A systematic search of literature and electronic databases such as PubMed, Cochrane Central Register of Controlled Trials, and Embase was conducted from their inception until June 15, 2024. The initial literature search was conducted on July 7, 2022. The strategy to develop search terms involved a blend of subject terms and freely used words. This includes terms like “electric stimulation therapy,” “intensive care units,” “critical illness,” and “ICU.” A detailed combination of free words and subject terms was utilized for retrieving literature, and the specifics of this search strategy can be found in Appendix 1.

The inclusion criteria were: (1) Study type: RCT, not limited to allocation concealment and blinding method; (2) Study population: ICU mechanically ventilated patients aged ≥18 years; (3) Interventions: Research electrical stimulation or combined conventional therapy in the observation group; (4) Comparisons: Usual care measures or comfort treatment in the control group; (5) Outcome: The primary outcome was the Medical Research Council (MRC) scale score, while the secondary outcomes were the duration of MV, mortality, ICU LOS, and hospital LOS; and (6) Language: Only articles published in English.

The exclusion criteria were: (1) conference abstracts; (2) case studies or Meta-analyses; and (3) studies where data were missing or could not be converted.

A pair of reviewers separately perused through the scholarly records using EndNote 20.0, concurrently validated the compiled data, and sought judgment from a third-party researcher during disparities. The process of screening the literature implied a thorough examination of the title, abstract, and the complete text. In situations where critical information, necessary for the study was missing, the original authors of the papers were reached out to, either via email or call. Excel was employed for data organization included several components such as the authors’ names, year of publication, Acute Physiology and Chronic Health Evaluation II (APACHE II), sample size, age demographics, gender, intervention strategies, and final outcomes.

Two reviewers conducted independent assessments of the risk of bias in the studies included in the review using the risk of bias assessment tool for RCTs as outlined in Cochrane Workbook 6.4, 2023 (14). In cases where there was a difference in their assessments, a third investigator was involved in discussions or arbitration to reach a consensus. The evaluation encompassed various aspects, including random sequence generation, allocation concealment, blinding of participants and investigators, blinding of outcome assessors, completeness of outcome data, selective reporting and other potential sources of bias. Each of these aspects was rated as “low risk of bias,” “unclear,” or “high risk of bias,” with a determination of “high risk of bias” made for each specific item where applicable.

We derived risk ratios (RRs) for categorical data and evaluated mean differences (MDs) for ongoing data, pairing them with their respective 95% confidence intervals (CIs) under a random-effect model. We gauged heterogeneity across studies using metric likeτ2, χ² (Cochrane Q), and I² statistics. According to the Cochrane handbook, the I² will be considered non-important (<30%), moderate (30–60%), and substantial (>60%) (14). Our findings were represented visually through forest plots. To evaluate the effect of individual studies on the overall results, we employed a sequential study removal method, iteratively excluding each study and recalculating the pooled effect size. For gauging publication bias, we used a funnel plot for Meta-analysis and employed Egger’s method for quantification, provided more than ten studies incorporated the results. We also executed subgroup evaluations focusing on the various electrical stimulation forms. Our statistical reviews were performed using the RevMan 5.4 software, establishing a statistical significance benchmark at a value below 0.05.

We identified a total of 7,768 studies related to our research. After checking for duplicates, screening titles, and browsing abstracts, we eliminated 7,745 studies that did not meet our inclusion criteria. The search process and study selection are illustrated in Figure 1. In the end, 23 randomized controlled trials (RCTs) (3, 12, 15–35) were included, covering a total of 1798 patients, of which 48.5% were females, conducted in 10 different countries and published between the years 2009 and 2023. The individual studies included a sample size range between 20 to 312 critical patients, and nearly 30.4% of the studies involved patients whose mean or median age was over 60 years.

The interventions in these trials were neuromuscular electrical stimulation (NMES) in 16 studies, electrical muscle stimulation (EMS) in five, and functional electrical stimulation (FES) in two. Notably, Abu-Khaber et al. (15) compared the effects of EMS with a placebo, Kho et al. (16) and Fischer et al. (17) compared it with a sham intervention, and Campos et al. (18) compared it with early mobilization. Table 1 provides additional details of the research and the clinical characteristics of the patients. This study employed a leave-one-out approach, removing the included studies one by one to observe changes in the results after the exclusion of specific studies, in order to examine the impact of any single study on the overall effect estimate. The results are shown in Table 2.

We appraised the included studies against seven domains for the risk of bias, which we categorized as ‘low’, ‘high’, or ‘unclear’. The results from these individual studies are summarized in Figure 2. We found that the method of randomized controlled allocation was flawed in three studies and was not expressly delineated in two studies. In six studies, the participants and personnel were not blinded, and in two studies, the outcome measures were not blinded.

Sixteen studies reported the effect of electrical stimulation on ICU LOS. The results showed that there was no statistical difference in the electrical stimulation group compared with the control group (MD = −2.41, 95% CI −6.23 to 1.42, p = 0.22, I² = 99%) (Figure 5). The subgroup analysis of EMS showed no differences in ICU LOS, but the sensitivity analysis reported the EMS could decrease the ICU LOS after excluding the study by Nakamura et al. (27) (MD = −11.0, 95% CI −21.12 to −0.88, p = 0.03). Nakamura et al. (27) primarily focused on patients with low APACHE II scores, and the ICU LOS was shorter than in other ICUs, which may be the main source of heterogeneity. The sensitivity analysis showed that the electrical stimulation could decrease the ICU LOS compared with the control group after excluding the study by Mahran et al. (25) (MD = −1.02, 95% CI −1.96 to −0.08, p = 0.03) (Table 2). Mahran et al. (25) focused on adult patients who were admitted to the ICU and required MV on the first day of admission, and aimed to evaluate the short-term outcomes of electrical stimulation in critically ill patients.

A total of nine studies, encompassing 451 patients, explored the impact of electrical stimulation. Statistical difference was not observed in hospital LOS when compared to the control group (MD = −1.45, 95% CI −4.12 to 1.23, p = 0.29, I² = 44%) (Figure 6). The subgroup analysis of EMS showed no differences in hospital LOS, but the sensitivity analysis was conducted and identified the study by Sumin et al. (33) as the primary source of heterogeneity. Excluding this study, the results indicated the electrical stimulation (MD = −2.38, 95% CI −3.81 to −0.94, p = 0.001) and the subgroup of NMES (MD = −2.36, 95% CI −3.85 to −0.88, p = 0.002) led to a reduction in hospital LOS in comparison to the control group (Table 2). The research by Sumin et al. (33), which assessed the efficacy of NMES during the initial rehabilitation phase for patients experiencing postoperative complications following cardiovascular surgery, did not blind the assessors of muscle strength, which is presumed to be a significant source of heterogeneity.

Eleven studies, including 989 patients, reported the effect of electrical stimulation on MRC scores. The results showed that the MRC scores were significantly improved in the electrical stimulation group compared with the control group (MD = 3.62, 95% CI 0.94 to 6.30, p = 0.008, I² = 87%) (Figure 7). The subgroup analysis of NEMS showed differences on MRC scores (MD = 5.12, 95% CI 1.36 to 8.87, p = 0.008, I² = 89%) (Figure 7).

Studies reporting on hospital LOS were less than ten, so we did not visualize the results using funnel plots. The funnel plots did not reveal obvious asymmetry for analyses on morality, duration of MV, MRC scores and the ICU LOS. Consistently, the Egger test suggested a lack of publication bias for morality (p = 0.25), duration of MV (p = 0.17), and ICU LOS (p = 0.15). Based on this analysis, the studies encompassed provided a broad coverage and produced statistically robust outcomes.

The primary strength of this study is its inclusion of numerous RCTs, along with the use of subgroup and sensitivity analyses to minimize biases, thereby ensuring the reliability of the data and the stability of the results. By strictly adhering to the guidelines of the Cochrane Handbook, the research achieves a high degree of standardization and scientific rigor. The findings indicate that electrical stimulation can significantly improve the MRC scores in ICU patients, offering a practical therapeutic foundation for clinical application and potentially benefiting the rehabilitation process of critically ill patients. However, it’s important to consider potential limitations and the overall quality of the included studies to fully assess the reliability of these findings. Nevertheless, there were some limitations to this study. The included literature is only published in English, which may lead to publication bias due to an incomplete literature search. In addition, the sample size of individual studies is small, which may affect the analysis results. Furthermore, some studies were not blinded in the implementation of electrical stimulation and measurement results, which may lead to some implementation and measurement bias. Finally, the wide age range of the adult ICU patients on mechanical ventilation included in this study introduces significant heterogeneity, which is a significant limitation of our work. Future studies with larger samples and longer follow-ups are needed to comprehensively explore the short and long-term effects of electrical stimulation on patient outcomes to understand the intervention’s effects further.