Neuromuscular electrical stimulation added to usual care, in comparison with usual care alone or sham stimulation, was associated with better muscle-strength outcomes in patients in the ICU, with moderate to strong evidence. However, the level of evidence was weaker and conflicting for outcomes related to muscle mass, with small to moderate effect sizes or no effect. These findings suggest that NMES may have the potential to prevent skeletal-muscle weakness in critically ill patients, which could confer many important physical, psychosocial, and economic benefits for these patients after discharge from ICU. However, it remains to be ascertained whether NMES therapy can also prevent the muscle wasting associated with critical illness.
The high inconsistency in ICU patient characteristics between studies was not unexpected (as attested by non-normal data distribution and lack of means and SDs), but it affected the methodological quality of the included studies, which prevented us from completing a meta-analysis. Therefore, the main results of this systematic review could only be interpreted with a thorough qualitative analysis. Although it is extremely challenging to perform large and well-controlled RCTs in this patient population, future NMES studies should consider stratifying patients for main diagnosis and eventually also for disease severity, as this latter feature has been identified as an independent risk factor for ICUAW incidence [10, 51]. It is conceivable that the benefits of NMES are greater for patients admitted to the ICU with respiratory complications (as suggested by the large effect sizes for patients with COPD) , or neurological complications, compared with patients with sepsis or trauma. For example, inflammation-mediated electrolyte changes and also edema may seriously affect conductivity and thus electrical current diffusion , which could lessen any systemic effect of NMES in these patient samples.
The questionable validity and heterogeneity of the NMES protocol characteristics adopted in the eight studies included in this systematic review further complicated the interpretation of the present results. The strength of the contraction induced by NMES (that is, evoked tension), which is the main determinant of NMES effectiveness , was not reported in any of the included studies. Quantifying this parameter, rather than stating simple current intensity/voltage, is crucial as it would also permit discrimination of responders from non-responders [54, 55], and eventually allows ascertainment of the optimal NMES characteristics for patients in the ICU on an individual basis. In addition, evoked tension should be maximized, whenever possible, by selecting appropriate current parameters (stimulation frequency of 50 to 100 Hz  and highest tolerable stimulation intensity, while minimizing fatigue with long relaxation phases), joint position (long muscle length), and methodological precautions such as the accurate determination of muscle motor points .
Assessing voluntary muscle strength in the ICU is extremely difficult. Despite potential limitations of manual muscle testing such as poor validity and inaccuracy of subjective ratings [58, 59], especially when assessors are not blinded, evaluation of voluntary strength using the MRC score was used in the majority of the included RCTs, and only one study used dynamometry . Considering the limited or absent cooperation of patients at admission into the ICU, and the considerable influence of central factors (including motivation) on maximal voluntary efforts , it would be preferable if evaluation of muscle function in these patients relied on artificially-evoked muscle responses. Therefore, alternative methods that are independent of patient cooperation such as peripheral magnetic stimulation (which can also be used to evaluate respiratory muscle function) , electrical impedance myography , myotonometry , and mechanomyography , would improve the validity of muscle testing in ICU.
The major risk factors for ICUAW are immobilization, multiple organ failure, systemic inflammatory response syndrome, gram-negative septicemia , hyperglycemia [3, 65], and medications such as aminoglycosides, colistin, and corticosteroids. All these elements should be viewed as important confounding factors that might distort accurate interpretations of our findings. For example, patients with a recent exacerbation of COPD (most of whom are prescribed corticosteroids, which can induce myopathy) were excluded in one instance , whereas another study examined the effects of NMES after COPD exacerbation (some patients received corticosteroids) . Studies must always carefully control for immobilization days, disease severity scores (organ specific and physiological), and medication use.
Even though physical-therapy practices vary widely between different ICUs, there is growing interest in early rehabilitation strategies that have the potential to prevent skeletal-muscle weakness and wasting in critically ill patients . These interventions range from passive stretching  and early mobilization therapy  to bedside cycling ergometry . Interestingly, NMES added to usual care has recently been shown to be effective in reducing ICUAW incidence . The present systematic review confirms these preliminary findings, highlighting the potential role of NMES as a preventive countermeasure against ICUAW. Compared with other rehabilitation strategies, the unique aspects of NMES are that it is relatively cost-effective (one multiple-user NMES unit costs less than €400), does not require patient cooperation (it can be applied to sedated patients) or stable cardiac or respiratory function, can be implemented during the first few days after ICU admission, and provokes considerable central effects, both acute and chronic , which could also contribute to preventing the occurrence of muscle weakness in critically ill patients. Moreover, in addition to muscle-related outcomes, NMES has been shown to be more effective than conventional care or sham stimulation for improving pulmonary function [36, 43, 47], including accelerated weaning from mechanical ventilation , physical function (6-minute walking distance  and bed to chair transfer ), and for reducing the incidence of critical illness polyneuromyopathy . However, the effects of NMES on the pathophysiological mechanisms of ICUAW are poorly known, and NMES cannot be easily used with all critically ill patients (for example, those with skin lesions, traumatic fractures, complete lower motor-neuron lesions and cardiac pacemakers), so that there is still no consensus among intensive care specialists about its real value.
The major limitation of the present review concerns the unavailability of outcome data (for example, baseline strength measurements) to allow a full meta-analysis to be conducted. At face value, this lack of data could be indicative of reporting bias at outcome level. However, rather than reporting bias, lack of outcome data should simply be seen as one of the many limitations inherent in studies conducted on patients admitted to the ICU. We factorized potential biases at study level by assessing the methodological quality of the studies, which allowed us to assess the reliability and validity of the data and to weigh the results of each study based on its methodological rigor. Unfortunately, because of the impossibility of calculating the effect sizes in many of the studies included in the review, the risk of publication bias could not be assessed.