To be included in this systematic review, studies had to be randomized controlled trials examining the effects of NMES combined with other therapies on gross motor function of children with CP. We included all trials that applied NMES on muscles of trunk or lower limbs, regardless of the ES dosage. The comparison group must not have received ES.

Clinical trials that did not provide information regarding the magnitude of the intervention effect, either in the experimental or in the control groups, were excluded from the meta-analysis.

Literature search was conducted in the following electronic databases (from the inception to December 2018): MEDLINE (accessed by PubMed), EMBASE, Cochrane Central Register of Controlled Trials (Cochrane CENTRAL), Physiotherapy Evidence Database (PEDro), and Scopus. The search terms used individually or combined included ‘cerebral palsy’, ‘electrical stimulation’ and a string of words previously proposed, which yielded a high sensibility in the search for randomized controlled trials.17 To enhance the sensitivity of the search, words related to the outcomes of interest were not included. There were no language restrictions in the strategy, but non-English studies were included only when the translation was possible. References used in the published articles were identified and used as an additional source to identify other clinical trials. The complete search strategy is shown in Appendix 1.

Two reviewers independently screened the titles and abstracts of the identified studies. A standard screening checklist based on the eligibility criteria was used for each study. Studies that did not meet the criteria were excluded. A third reviewer solved all disagreements related to trial eligibility and assisted in the decision process of including or excluding studies. For studies without sufficient information to evaluate the eligibility, authors were contacted via email to obtain clarifications. Studies with insufficient information, even after contact with authors, were excluded.

We extracted the following data from the studies: methodological design, number of subjects, comparison groups, intervention protocol, frequency, intensity and duration of stimulation, as well as results of the outcomes. The primary outcome extracted was the gross motor function, assessed by means and functional dimensions of GMFM-88 or -66.

The same two reviewers also assessed the risk of bias of studies using the PEDro scale. We collected the PEDro scores directly from PEDro database when they were available. The PEDro scale assesses the methodological quality of a study based on important criteria, such as concealed allocation, intention-to-treat analysis, and the adequacy of blinding. The scale consists of 11 items. One item of the PEDro scale (regarding eligibility criteria) is related to external validity and generally is not used to calculate the method score. Then, the score ranges from 0 to 10.18 Studies were classified as follows: scores 9 or 10, excellent; scores between 6 and 8, good; scores 4 or 5, fair; and scores <4, poor. Scores between 0 and 5 correspond to a high risk of bias; scores ranging from 6 to 10 correspond to a low risk of bias.19

The overall quality of evidence was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE). This approach comprises five items: (1) presence of within-study limitations (risk of bias); (2) inconsistency of results (unexplained heterogeneity of results); (3) indirectness of evidence (target population, intervention, and outcomes of interest); (4) imprecision of the effect estimates (regarding the confidence interval, i.e. confidence in the estimates of effect); (5) Publication bias (systematic underestimation or an over-estimation of the underlying beneficial or harmful effect due to the selective publication of studies). Each non-satisfied item downgraded the overall quality of evidence for each outcome. The quality of the evidence was classified into four categories: high, moderate, low, and very low.19 We created the Evidence Profile and Summary of Findings Table for each population using the GRADE's electronic tool GRADEpro GDT (available at www.gradepro.org).

For quantitative synthesis, pooled-effect estimates were obtained comparing the change between baseline and follow-ups of studies in each group. For those studies with more than one time-point, we used the time point immediately after the intervention. If the change score was not available, we calculated the delta score (change) based on data from post and baseline evaluations. Results were presented as the weighted mean difference (MD with 95% confidence intervals (95% CIs). Statistical heterogeneity was assessed using I2 statistic. I2 values below 25% were considered indicative of low heterogeneity, between 25% and 50% moderate heterogeneity and I2 values above 50% were considered as high heterogeneity. Calculations were performed using the random effects method.

Effect sizes were based on thresholds defined by Cohen, which categorize effect size as small (Standard mean difference (SMD) between 0.2 and 0.5), medium (SMD ranges from 0.5 to 0.8), and large (SMD greater than 0.8).20 A p-value ≤0.05 was considered statistically significant. All analyses were conducted using the R statistical software (version 1.0.153, package metaphor version 2.0-0).20

A total of 2234 articles were identified by the search strategy. After screening titles and abstracts, 16 studies were considered potentially relevant. The majority of the screened studies were not randomized controlled trials and were excluded. Six randomized trials met the eligibility criteria and were included in the meta-analysis,14,21–25 providing data from 174 children with CP, of which 90 were included in the intervention group and 84 in the control group. Fig. 1 shows the flow diagram of the studies.

Figure 1.

The flow diagram of studies included in the review.

(0.24MB).

The mean age (years) of participants ranged from 1.0425 to 8.6.24 Studies included only children with spastic CP.14,21–25 Regarding the topographic distribution, papers included children with unilateral asymmetric spasticity (hemiplegic), bilateral symmetric spasticity (diplegic) and bilateral symmetric spasticity (quadriplegic).26

Characteristics of studies and their protocols are summarized in Table 1. All studies used NMES combined with other therapies (conventional physical therapy14,24; neurodevelopmental therapy (NDT)25; conventional physical therapy and NDT22,23; treadmill21).

Considering the stimulation parameters, the frequencies ranged from 10Hz24 to 50Hz14 and the pulse width from 75μs24 to 300μs.14 The intensity of stimulation ranged from 20mA to 30mA.22,23,25 Two studies reported that intensity was set as the highest tolerable intensity to produce muscle contraction.14,21 One trial did not report the pulse width21 and another one did not present the intensity of stimulation.24 Three studies applied ES on trunk muscles (abdominal and paravertebral)22,23,25 and three on lower limbs muscles (triceps surae,21 quadriceps,14 and gluteus14,24).

The time of intervention ranged from 1521 to 60min24 of stimulation. Weekly frequency ranged from three21 to six times per week,25 during four,21–23 six14,25 or eight24 weeks.

The median PEDro score of studies was 6 (ranged from 4 to 8) (Table 2). All studies were randomized, had similar groups at baseline, showed between-group differences, reported and pointed out estimate and variability. The majority of studies conducted adequate follow-up (83%) and assessors were blinding (67%). However, many studies did not conceal the allocation list (83%) or carry out intention-to-treat analysis (67%). None of the studies blinded neither participants nor therapists.

The overall quality of evidence was downgraded in two GRADE domains: (a) inconsistency of results due to high heterogeneity (87%) evidenced in the meta-analysis; (b) imprecision of the effect estimates because confidence intervals were large (ranged from 1.57 to 7.79). For these reasons, the quality of evidence was downgraded to low-quality evidence for the effectiveness of NMES on gross motor function. The evidence profile is presented in (Table 3).

NMES combined with other therapies presented medium effect size (SMD=0.76) to improve gross motor function in children with CP in comparison with conventional physical therapy or NDT (MD=4.68 [95% CI, 1.57–7.79; I2 73%, p=0.003]) (Fig. 2).

Figure 2.

Effect of NMES on gross motor function (GMFM dimensions) in children with CP compared with conventional physical therapy or neurodevelopmental therapy. NMES=neuromuscular electrical stimulation; CP=cerebral palsy.

(0.25MB).

We also performed a sensitivity analysis grouping studies according to the GMFM dimensions: sitting, standing or walking. Other dimensions were not studied because they were not included in the articles. NMES showed large effect size (SMD=1.26) for improving GMFM-sitting dimension (MD=.36 [95% CI, 5.21–9.51; I2 0%, p<0.0001]) and medium effect size for GMFM-standing dimension (standing: MD=1.68 [95% CI, 0.26–3.09; I2 0%, p=0.02]) when compared to conventional physical therapy or NDT (Fig. 2). On the other hand, NMES did not differ from conventional physical therapy or NDT to improve GMFM-walking dimension (walking: MD=1.31 [95% CI, −0.63 to 3.26; I2 0%, p=0.19]) (effect size SMD=0.29) (Fig. 2).

When analyzing our results, it is important to consider some limitations. For instance, stimulus parameters applied by the included studies were very heterogeneous (i.e. intensity, duration, frequency and number of sessions). The frequency of stimulation ranged between low- and medium-frequencies; time of intervention varied from 15min21 to 60min24; and the duration of treatment ranged between 421–23 and 16 weeks. Chiu et al., also reported an important heterogeneity related to the time of intervention which ranged from 180 to 2880min across the studies included in their research.10

Our review was restricted to the highest standard of evidence, once only randomized controlled trials were included. However, evidence is limited considering the small number of studies included, the wide variety of protocols applied and the lack of precise information about the scales used for gross motor evaluation (if GMFM-88 or GMFM-66). Further controlled studies are still needed to help clinicians make decisions about the effectiveness of ES on GMFM improvement in children with spastic CP.