Effects of Robotic Therapy Associated With Noninvasive Brain Stimulation on Upper-Limb Rehabilitation After Stroke: Systematic Review and Meta-analysis of Randomized Clinical Trials

Abstract

Background

Robot-assisted therapy and noninvasive brain stimulation (NIBS) are promising strategies for stroke rehabilitation.

Objective

This systematic review and meta-analysis aims to evaluate the evidence of NIBS as an add-on intervention to robotic therapy in order to improve outcomes of upper-limb motor impairment or activity in individuals with stroke.

Methods

This study was performed according to the PRISMA Protocol and was previously registered on the PROSPERO Platform (CRD42017054563). Seven databases and gray literature were systematically searched by 2 reviewers, and 1176 registers were accessed. Eight randomized clinical trials with upper-limb body structure/function or activity limitation outcome measures were included. Subgroup analyses were performed according to phase poststroke, device characteristics (ie, arm support, joints involved, unimanual or bimanual training), NIBS paradigm, timing of stimulation, and number of sessions. The Grade-Pro Software was used to assess quality of the evidence.

Results

A nonsignificant homogeneous summary effect size was found both for body structure function domain (mean difference [MD] = 0.15; 95% CI = −3.10 to 3.40; P = 0.93; I² = 0%) and activity limitation domain (standard MD = 0.03; 95% CI = −0.28 to 0.33; P = 0.87; I² = 0%).

Conclusions

According to this systematic review and meta-analysis, at the moment, there are not enough data about the benefits of NIBS as an add-on intervention to robot-assisted therapy on upper-limb motor function or activity in individuals with stroke.

Keywords

stroke robotics upper extremity transcranial direct current stimulation transcranial magnetic stimulation meta-analysis

Introduction

The mean global lifetime risk of stroke increased from 22.8% in 1990 to 24.9% in 2016.¹ Only 38% of individuals who present flaccid arm plegia after stroke recover dexterity to some extent, and only 11% achieve full functional recovery.² Upper-limb function is directly related to the satisfactory performance of self-care activities, leisure, and labor.³ Upper-limb function also affects mental health⁴ and perception of quality of life.⁵

The identification of more effective interventions to enhance arm and hand function is a priority of neurorehabilitation research, reflected by an increasing number of randomized clinical trials.⁶ In this context, robotic therapy (RT) has evolved rapidly.⁷ Intensity of practice and task-specific training are key elements of successful RT.^8,9 Although a Cochrane review (45 trials, 1619 participants) reported that RT improved arm function, activities of daily living, and muscle strength scores,¹⁰ a recent multicenter trial (770 participants) suggested that further research is needed to find ways, or combined therapy strategies, to translate the improvements in upper-limb impairment after RT training into improvements in upper-limb function and activities of daily living.¹¹

In this context, noninvasive brain stimulation (NIBS) has emerged as a potential add-on tool to rehabilitation interventions by boosting adaptive plasticity mechanisms.¹² NIBS is based on the principle of application of external stimuli to modify brain activity. NIBS techniques include transcranial direct current stimulation (tDCS) and transcranial magnetic stimulation (TMS), among other interventions.^13,14

This systematic review and meta-analysis aims, for the first time, to evaluate the evidence of the association of RT and NIBS for upper-limb rehabilitation in stroke. We analyzed randomized clinical trials that compared the association of RT and active NIBS or sham for the improvement of upper-limb motor recovery on body function and structure as well as on activity according to the International Classification of Functioning, Disability and Health (ICF).¹⁵

Methods

This study was performed according to the PRISMA Protocol¹⁶ and was approved by the Sao Paulo University Medical School Research Ethics Committee in March 2017 (No. 0085/17). The protocol was registered on the PROSPERO Platform (CRD42017054563).

Search Strategy

Seven scientific databases were systematically searched: MEDLINE (Medical Literature Analysis and Retrieval System Online; through the PubMed interface), EMBASE (Excerpt Medical Database), Cochrane Central Register of Controlled Trials (CENTRAL), LILACS (Latin American & Caribbean Health Sciences Literature; through the Virtual Health Library—Bireme interface), CINAHL (Cumulative Index to Nursing and Allied Health Literature through the EBSCO interface), DORIS (Database of Research in Stroke), and PEDro (Physiotherapy Evidence Database). The articles were manually retrieved. In addition, the following online archives of theses or trial registers were searched: ClinicalTrials, Digital Library of Theses and Dissertations of São Paulo University, DORIS, Digital Brazilian Library of Theses and Dissertations, Public Domain Portal, and CAPES (Coordination for the Improvement of Higher Education Personnel) Thesis and Dissertation Bank.

The following keywords were used: stroke, robot, transcranial direct current stimulation, and transcranial magnetic current stimulation. The full search strategy can be found in Supplementary Material (SM), Table SM1. No publication date or language restrictions were imposed. The search included all studies published until July 2019.

Study Selection

Titles and/or abstracts of studies retrieved using the search strategy and those from additional sources were independently assessed by 2 reviewers. The 2 reviewers read the full-text articles and selected studies according to the inclusion and exclusion criteria. Disagreements between the 2 reviewers were resolved by consensus.

Eligibility Criteria

Types of studies: Randomized clinical trials with parallel or crossover design were included. Authors were contacted to provide missing data of abstracts and nonpublished studies.

Types of participants: People with upper-limb paresis caused by stroke were included. No restrictions were imposed regarding age, residual upper-limb motor-function, time since last stroke, type of stroke, or history of previous strokes. Studies that included individuals with cerebellar strokes or strokes in cerebellar pathways were not included.

Types of interventions: Trials that added active NIBS before, during, or after RT to improve upper-limb outcomes were included. Four types of NIBS were assessed: tDCS, transcranial alternating current stimulation, transcranial random noise stimulation, and TMS.

Types of controls: RT associated with sham NIBS or RT alone.

Types of outcome measures: Upper-limb performance measured by the main standard scales for the ICF body structure/body function domain (eg, Fugl-Meyer Assessment Scale [FMA], Ashworth or modified Ashworth Scale, force and range of motion) or activity level (eg, Wolf Motor Function Test [WMFT], Action Research Arm Test [ARAT], Motor Activity Log [MAL], Box and Blocks Test [BBT], Jebsen-Taylor Hand Function Test, Nine Hole Peg Test, and Motor Assessment Scale).¹⁷

Data Extraction

Data were extracted from the records to a standard form by 2 reviewers. Disagreements were resolved by consensus. If information was missing/unclear in articles, their authors were contacted. In case of crossover studies, only data of the first intervention were extracted for metanalyses.

If more than 1 outcome was assessed to measure body function/structure, the priority for analysis was the Fugl-Meyer according to recommendations of the measurement working group of the Stroke Recovery and Rehabilitation Roundtable.¹⁸ To measure activity limitation, priority of analysis was given to the ARAT according to recommendations of the same panel.¹⁸ If these scales were not present in the study, the choice of outcome was based on the most frequent outcomes across the selected studies.

Methodological Quality

Risks of bias were assessed independently by 2 reviewers. Disagreements were settled by consensus, and no consultation with a third reviewer was necessary. If there was a lack of, or unclear, information, authors were asked for clarification, or requests were made for the provision of the missing information.

The following domains were assessed for each study: randomization,¹⁹ concealment allocation,¹⁹ blinding of outcome assessment,¹⁹ blinding of participants and personnel,¹⁹ description or implicit intention-to-treat analysis, extent of loss,²⁰ sample homogeneity (similarity between characteristics data of active and control groups), sample representativeness (absence or presence of exclusion criteria other than those usually present on RT or NIBS trials in general), presence of description of sample calculation, information regarding early cessation of trials, and selective reporting.¹⁹ For each domain, the study was classified as having high, low, or unclear risk of bias.

The classification of bias was not used as a criterion to exclude studies from a possible meta-analysis but was used for analysis of quality of the evidence according to GRADEpro GDT (Grading of Recommendations Assessment, Development and Evaluation),¹⁶ by a GRADE Evidence Profile across the domain study design, risk of bias, inconsistency, indirectness, and imprecision. The domain risk of bias was classified prioritizing randomization, allocation, blinding, and extent of loss. Other domains contributed with the percentage of high or unclear risk of bias (ie, <50% not serious; 50%-75%, serious; >75%, very serious). The domain inconsistency was classified according to I² of metanalysis (<50% not serious; 50%-75%, serious; >75%, very serious). The domain imprecision was classified according to the Z test for overall effect (<50% not serious; 50%- 75%, serious; >75%, very serious).

The quality was stated as high (further research is very unlikely to change our confidence in the estimate of the effect), moderate (further research is likely to have an important impact on our confidence in the estimate of the effect and may change the estimate), low (further research is very likely to have an important impact on the confidence in the estimate of effect and is likely to change the estimate), or very low (very uncertain about the estimate of the effect).¹⁶

Quantitative Data Analysis

For studies in which the same scale was used to evaluate the outcome (ie, Fugl-Meyer for body structure/function), the number of participants in each group, mean scores, and SDs after interventions in the active and control groups were analyzed in RevMan 5.3. (Review Manager 5.3). For the Fugl-Meyer, a higher score was regarded as positive. The mean difference (MD) was established in individual studies by calculating the difference between the means of the active and control groups.

For types of outcomes assessed with different scales (ie, Activity Limitation), measures of postintervention and preintervention of each subject were assessed after contacting the authors and requesting individual data. The individual relative difference between postintervention and preintervention (Postintervention − Preintervention/Preintervention) was calculated for each participant. The means and SDs of relative differences in active and control groups were analyzed in RevMan 5.3. In individuals who presented a baseline score of 0, the relative difference was also considered to be 0. We opted for this approach because none of the individuals who scored 0 at baseline presented a postintervention to preintervention difference above the minimal detectable changes of the respective scale.

For scales in which a lower score was regarded as positive compared with a higher score, the mean was multiplied by −1. The standard mean difference (SMD) was established in individual studies by calculating the difference between the means of the active and control groups.

The summary effect size (SES) was calculated by the average of the MD or SMD of individual studies, with the corresponding 95% CI. Heterogeneity among studies was assessed by I². In case of statistical heterogeneity, defined as an I² ≥50%, a random-effect model was applied, whereas a fixed-effect model was applied for I² <50%.²¹ All analyses were performed using RevMan 5.3, and P values ≤0.05 were considered statistically significant.

In addition, the following subgroup analyses were conducted^7,12,22,23:

According to participants’ characteristics: Phase poststroke (acute, 1-7 days; early subacute, >7 days to 3 months; late subacute, 3-6 months; chronic, >6 months)²⁴

According to robotic device characteristics: Arm support (end-effector or exoskeleton),²⁵ joints involved (shoulder, elbow, wrist, hand), and unimanual or bimanual training

According to NIBS characteristics: NIBS paradigm (aiming to increase or decrease cortical excitability), timing of stimulation (preintervention, postintervention, or during RT)

According to general characteristics: Number of sessions (1-6, 7-12, 13-24, 25-36)²⁶

Results

Search Results

At the last search in July 2019, a total of 1266 records were identified from databases, gray literature, and hand search. After removing duplicates, 1176 articles were identified, and 14 abstracts were selected for full-text reading. The study selection process is shown in Supplementary Material, Figure SM1. At the end of the selection, 6 trials were excluded for the following reasons: 3 did not attempt criteria for control groups,^22,27,28 and 4 did not use standard clinical scales as outcome measures.^23,28-30

Characteristics of the Trials

The characteristics of participants of trials selected for the systematic review and meta-analysis are shown in Table 1, and the characteristics of interventions are shown in Table 2. One study³¹ applied 3 different interventions (sham tDCS, anodal tDCS [atDCS], cathodal tDCS [ctDCS]), and in this review, was divided into 2 (Hesse et al,³¹ A: anodal stimulation; and Hesse et al,³¹ C: cathodal stimulation). Thus, considering Hesse et al, 2011 as 2 different trials for statistical analysis, the review included 9 trials, with a total of 324 participants (161, active; 163, control).

Table 1.

Characteristics of the Participants of the Studies Included in the Meta-analysis.

Study	n (Sex)	Age Mean (SD)	Months/weeks/days poststroke, mean (SD)	Motor basal impairment FMA mean (SD)	Type of stroke	Affected hemisphere	Site of the lesion
Study	Active control (M/F)	Active control	Active control	Active control	Active control	Active control	Active control
Ang et al,³² 2015	10 (6/4)	52.1 (11.7)	1052 (722), D	35.3 (7.8)	6 I/4 H	5 L/5 R	1 C/9 SubC
Ang et al,³² 2015	9 (8/1)	56.3 (9.5)	1021 (465), D	32.6 (8.1)	7 I/2 H	6 L/3 R	0 C/9 SubC
Dehem et al,³⁸ 2018	11 (7/4)	62.73 (8.0)	17.47 (15.8), Mo	—	7 I/4 H	6 L/5 R	3 C/8 SubC
Dehem et al,³⁸ 2018	10 (8/2)	58.1 (10.8)	58.5 (75.5) – Mo	—	8 I/2 H	5 L/5 R	4 C/6 SubC
Di Lazzaro et al,³³ 2016	8 (4/4)	57.9 (12.4)	63.3 (72), Mo	14.5 (2.4)	8 I/0 H	NR	NR
Di Lazzaro et al,³³ 2016	9 (4/5)	56.8 (9.6)	61.3 (44.2), Mo	12.6 (2.2)	9 I/0 H	NR	NR
Edwards et al,³⁴ 2019	41 (25/16)	(See footnote)^a	(See footnote)^a	(See footnote)^a	41 I	41 R	26 C/15 SubC
Edwards et al,³⁴ 2019	41 (25/16)	(See footnote)^a	(See footnote)^a	(See footnote)^a	41 I	41 R	27 C/14 SubC
Hesse et al,³¹ 2011: A	32 (20/12)	63.9 (10.5)	3.4 (1.8), W	7.8 (3.8)	32 I/0 H	18 L/14 R	NR
Hesse et al,³¹ 2011: A	32 (21/11)	65.6 (10.3)	3.8 (1.5), W	8.2 (4.4)	32 I/0 H	16 L/16 R	NR
Hesse et al,³¹ 2011: C	32 (18/14)	65.4 (8.6)	3.8 (1.4), W	7.9 (3.4)	32 I/0 H	17 L/14 R	NR
Hesse et al,³¹ 2011: C	32 (21/11)	65.6 (10.3)	3.8 (1.5), W	8.2 (4.4)	32 I/0 H	16 L/16 R	NR
Panker et al, 2011	9 (6/3)	60.4 (16.3)	55.3 (42.2), Mo	33.4 (16.3)	4 I/3 H/2 unk	3 L/6 R	0 C/5 SubC/4 C + SubC
Panker et al, 2011	9 (6/3)	59.1 (10.9)	50.8 (15.1), Mo	22.2 (18.2)	8 I/0 H/1 unk	7 L/2 R	0 C/3 SubC/6 C + SubC
Straudi et al,³⁶ 2016	12 (5/7)	52.7 (16.0)	40.7 (35.1), W	24.1 (16.6)	10 I/2 H	9 L/3 R	9 C/3 SubC
Straudi et al,³⁶ 2016	11 (7/4)	64.3 (9.7)	78.2 (61.9), W	21.4 (13.2)	9 I/2 H	6 L/5 R	5 C/6 SubC
Triccas et al,³⁷ 2015	11 (6/5)	64.7 (9.2)	22.1 (31.2), Mo	24.9 (16.0)	9 I/2 H	6 L/5 R	3 C/7 SubC/1 unk
Triccas et al,³⁷ 2015	11 (7/4)	62.5 (14.3)	13.4 (16.3), Mo	37.1 (1.6)	9 I/2 H	5 L/6 R	4 C/6 SubC/1 C + SubC

Abbreviations: C, cortical; D, days; F, female; H, hemorrhagic; I, ischemic; M, male; L, left; Mo, months; NR, not reported; R, right; SubC, subcortical; unk, unknown; W, weeks.

Edward et al, 2019, provided median [interquartile range] of continuous variables: age, 70.0 [64.0, 77.0] active, 66.0 [61.0, 73.0] control; days poststroke, 654.0 [365.0, 1445.0] active, 1201.0 [425.0, 1693.0] control; motor basal impairment (FMA), 22.0 [9.0, 42.0] active; 22.0 [11.0, 43.0] control.

Table 2.

Interventions Reported in the Meta-analysis.

Study	Robotic device	NIBS	Therapy protocol	Outocome measure
Study	Robotic device	NIBS	Therapy protocol	Body function and Structure	Activity limitation
Ang et al,³² 2015	MIT Manus	Bilateral tDCS	10 Sessions of 20 minutes of bilateral tDCS (active or sham), followed by 1-hour RT session (MI-BCI) during 2 weeks	FMA	—
Dehem et al,³⁸ 2018	REAplan robot	Bilateral tDCS	1 Session of 20 minutes of bilateral tDCS (active or sham) during the 20 minutes of RT	—	BB
Di Lazzaro et al,³³ 2016	InMotion2 (commercial version MIT-Manus)	Theta burst	10 Sessions of inhibitory cTMS on affected hemisphere (active or sham), followed by 960 movements (active assistive) of RT during 2 weeks	FMA	—
Edwards et al,³⁴ 2019	MIT Manus	Anodal tDCS	36 Sessions of 20 minutes of anodal tDCS (active or sham), followed by ~1 hour RT session (1024 movement repetitions) during 12 weeks	FMA	WMFT
Hesse et al,³¹ 2011: A	Bi-Manu Track	Anodal tDCS	30 Sessions of anodal tDCS (active or sham) during the first 20 minutes of RT (400 movements) plus conventional therapy (physical and occupational therapy) during 6 weeks	FMA	BB
Hesse et al,³¹ 2011: C	Bi-Manu Track	Cathodal tDCS	30 Sessions of anodal tDCS (active or sham) during the first 20 minutes of RT (400 movements) plus conventional therapy (physical and occupational therapy) during 6 weeks	FMA	—
Panker et al, 2011	ReoGo	Anodal tDCS	22 Sessions of anodal tDCS (active or sham) during the first 20 of 60 minutes of RT during 2.5 weeks	FMA	WMFT
Straudi et al,³⁶ 2016	ReoGo	Bilateral tDCS	10 Sessions of bilateral tDCS (active or sham) during the 30 minutes of RT for 2 weeks	FMA	BB
Triccas et al,³⁷ 2015	Armeo®Spring	Anodal tDCS	18 Sessions of anodal tDCS (active or sham) during the first 25 of 75 minutes of RT during 8 weeks	FMA	ARAT

Abbreviations: ARAT, Action Research Arm Test; BB, Box and Blocks Test; FMA, Fugl Meyer Assessment; MI-BCI, Motor Imagery—Brain Computer Interface; NIBS, noninvasive brain stimulation; RT, robotic therapy; tDCS, transcranial direct current stimulation; WMFT, Wolf Motor Function Test.

The robot devices used in the studies were the following: InMotion2, Mit Manus, BiManu Track, ReoGo, Armeo®Spring, and REAplan robot. Their characteristics are shown in Supplementary Material, Table SM2.

Methodological Quality

Risk-of-bias assessment can be found in Supplementary Material, Figure SM2.

Main Analysis

Body structure/body function: 7 of the included studies^31-37 chose the FMA as an outcome measure of body structure/function. Overall, a nonsignificant homogeneous SES (MD = 0.15; 95% CI = −3.10 to 3.40; P = .93; I² = 0%) was found (Figure 1).

Figure 1.

Forest plot of all trials comparing RT associated with NIBS (active) versus RT associated with sham NIBS (control) for body structure/function outcomes.

Activity Limitation

Six of the included studies^31,34-38 evaluated eligible scales of activity limitation: ARAT,³⁷ BBT,^31,36,38 MAL,³⁶ and WMFT.^34,35 Of the 2 studies that assessed the WMFT, one reported the WMFT-Time score (time required to complete the tasks: a lower score was regarded as positive compared with a higher score).³⁵ The other study described the WMFT Functional Ability Scale score (a higher score was regarded as positive compared with a lower score).³⁴ Data from Hesse et al³¹ were not available. Therefore, results of 5 studies^34-38 were analyzed. Overall, a nonsignificant homogeneous SES (SMD = 0.03; 95% CI = −0.28 to 0.33; P = .87; I² = 0%) was found (Figure 2).

Figure 2.

Forest plot of trials comparing RT associated with NIBS (active) versus RT associated with sham NIBS (control) for activity limitation outcomes.

Subgroup Analyses

Subgroup analyses are shown in Table 3.

Table 3.

Paradigms of Studies Included in Subgroup Analyses.

Study	Participants, phase poststroke	Robot device			NIBS		General, number of sessions
Study	Participants, phase poststroke	Arm support	Joints involved	Unimanual or bimanual training	NIBS paradigm	Timing of stimulation	General, number of sessions
Ang et al,³² 2015	Chronic	End effector	Shoulder/elbow	Unimanual	Bilateral atDCS and ctDCS	Before RT	7-12
Dehem et al,³⁸ 2018	Chronic	End effector	Shoulder/elbow	Unimanual	Bilateral atDCS and ctDCS	During RT	1-6
Di Lazzaro et al,³³ 2016	Chronic	End effector	Shoulder/elbow	Unimanual	cTBS on affected hemisphere	Before RT	7-12
Edwards et al,³⁴ 2019	Chronic	End effector	Whole arm	Unimanual	atDCS on affected hemisphere	Before RT	25-36
Hesse et al,³¹ 2011: A	Subacute	End effector	Wrist/hand	Bimanual	atDCS on affected hemisphere	During RT	25-36
Hesse et al,³¹ 2011: C	Subacute	End effector	Wrist/hand	Bimanual	ctDCS on unaffected hemisphere	During RT	25-36
Panker et al, 2011	Chronic	End effector	Shoulder/elbow	Unimanual	atDCS on affected hemisphere	During RT	13-24
Straudi et al,³⁶ 2016	Chronic, subacute	End effector	Shoulder/elbow	Unimanual	Bilateral atDCS and ctDCS	During RT	7-12
Triccas et al,³⁷ 2015	Chronic, subacute	Exoskeleton	Whole arm	Unimanual	atDCS on affected hemisphere	During RT	13-24

Abbreviations: atDCS, anodal tDCS; ctDCS, cathodal tDCS; cTBS, continuous theta burst stimulation; NIBS, noninvasive brain stimulation; RT, robotic therapy; tDCS, transcranial direct current stimulation.

Phase Poststroke

Body structure/body function: Nonsignificant homogeneous SESs were found for the chronic (MD = 1.18; 95% CI = −3.13 to 5.50; P = .59; I² = 0%) and subacute (MD = −1.60; 95% CI = −6.14 to 2.94; P = .49; I² = 36%) phases (Supplementary Material, Figure SM3).

Activity limitation: Nonsignificant homogeneous SESs were found for participants in the chronic (SMD = 0.10; 95% CI = −0.23 to 0.43; P = .56; I² = 30%) and subacute phases (SMD = −0.28, 95% CI = −1.14 to 0.59, P = .53, I² = 0%; Supplementary Material, Figure SM4).

Arm Support

Body structure/body function: Nonsignificant homogeneous SESs were found for end-effector (MD = 0.83, 95% CI = −2.51 to 4.18; P = .62; I² = 0%) and exoskeleton (MD = −11.20; 95% CI = −24.82 to 2.42; P = .11) devices (Supplementary Material, Figure SM5).

Activity limitation: Nonsignificant homogeneous SESs were found for end-effector (SMD 0.07, 95% CI -0.26 to 0.40; P = .69, I² = 7%) and exoskeleton (SMD = −0.24; 95% CI = −1.07 to 0.60; P = .58) devices (Supplementary Material, Figure SM6).

Joints Involved

Body structure/body function: Nonsignificant homogeneous SESs were found for the comparisons involving robots with only shoulder/elbow movements (MD = 3.28; 95% CI = −2.35 to 8.91; P = .25; I² = 0%), robots with wrist movements only (MD = −0.21; 95% CI = −4.97 to 4.55; P = .93; I² = 0%), and robots with whole arm movements (MD = −4.14, 95% CI = −11.35 to 3.06, P = .26, I² = 30%; Supplementary Material, Figure SM7).

Activity limitation: Nonsignificant homogeneous SESs were found for the comparison involving robots with only shoulder/elbow movements (SMD = 0.01; 95% CI = −0.50 to 0.52; P = .98; I² = 36) and for the comparison involving robots with whole arm movements (SMD = −0.04, 95% CI = −0.35 to 0.42, P = .85, I² = 0%; Supplementary Material, Figure SM8).

Unimanual or Bimanual Training

Only 1 trial³¹ used a robotic device with bimanual training. The results of this trial were split according to the intervention performed: Hesse et al,³¹ A (atDCS) and Hesse et al, C (ctDCS).

Body structure/body function: For unimanual (MD = 0.47; 95% CI = −3.97 to 4.90; P = .84; I² = 7%) and bimanual (MD = −0.21; 95% CI = −4.97 to 4.55; P = .93; I² = 0%) training, nonsignificant homogeneous SESs were found.

Activity limitation: A nonsignificant homogeneous SES (SMD = 0.03; 95% CI = −0.28 to 0.33; P = .87; I² = 0%) was found.

NIBS Paradigm

Three studies used atDCS to improve cortical excitability of the affected hemisphere^34,35,37; 3 studies placed the tDCS anode over the affected hemisphere and the cathode over the unaffected hemisphere^32,36,38; and 1 study used atDCS and ctDCS in different groups.³¹ Just 1 study designed a proof-of-principle trial, aiming to induce long-term depression-like changes on the affected hemisphere by continuous theta burst stimulation (cTBS), a specific paradigm of repetitive TMS.³³

Body structure/body function: For atDCS, a nonsignificant heterogeneous SES (MD = −0.78; 95% CI = −5.62 to 4.06; P = .75; I² = 37%) was found. For ctDCS, a nonsignificant SES (MD = −0.30; 95% CI = −6.64 to 6.04; P = .93) was found. For bilateral atDCS and ctDCS, a nonsignificant homogeneous SES (MD = 1.18; 95% CI = −7.22 to 9.58; P = .78; I² = 0%) was found. For cTBS of the affected hemisphere, a nonsignificant SES (MD = 2.90; 95% CI = −5.80 to 11.60; P = .51) was found (Supplementary Material, Figure SM9).

Activity limitation: For atDCS of the affected hemisphere, a nonsignificant homogeneous SES was found (SMD = −0.04; 95% CI = −0.39 to 0.32; P = .84; I² = 0%). For bilateral atDCS and ctDCS, a nonsignificant homogeneous SES was found (SMD = −0.21, 95% CI = −0.40 to 0.82, P = .50, I² = 40%; Supplementary Material, Figure SM10).

Timing of Stimulation

Body structure/body function: Nonsignificant homogeneous SESs were found for NIBS applied during RT (MD = −0.21; 95% CI = −4.34 to 3.93; P = .92; I² = 17%) and for NIBS applied before RT (MD = 0.72, 95% CI = −4.52 to 5.97, P = .79, I² = 0%; Supplementary Material, Figure SM11).

Activity limitation: Nonsignificant homogeneous SESs were found for both NIBS applied during RT (SMD = −0.06, 95% CI = −0.50 to 0.38, P = .79, I² = 11%) and before RT (SMD = 0.11, 95% CI = −0.32 to 0.54, P = .062; Supplementary Material, Figure SM12).

Number of Sessions

In 1 crossover trial with 1 active and 1 sham session of treatment, the results of the first session were included in the analysis.³⁸ The other trials had a parallel design.

Body structure/body function: Nonsignificant homogeneous SESs were found for trials which had 7 to 12 sessions (MD = 2.01; 95% CI = −4.03 to 8.05; P = .51; I² = 0%) and from 25 to 36 sessions (MD = −0.50; 95% CI = −4.65 to 3.66; P = .81; I² = 0%). A nonsignificant heterogeneous SES was found for trials with 13 to 24 sessions (MD = −0.07, 95% CI = −22.50 to 22.36, P = 1.00, I² = 79%; Supplementary Material, Figure SM13).

Activity limitation: Nonsignificant homogeneous SESs were found for the first session of the crossover trial (SMD = 0.65; 95% CI = −0.25 to 1.56; P = .16) and for trials with 7 to 12 sessions (SMD = −0.15; 95% CI = −0.97 to 0.67; P = .71), 13 to 24 sessions (SMD = −0.35; 95% CI = −0.97 to 0.28; P = .28; I² = 0%), and 25 to 36 sessions (SMD = 0.11, 95% CI = −0.32 to 0.54, P = .62; Supplementary Material, Figure SM14).

Quality of the Evidence

An evidence profile was performed on GRADEpro GDT. Quality was considered high for both body structure/function and activity limitation outcome measures, suggesting that further research is very unlikely to change the confidence in the estimate of the effect with the paradigms used in these studies¹⁶ (Supplementary Material, Table SM3). However, quality of the evidence ranges from very low to high scores according to subgroup analysis (Supplementary Material, Table SM4).

Discussion

A total of 324 individuals were included in this meta-analysis, and there was no statistical heterogeneity in the results. Overall, the quality of the evidence was high. Despite the increase in the number of clinical trials that assessed NIBS as an add-on intervention of RT over the past years, there is no evidence to support the hypothesis that effects of RT on outcomes of body structure and function or on outcomes of activity limitation can be enhanced by NIBS. This conclusion is in line with the conclusion of a previous narrative review.¹² Despite the lack of statistical heterogeneity, however, the NIBS paradigms and motor training protocols were very diverse across studies.

Four different NIBS paradigms were used in the included studies, and all, except one, were based on the hypothesis of the interhemispheric competition model. This hypothesis seems not to apply to patients with stroke with different levels of impairment^39,40 or lesion locations.⁴¹ Other variables that raise concerns about the potential of NIBS as a therapeutic tool are intersubject and intrasubject variability that can be caused by individual brain anatomy, level of ongoing cortical activity, muscle precontraction, participant attentional focus, and even menstrual cycle and circadian rhythms.⁴² Neuronavigated NIBS and electroencephalogram monitoring of stimulation effects in real time may be useful to decrease variability of effects.⁴² To select patients for clinical trials or proof-of-principle studies involving NIBS and motor training, understanding of mechanisms and biomarkers of responsiveness may be key to enhance the probability of success.⁴³

The need to tailor NIBS interventions to particular characteristics has been underscored by the negative results of the NICHE study, the largest clinical trial to date in which low-frequency rTMS was administered to the unaffected hemisphere of patients in the chronic phase after stroke with the goal of downregulating excessive inhibition of the affected hemisphere, prior to sessions of motor training.⁴⁴ Another plausible explanation of the negative results of the NICHE trial as well as those of this meta-analysis is a ceiling effect: the impact of high-intensity motor practice on improvement of performance may exceed the magnitude of NIBS effects.

Alternatively, the association of NIBS and RT may have effects on retention of improved motor performance rather than on immediate effects of motor training. Results of a trial that randomized 164 patients to active or sham atDCS of the motor cortex of the affected hemisphere prior to RT,³⁴ included in this review, contradict this hypothesis. After 36 sessions of treatment, the benefits of RT were unchanged by add-on atDCS. However, at 6 months posttreatment, the percentage of “responders” (improvement in FM >5 points) was significantly greater in the control group than in the active group. Finally, it is necessary to consider that NIBS objectives are not only to enhance effects of training, but also to yield comparable results of a training intervention when administered for a shorter period of time. No studies addressing this issue were identified.

Another point that deserves attention is that an optimal training RT paradigm to be associated with NIBS has not yet been determined. Giacobbe et al²³ demonstrated that a single session of tDCS can enhance effects of motor training when applied immediately prior to RT but not during or after this intervention. Also, alternating sessions between proximal and distal joints seem to be a more efficient strategy to improve upper-limb motor performance than isolated block sessions of proximal and distal joints.^11,34 It remains to be determined if this result also applies to NIBS + robotic training because no head-to-head comparisons of different paradigms of training associated with the same NIBS intervention were found. Finally, the role of robotic bimanual training remains to be clarified. This intervention may enhance the output to the paretic limb in some individuals with stroke,⁴⁵ but we only included in this review 1 study that applied this approach.³¹

Clinical trials that test a large number of combinations of NIBS and training paradigms may be challenging as a result of resource and time constraints. Kinematic measurements are sensitive and correlate well with clinical assessment in patients with stroke.⁴⁶ Therefore, proof-of-principle studies that assess responsiveness to different paradigms may inform the design of larger clinical trials.

None of the studies included in this meta-analysis chose outcomes of activity limitation as primary end points, and only 1 study assessed the ARAT,³⁷ an outcome recommended by the Stroke Recovery and Rehabilitation Roundtable.¹⁸ Future studies may include this outcome measure as performed in the RATULS (Robot Assisted Training for the Upper Limb after Stroke) study.¹¹ No significant changes were reported for the ARAT in RATULS,¹¹ a trial in which RT was not associated with functional training. On the other hand, a smaller trial compared RT training (n = 22) with RT associated with therapist-assisted functional training (n = 23) and found significant improvement on activity limitation favoring the experimental group.⁴⁷ Overall, these results suggest that changes in measures of body structure and function or activity limitation may vary according to the specific model of motor training. It is possible that translation-to-task interventions as add-on therapies to RT are required to ameliorate activity limitation. In trials that plan to address the possibility of boosting the effect of RT by adding NIBS to enhance activity, inclusion of translation-to-task interventions should be considered.

Limitations

The number of included studies in this systematic review was relatively low, but the sample size (n = 324) was greater than those reported in other meta-analyses about effects of specific rehabilitation interventions in stroke. For instance, a meta-analysis⁴⁸ that evaluated the effectiveness of a combination of NIBS and virtual reality included 213 individuals. Nevertheless, one needs to interpret our result with the appropriate caveats: the small pooled sample size, the large number of combinations of NIBS and RT, and the lack of details about baseline motor impairment and intensity of robot training protocols. Yet one could argue that the fact that our approach led to an inconclusive result might be an indication that the benefit of NIBS added on top of RT might be small, if any, and represent the natural limitation of the remaining neuronal substrate. This systematic review and meta-analysis points to directions of future research: to define whether one or more types of NIBS can be useful as add-on therapies to RT, we might track the profile and speed of the improvements rather than simply register the changes from admission to completion of the protocols.

Conclusion

According to this systematic review and meta-analysis, at the moment, there are not enough data about the benefits of NIBS as an add-on intervention to RT on upper-limb motor function or activity in individuals with stroke. There is a need for homogeneous clinical trials regarding participants’ characteristics, kind of robotic device, NIBS paradigm, duration of treatment, and primary end points.

Supplemental Material

sj-docx-1-nnr-10.1177_1545968321989353 – Supplemental material for Effects of Robotic Therapy Associated With Noninvasive Brain Stimulation on Upper-Limb Rehabilitation After Stroke: Systematic Review and Meta-analysis of Randomized Clinical Trials

Supplemental material, sj-docx-1-nnr-10.1177_1545968321989353 for Effects of Robotic Therapy Associated With Noninvasive Brain Stimulation on Upper-Limb Rehabilitation After Stroke: Systematic Review and Meta-analysis of Randomized Clinical Trials by Suzana Bleckmann Reis, Wanderley Marques Bernardo, Carlos Andre Oshiro, Hermano Igo Krebs and Adriana Bastos Conforto in Neurorehabilitation and Neural Repair

Footnotes

Declaration of Conflicting Interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: SBR, WMB, CAO, and ABC have no conflicts of interest. HIK has a patent Bionik Laboratories with royalties paid to Massachusetts Institute of Technology.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported in part by the MIT International Science and Technology Initiatives (MISTI Brazil TVLM), under grant number 2160714.

ORCID iDs

Suzana Bleckmann Reis

Adriana Bastos Conforto

Supplementary material for this article is available on the Neurorehabilitation & Neural Repair website at .

References

Feigin

Nguyen

Cercy

, et al. Global, regional, and country-specific lifetime risks of stroke, 1990 and 2016. N Engl J Med. 2018;379:2429-2437.

Kwakkel

Kollen

van der Grond

Prevo

AJ.

Probability of regaining dexterity in the flaccid upper limb: impact of severity of paresis and time since onset in acute stroke. Stroke. 2003;34:2181-2186.

Sveen

Bautz-Holter

Sødring

Wyller

Laake

Association between impairments, self-care ability and social activities 1 year after stroke. Disabil Rehabil. 1999;21:372-377.

Morris

van Wijck

Joice

Donaghy

Predicting health related quality of life 6 months after stroke: the role of anxiety and upper limb dysfunction. Disabil Rehabil. 2013;35:291-299.

Franceschini

La Porta

Agosti

Massucci

; ICR2 Group. Is health-related-quality of life of stroke patients influenced by neurological impairments at one year after stroke? Eur J Phys Rehabil Med. 2010;46:389-399.

Pollock

Farmer

Brady

, et al. Interventions for improving upper limb function after stroke. Cochrane Database Syst Rev. 2014;2014(11):CD010820.

Veerbeek

Langbroek-Amersfoort

van Wegen

Meskers

Kwakkel

Effects of robot-assisted therapy for the upper limb after stroke: a systematic review and meta-analysis. Neurorehabil Neural Repair. 2017;31:107-121.

Veerbeek

van Wegen

van Peppen

, et al. What is the evidence for physical therapy poststroke? A systematic review and meta-analysis. PLoS One. 2014;9:e87987.

Lohse

Hilderman

Cheung

Tatla

Van der Loos

HF.

Virtual reality therapy for adults post-stroke: a systematic review and meta-analysis exploring virtual environments and commercial games in therapy. PLoS One. 2014;9:e93318.

10.

Mehrholz

Pohl

Platz

Kugler

Elsner

Electromechanical and robot-assisted arm training for improving activities of daily living, arm function, and arm muscle strength after stroke. Cochrane Database Syst Rev. 2018;(9):CD006876.

11.

Rodgers

Bosomworth

Krebs

, et al. Robot assisted training for the upper limb after stroke (RATULS): a multicentre randomised controlled trial. Lancet. 2019;394:51-62.

12.

Simonetti

Zollo

Milighetti

, et al. Literature review on the effects of tDCS coupled with robotic therapy in post stroke upper limb rehabilitation. Front Hum Neurosci. 2017;11:268.

13.

Fertonani

Miniussi

Transcranial electrical stimulation: what we know and do not know about mechanisms. Neuroscientist. 2017;23:109-123.

14.

Kubis

Non-invasive brain stimulation to enhance post-stroke recovery. Front Neural Circuits. 2016;10:56.

15.

Langhorne

Bernhardt

Kwakkel

Stroke rehabilitation. Lancet. 2011;377:1693-1702.

16.

Shamseer

Moher

Clarke

, et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015: elaboration and explanation. BMJ. 2015;350:g7647.

17.

Santisteban

Térémetz

Bleton

Baron

Maier

Lindberg

PG.

Upper limb outcome measures used in stroke rehabilitation studies: a systematic literature review. PLoS One. 2016;11:e0154792.

18.

Kwakkel

Lannin

Borschmann

, et al. Standardized measurement of sensorimotor recovery in stroke trials: consensus-based core recommendations from the Stroke Recovery and Rehabilitation Roundtable. Int J Stroke. 2017;12:451-461.

19.

Carvalho

APV

Silva

Grande

. Avaliação do risco de viés de ensaios clínicos randomizados pela ferramenta da colaboração Cochrane. Diagn Tratamento. 2013;18:38-44.

20.

Nobre

MRC

Bernardo

Jatene

. Evidence based clinical practice. Part III—critical appraisal of clinical research [in Portuguese]. Rev Assoc Méd Bras (1992). 2004;50:221-228.

21.

Higgins

Thomas

Chandler

, et al. Cochrane Handbook for Systematic Reviews of Interventions. 2nd ed. John Wiley & Sons; 2019.

22.

Ochi

Saeki

Oda

Matsushima

Hachisuka

Effects of anodal and cathodal transcranial direct current stimulation combined with robotic therapy on severely affected arms in chronic stroke patients. J Rehabil Med. 2013;45:137-140.

23.

Giacobbe

Krebs

Volpe

, et al. Transcranial direct current stimulation (tDCS) and robotic practice in chronic stroke: the dimension of timing. NeuroRehabilitation. 2013;33:49-56.

24.

Bernhardt

Hayward

Kwakkel

, et al. Agreed definitions and a shared vision for new standards in stroke recovery research: The Stroke Recovery and Rehabilitation Roundtable Taskforce. Neurorehabil Neural Repair. 2017;31:793-799.

25.

Chang

Kim

YH.

Robot-assisted therapy in stroke rehabilitation. J Stroke. 2013;15:174-181.

26.

Guarino

Richards

, et al. Robot-assisted therapy for long-term upper-limb impairment after stroke. N Engl J Med. 2010;362:1772-1783.

27.

Hesse

Werner

Schonhardt

Bardeleben

Jenrich

Kirker

SG.

Combined transcranial direct current stimulation and robot-assisted arm training in subacute stroke patients: a pilot study. Restor Neurol Neurosci. 2007;25:9-15.

28.

Edwards

Krebs

Rykman

, et al. Raised corticomotor excitability of M1 forearm area following anodal tDCS is sustained during robotic wrist therapy in chronic stroke. Restor Neurol Neurosci. 2009;27:199-207.

29.

Ang

Guan

Phua

, et al. Transcranial direct current stimulation and EEG-based motor imagery BCI for upper limb stroke rehabilitation. Annu Int Conf IEEE Eng Med Biol Soc. 2012;2012:4128-4131.

30.

Buetefisch

Heger

Schicks

Seitz

Netz

Hebbian-type stimulation during robot-assisted training in patients with stroke. Neurorehabil Neural Repair. 2011;25:645-655.

31.

Hesse

Waldner

Mehrholz

Tomelleri

Pohl

Werner

Combined transcranial direct current stimulation and robot-assisted arm training in subacute stroke patients: an exploratory, randomized multicenter trial. Neurorehabil Neural Repair. 2011;25:838-846.

32.

Ang

Guan

Phua

, et al. Facilitating effects of transcranial direct current stimulation on motor imagery brain-computer interface with robotic feedback for stroke rehabilitation. Arch Phys Med Rehabil. 2015;96(3, suppl):S79-S87.

33.

Di Lazzaro

Capone

Di Pino

, et al. Combining robotic training and non-invasive brain stimulation in severe upper limb-impaired chronic stroke patients. Front Neurosci. 2016;10:88.

34.

Edwards

Cortes

Rykman-Peltz

, et al. Clinical improvement with intensive robot-assisted arm training in chronic stroke is unchanged by supplementary tDCS. Restor Neurol Neurosci. 2019;37:167-180.

35.

Panker

Nicole

. The Effects of Robotic Training and Cortical Stimulation on Reaching Skill After Chronic Stroke. Dissertation. Georgetown University; 2011.

36.

Straudi

Fregni

Martinuzzi

Pavarelli

Salvioli

Basaglia

tDCS and robotics on upper limb stroke rehabilitation: effect modification by stroke duration and type of stroke. Biomed Res Int. 2016;2016:5068127.

37.

Triccas

Burridge

Hughes

Verheyden

Desikan

Rothwell

A double-blinded randomised controlled trial exploring the effect of anodal transcranial direct current stimulation and uni-lateral robot therapy for the impaired upper limb in sub-acute and chronic stroke. NeuroRehabilitation. 2015;37:181-191.

38.

Dehem

Gilliaux

Lejeune

, et al. Effectiveness of a single session of dual-transcranial direct current stimulation in combination with upper limb robotic-assisted rehabilitation in chronic stroke patients: a randomized, double-blind, cross-over study. Int J Rehabil Res. 2018;41:138-145.

39.

Di Pino

Pellegrino

Assenza

, et al. Modulation of brain plasticity in stroke: a novel model for neurorehabilitation. Nat Rev Neurol. 2014;10:597-608.

40.

Sankarasubramanian

Machado

Conforto

, et al. Inhibition versus facilitation of contralesional motor cortices in stroke: deriving a model to tailor brain stimulation. Clin Neurophysiol. 2017;128:892-902.

41.

Thickbroom

Cortes

Rykman

, et al. Stroke subtype and motor impairment influence contralesional excitability. Neurology. 2015;85:517-520.

42.

Terranova

Rizzo

Cacciola

, et al. Is there a future for non-invasive brain stimulation as a therapeutic tool? Front Neurol. 2018;9:1146.

43.

Ward

NS.

Non-invasive brain stimulation for stroke recovery: ready for the big time?

J Neurol Neurosurg Psychiatry. 2016;87:343-344.

44.

Harvey

Edwards

Dunning

, et al. Randomized sham-controlled trial of navigated repetitive transcranial magnetic stimulation for motor recovery in stroke. Stroke. 2018;49:2138-2146.

45.

Cunningham

Knutson

Sankarasubramanian

Potter-Baker

Machado

Plow

EB.

Bilateral contralaterally controlled functional electrical stimulation reveals new insights into the interhemispheric competition model in chronic stroke. Neurorehabil Neural Repair. 2019;33:707-717.

46.

Krebs

Krams

Agrafiotis

, et al. Robotic measurement of arm movements after stroke establishes biomarkers of motor recovery. Stroke. 2014;45:200-204.

47.

Conroy

Wittenberg

Krebs

Zhan

Bever

Whitall

Robot-assisted arm training in chronic stroke: addition of transition-to-task practice. Neurorehabil Neural Repair. 2019;33:751-761.

48.

Subramanian

Prasanna

SS.

Virtual reality and noninvasive brain stimulation in stroke: how effective is their combination for upper limb motor improvement?—A meta-analysis. PM R. 2018;10:1261-1270.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.27 MB