk of bias judgement of 'high', 'low' or 'unclear' was determined for each of these main domains. The strength of evidence was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach.|34| The five domains considered by the GRADE approach included risk of bias, inconsistencies between studies, indirectness, imprecision and publication bias. The quality of the evidence was ranked high, medium, low or very low by two review authors independently (MM and JJ).

Data analysis
As stated in the systematic review protocol, it was decided that if more than five adequately powered studies demonstrate homogeneity in terms of rehabilitation interventions and outcomes, results for individual outcomes would be pooled quantitatively using metaanalysis. Due to the limited number and marked heterogeneity in rehabilitation interventions and outcomes of the selected studies, it was not appropriate to undertake a meta-analysis. Therefore, a descriptive review of results was performed. As there may be differences in recovery rates and outcomes according to the time post-stroke, studies were grouped into three timeframes post-stroke determined on the basis of when participants were recruited to the study and when the study finished. These timeframes were the acute to early subacute stage (up to 3 months post-stroke), acute to late-subacute stage (up to 6 months post-stroke) and chronic stage (greater than 6 months post-stroke). These timeframes were chosen based on recommendations for the standardised measurement of sensorimotor recovery in stroke trials.|35| Study findings were presented according to these three timeframes.  high quality of evidence are reported in this section. Forest plots are included for key outcomes, although effect sizes were not estimable for studies that did not provide raw data and were not pooled due to heterogeneity in rehabilitation outcomes.

Sensorimotor Function
Seventeen studies evaluated changes in sensorimotor function. Ten studies were completed In the acute to early subacute phase post-stroke, there was moderate quality evidence from one study that a 6-week course of neuromuscular electrical stimulation (NMES) applied to the wrist and finger extensors in conjunction with usual therapy resulted in no improvement in wrist active movement compared to usual therapy.|49| Wrist strength and grip strength improved in the NMES group during the treatment period although these improvements were not evident at the 9-month follow-up.

Activities of Daily Living
Twenty studies explored independence and ability to perform activities of daily living (ADLs).
Eleven studies were completed in the acute to early subacute phase,|36,37,41-43,45-50| seven studies were completed in acute to late subacute phase|52,54-57,60-63| and two studies were completed in the chronic phase.|65,67| Eighteen studies used the Barthel Index as the main outcome measure to assess independence in ADLs. Four studies used the Modified Rankin Scale and three studies used the Functional Independence Measure. Figure   4 provides a visual representation of studies' effect sizes.
In the acute to early subacute phase, there was high quality evidence that frequent, higher dose, very early mobilisation commencing within 24 hours post-stroke did not result in more patients being less dependent in ADLs at 3 months post-stroke compared to usual care, which traditionally started more than 24 hours post-stroke.|36| However, caution is required with interpreting this finding as the sub-group analysis of patients with severe stroke was not powered for this outcome. There was moderate quality evidence that a 6-week course of NMES applied to the wrist and finger extensors in conjunction with usual therapy resulted in no difference in ADL independence compared to usual care.|49| In the acute to late subacute phase, there was moderate quality evidence that additional LL therapy in conjunction with regular physical rehabilitation performed in the first 20 weeks post-stroke improved ADL independence whilst the intervention was being delivered when compared to regular physical rehabilitation alone.|57| However, these improvements were not seen 6 months post-stroke.
In the chronic phase, there was high quality evidence that a 3-month OT intervention provided to residents in care homes resulted in no difference in ADL independence compared to usual care.|65| Similar caution is required with interpreting this finding as the sub-group analysis of patients who were severely or very severely disabled was not powered for this outcome.

Gait
Nine studies investigated gait, which included gait ability and gait speed. Six studies were performed in the acute to early subacute phase,|38-40,43,45,48| two studies were performed in the acute to late subacute phase|57,60| and one study was performed in the chronic phase.|64| The Functional Ambulation Classification was used in eight studies, making it the most frequently used outcome measure of gait ability. The 10-metre walk test was used in five studies, making it the most frequently used outcome measure of gait speed. Figure 5 provides a visual representation of studies' effect sizes.
Only one study demonstrated moderate quality evidence. In the acute to late subacute phase, additional LL therapy in conjunction with regular physical rehabilitation performed in the first 20 weeks post-stroke improved gait ability and speed when compared to regular physical rehabilitation alone.|57| However, these improvements were not seen 6 months post-stroke.

Mortality
One study investigated the effect of very early mobilisation on mortality.|36| There was high quality evidence that frequent, higher dose, very early mobilisation commencing within 24 hours post-stroke did not result in more patients dying at 3 months when compared to usual care, which traditionally started more than 24 hours post-stroke.

Other Outcomes
There was low quality of evidence for cardiorespiratory function (2 studies

Implications for Practice and Research
In light of these findings, it may be necessary to re-evaluate the design of future trials investigating rehabilitation interventions in severe stroke.

Strengths and Limitations
In terms of strengths, this is the first systematic review to investigate rehabilitation interventions specifically to survivors of severe stroke, who tend to be underrepresented in stroke rehabilitation research, and the identification of topics for future rehabilitation research will hopefully guide much needed research for this cohort of the stroke population.
As well, the outcomes of the review focussed on not just physical function but immobilityrelated post-stroke complications, which are known to be higher in the severe stroke population and contribute to high levels of caregiver burden.|28-30| In terms of limitations, it has been reported that the defining severe stroke is difficult due to different criteria used to classify severity.|71| The use of objective scores on validated outcome measures to classify stroke severity in our systematic review, necessary to ensure that participants had actually sustained a severe stroke, may have precluded the inclusion of studies that either used different scoring systems or outcome measures to classify stroke severity. However, these studies were discussed in detail amongst three review authors to determine suitability for inclusion and therefore it is likely that the number of relevant studies excluded from the review was minimal. Another limitation is the use of data from subgroups within larger clinical trials. As subgroup analyses may not be powered to detect changes between groups, caution is required in the interpretation of findings from these trials. In addition, raw subgroup data were not fully reported in some studies preventing estimation of effect sizes.

CONCLUSION
There was a paucity of high-quality evidence to support the use of rehabilitation

METHODS
Protocol and registration 5 Indicate if a review protocol exists, if and where it can be accessed (e.g., Web address), and, if available, provide registration information including registration number.
Abstract page, 2 Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, language, publication status) used as criteria for eligibility, giving rationale.

3
Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched.

4
Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated.

Supplementary file
Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis).

3, 4
Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

3, 4
Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made. 3

Risk of bias in individual studies
12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis.

DISCUSSION
Summary of evidence 24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers).

9-12
Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias).     Usual care group-BI 0-4 n=234 BI 5-9 n=104 BI, RMI, GDS, EQ-5D-3L No differences between the groups on any outcome measure at 3-, 6-and 12months post-randomisation. Higher fall rate per resident in OT intervention group at 3 months.   Two studies explored participants' cardiorespiratory response to different types of treadmill gait training within the acute to early subacute phase post-stroke. 4,8 There was low-quality evidence that 2 weeks of robot-assisted bodyweight supported treadmill gait training delivered in the first 6 weeks post-stroke improved peak VO 2 compared to conventional gait training. 4 There was low-quality evidence that a 4-week course of bodyweight supported treadmill training delivered in the first 3 months post-stroke was not perceived to be more effortful than conventional gait training. 8 Neurological Impairment Three studies evaluated changes in neurological function. 6,25,27 In the acute to early subacute phase post-stroke, there was very low-quality evidence that there was no difference in an intensive or ordinary 2-week acute physical rehabilitation programme on reducing neurological impairment at 2 weeks and 6 months post-stroke. 6 In the acute to late subacute phase post-stroke, there was very low-quality evidence that a 3-month course of robot-assisted bodyweight supported treadmill gait training commenced within the first 6 weeks post-stroke was just as effective as conventional gait training on improving neurological function. 25 There was very low-quality evidence that an 8-week course of acupuncture provided in conjunction with rehabilitation during the subacute phase of stroke reduced neurological impairment compared to rehabilitation alone. 27 Sensorimotor Function Sixteen studies evaluated changes in sensorimotor function. Nine studies were performed in the acute to early subacute phase post-stroke, [3][4][5]7,8,[10][11][12][13] five studies in the acute to late subacute phase post-stroke, 18,21,25,27,28 and two studies in the chronic phase post-stroke. 29,31 In the acute to early subacute phase post-stroke, there was low quality evidence from two studies that thermal stimulation in conjunction with standard rehabilitation resulted in improvements in lower limb sensorimotor function and strength when compared to standard rehabilitation alone. 5,10 Improvements in lower limb sensorimotor function were maintained at 12 months post-intervention. There was low quality evidence that 2 weeks of robot-assisted bodyweight supported treadmill gait training resulted in improvements in lower limb sensorimotor function but not strength compared to conventional gait training. 4 There was low quality evidence that there was no difference between: 4 weeks of robotassisted treadmill gait training and conventional gait training on improving lower limb sensorimotor function; 13 wearing a cueing wristwatch and wearing a sham wristwatch for 3 hours per weekday for 3 weeks during rehabilitation on improving upper limb sensorimotor function and number of arm movements; 7 a 4-week course of bodyweight supported treadmill training and conventional overground gait training on improving lower limb strength; 8 and a 5-week course of additional upper limb therapy provided by a qualified physiotherapist or a physiotherapy assistant and standard physiotherapy on improving upper limb motor activity and grip strength. 11 There was very low-quality evidence that a thrice weekly, 6-week course of electromyography (EMG) biofeedback combined with conventional physiotherapy had no  3 There was very low-quality evidence that a 3-month course of acupuncture in conjunction with rehabilitation resulted in better upper and lower limb sensorimotor function when compared to rehabilitation alone. 12 In the acute to late subacute phase post-stroke, there was very low quality evidence that a 6-week course of robotic tilt-table verticalisation that combines cyclic leg movements and FES and used in conjunction with standard physiotherapy resulted in better lower limb strength and sensorimotor function compared to physiotherapy-assisted verticalisation using a standard tilt-table and used in conjunction with standard physiotherapy. 18 There was very low-quality evidence that an 8-week course of acupuncture provided in conjunction with rehabilitation resulted in improvements in upper and lower limb sensorimotor function compared to rehabilitation alone. 27 There was very low-quality evidence that a 3-month course of nurse-led acupressure resulted in improvements in upper and lower limb motor function compared to routine care. 28 There was very low quality evidence that there was no difference between: a functionally-orientated and a sensorimotor integrative occupational therapy treatment approach delivered over 8 weeks on improving upper limb sensorimotor function; 21 and a 3-month course of robot-assisted bodyweight supported treadmill gait training and conventional gait training on improving lower limb power. 25 In the chronic phase post-stroke, there was very low-quality evidence that a 6-week course of robot-assisted bodyweight supported treadmill gait training using slower treadmill speeds resulted in improvements in lower limb sensorimotor function compared to similar treadmill training using faster treadmill speeds. 29 There was very low-quality evidence that either an intensive therapist-driven UL protocol or an intensive robotic-driven UL protocol delivered thrice weekly for 6 weeks resulted in an improvement in shoulder and elbow sensorimotor function. 31 Activity Activities of Daily Living Sixteen studies explored independence and ability to perform activities of daily living (ADLs). Nine studies were completed in the acute to early subacute phase, 2,6-8,10-13,15 six studies were completed in acute to late subacute phase 17,[19][20][21]25,27,28 and one study was completed in the chronic phase. 32 In the acute to early subacute phase, there was low quality evidence that a 6-week course of thermal stimulation used in conjunction with standard rehabilitation resulted in improvements in ADL independence 3 months post-stroke compared to standard rehabilitation alone, although improvements were not seen at 6 months post-stroke. 10 There was low quality evidence that there was no difference between: regular physiotherapy and regular physiotherapy in conjunction with use of an Oswestry standing frame delivered over 14 consecutive weekdays in the first 3 months post-stroke on ADL independence; 2 wearing a cueing wristwatch and wearing a sham wristwatch for 3 hours per weekday for 3 weeks during rehabilitation on ADL independence; 7 a 4-week course of bodyweight supported treadmill training and conventional overground gait training on improving ADL independence; 8 a 5-week course of additional upper limb therapy provided by a qualified physiotherapist or a physiotherapy assistant and standard physiotherapy on improving ADL independence; 11 and 4 weeks of robot-assisted treadmill gait and conventional overground gait training on ADL independence. 13 There was very low-quality evidence that there was no difference in an intensive or ordinary 2-week acute physical rehabilitation programme in improving ADL independence at 2 weeks and 6 months post-stroke. 6 There was very low-quality evidence that a 3-month course of acupuncture in conjunction with rehabilitation resulted in better ADL independence when compared to rehabilitation alone. 12 There was very low-quality evidence that providing additional physiotherapy in conjunction to regular rehabilitation in the first few weeks poststroke resulted in improvements in ADL independence at 6 months post-stroke compared to regular rehabilitation alone. 15 In the acute to late subacute phase, there was low quality evidence that a 6-month course of a staged physical rehabilitation programme resulted in greater improvements in ADL independence compared to usual care that did not involve formal rehabilitation. 17 There was very low-quality evidence that a monthly home-based physiotherapy programme delivered over 6 months resulted in improvements in ADL independence compared to standard care. 19,20 There was very low-quality evidence that there was no difference between a functionally orientated or a sensorimotor integrative occupational therapy treatment approach delivered over 8 weeks on ADL independence. 21 There was very lowquality evidence that a 3-month course of robot-assisted bodyweight supported treadmill gait training resulted in improvements in ADL independence compared to conventional gait training. 25 Improvements were only seen in the cohort of participants who demonstrated significant motor impairment. Improvements were maintained at the 2-year follow-up. 26 There was very low-quality evidence that an 8-week course of acupuncture provided in conjunction with rehabilitation during the subacute phase of stroke improved ADL independence compared to rehabilitation alone. 27 There was very low-quality evidence that a 3-month course of nurse-led acupressure resulted in improvements in ADL independence compared to routine care. 28 In the chronic phase, there was very low-quality evidence that a 16-week course of trunk acupuncture combined with rehabilitation training resulted in greater improvements in ADL independence compared to rehabilitation training alone. 32 Balance and Postural Control Eight studies investigated balance and postural control. Four studies were completed in the acute to early subacute phase, 2,5,10,16 two studies were completed in the acute to late subacute phase 18,25 and two studies were completed in the chronic phase. 29,32 In the acute to early subacute phase, there was low quality evidence that a 6-week course of thermal stimulation in conjunction with standard rehabilitation resulted in improvements in trunk postural control but not balance compared to standard rehabilitation alone. 5 In a separate study, there was low quality evidence that a 6-week course of thermal stimulation in conjunction with standard rehabilitation resulted in improvements in balance 3 months post-stroke compared to standard rehabilitation alone, although improvements were not seen at 6 months post-stroke. 10 There was low quality evidence that there was no difference between regular physiotherapy and regular physiotherapy in conjunction with use of an Oswestry standing frame delivered over 14 consecutive weekdays in the first 3 months poststroke on trunk postural control. 2 There was low quality evidence that an 8-week course of physiotherapy involving early mobilisation combined with the Bobath approach resulted in improvements in balance when compared to physiotherapy just involving the Bobath approach. 16 In the acute to late subacute phase, there was very low quality evidence that a 6-week course of robotic tilt-table verticalisation that combines cyclic leg movements and FES and used in conjunction with standard physiotherapy resulted in improved postural control during different activities compared to physiotherapy-assisted verticalisation using a standard tilt-table and used in conjunction with standard physiotherapy. 18 There was very low-quality evidence that a 3-month course of robot-assisted bodyweight supported treadmill gait training resulted in improvements in trunk control compared to conventional gait training. 25 Improvements were only seen in the cohort of participants who demonstrated significant motor impairment.
In the chronic phase, there was very low-quality evidence that a 6-week course of robotassisted bodyweight supported treadmill gait training resulted in improvements in balance regardless if slower or faster treadmill training speeds were used. 29 There was very lowquality evidence that a 16-week course of trunk acupuncture combined with rehabilitation training resulted in greater improvements in balance compared to rehabilitation training alone. 32 Gait Eight studies investigated gait, which included gait ability and gait speed. Six studies were performed in the acute to early subacute phase, [3][4][5]8,10,13 one study was performed in the acute to late subacute phase 25 and one study was performed in the chronic phase. 29 In the acute to early subacute phase, there was low quality evidence from two studies that a 6week course of thermal stimulation in conjunction with standard rehabilitation resulted in improvements in gait ability compared to standard rehabilitation alone. 5,10 There was low quality evidence that 4 weeks of robot-assisted treadmill gait training resulted in better gait ability than conventional gait training. 13 There was low quality evidence that there was no difference between: a 2-week course of robot-assisted bodyweight supported treadmill gait training and conventional gait training delivered in the first 6 weeks post-stroke on improving gait ability; 4 a 4-week course of bodyweight supported treadmill training and conventional overground gait training on improving gait ability; 8 and a thrice weekly, 6-week course of EMG biofeedback combined with conventional physiotherapy and conventional physiotherapy alone in improving gait speed. 3 In the acute to late subacute phase, there was very low-quality evidence that a 3-month course of robot-assisted bodyweight supported treadmill gait training resulted in improvements in gait ability compared to conventional gait training. 25 Improvements were only seen in the cohort of participants who demonstrated significant motor impairment. Improvements were maintained at the 2-year follow-up. 26 In the chronic phase, there was very low-quality evidence that a 6-week course of robotassisted bodyweight supported treadmill gait training using slower treadmill speeds resulted in improvements gait ability compared to similar treadmill training using faster treadmill speeds. 29 General Physical Activity Seven studies examined the effects of different interventions on improving general physical activity. Six studies were performed in the acute to early subacute phase 2,3,5,10,11,16 and one study was performed in the acute to late subacute phase. 25 In the acute to early subacute phase, there was low quality evidence from two studies that thermal stimulation in conjunction with standard rehabilitation resulted in improvements in physical activity when compared to standard rehabilitation alone. 5,10 Improvements were seen up until 3 months post-intervention but disappeared at the 6-month follow-up. There was low quality evidence that an 8-week course of physiotherapy involving early mobilisation combined with the Bobath approach resulted in improvements in physical activity when compared to physiotherapy just involving the Bobath approach. 16 There was low quality evidence that there was no difference between: regular physiotherapy and regular physiotherapy in conjunction with use of an Oswestry standing frame delivered over 14 consecutive weekdays in the first 3 months post-stroke on physical activity; 2 and a 5-week course of additional upper limb therapy provided by a qualified physiotherapist or a physiotherapy assistant and standard physiotherapy on improving physical activity. 11 There was very lowquality evidence that there was no difference between a thrice weekly, 6-week course EMG biofeedback combined with conventional physiotherapy and conventional physiotherapy alone on improving physical activity. 3 In the acute to late subacute phase, there was very low-quality evidence that a 3-month course of robot-assisted bodyweight supported treadmill gait training resulted in improvements in physical activity compared to conventional gait training. 25 Improvements were only seen in the cohort of participants who demonstrated significant motor impairment. Improvements were maintained at the 2-year follow-up. 26 Upper Limb Function Two studies investigated changes in upper limb function. 11,31 In the acute to early subacute phase, there was low quality evidence that a 5-week course of additional upper limb therapy provided by a qualified physiotherapist was no more effective at improving upper limb function than additional upper limb therapy provided by a physiotherapy assistant or to standard physiotherapy. 11 In the chronic phase, there was very low-quality evidence that there was no improvement in upper limb function with either an intensive therapist-driven UL protocol or an intensive robotic-driven UL protocol delivered thrice weekly for 6 weeks. 31 Participation Extended Activities of Daily Living Four studies investigated the effect of different interventions on extended ADLs. 2,3,9,21 In the acute to early subacute phase, there was low quality evidence that there was no difference between: regular physiotherapy and regular physiotherapy in conjunction with use of an Oswestry standing frame delivered over 14 consecutive weekdays on ability to perform extended ADLs at 6 months post-stroke, 2 and an 8-week course of rehabilitation with the addition of a leg cycling machine compared to regular rehabilitation alone on extended ADLs 6 months post stroke. 9 There was very low-quality evidence that there was no difference between a thrice weekly, 6-week course of electromyography (EMG) biofeedback combined with conventional physiotherapy and conventional physiotherapy alone in improving performance in extended ADLs time. 3 In the acute to late subacute phase, there was very low-quality evidence that there was no difference between a functionally orientated or a sensorimotor integrative occupational therapy treatment approach delivered over 8 weeks on the ability to prepare meals. 21 Perceived Health Status Two studies explored carers' and patients' perceived health status. 2,31 In the acute to early subacute phase, there was low quality evidence that there was no difference between regular physiotherapy and regular physiotherapy in conjunction with use of an Oswestry standing frame delivered over 14 consecutive weekdays on carer's perceived health status at 12 weeks and 6 months post-stroke. 2 In the chronic phase, there was very low-quality evidence that there was no change in patient's perceived health status with the provision of either an intensive therapist-driven UL protocol or an intensive robotic-driven UL protocol delivered thrice weekly for 6 weeks. 31 Quality of Life There was very low-quality evidence that a monthly home-based physiotherapy programme delivered over 6 months resulted in an improvement in quality of life compared to standard care. 19 Complications Caregiver Burden There was low quality evidence that there was no difference between regular physiotherapy and regular physiotherapy in conjunction with use of an Oswestry standing frame delivered over 14 consecutive weekdays in the first 3 months post-stroke on caregiver strain and psychological well-being at 12 weeks and 6 months post-stroke. 2 Depression Three studies explored changes in depression. 2,20,31 In the acute to early subacute phase, there was low quality evidence that there was no difference between regular physiotherapy and regular physiotherapy in conjunction with use of an Oswestry standing frame delivered over 14 consecutive weekdays on depression at 12 weeks and 6 months post-stroke. 2 In the acute to late subacute phase, there was very low-quality evidence that a monthly homebased physiotherapy programme delivered over 6 months resulted in a reduction in level of depression compared to standard care. 20 In the chronic phase, there was very low-quality evidence that there was no difference between an intensive therapist-driven UL protocol and an intensive robotic-driven UL protocol delivered thrice weekly for 6 weeks in reducing depression. 31 Shoulder Pain/Dislocation There was very low-quality evidence that either an intensive therapist-driven UL protocol or an intensive robotic-driven UL protocol delivered thrice weekly for 6 weeks had no effect on shoulder pain nor caused any shoulder dislocation when delivered to participants in the chronic phase post-stroke. 31 Spasticity Six studies explored the effect of different interventions on spasticity. 3,8,11,17,25,31 In the acute to early subacute phase, there was low quality evidence that there was no difference between: bodyweight supported treadmill training and conventional overground gait training delivered over 4 weeks on reducing lower limb spasticity; 8 and a 5-week course of additional upper limb therapy provided by a qualified physiotherapist or a physiotherapy assistant and standard physiotherapy on reducing upper limb spasticity. 11 There was very low-quality evidence that there was no reduction in spasticity with a 6-week course of conventional physiotherapy with or without EMG biofeedback. 3 In the acute to late subacute phase, there was low quality evidence that a 6-month course of a staged physical rehabilitation programme resulted in a lower incidence of upper and lower limb spasticity compared to usual care that did not involve formal rehabilitation. 17 There was very low-quality evidence that a 3-month course of either robot-assisted bodyweight supported treadmill training or conventional gait training had no effect on reducing lower limb spasticity. 25 In the chronic phase, there was very low-quality evidence that there was no difference between an intensive therapist-driven UL protocol and an intensive robotic-driven UL protocol delivered thrice weekly for 6 weeks in reducing UL spasticity. 31 For peer review only -http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59

Strengths and Limitations of this Study
 This is the first systematic review to investigate rehabilitation interventions specifically to survivors of severe stroke  The review included outcomes on physical function and immobility-related poststroke complications, of which the latter contribute to high levels of caregiver burden and are less commonly reported outcomes in stroke rehabilitation research  Marked heterogeneity of included studies prevented meta-analysis  Most included studies were rated as low or very low-quality evidence due to unclear or high risk of bias as well as recruitment of very small samples  Severe stroke can be understood as a stroke resulting in a significant amount of brain tissue damage and multiple neurological impairments, which leads to a significant loss of function and residual disability.|12| Dependent upon how it is measured, 14 -31% of people who sustain a stroke globally are classified as having a severe stroke,|13-18| a cohort of the stroke population that experiences worse outcomes compared to survivors of less severe stroke.|19-30| In the initial hospitalisation phase post-stroke, they are more likely to  This systematic review aims to establish the effectiveness of rehabilitation interventions on physical function and immobility-related complications for survivors of severe stroke and identify areas for future rehabilitation research for these patients.

METHODS
The systematic review has been reported according to the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) statement (see Supplementary File 1).|32| The protocol for the systematic review has been published previously.|33|

Study design
The systematic review included randomised controlled trials (RCTs). The systematic review excluded quasi-experimental, correlational and descriptive study designs. Studies were selected according to the PICO (participant, intervention, comparator and outcome) format.
The systematic review protocol provides full details of the PICO components|33| and a brief summary of the components is reported below. There were no deviations from the protocol PICO.

Participants
The review included studies of adult (≥ 18 years) stroke patients with severe stroke. Stroke severity was defined using a score on a validated and routinely used outcome measure (e.g.

Comparators
The review included studies that had a comparator, which included any of the following: another type of rehabilitation intervention, usual care or no intervention. Usual care was defined as the rehabilitation that the patient would normally receive as part of undergoing stroke rehabilitation.

Outcomes
The review included studies that focused on the primary outcomes of physical function and post-stroke complications. As per the definition of function in the International Classification of Functioning, Disability and Health, physical function was assessed using measures of body function (e.g. Fugl-Meyer Assessment), activity (e.g. BI), and participation (e.g. Stroke  or cultural bias, studies in any language or geographical location were included.

Data management and study selection
The results from the literature search were uploaded to a reference management programme (Refworks) and duplicate references were removed. A final list of non- independently performed data extraction for all eligible articles using a data extraction proforma previously piloted. Any differences in opinion between the two authors at any stage of the study selection and data extraction process were resolved by a third review author (CS). present. The quality of the evidence was ranked 'high', 'medium', 'low' or 'very low' by two review authors independently (MM and JJ). Any differences in opinion between the two authors at any stage of the study selection and data extraction process were resolved by a third reviewer (CS).

Data analysis
Due to the limited number of studies investigating each individual intervention and the marked heterogeneity of the selected studies, it was not appropriate to undertake a metaanalysis. Heterogeneity was seen in the rehabilitation interventions (type, dosage, method of delivery, timeframe completed post-stroke) as well as outcomes (type and timeframe completed post-stroke). Therefore, a descriptive review of results was performed. As there may be differences in recovery rates and outcomes according to the time post-stroke, studies were grouped into three timeframes post-stroke based on when participants were recruited to the study and when the study finished. These timeframes were the acute to early subacute stage (up to 3 months post-stroke), acute to late-subacute stage (up to 6 months post-stroke) and chronic stage (greater than 6 months post-stroke). These timeframes were chosen based on recommendations for the standardised measurement of sensorimotor recovery in stroke trials.|41| Study findings were presented according to these three timeframes.

Patient and public involvement
There was no patient involvement in this study.

RESULTS
The initial literature review identified 7589 articles (Figure 1). After removing duplicates and screening titles and abstracts, 1083 full text articles were assessed for eligibility. 28 studies were included in the systematic review.|42-73| 2677 participants were recruited to these studies-mean participant age was 72.7 years, 49.3% were male and 87% of patients  Table 4).  were not evident at the 9-month follow-up.

Activities of Daily Living
Twenty studies explored independence and ability to perform activities of daily living (ADLs). In the acute to early subacute phase, there was high quality evidence that frequent, very early mobilisation (median of 6.5 times per day) commencing within 24 hours post-stroke did not result in more patients being less dependent in ADLs at 3 months post-stroke compared to usual care, which traditionally started more than 24 hours post-stroke and averaged 3 times per day.|42| However, caution is required with interpreting this finding as the sub-group analysis of patients with severe stroke was not powered for this outcome.
There was moderate quality evidence that a 6-week course of NMES applied to the wrist and finger extensors in conjunction with usual therapy resulted in no difference in ADL independence compared to usual care.|55| In the acute to late subacute phase, there was moderate quality evidence that additional lower limb (LL) therapy in conjunction with regular physical rehabilitation performed in the first 20 weeks post-stroke improved ADL independence whilst the intervention was being delivered when compared to regular physical rehabilitation alone.|63| However, these improvements were not seen 6 months post-stroke.
In the chronic phase, there was high quality evidence that a 3-month occupational therapy (OT) intervention provided to residents in care homes resulted in no difference in ADL independence compared to usual care.|71| Similar caution is required with interpreting this finding as the sub-group analysis of patients who were severely or very severely disabled was not powered for this outcome.

Gait
Nine studies investigated gait, which included gait ability and gait speed.

Main Findings
Although 28 RCTs investigating 20 different rehabilitation interventions were identified in this review, there was a paucity of high-quality evidence to support the use of these interventions to improve physical function and reduce immobility-related complications after severe stroke. Most studies were rated as low or very low-quality evidence due to unclear or high risk of bias as well as recruitment of very small samples (refer to Supplementary Table 1). However, compared to data from national (United Kingdom) and global estimates of stroke incidence and prevalence, participants recruited to these studies were similar in terms of stroke type and gender but slightly younger (median age of stroke in the United Kingdom is 77 years). |1,2,18|Therefore, participants were generally representative of the wider stroke population.

Physical Function
Two large, multi-centre studies provided high quality evidence that their respective treatment interventions were no more effective at improving different aspect of physical function than usual care.|42,71| However, patients with severe stroke or severe disability post-stroke comprised a smaller sample within these larger trials. Analyses of data from these sub-groups may not be powered to detect changes between the treatment and usual care interventions and therefore caution is required in interpreting the studies' findings.
In AVERT (A Very Early Rehabilitation Trial),|42| very early and frequent mobilisation commencing within 24 hours post-stroke did not result in more patients being less dependent in ADLs 3 months post-stroke compared to usual care, which traditionally started more than 24 hours post-stroke. Although the data seemed to favour usual care practice for patients with severe stroke, this finding did not achieve statistical significance. It could be argued that patients with severe stroke may be less likely to tolerate very early and intensive therapy in the first few days after stroke due to fatigue and reduced exercises tolerance. |74|This would suggest that mobilising patients less intensively after 24 hours   investigating the effectiveness of rehabilitation interventions may be on improving functional recovery post-stroke rather than reducing immobility-related complications.

Immobility-Related Complications
Only two high-quality studies investigated the effectiveness of their respective interventions at reducing immobility-related complications. In AVERT, very early and frequent mobilisation commencing within 24 hours post-stroke did not result in more patients dying at 3 months post-stroke compared to usual care. |42|Whilst this finding is obviously positive, very early and frequent mobilisation did not result in less patient dependency as reported earlier in the discussion. Therefore, the optimal time and frequency to commence the mobilisation of patients with severe stroke is not clear.
In the OT in care home trial,|71| a 3-month, goal-orientated OT intervention for stroke survivors living in care homes did not result in reduced depression up to 1-year postintervention. Whilst post-stroke depression has a multi-factorial cause, it has been reported that mental distress associated with residual disability may contribute to the development of post-stroke depression.|75| Therefore, reductions in residual disability may alleviate depressive symptoms post-stroke. As the OT intervention did not result in improved ADL ability, it is possible that depression did not significantly change due to the lack of improvement in ADL ability.

Strengths and Limitations
In terms of strengths, this is the first systematic review to investigate rehabilitation interventions specifically to survivors of severe stroke, who tend to be underrepresented in stroke rehabilitation research, and the identification of topics for future rehabilitation research will hopefully guide much needed research for this cohort of the stroke population.
As well, the outcomes of the review focussed on not just physical function but immobilityrelated post-stroke complications, which are known to be higher in the severe stroke population and contribute to high levels of caregiver burden.|28-30| In terms of limitations, it has been reported that the defining severe stroke is difficult due to different criteria used to classify severity.|79| The use of objective scores on validated outcome measures to classify stroke severity in our systematic review was deemed necessary to ensure that participants had actually sustained a severe stroke. In our review, the BI was the most commonly used measure to classify stroke severity, reported in 17 out of 28 studies. Using a pre-specified score on the BI to classify severe stroke (≤9/20 or ≤45/100)|33| enabled the identification of patients with severely disabling stroke. However, the use of an alternative measure of stroke severity, such as the NIHSS, may have resulted in the inclusion of a study with participants with a slightly different clinical presentation than participants measured with the BI. Alternatively, we may have excluded studies that used a different scoring system to classify stroke severity. However, these studies were discussed in detail amongst three review authors to determine suitability for inclusion and therefore it is likely that the number of relevant studies excluded from the review was minimal. Another limitation is the use of data from subgroups within larger clinical trials. As subgroup analyses may not be powered to detect changes between groups, caution is required in the interpretation of findings from these trials.

METHODS
Protocol and registration 5 Indicate if a review protocol exists, if and where it can be accessed (e.g., Web address), and, if available, provide registration information including registration number.
Abstract page, 2 Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, language, publication status) used as criteria for eligibility, giving rationale.

3
Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched.

4
Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated.

Supplementary file
Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis).

3, 4
Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

3, 4
Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made. 3

DISCUSSION
Summary of evidence 24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers).

9-12
Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias).     Two studies explored participants' cardiorespiratory response to different types of treadmill gait training within the acute to early subacute phase post-stroke. 4,8 There was low-quality evidence that 2 weeks of robot-assisted bodyweight supported treadmill gait training delivered in the first 6 weeks post-stroke improved peak VO2 compared to conventional gait training. 4 There was low-quality evidence that a 4-week course of bodyweight supported treadmill training delivered in the first 3 months post-stroke was not perceived to be more effortful than conventional gait training. 8 Neurological Impairment Three studies evaluated changes in neurological function. 6,25,27 In the acute to early subacute phase post-stroke, there was very low-quality evidence that there was no difference in an intensive or ordinary 2-week acute physical rehabilitation programme on reducing neurological impairment at 2 weeks and 6 months post-stroke. 6 In the acute to late subacute phase post-stroke, there was very low-quality evidence that a 3-month course of robot-assisted bodyweight supported treadmill gait training commenced within the first 6 weeks post-stroke was just as effective as conventional gait training on improving neurological function. 25 There was very low-quality evidence that an 8-week course of acupuncture provided in conjunction with rehabilitation during the subacute phase of stroke reduced neurological impairment compared to rehabilitation alone. 27 Sensorimotor Function Sixteen studies evaluated changes in sensorimotor function. Nine studies were performed in the acute to early subacute phase post-stroke, [3][4][5]7,8,[10][11][12][13] five studies in the acute to late subacute phase post-stroke, 18,21,25,27,28 and two studies in the chronic phase post-stroke. 29,31 In the acute to early subacute phase post-stroke, there was low quality evidence from two studies that thermal stimulation in conjunction with standard rehabilitation resulted in improvements in lower limb sensorimotor function and strength when compared to standard rehabilitation alone. 5,10 Improvements in lower limb sensorimotor function were maintained at 12 months post-intervention. There was low quality evidence that 2 weeks of robot-assisted bodyweight supported treadmill gait training resulted in improvements in lower limb sensorimotor function but not strength compared to conventional gait training. 4 There was low quality evidence that there was no difference between: 4 weeks of robotassisted treadmill gait training and conventional gait training on improving lower limb sensorimotor function; 13 wearing a cueing wristwatch and wearing a sham wristwatch for 3 hours per weekday for 3 weeks during rehabilitation on improving upper limb sensorimotor function and number of arm movements; 7 a 4-week course of bodyweight supported treadmill training and conventional overground gait training on improving lower limb strength; 8 and a 5-week course of additional upper limb therapy provided by a qualified physiotherapist or a physiotherapy assistant and standard physiotherapy on improving upper limb motor activity and grip strength. 11 There was very low-quality evidence that a thrice weekly, 6-week course of electromyography (EMG) biofeedback combined with conventional physiotherapy had no effect on improving lower limb active range of movement when compared to conventional  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n l y physiotherapy alone. 3 There was very low-quality evidence that a 3-month course of acupuncture in conjunction with rehabilitation resulted in better upper and lower limb sensorimotor function when compared to rehabilitation alone. 12 In the acute to late subacute phase post-stroke, there was very low quality evidence that a 6-week course of robotic tilt-table verticalisation that combines cyclic leg movements and FES and used in conjunction with standard physiotherapy resulted in better lower limb strength and sensorimotor function compared to physiotherapy-assisted verticalisation using a standard tilt-table and used in conjunction with standard physiotherapy. 18 There was very low-quality evidence that an 8-week course of acupuncture provided in conjunction with rehabilitation resulted in improvements in upper and lower limb sensorimotor function compared to rehabilitation alone. 27 There was very low-quality evidence that a 3-month course of nurse-led acupressure resulted in improvements in upper and lower limb motor function compared to routine care. 28 There was very low quality evidence that there was no difference between: a functionally-orientated and a sensorimotor integrative occupational therapy treatment approach delivered over 8 weeks on improving upper limb sensorimotor function; 21 and a 3-month course of robot-assisted bodyweight supported treadmill gait training and conventional gait training on improving lower limb power. 25 In the chronic phase post-stroke, there was very low-quality evidence that a 6-week course of robot-assisted bodyweight supported treadmill gait training using slower treadmill speeds resulted in improvements in lower limb sensorimotor function compared to similar treadmill training using faster treadmill speeds. 29 There was very low-quality evidence that either an intensive therapist-driven UL protocol or an intensive robotic-driven UL protocol delivered thrice weekly for 6 weeks resulted in an improvement in shoulder and elbow sensorimotor function. 31 Activity Activities of Daily Living Sixteen studies explored independence and ability to perform activities of daily living (ADLs). Nine studies were completed in the acute to early subacute phase, 2,6-8,10-13,15 six studies were completed in acute to late subacute phase 17,[19][20][21]25,27,28 and one study was completed in the chronic phase. 32 In the acute to early subacute phase, there was low quality evidence that a 6-week course of thermal stimulation used in conjunction with standard rehabilitation resulted in improvements in ADL independence 3 months post-stroke compared to standard rehabilitation alone, although improvements were not seen at 6 months post-stroke. 10 There was low quality evidence that there was no difference between: regular physiotherapy and regular physiotherapy in conjunction with use of an Oswestry standing frame delivered over 14 consecutive weekdays in the first 3 months post-stroke on ADL independence; 2 wearing a cueing wristwatch and wearing a sham wristwatch for 3 hours per weekday for 3 weeks during rehabilitation on ADL independence; 7 a 4-week course of bodyweight supported treadmill training and conventional overground gait training on improving ADL independence; 8 a 5-week course of additional upper limb therapy provided by a qualified physiotherapist or a physiotherapy assistant and standard physiotherapy on improving ADL independence; 11 and 4 weeks of robot-assisted treadmill gait and conventional overground gait training on ADL independence. 13 There was very low-quality evidence that there was no difference in an intensive or ordinary 2-week acute physical rehabilitation programme in improving ADL independence at 2 weeks  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y and 6 months post-stroke. 6 There was very low-quality evidence that a 3-month course of acupuncture in conjunction with rehabilitation resulted in better ADL independence when compared to rehabilitation alone. 12 There was very low-quality evidence that providing additional physiotherapy in conjunction to regular rehabilitation in the first few weeks poststroke resulted in improvements in ADL independence at 6 months post-stroke compared to regular rehabilitation alone. 15 In the acute to late subacute phase, there was low quality evidence that a 6-month course of a staged physical rehabilitation programme resulted in greater improvements in ADL independence compared to usual care that did not involve formal rehabilitation. 17 There was very low-quality evidence that a monthly home-based physiotherapy programme delivered over 6 months resulted in improvements in ADL independence compared to standard care. 19,20 There was very low-quality evidence that there was no difference between a functionally orientated or a sensorimotor integrative occupational therapy treatment approach delivered over 8 weeks on ADL independence. 21 There was very lowquality evidence that a 3-month course of robot-assisted bodyweight supported treadmill gait training resulted in improvements in ADL independence compared to conventional gait training. 25 Improvements were only seen in the cohort of participants who demonstrated significant motor impairment. Improvements were maintained at the 2-year follow-up. 26 There was very low-quality evidence that an 8-week course of acupuncture provided in conjunction with rehabilitation during the subacute phase of stroke improved ADL independence compared to rehabilitation alone. 27 There was very low-quality evidence that a 3-month course of nurse-led acupressure resulted in improvements in ADL independence compared to routine care. 28 In the chronic phase, there was very low-quality evidence that a 16-week course of trunk acupuncture combined with rehabilitation training resulted in greater improvements in ADL independence compared to rehabilitation training alone. 32 Balance and Postural Control Eight studies investigated balance and postural control. Four studies were completed in the acute to early subacute phase, 2,5,10,16 two studies were completed in the acute to late subacute phase 18,25 and two studies were completed in the chronic phase. 29,32 In the acute to early subacute phase, there was low quality evidence that a 6-week course of thermal stimulation in conjunction with standard rehabilitation resulted in improvements in trunk postural control but not balance compared to standard rehabilitation alone. 5 In a separate study, there was low quality evidence that a 6-week course of thermal stimulation in conjunction with standard rehabilitation resulted in improvements in balance 3 months post-stroke compared to standard rehabilitation alone, although improvements were not seen at 6 months post-stroke. 10 There was low quality evidence that there was no difference between regular physiotherapy and regular physiotherapy in conjunction with use of an Oswestry standing frame delivered over 14 consecutive weekdays in the first 3 months poststroke on trunk postural control. 2 There was low quality evidence that an 8-week course of physiotherapy involving early mobilisation combined with the Bobath approach resulted in improvements in balance when compared to physiotherapy just involving the Bobath approach. 16 In the acute to late subacute phase, there was very low quality evidence that a 6-week course of robotic tilt-table verticalisation that combines cyclic leg movements and FES and used in conjunction with standard physiotherapy resulted in improved postural control  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y during different activities compared to physiotherapy-assisted verticalisation using a standard tilt-table and used in conjunction with standard physiotherapy. 18 There was very low-quality evidence that a 3-month course of robot-assisted bodyweight supported treadmill gait training resulted in improvements in trunk control compared to conventional gait training. 25 Improvements were only seen in the cohort of participants who demonstrated significant motor impairment.
In the chronic phase, there was very low-quality evidence that a 6-week course of robotassisted bodyweight supported treadmill gait training resulted in improvements in balance regardless if slower or faster treadmill training speeds were used. 29 There was very lowquality evidence that a 16-week course of trunk acupuncture combined with rehabilitation training resulted in greater improvements in balance compared to rehabilitation training alone. 32 Gait Eight studies investigated gait, which included gait ability and gait speed. Six studies were performed in the acute to early subacute phase, [3][4][5]8,10,13 one study was performed in the acute to late subacute phase 25 and one study was performed in the chronic phase. 29 In the acute to early subacute phase, there was low quality evidence from two studies that a 6week course of thermal stimulation in conjunction with standard rehabilitation resulted in improvements in gait ability compared to standard rehabilitation alone. 5,10 There was low quality evidence that 4 weeks of robot-assisted treadmill gait training resulted in better gait ability than conventional gait training. 13 There was low quality evidence that there was no difference between: a 2-week course of robot-assisted bodyweight supported treadmill gait training and conventional gait training delivered in the first 6 weeks post-stroke on improving gait ability; 4 a 4-week course of bodyweight supported treadmill training and conventional overground gait training on improving gait ability; 8 and a thrice weekly, 6-week course of EMG biofeedback combined with conventional physiotherapy and conventional physiotherapy alone in improving gait speed. 3 In the acute to late subacute phase, there was very low-quality evidence that a 3-month course of robot-assisted bodyweight supported treadmill gait training resulted in improvements in gait ability compared to conventional gait training. 25 Improvements were only seen in the cohort of participants who demonstrated significant motor impairment. Improvements were maintained at the 2-year follow-up. 26 In the chronic phase, there was very low-quality evidence that a 6-week course of robotassisted bodyweight supported treadmill gait training using slower treadmill speeds resulted in improvements gait ability compared to similar treadmill training using faster treadmill speeds. 29 General Physical Activity Seven studies examined the effects of different interventions on improving general physical activity. Six studies were performed in the acute to early subacute phase 2,3,5,10,11,16 and one study was performed in the acute to late subacute phase. 25 In the acute to early subacute phase, there was low quality evidence from two studies that thermal stimulation in conjunction with standard rehabilitation resulted in improvements in physical activity when compared to standard rehabilitation alone. 5,10 Improvements were seen up until 3 months post-intervention but disappeared at the 6-month follow-up. There was low quality evidence that an 8-week course of physiotherapy involving early mobilisation combined with the Bobath approach resulted in improvements in physical activity when compared to  16 There was low quality evidence that there was no difference between: regular physiotherapy and regular physiotherapy in conjunction with use of an Oswestry standing frame delivered over 14 consecutive weekdays in the first 3 months post-stroke on physical activity; 2 and a 5-week course of additional upper limb therapy provided by a qualified physiotherapist or a physiotherapy assistant and standard physiotherapy on improving physical activity. 11 There was very lowquality evidence that there was no difference between a thrice weekly, 6-week course EMG biofeedback combined with conventional physiotherapy and conventional physiotherapy alone on improving physical activity. 3 In the acute to late subacute phase, there was very low-quality evidence that a 3-month course of robot-assisted bodyweight supported treadmill gait training resulted in improvements in physical activity compared to conventional gait training. 25 Improvements were only seen in the cohort of participants who demonstrated significant motor impairment. Improvements were maintained at the 2-year follow-up. 26 Upper Limb Function Two studies investigated changes in upper limb function. 11,31 In the acute to early subacute phase, there was low quality evidence that a 5-week course of additional upper limb therapy provided by a qualified physiotherapist was no more effective at improving upper limb function than additional upper limb therapy provided by a physiotherapy assistant or to standard physiotherapy. 11 In the chronic phase, there was very low-quality evidence that there was no improvement in upper limb function with either an intensive therapist-driven UL protocol or an intensive robotic-driven UL protocol delivered thrice weekly for 6 weeks. 31 Participation Instrumental Activities of Daily Living Four studies investigated the effect of different interventions on instrumental ADLs. 2,3,9,21 In the acute to early subacute phase, there was low quality evidence that there was no difference between: regular physiotherapy and regular physiotherapy in conjunction with use of an Oswestry standing frame delivered over 14 consecutive weekdays on ability to perform instrumental ADLs at 6 months post-stroke, 2 and an 8-week course of rehabilitation with the addition of a leg cycling machine compared to regular rehabilitation alone on instrumental ADLs 6 months post stroke. 9 There was very low-quality evidence that there was no difference between a thrice weekly, 6-week course of electromyography (EMG) biofeedback combined with conventional physiotherapy and conventional physiotherapy alone in improving performance in instrumental ADLs. 3 In the acute to late subacute phase, there was very low-quality evidence that there was no difference between a functionally orientated or a sensorimotor integrative occupational therapy treatment approach delivered over 8 weeks on the ability to prepare meals. 21 Perceived Health Status Two studies explored carers' and patients' perceived health status. 2,31 In the acute to early subacute phase, there was low quality evidence that there was no difference between regular physiotherapy and regular physiotherapy in conjunction with use of an Oswestry standing frame delivered over 14 consecutive weekdays on carer's perceived health status at 12 weeks and 6 months post-stroke. 2 In the chronic phase, there was very low-quality evidence that there was no change in patient's perceived health status with the provision of  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y either an intensive therapist-driven UL protocol or an intensive robotic-driven UL protocol delivered thrice weekly for 6 weeks. 31 Quality of Life There was very low-quality evidence that a monthly home-based physiotherapy programme delivered over 6 months resulted in an improvement in quality of life compared to standard care. 19 Complications Caregiver Burden There was low quality evidence that there was no difference between regular physiotherapy and regular physiotherapy in conjunction with use of an Oswestry standing frame delivered over 14 consecutive weekdays in the first 3 months post-stroke on caregiver strain and psychological well-being at 12 weeks and 6 months post-stroke. 2 Depression Three studies explored changes in depression. 2,20,31 In the acute to early subacute phase, there was low quality evidence that there was no difference between regular physiotherapy and regular physiotherapy in conjunction with use of an Oswestry standing frame delivered over 14 consecutive weekdays on depression at 12 weeks and 6 months post-stroke. 2 In the acute to late subacute phase, there was very low-quality evidence that a monthly homebased physiotherapy programme delivered over 6 months resulted in a reduction in level of depression compared to standard care. 20 In the chronic phase, there was very low-quality evidence that there was no difference between an intensive therapist-driven UL protocol and an intensive robotic-driven UL protocol delivered thrice weekly for 6 weeks in reducing depression. 31 Shoulder Pain/Dislocation There was very low-quality evidence that either an intensive therapist-driven UL protocol or an intensive robotic-driven UL protocol delivered thrice weekly for 6 weeks had no effect on shoulder pain nor caused any shoulder dislocation when delivered to participants in the chronic phase post-stroke. 31 Spasticity Six studies explored the effect of different interventions on spasticity. 3,8,11,17,25,31 In the acute to early subacute phase, there was low quality evidence that there was no difference between: bodyweight supported treadmill training and conventional overground gait training delivered over 4 weeks on reducing lower limb spasticity; 8 and a 5-week course of additional upper limb therapy provided by a qualified physiotherapist or a physiotherapy assistant and standard physiotherapy on reducing upper limb spasticity. 11 There was very low-quality evidence that there was no reduction in spasticity with a 6-week course of conventional physiotherapy with or without EMG biofeedback. 3 In the acute to late subacute phase, there was low quality evidence that a 6-month course of a staged physical rehabilitation programme resulted in a lower incidence of upper and lower limb spasticity compared to usual care that did not involve formal rehabilitation. 17 There was very low-quality evidence that a 3-month course of either robot-assisted  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 F o r p e e r r e v i e w o n l y bodyweight supported treadmill training or conventional gait training had no effect on reducing lower limb spasticity. 25 In the chronic phase, there was very low-quality evidence that there was no difference between an intensive therapist-driven UL protocol and an intensive robotic-driven UL protocol delivered thrice weekly for 6 weeks in reducing UL spasticity. 31 For peer review only -http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60