Introduction

Neurostimulation is fast growing in the field of neurological rehabilitation, where many types of patients such as those with stroke, Parkinson’s disease and multiple sclerosis are benefitting from it [1,2,3,4,5,6,7,8,9,10]. It is defined as the use of electric, electromagnetic, chemical or optogenetic methods to stimulate or block the flow of action potential through the central nervous system (CNS) [11,12,13,14,15]. In patients with stroke, it is used to help with recovery of brain functions such as sensory, motor and cognitive functions [1, 16].

There are basically two methods of application of neurostimulation, invasive (where the stimulation is achieved by surgically implanting electrodes in the stimulation sites) and non-invasive (where the stimulation is achieved by connecting electrodes to the external parts of the stimulation sites such as the skin) techniques. The invasive type of neurostimulation includes techniques such as the invasive vagus nerve stimulation (VNS) and deep brain stimulation [17, 18]; whereas, the non-invasive type of neurostimulation includes techniques such as the transcutaneous electrical nerve stimulation (TENS), neuromuscular electrical stimulation, orthosis-supported neuromuscular electrical stimulation, transcranial direct current stimulation (tDCS), transcranial alternating current simulation (tACS), transcranial pulse simulation (tPS), transcranial random noise stimulation (tRNS), transcranial magnetic stimulation (TMS), radio-electric asymmetry conveyer (REAC) and non-invasive VNS [19,20,21,22,23,24,25,26]. However, functional electrical stimulation can be used as either non-invasive or invasive type of neurostimulation [10].

Neurostimulation techniques can be delivered in the clinic or at home [27,28,29]. A home-based mode of rehabilitation is a healthcare delivery model employed to enhance easy access of rehabilitation services for patients with various conditions [30,31,32,33]. Its sole aims are to help reduce the cost of healthcare, and improve patients’ confidence and motivation, and compliance with the rehabilitation [34, 35]. This is because aside from the effectiveness of an intervention based on behavioural and neurophysiological outcomes, its cost is equally important; and a recent suggestion seeks for the use of the most cost-effective interventions [36]. In addition, home-based rehabilitation seems to offer more opportunity for increased intensity of rehabilitation, which is an important factor for recovery of function after stroke [37]. Similarly, it affords the patients with the opportunity to save money on transport, and reduce or prevent the risk of hospital-acquired infections and other communicable diseases, especially during epidemics or pandemics [38,39,40].

Furthermore, what is very interesting in stroke rehabilitation is that, home-based rehabilitation using exercises produces similar positive results as clinic-based rehabilitation [41]. The aim of this study is to carry out a systematic review and meta-analysis to determine from the literature, the effects of home-based neurostimulation in patients with stroke. In addition, the study is aimed at investigating its reported feasibility by summarizing reports of serious adverse events, and participants’ compliance with the protocols.

Materials and methods

We conducted a systematic review and meta-analysis, which was registered in PROSPERO database (registration number, CRD42023401257) using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline.

Inclusion and exclusion criteria

In the study, only randomized controlled trials (RCT) that compared the effects of home-based neurostimulation with sham neurostimulation or a control intervention on outcomes such as upper limb function, lower limb function, neurophysiological changes, spasticity and adverse events after stroke were included. The studies must also include participants who were 18 years old or more. However, studies that were not published in English language were excluded.

To make effective syntheses of the included studies, they were grouped based on the body part treated (upper and lower limbs) and the outcomes they assessed.

Procedure for literature search

The following databases: PubMED, Embase, Web of Science, Scopus, and CENTRAL were searched from their earliest dates until July, 2023. In addition, manual search of the references of the included studies and relevant systematic reviews was also carried out [29, 42]. The search was carried out using strategies adapted to the particular database by one of the researchers (AA); however, it was independently verified by another researcher (TWLW). The search terms used include stroke, brain infarction, cerebrovascular accident, electrical stimulation, transcutaneous electrical nerve stimulation, transcranial direct current stimulation, transcranial magnetic stimulation, deep brain stimulation, transcranial alternating current stimulation, transcranial random noise stimulation, telerehabilitation, virtual rehabilitation and remote rehabilitation. See Appendix 1 for the details of the search strategy used.

Selection of studies and extraction of data

Eligible studies were selected manually and by using Endnote software. The selection was carried out independently by two of the researchers (AA & TWLW).

At first, some of the studies that were ineligible based on the information from their titles and abstract were excluded. However, when the information in their titles and abstract was not sufficient to decide on their eligibility, their full texts were read to decide for their inclusion or otherwise. Moreover, in case of disagreement on the selection decisions between the two researchers (AA & TWLW), a third researcher (SSMN) was consulted for consensus.

Similarly, the data was extracted by one of the researchers (AA); however, it was verified by the other two researchers (TWLW & SSMN). The data extracted include information on the sociodemographic and clinical characteristics of the study participants such as the study authors, participants mean age, time since stroke, sample size, type of stroke and side affected; and the mean scores on the outcomes of interest (primary and secondary outcomes).

The primary outcomes are upper limb function (level of motor impairment, motor function, real world arm use and manual dexterity), lower limb function (walking speed, walking endurance, number of steps, cadence and functional mobility), neurophysiological changes such cortical activation or electrical activity of the muscles, muscle strength, trunk impairment, muscle thickness, spasticity, balance, range of motion, disability and cognitive function. The secondary outcomes are adverse events and caregiver stress.

Since we extracted sufficient information from the studies, no additional information was sought from the authors of the included studies. However, in case of any missing or unreported data, it was designated as ‘not reported.’

Risks of bias and methodological quality assessments

We used Cochrane Risks of Bias Assessment tool and PEDro scale to assess the risks of bias and methodological quality of the included studies. Both the tool and the scale are known to be valid and reliable [43, 44].

The Cochrane Risk of Bias Assessment tool assesses bias due to the selection of participants, blinding of participants and personnel and outcome assessors, attrition and reporting. The result of the assessment is presented in a risk of bias graph.

The PEDro scale consists of 11 items that assess external and internal validity of a study. The external validity is assessed using the first item; whereas, the internal validity is assessed using the remaining 10 items [44]. In addition, a two-point scale, 0 (no) to 1 (yes) is used to rate the responses to the items that assess the internal validity. In this regard, since the scale has 10 items, the possible scores for methodological quality of a study will range between 0 and 10. When the total score ranges between zero and three; or four and five; or six and ten, the methodological quality is said to be low, moderate or high respectively. [45,46,47] The result of the assessment is presented in a table.

All assessments were carried out independently by two of the researchers (AA & TWLW); however, any disagreements arising from the assessments were managed by consulting the other researcher (SSMN).

Qualitative and quantitative syntheses of the extracted data

In the qualitative synthesis, a summary of the characteristics, risk of bias and methodological quality, and findings of the included studies was carried out. In the quantitative synthesis, a random effect model meta-analysis was carried out.

In the meta-analysis, the data used were the study sample size, the group mean and standard deviation of the scores on the outcomes of interest at post intervention and follow-up. However, when a study provided median scores and interquartile range on the outcomes of interest, the formulae, mean =\(\frac{a+2m+b}{4}\) [where a = the smallest value (minimum), b = the largest value (maximum), and m = median]; and standard deviation = \(\frac{IQR}{4}\)[where IQR = interquartile range] were used to determine the mean and the standard deviation respectively [48]. Furthermore, percentage of variation across the studies due to heterogeneity (I2) was deemed significant when the value is between 50 and 90% at p < 0.05.

Eligibility for inclusion in the syntheses was determined using a table of characteristics of the included studies to check which studies assessed similar outcomes. The meta-analysis was carried out using RevMan version 5.4.1; [49] and all the results of the meta-analyses were visually displayed using forest plots. In addition, sensitivity analyses of the findings of the included studies were carried out based on the period of the outcomes’ assessments (post intervention and follow-up). In addition, an adapted body of evidence matrix of the Australian National Health and Medical Research Council's (NHMRC) was used to interpret the findings of the study [50].

Result

The qualitative synthesis

Selection of the studies

The search provided a total of 11,380 studies. Following screening of the studies, only 14 studies were eligible for inclusion [51,52,53,54,55,56,57,58,59,60,61,62,63,64]. However, two other studies seemed to be eligible for inclusion, but they were excluded following careful scrutiny. [65, 66]

Among the included studies, two of them contain two experimental and two control groups each [54, 55]; whereas, in one study, there were two experimental groups and one control group [52]. Figure 1 provides the details of the literature search and the process of selection of the eligible studies.

Fig. 1
figure 1

The study flowchart showing the process of selection of the included studies

Full size image

Characteristics of the included studies

In total, the number of participants in the included studies was 558. In addition, although two studies did not provide information on sex [51, 55]; 291 and 139 of the participants in the included studies are men and women respectively. Furthermore, only one study included participants in the acute and subacute stages of stroke [59]; all the other studies included participants who were in the chronic stage.

In addition, information on types of stroke was provided in only seven studies [51, 52, 57, 59,60,61, 63]. In these studies, 212 and 63 participants had ischaemic and haemorraghic stroke respectively. Furthermore, only nine studies provided information on the side affected [53,54,55, 57, 58, 60, 61, 63, 64]. In these studies, there were 204 and 136 participants who had left and right sided hemiplegia respectively. Similarly, only two studies provided information on handedness before the stroke, wherein one and 43 participants were left and right handed respectively [57, 61].

The inclusion criteria used in the studies include mild to moderate impairment in motor ability [51,52,53, 55,56,57,58,59, 61, 62, 64]; ability to walk several meters independently [51, 54, 57]; no joint deformity [51]; tolerance for electrical stimulation [51]; impaired sitting balance [60]; no significant cognitive impairment [54,55,56,57, 59, 60, 62, 64]; and no significant spasticity [54, 57].

The exclusion criteria used in the studies include presence of severe joint deformity [57, 62, 64]; a debilitating medical or any chronic condition [51, 52, 54,55,56,57,58,59,60,61, 64]; use of chemotherapy, use of anti-spasticity medication or a medication that can impair neuromuscular performance [51, 58, 64]; pregnancy or lactation [59, 64]; having a pacemaker or other implants [51, 56, 57, 59, 60, 62, 64, 67]; excessive pain [57, 58, 61, 62, 64]; presence of aphasia or dysphasia [52, 54]; having severe sensory deficit or neglect [60]; skin infection [57, 59]; hearing or visual impairment [52]; left-handedness before the stroke [61]; and contraindication to stimulation [58].

The result showed that, home-based neurostimulation is feasible and improves outcomes such as level of motor impairment, motor function, real-world arm use, manual dexterity, walking speed and endurance, functional mobility, joint range of motion, cortical activity, cognitive function and spasticity. Further details on the study participants, intervention protocols including intensity for the experimental and the control groups, and the outcomes assessed are presented in Table 1.

Table 1 Characteristics of the included studies
Full size table

Risks of bias and methodological quality of the included studies

Risks of bias

Some of the studies have high risk of bias in allocation concealment (selection bias) [51, 53, 61]; blinding of participants and personnel (performance bias) [51,52,53,54,55, 58,59,60,61,62]; blinding of outcome assessment (detection bias) [51]; and incomplete outcome data (attrition bias) [54,55,56,57,58, 62, 63].

Similarly, some of the studies have unclear risks of bias in random sequence generation (selection) [51, 57]; allocation concealment (selection bias [54,55,56]; blinding of participants and personnel (performance bias) [56, 63]]; and blinding of outcome assessment (detection bias) [56, 63]. See Fig. 2 and Supplementary File 1 for the risk of bias graph and summary table of the included studies respectively.

Fig. 2
figure 2

Risks of bias graph of the included studies

Full size image

Methodological quality

The methodological quality of the included studies is either moderate [56, 63, 64]; or high [51,52,53,54,55, 57,58,59,60,61,62]. See Table 2 for the met.

Table 2 Methodological quality of the included studies
Full size table

The quantitative synthesis

Only eight studies were used in the meta-analysis for the post intervention outcomes [51,52,53,54,55,56,57,58]. Out of this number, five studies were used for the meta-analysis of upper limb function [52,53,54, 56,57,58]; and four studies were used for the meta-analysis of lower limb function [51, 54, 55, 57]. However, for the upper limb, only two studies were included for the meta-analysis at follow-up [52, 57].

For one of the studies, the scores for the outcome of interest were given in median and interquartile range [52]. Consequently, the formulae already explained in the method sections were used to convert them to mean and standard deviation respectively [48].

Upper limb function

Post intervention, the result showed that, home-based neurostimulation compared to the control, was only superior at improving muscle strength (SMD = 0.72, 95% CI = 0.08 to 1.32, p = 0.03). In addition, there was no significant heterogeneity between the included studies (I2 = 0%, p = 0.85). See Fig. 3 for the forest plot detailing the result. Furthermore, sensitivity analysis carried out by considering motor activity log (MAL) amount of use (AOU) subscale and MAL quality of movement (QOU) separately, did not reveal any significant difference between groups for the two subscales respectively, (SMD = 0.58, 95% CI = -0.26 to 1.41, p = 0.18) and (SMD = 0.70, 95% CI = -0.14 to 1.55, p = 0.10). See Fig. 4 for the details of the result.

Fig. 3
figure 3

A forest plot showing effects of neuromodulation on upper limb function post intervention

Full size image
Fig. 4
figure 4

A forest plot showing effects of neuromodulation on upper limb function post intervention (sensitivity analyses)

Full size image

At follow-up, only two studies assessed one outcome, motor function [52, 57]. The result showed that, home-based neurostimulation was superior to the control at improving motor function (SMD = 0.9, 95% CI = 0.10 to 1.70, p = 0.03). However, there was significant heterogeneity between the included studies (I2 = 80%, p = 0.007). See Fig. 5 for the forest plot detailing the result.

Fig. 5
figure 5

A forest plot showing effects of neuromodulation on upper limb function at follow-up

Full size image

Lower limb function

Post intervention, the result showed that, home-based neurostimulation compared to the control, was only superior at improving functional mobility (SMD = -0.39, 95% CI = -0.65 to 0.14, p = 0.003), and walking endurance (SMD = 0.33, 95% CI = 0.08 to 0.59, p = 0.01). In addition, there was no significant heterogeneity between the included studies (I2 = 0%, p = 0.49) and (I2 = 0%, p = 0.92), respectively. See Fig. 6 for the forest plot detailing the result.

Fig. 6
figure 6

A forest plot showing effects of neuromodulation on lower limb function post intervention

Full size image

At follow-up, the result showed that, still home-based neurostimulation compared to the control, maintained its superiority at improving functional mobility (SMD = -0.30, 95% CI = -0.56 to -0.05, p = 0.02), and walking endurance (SMD = 0.33, 95% CI = 0.08 to 0.59, p = 0.01). In addition, there was no significant heterogeneity between the included studies (I2 = 0%, p = 0.80) and (I2 = 0%, p = 0.86), respectively. See Fig. 7 for the forest plot detailing the result.

Fig. 7
figure 7

A forest plot showing effects of neuromodulation on lower limb function at follow-up

Full size image

Interpretation of the evidence

It is difficult to be very sure of the evidence since there is variation between studies especially in the use of outcome measures, intensity of rehabilitation used and the types of neurostimulation and devices used. However, the evidence seems excellent, appreciably consistent, with satisfactory clinical impact and excellent generalizability and applicability, and as such, it may be used in clinical practice. See Table 3 for more details.

Table 3 Body of evidence matrix
Full size table

Discussion

The aim of this study is to determine the effects of home-based neurostimulation on outcomes after stroke. The result showed that, home-based neurostimulation is feasible and is superior to the control at improving upper limb muscle strength post intervention, and motor function at follow-up. In addition, it is also superior to the control at improving functional mobility and walking endurance both post intervention and at follow-up. This is not surprising since home-based rehabilitation has been reported to be feasible and effective at improving outcomes such as motor function following the use of various interventions such as the constraint induced movement, mirror therapy and therapeutic exercise [29, 68].

Concerning the findings of this study, improvement in muscle strength (an important aspect of motor function), motor function, walking endurance and functional mobility is important for independence in carrying out ADL [69, 70]. For instance, the upper limb is used for eating, washing and grooming oneself. In addition, independence in carrying out ADL is important for overall well-being and good quality of life [71, 72]. Furthermore, it is important for return to work, and by extension economic opportunities and sustainable development [73].

Similarly, impairment in lower limb function may result in sedentary lifestyle and its attendant muscle weakness [74, 75]. Sedentary life is a risk factor for various non-communicable diseases such diabetes, heart disease and depression [76, 77]. Moreover, impaired limb motor function is a significant risk factor for not returning to work after stroke [78]. Thus, finding an intervention such as home-based neurostimulation that will help improve the above outcomes and eventually the patients’ quality of life is important. In particular, home-based neurostimulation, being a home-based intervention may be more cost-effective and acceptable to patients.

However, home-based neurostimulation also has its own limitations like any other home-based rehabilitation. These include problems with the ability of patients and/ or their caregivers to operate the devices and frustration with the use of the devices [79]. In addition, it may be difficult to administer some neurostimulation techniques such as TMS without medical supervision. Furthermore, the cost and size of devices can limit the home-based procedure. However, to help solve some of those problems, we suggest maintaining a regular communication between patients, their caregivers and the clinicians. This can be achieved by using tele-supervision such as via video conference, where the clinicians can observe what the patients are doing [80, 81]. Similarly, community-based rehabilitation can also be used where the clinicians supervise the sessions in person at the patient’s home [82]. In addition, a hybrid model of rehabilitation can be adopted, where in-clinic and home-based sessions are combined to help supplement each other.

Furthermore, the types of neurostimulation and the devices used differ between studies. In particular, five out of the eight studies included in the meta-analysis used neuromuscular electrical stimulation [53, 58, 59, 61, 64]; four studies used TENS which is a weaker form of neuromuscular electrical stimulation [52, 54, 55, 60]; three studies used FES [51, 56, 62]; and two used tDCS [57, 63]. These techniques of neurostimulation have different mechanisms through which they modulate the nervous system. The neuromuscular electrical stimulation is used to stimulate the peripheral nervous which will indirectly help to modulate the central nervous system (CNS) [16, 83]. The tDCS works to directly modulate the CNS [84, 85]. Thus, the findings of this study may only be limited to the effects of neuromuscular electrical stimulation. However, the findings are still very significant since neuromuscular electrical stimulation is easier to administer compared to other forms of neurostimulation such as the tDCS and TMS.

Similarly, in most of the studies, neurostimulation was combined with other rehabilitation techniques such as functional exercises. Thus, it is difficult to confidently say the effects were exclusively due to the neurostimulation. However, the findings are still a significant milestone since providing rehabilitation at home has so may merits such as the opportunity to increase the intensity of rehabilitation [37]. Therefore, further well controlled studies should be carried out to determine the effects of different forms of home-based neurostimulation on outcomes after stroke. In addition, the process of our review is limited in terms of the language in which the included studies were published. Therefore, the findings of the review should be interpreted bearing all the above discussed limitations in mind.

Conclusion

Home-based neuromuscular electrical stimulation, TENS, FES, and tDCS are feasible and effective at improving many outcomes after stroke. These findings represent a significant milestone since providing rehabilitation at home has so many merits such as the opportunity to increase the intensity of rehabilitation. However, further well controlled studies should be carried out to determine the effects of home-based neurostimulation on outcomes after stroke.