Brain‐computer interface training for motor recovery after stroke

Objectives

This is a protocol for a Cochrane Review (intervention). The objectives are as follows:

To evaluate the effectiveness of brain‐computer interface training for motor recovery after stroke.

Background

Description of the condition

Stroke is defined as “rapidly developed clinical signs of focal or global disturbance of cerebral function, lasting more than 24 hours or until death, with no apparent non‐vascular cause" (WHO 2021). The interruption of blood supply to the brain is usually the direct and most important cause of stroke. When a blood vessel inside the brain ruptures or a clot blocks a vessel, blood supply is interrupted. Currently, stroke is the second leading cause of death and the third leading cause of death and disability combined around the world, after ischaemic heart disease (GBD 2021). In 2016, more than 5.5 million people died from stroke worldwide (GBD 2019). In 2019, the number of deaths due to stroke increased to more than 6.55 million (GBD 2021). From 1990 to 2019, the absolute number of incident strokes globally increased by 70%, and deaths from stroke increased by 43% (GBD 2021). The increase in absolute numbers is mainly due to population growth and ageing. Most of the burden caused by stroke (including 68.6% of incident strokes, 52.2% of prevalent strokes, 70.9% of stroke deaths, and 77.7% of disability‐adjusted life years (DALYs) lost) was in low‐income and middle‐income countries (Feigin 2014). As the largest developing country in the world, China has the highest incidence of stroke. Stroke is the third‐leading cause of death in China, behind only malignant tumours and heart disease (Zhou 2019). 

Stroke has the characteristics of high morbidity, high mortality, high disability rate, and high recurrence rate (Feigin 2014). Motor dysfunction is regarded as a loss or limitation of function in muscle control or movement, or a limitation in mobility. It is the most common and widely recognised impairment caused by stroke, and affects about 80% of patients (Langhorne 2009).  Other effects, such as dysphagia, aphasia, cognitive impairment, memory impairment, deep‐vein thrombosis, pain, etc, usually depend on the location in the brain where the stroke occurred, and the post‐stroke treatment. An analysis based on the Framingham Study showed that 26% of the population in the study had become disabled in basic activities of daily living; 50% had reduced mobility due to hemiparesis (Kelly‐Hayes 2003). As one of the leading causes of long‐term disability, stroke was the largest contributor to global neurological DALYs, responsible for 42.2% of all neurological disorder DALYs in 2016 (GBD 2019). From 1990 to 2019, DALYs due to stroke increased 32%, with an increase of 23% from 2016 to 2019 (GBD 2021). 

Motor dysfunction is one of the greatest post‐stroke functional impairments. It presents as a decrease in or loss of limb movement, balance, and mobility. Motor dysfunction always reduces one’s ability to take part in activities of daily living and participate in society. This leads to an increased burden on families and on society itself. Because of the high incidence and disability of stroke, it is essential to find effective ways to promote the recovery of motor impairment in stroke patients.

Description of the intervention

With the development of neurophysiology and engineering, neurobionics is increasingly used in medicine. Neurobionics is the science of repairing or substituting impaired nervous functions by directly integrating electronics with the nervous system (Rosenfeld 2017). Cochlear implants were one of the early neurobiology‐based products used in medicine in the 1960s and 1970s to restore hearing (Mudry 2013). These devices are now widely accepted, and have helped many deaf people restore their hearing. The brain‐computer interface (BCI), sometimes called 'brain‐machine interface', is the linkage of the brain to computers through scalp, subdural or intracortical electrodes (Rosenfeld 2017). BCI has become one of the most important advances in neurosciences and engineering in recent years. BCI training is a type of treatment based on BCI techniques. 

BCI is a direct connection or pathway between the brain and computers or other external devices, made without the use of the normal neuromuscular pathways (Soekadar 2015). Compared to other new devices such as upper limb robots, BCI training helps patients participate in exercise more actively, and better facilitate their motor function recovery (Yang 2021). Constraint‐induced movement therapy (CIMT), occupational therapy (OT), virtual reality (VR), and other exercise interventions have been proven to have some positive short‐term effects in patients with mild to moderate stroke (Raffin 2018). However, the long‐term efficacy of these interventions is still controversial (Wu 2016). Generally, some interventions, such as CIMT, cannot be used in stroke survivors with severe motor impairment, because these patients are usually immobile for exercise training (Guggisberg 2017). Even severe stroke patients can also be guided to attend activities through BCI training, relying on the plasticity of the brain. Thus, BCI training is a very promising potential neurorehabilitative intervention.

How the intervention might work

BCI interventions can establish new pathways between the brain and computers or other electronic devices that do not rely on normal neuromuscular pathways (Daly 2008). BCI converts brain signals into computer commands, bypassing the damaged area and re‐establishing the connection between brain and limb movements (Coscia 2019).

Neural plasticity provides a theoretical basis for recovery from neurological damage. It means that neurons adjust structures, functions, and connections through learning under the influence of internal and external environmental factors, leading to the reorganisation of neural function (Gittler 2018). The type and extent of neural plasticity are often task‐oriented. Neural plasticity involves the germination of axon and dendrite, the regeneration of nerve cells, the activation of alternate pathways, and an increase in the number of synapses (Wang 2010). Methods for activating neural plasticity and neurofunctional reorganisation is a hot topic of research in functional rehabilitation after neurological damage.

BCI interventions typically promote neural plasticity for stroke patients through two methods, 'recovery' and 'substitution' (Chamola 2020). 'Recovery' means that BCI training activates neural plasticity and produces normal brain activity in stroke patients. Evidence from a 2004 study suggests that BCI treatment can alter brain signal characteristics, change neuronal function (e.g. synaptic growth), increase the number of synapses and increase axonal germination in intact peripheral regions (Leuthardt 2004). 'Substitution' refers to generating sensory input and inducing neural plasticity with motor‐assistive devices, such as functional electrical stimulation (FES), exoskeletons, and robots. Substitution facilitates the improvement of motor control among stroke patients.

Why it is important to do this review

Apart from death, long‐term disability is the primary burden of stroke. Interventions that can speed up recovery after stroke and reduce long‐term disability have an important impact on patients, their families, and society. The booming development of computer science has provided new opportunities and challenges to stroke rehabilitation research. BCI training has been reported in clinical studies to be safe, and potentially beneficial for patients after stroke (Ang 2015; Frolov 2017; Kim 2016; Lee 2020). BCI training may enhance the motor recovery of people with stroke. Our review will systematically summarise the currently available clinical evidence on the effectiveness of BCI training for improving recovery in stroke patients.

Objectives

To evaluate the effectiveness of brain‐computer interface training for motor recovery after stroke.

Methods

Criteria for considering studies for this review

Types of studies

We will include randomised controlled trials (RCTs) only.

Types of participants

We will include participants aged 18 years and over with a definite clinical diagnosis of stroke accompanied by motor dysfunction, including in the upper limbs, lower extremities, and hands.

Types of interventions

BCI is the direct linkage of the brain to computers or other external devices through scalp, subdural, or intracortical electrodes without the use of the normal neuromuscular pathways (Rosenfeld 2017; Soekadar 2015).

We will include experiments with BCI training as an intervention and focused on motor impairment. Comparisons can be usual care or other interventions, including FES or sham interventions.  

Types of outcome measures

We will collect primary and secondary outcome measure data taken by trial investigators before and after BCI training programmes, and at the end of programme follow‐up.

We are primarily interested in measures of motor function and activities of daily living, as these outcomes are likely to be most meaningful to stroke survivors. We are also interested in measures of balance, muscle function, and neurological function, as they will indirectly indicate the improvement of motor function.

Primary outcomes

• Motor function, measured by any scales with reported validated studies including, but not limited to:

  • overall function measures, such as Motor Assessment Scale (MAS, Carr 1985), Fugl‐Meyer assessment (FMA, Fugl‐Meyer 1975);

  • lower limb function measures, such as Wolf Motor Function Test (WMFT, Wolf  2001);

  • upper limb function measures, such as Action Research Arm Test (ARAT, Lyle 1981),  Box and Block Test (BBT, Mathiowetz 1985);

  • hand function measures, such as Jebsen‐Taylor hand function test (JTHF, Jebsen 1969).

• Activities of daily living, measured by Functional Independence Measure (FIM, Keith 1987), Barthel index (BI, Mahoney 1965), or Modified Barthel Index (MBI, Sulter 1999).

As we plan to pool outcome data, we will select the measure for which studies have collected the most data, in order to avoid double‐counting if studies have used more than one measure. If there are measures in a study with equal amounts of data, we will select the measure listed first in the study.

Some scales include both upper and lower extremity function assessments (i.e. FMA). If a study reports only the upper or limb components of these scales, we will prefer other specific upper or lower limb function measures used in the study for analysis. If the part of these measures is the only used outcome in the study, we will use it for analysis directly.

Secondary outcomes

• Measures of balance, such as Berg Balance Scale (BBS, Berg 1992).

• Measures of muscle function, such as Medical Research Council (MRC) scale (James 2007) for muscle strength, or Modified Ashworth Scale (MAS, Ashworth 1964) for muscle tension.

• Validated scales of neurological function, such as Brunnstrom Stages (Brunnstrom 1966), Canadian Neurological Scale (CNS, Cote 1986) or National Institutes of Health Stroke Scale (NIHSS, Brott 1989).

• Adverse events such as infections, cardiac arrhythmias, intra and extracranial bleeding, death. We will collect adverse events reported by trial investigators.

Search methods for identification of studies

See search methods for the Specialised Register of the Cochrane Stroke Group (apps.ccbs.ed.ac.uk/csrg/entity/searchmethods.pdf). We will not apply any date or language restrictions in the search for trials and we will arrange translation of trials published in languages other than English and Chinese. Two review authors (QY and LMX) will be responsible for searching the literature.

Electronic searches

We will search the Cochrane Stroke Group's Specialised Register and the following electronic databases.

• Cochrane Central Register of Controlled Trials (CENTRAL) (Cochrane Library; latest issue) in the Cochrane Library.

• MEDLINE Ovid (from 1946 to the search date) (Appendix 1).

• Embase Ovid (from 1974 to the search date).

• Conference Proceedings Citation Index‐ Science (CPCI‐S) (from 1990 to the search date) and Science Citation Index Expanded (SCI‐EXPANDED) (from 1900 to the search date) in Web of Science.

• BIOSIS Citation Index in Web of Science

• Cumulative Index to Nursing and Allied Health Literature (CINAHL EBSCO) (from 1982 to the search date).

• AMED Ovid (from 1985 to the search date).

• PEDro (Physiotherapy Evidence Database) (www.pedro.org.au/).

• Compendex and Inspec (Elsevier Engineering Village)

• Scopus (Elsevier)

• IEEE Xplore Digital Library

We will modify the subject strategies for databases modelled on the search strategy designed for MEDLINE by the Cochrane Stroke Group’s Information Specialist (Appendix 1). We will combine all search strategies deployed with subject strategy adaptations of the highly sensitive search strategy designed by Cochrane for identifying randomised controlled trials and controlled clinical trials, as described in the Cochrane Handbook for Systematic Reviews of Interventions (Lefebvre 2021).

We will search the following registers of ongoing trials:

• ClinicalTrials.gov at the US National Institutes of Health (www.clinicaltrials.gov/);

• International Clinical Trials Registry Platform (ICTRP) at the World Health Organization (who.int/ictrp/en/).

Searching other resources

In an effort to identify further published, unpublished and ongoing trials, we will:

• check the bibliographies of included studies and any relevant systematic reviews identified for further references to relevant trials and use Google Scholar (scholar.google.co.uk/) to forward track relevant references;

• contact original study authors for clarification and further data if trial reports are unclear;

• where necessary, contact experts, trialists and organisations in the field to obtain additional information on relevant trials.

Data collection and analysis

Selection of studies

Two review authors (QY and LMX) will independently screen titles and abstracts of the references obtained as a result of our searching activities and will exclude clearly irrelevant reports. We will retrieve the full‐text articles for the remaining references. Two review authors (QY and LMX) will independently screen the full‐text articles and identify studies for inclusion, and will identify and record reasons for exclusion of the ineligible studies. We will resolve any disagreements through discussion or, if required, we will consult a third review author (YKH). We will collate multiple reports of the same study so that each study, not each reference, is the unit of interest in the review. We will record the selection process and complete a PRISMA flow diagram (Moher 2009).

Data extraction and management

Four review authors (QY, LMX, LYF and CGC) will independently extract data from included studies. We will resolve any disagreements through discussion or, if required, we will consult a third review author (YKH).

We will extract information for the following items.

• Basic trial characteristics: name, year, funding, conflicts  of interest, and publication type.

• Participant characteristics: age, gender, number, type and location of stroke, severity, time since the onset of stroke symptoms, type and severity of stroke‐related motor function impairments.

• Intervention type: method, onset, duration, and frequency.

• Comparison intervention characteristics.

• Outcomes data.

• Information needed for risk of bias assessment.

Where missing information or unclear methods of reporting are found, we will contact authors to gain the missing data.

Assessment of risk of bias in included studies

Four review authors (QY, LMX, LYF and ZHT) will independently assess the risk of bias in each study, using the criteria of the original Cochrane 'Risk of bias' tool (RoB1) for randomised trials, outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We will resolve any disagreements by discussion or by involving a third review author (YKH). 

We will evaluate the following domains.

• Random sequence generation: ‘low risk’ (random number table, computer‐generated random numbers, coin toss, drawing lots, throwing dice or shuffling cards), ‘high risk’ (sequence generation by date of admission, clinical record number) or ‘unclear risk’.

• Allocation concealment: ‘low risk’ (telephone or web‐based central allocation or with opaque, sealed envelopes with sequential numbers), ‘high risk’ (predictable methods such as an open allocation schedule or the use of non‐opaque, non‐sealed or non‐sequentially numbered envelopes) or ‘unclear risk’.

• Blinding of participants. 

• Blinding of  personnel.

• Blinding of outcome assessment.

• Incomplete outcome data: ‘low risk' (detailed description of missing data or imputation of data), ‘high risk’ (improper description of missing data), or ‘unclear risk’.

• Selective outcome reporting: ‘low risk' (complete description of outcomes in a predefined way), ‘high risk’ (incomplete description of predefined outcomes or reports of newly added outcomes), or ‘unclear risk’.

• Other bias.

The assessment of risk of bias in the blinding of participants and personnel will be dependent on whether the lack of blinding would have influenced the outcome. If there was no blinding or incomplete blinding, but the review authors judge that the outcome is not likely to have been influenced by the lack of blinding, we will assign a low risk of bias. If participants and key study personnel were blinded, and it was unlikely that the blinding could have been broken, we will assign a low risk of bias. However, if an outcome was likely to have been influenced by lack of blinding or incomplete blinding, or the blinding could have been broken, we will assign a high risk of bias.

The assessment of risk of bias of blinding of outcome assessment will be dependent on the potential influence that lack of blinding may have had. If the outcome assessor was not blinded and we judge that the outcome measure could have been influenced by the assessor, we will assign a high risk of bias. If we judge that the outcome measure could not have been influenced by the assessor, we will assign a low risk of bias, regardless of whether or not the outcome assessor was blinded.

In the 'Risk of bias' tables, we will assess the risk of bias for each domain as high, low, or unclear, and will give a justification for our judgements, along with relevant information from the study report.

Review authors will not evaluate the risk of bias for studies in which they participated as an author.

Measures of treatment effect

For dichotomous data, we will calculate and report risk ratios (RRs) with 95% confidence intervals (CIs). For continuous outcomes, we will calculate standardised mean differences (SMDs) with 95% CIs, if studies measure the same outcome using different tools. Where all studies use the same measurement tool, we will use mean differences (MDs) and 95% CIs.

Unit of analysis issues

In standard RCTs, we will treat the participant as the unit of analysis. In the case of cluster‐randomised controlled trials we will treat the cluster as the unit of analysis. Studies with multiple time‐point observations will be subjected respectively to follow‐up and analysis.

Dealing with missing data

We will directly contact study authors to acquire missing data.

If we receive no response, we will explore the effects of excluding the study in a sensitivity analysis and use the pooled correlation coefficient to calculate standard deviations (SDs).

Assessment of heterogeneity

We will choose the random‐effects model for the meta‐analysis and use the I² statistic to measure heterogeneity among the trials. The cut‐off point of meaningful heterogeneity will be 50% (Higgins 2021). If meaningful heterogeneity is observed, we will conduct sensitivity analyses to explore the possible causes.

Assessment of reporting biases

If 10 or more studies are included, we will assess the potential for reporting bias using funnel plots and Egger's regression tests, as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2021). The performance of Egger's and related tests has been extensively studied for binary outcomes, but not for continuous ones.

Data synthesis

Where we consider studies to be sufficiently similar, we will conduct a meta‐analysis by pooling the appropriate data using Cochrane’s Review Manager Web ( RevMan Web).

We will pool data using the random‐effects model. We will also analyse data with the fixed‐effect model to explore the influence of small‐study effects. 

Subgroup analysis and investigation of heterogeneity

We will conduct our pre‐planned subgroup analyses if there are at least six studies within an analysis. We will investigate heterogeneity by performing the following prespecified subgroup analyses for the primary outcomes.

• The phase of stroke: according to the duration of the stroke; the phase was divided into: (hyper) acute phase (the first 24 hours up to seven days); early subacute phase (more than seven days up to three months); late subacute phase (more than three months up to six months); and chronic phase (more than six months).

• Different type of stroke (dichotomised ischaemic stroke versus haemorrhagic or unknown).

• Baseline motor function score.

• Different support of brain‐computer interface training.

• Different type of control groups.

We will report the test for subgroup differences, which can be calculated when we do the subgroup analyses using RevMan Web.

Sensitivity analysis

If heterogeneity is high (I² > 50%), we will conduct sensitivity analyses for primary outcomes.

We will carry out sensitivity analyses with the following parameters.

• Excluding studies with high or unclear risk of bias for any domain.

• Excluding studies with missing data, where this cannot be supplied by the study authors.

Summary of findings and assessment of the certainty of the evidence

We will create a summary of findings table using the following outcomes: motor function, activities of daily living, balance, muscle function, neurological function, adverse events, and death (Table 1). We will use the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) to assess the certainty of a body of evidence as it relates to the studies that contribute data to the meta‐analyses for the prespecified outcomes (Atkins 2004). We will use methods and recommendations described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2021), and the GRADE handbook (Schünemann 2013). We will use the GRADEproGDT software to generate the summary of findings table (GRADEpro GDT 2015). We will justify all decisions to downgrade the certainty of evidence using footnotes, and we will make comments where necessary to aid the reader's understanding of the review.

1. Template for summary of findings table.

Brain‐computer interface training compared with control intervention for people after stroke
Patient or population: participants with stroke Settings: hospital, clinic, inpatient rehabilitation centre Intervention: all types of brain‐computer interface training Comparison: usual care, other exercise or training
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	control intervention	brain‐computer interface training
Motor function
Activities of daily living
Balance
Muscle function
Neurological function
Adverse event
Death
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; GRADE: Grading of Recommendations, Assessment, Development and Evaluation
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited; the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate; the true effect is likely to be substantially different from the estimate of effect.

Appendices

Appendix 1. MEDLINE search strategy

1. cerebrovascular disorders/ or exp basal ganglia cerebrovascular disease/ or brain ischemia/ or ischemic attack, transient/ or vertebrobasilar insufficiency/ or carotid artery diseases/ or carotid artery injuries/ or carotid artery thrombosis/ or carotid stenosis/ or carotid artery, internal, dissection/ or vertebral artery dissection/ or intracranial arterial diseases/ or cerebral arterial diseases/ or intracranial aneurysm/ or intracranial arteriosclerosis/ or intracranial arteriovenous malformations/ or "exp intracranial embolism and thrombosis"/ or intracranial hemorrhages/ or cerebral hemorrhage/ or exp basal ganglia hemorrhage/ or cerebral intraventricular hemorrhage/ or intracranial hemorrhage, hypertensive/ or subarachnoid hemorrhage/ or stroke/ or brain infarction/ or brain stem infarctions/ or cerebral infarction/ or infarction, anterior cerebral artery/ or infarction, middle cerebral artery/ or infarction, posterior cerebral artery/ or hemorrhagic stroke/ or exp ischemic stroke/ or vasospasm, intracranial/
2. exp neurological rehabilitation/
3. (stroke or poststroke or post‐stroke or cerebrovasc$ or (cerebr$ adj3 vasc$) or CVA$ or apoplectic or apoplex$ or (transient adj3 isch?emic adj3 attack) or tia$ or SAH or AVM or (cerebral small vessel adj3 disease)).tw.
4. ((cerebr$ or cerebell$ or arteriovenous or vertebrobasil$ or interhemispheric or hemispher$ or intracran$ or intracerebral or infratentorial or supratentorial or MCA$ or ((anterior or posterior) adj3 circulat$) or lenticulostriate or ((basilar or brachial or vertebr$) adj3 arter$)) adj3 (disease or damage$ or disorder$ or disturbance or dissection or lesion or syndrome or arrest or accident or lesion or vasculopathy or insult or attack or injury or insufficiency or malformation or obstruct$ or anomal$)).tw.
5. ((cerebr$ or cerebell$ or vertebrobasil$ or interhemispheric or hemispher$ or intracran$ or corpus callosum or intracerebral or intracortical or intraventricular or periventricular or posterior fossa or infratentorial or supratentorial or MCA$ or ((anterior or posterior) adj3 circulation) or basal ganglia or ((basilar or brachial or vertebr$) adj3 arter$) or space‐occupying or brain ventricle$ or subarachnoid$ or arachnoid$) adj3 (h?emorrhage or h?ematoma or bleed$ or microh?emorrhage or microbleed or (encephalorrhagia or hematencephal$))).tw.
6. ((cerebr$ or cerebell$ or arteriovenous or vertebrobasil$ or interhemispheric or hemispher$ or intracran$ or corpus callosum or intracerebral or intracortical or intraventricular or periventricular or posterior fossa or infratentorial or supratentorial or MCA$ or ((anterior or posterior) adj3 circulation) or basal ganglia or ((basilar or brachial or vertebr$) adj3 arter$) or space‐occupying or brain ventricle$ or lacunar or cortical or ocular) adj3 (isch?emi$ or infarct$ or thrombo$ or emboli$ or occlus$ or hypoxi$ or vasospasm or obstruct$ or vasculopathy or vasoconstrict$)).tw.
7. ((carotid or cerebr$ or cerebell$ or intracranial or basilar or brachial or vertebr$) adj3 (aneurysm or malformation$ or dysplasia or disease or bruit or injur$ or obstruct$ or occlusion or constriction or presclerosis or scleros$ or stenos$ or atherosclero$ or arteriosclero$ or plaque$ or thrombo$ or embol$ or arteriopathy)).tw.
8. hemiplegia/ or paresis/ or exp gait disorders, neurologic/
9. (hemipleg$ or hemipar$ or paresis or paraparesis or paretic).tw.
10. or/1‐9
11. brain‐computer interfaces/ or electrodes/ or electrodes, implanted/ or implantable neurostimulators/ or neural prostheses/ or auditory brain stem implants/ or ion‐selective electrodes/ or microelectrodes/ or exp micro‐electrical‐mechanical systems/ or wearable electronic devices/
12. electroencephalography/
13. imagination/
14. neural networks, computer/ or computer simulation/ or augmented reality/ or virtual reality/ or computers/ or user‐computer interface/ or man‐machine systems/ or signal processing, computer‐assisted/ or therapy, computer‐assisted/
15. (((brain$ or man or cerebral$ or cerebell$ or mind$ or neur$ or scalp or subdural or intracort$) adj5 (machine$ or comput$ or interface$ or integrated$ or interact$ or system$ or transmit$ or transmission$ or technolog$ or communicat$ or sens$ or applicat$ or programm$ or intervention$ or augment$ or process$ or implant$ or prosthe$ or devic$ or neuroprosthet$ or electrod$ or biofeedback or feedback)) or BMI or BCI or NCI or DNI or MMI).tw.
16. (((signal or (event adj3 related) or (motor adj2 evoke$) or somatosens$) adj3 (acquisition or process$ or extract$ or preprocess$ or classif$)) or P300 or electro‐cerebral or ECoG or electro‐corticogra$ or electro‐encephalogra$).tw.
17. ((neuroprosthes$ or neuroprosthetic$ or neurobionic$ or neurobionic$ or (neuro$ or neural$ or ((electromyograph$ or emg) adj3 trigger$))) adj3 (orthos$ or orthotic$ or prosthes$ or prosthetic$ or devic$ or technolog$ or implant$ or interface or sensor$ or electrod$)).tw.
18. or/11‐17
19. randomized controlled trial.pt.
20. controlled clinical trial.pt.
21. randomized.ab.
22. placebo.ab.
23. clinical trials as topic.sh.
24. random$.ab.
25. trial.ti.
26. or/19‐25
27. exp animals/ not humans.sh.
28. 26 not 27
29. 10 and 18 and 28

Contributions of authors

Yu Qin: wrote the protocol and designed search strategies.

Meixuan Li: provided general advice on the protocol.

Yanfei Li: provided general advice on the protocol.

Yaqin Lu: provided general advice on the protocol.

Xiue Shi: provided general advice on the protocol.

Gecheng Cui: provided general advice on the protocol.

Haitong Zhao: provided general advice on the protocol.

Kehu Yang: wrote the protocol.

Sources of support

Internal sources

• Key Laboratory of Evidence Based Medicine and Knowledge Translation of Gansu Province funding, China

  GSEBMKT‐2021CC01

External sources

• No sources of support provided

Declarations of interest

Yu Qin: none known
Meixuan Li: none known
Yanfei Li: none known
Yaqin Lu: none known
Xiue Shi: none known
Gecheng Cui: none known
Haitong Zhao: none known
KeHu Yang: none known

New