Differential contractile response of critically ill patients to neuromuscular electrical stimulation

Critical illness and intensive care treatment can lead to severe impairments of physical function, cognitive capabilities and mental health, which directly affect the individuals’ daily living, e.g. return to work [1‐8]. The physical impairments, in terms of muscular weakness and reduced walking distance, directly affect short-term outcome, as patients suffering from weakness are less likely to be weaned from mechanical ventilation, discharged from the hospital or survive [4]. These impairments can persist at least up to 5 years after discharge from the intensive care unit (ICU) [9]. Therapeutic options that have been the focus of investigations throughout the last decade are early mobilisation and neuromuscular electrical stimulation (NMES). Both of these options have yielded positive as well as negative results in recent randomized controlled trials [10‐14]. NMES is able to elicit muscle contractions in unconscious or sedated and therefore uncooperative patients. As it is possible to initiate this form of treatment in the early phase of critical illness, even if the patient has to remain heavily sedated, it was reasoned to be a promising therapeutic option to target muscular pathophysiological processes in their beginning stages [15, 16]. While some trials were able to corroborate this hypothesis, others yielded negative results [13, 14, 17]. Fossat and colleagues found no improvement in muscle strength or mobility in 314 patients receiving NMES of the M. quadriceps femoris in conjunction with in-bed-cycling [13]. An underrecognised aspect throughout recent NMES trials is the fact that not every electrical impulse necessarily elicits an appropriate muscle contraction, as Segers et al. showed [18]. Sepsis, use of vasopressors and lower limb edema were independent predictors for a missing contractile response to NMES [18]. It remains unclear whether neuromuscular electrical stimulation per se has a beneficial effect on the patient or if the beneficial effect is a causal effect of the evoked muscle contraction, which would be in line with current knowledge about muscle physiology. In this case, the lack of contraction from the electrical stimulation would leave study results prone to bias. Furthermore, it remains unclear whether the electrical stimulus during NMES has harmful effects when not properly translated into a muscle contraction. Our aim was, therefore, to characterise the response of critically ill patients to NMES, determine predictors for an adequate or inadequate therapeutic response and investigate potential benefits or detriments of an adequate respectively inadequate muscular response to NMES.

During the 2-year inclusion period, 468 out of 3147 patients that were admitted to two ICUs within the Charité – Universitätsmedizin Berlin fulfilled the inclusion criteria. Fifty of these patients were successfully enrolled in the trial and randomised. Further information regarding the enrolment process has been published before [19]. For this sub-analysis, only patients receiving NMES (n = 21) were considered (Table 1).

In the 21 patients treated with NMES, a total of 1824 muscle groups were stimulated during the first 7 days of treatment. These stimulations were equally distributed between upper (n = 886) and lower (n = 938) extremities (p = 0.522). On day 1 of NMES, only 64.4% of stimulations led to a contractile response with a significant difference between upper (100%) and lower (41.7%) extremities (p = 0.001). This difference could consistently be observed until day 7 (Fig. 1a + b). Nevertheless, a significant correlation for contractile response between upper and lower extremities was observed (k = 0.687, r² = 0.472, p = 0.001). Contractile response declined throughout the 7-day observation period from 64.4 to 25.0% overall. We similarly observed a decrease in contractile response for upper (100.0 to 58.3%) and lower (41.7 to 12.5%) extremities (Fig. 1a + b) (Additional file 1: Table S2b).

Furthermore, the results showed that the proximal muscle groups overall are less likely to respond to NMES than the distal muscle groups (median [IQR] proximal 40.7 [22.5/60.4] vs. distal 75.0 [38.0/91.7] %, p < 0.001). This finding is evident in both upper (median [IQR] proximal 73.2 [42.1/88.7] vs. distal 79.1 [72.5/94.6] %, p = 0.047) as well as the lower extremities (median [IQR] proximal 8.3 [0.0/45.8] vs. distal 52.5 [3.6/87.9] %, p = 0.003). The M. vastus lateralis was the least likely to respond to neuromuscular electrical stimulation (Fig. 1c) (Additional file 1: Table S2b).

The threshold of electrical current that led to a contractile response did not significantly change between day 1 and 7 for all muscle groups taken together (median [IQR] day 1, 50.2 [31.3/58.8] mA; day 7, 45.3 [38.0/57.5] mA) as well as for upper (median [IQR] day 1, 47.9 [24.4/60.5] mA; day 7, 44.4 [32.1/57.9] mA) and lower extremities (median [IQR] day 1, 49.3 [41.8/57.3] mA; day 7, 45.9 [37.1/49.8] mA) separately (Fig. 1d + e). Nevertheless, there was a significant difference between the upper and lower extremities observable on days 2, 3, 4, 5 and 6 with a higher electrical current mostly necessary in the lower extremities. The applied electrical current between upper and lower extremities correlated significantly (k = 0.710, r² = 0.505, p = 0.002) (Additional file 1: Table S2b).

At first adequate awakening, all patients presented with significant weakness overall (median [IQR] 3.2 [2.5/3.8] MRC score) as well as for upper (median [IQR] 3.1 [2.4/3.8] MRC score) and lower extremities (median [IQR] 3.0 [2.9/3.8] MRC score) individually (Fig. 1c, f, g). Until discharge, an increase was observable overall (median [IQR] 3.5 [3.0/3.9] MRC score), for upper (median [IQR] 3.8 [2.9/4.0] MRC score) and for lower extremities (median [IQR] 3.8 [3.1/4.0] MRC score), while statistical significance was only reached for the lower extremities (p = 0.042) (Fig. 1f, g). Muscle strength between upper and lower extremities at first awakening as well as at discharge showed a significant correlation (first awakening k = 0.864, r² = 0.746, p < 0.001; ICU discharge k = 0.751, r² = 0.564, p < 0.001).

Using the contractile response cut-off value, 8 patients were classified as responders (> 50% contractile response during the first 7 days) and 13 were classified as non-responders (≤ 50% contractile response during the first 7 days). Univariate analysis revealed a significantly higher SOFA score in non-responders at admission to the ICU. Baseline characteristics were otherwise balanced including norepinephrine treatment (Table 2).

When comparing the contractile response from responders to non-responders, we see a significantly greater proportion of stimulations leading to an adequate contractile response in responders (responders vs. non-responders median [IQR] both, 83.7 [73.4/93.5] vs. 35.0 [20.2/44.2] %, p < 0.001; upper extremities, 91.1 [86.6/99.1] vs. 67.0 [40.9/74.1] %, p < 0.001; lower extremities, 77.7 [62.5/90.2] vs. 7.1 [1.8/23.2] %, p = 0.002), indicating that our cut-off value was sufficient in distinguishing between patients that respond well to those who do not (Fig. 2a–d). Interestingly, no decrease in contractile response over time was observed in responders as opposed to non-responders (Additional file 2: Figure S1). When evaluating contractile response for all muscle groups separately, we also observed a consistent, remarkable and significant difference between responders and non-responders. Furthermore, it became evident that except for the dorsal thigh and M. vastus lateralis, the difference in response between upper and lower extremities is only significant for non-responders (Fig. 2d). This difference in contractile response is also reflected by the electrical current that was necessary to elicit a contraction if possible at all, as it was significantly higher in non-responders overall (responders vs. non-responders median [IQR] 38.0 [32.8/42.9] vs. 54.7 [51.3/56.0] mA, p < 0.001) and for upper extremities (responders vs. non-responders median [IQR] 32.6 [30.6/37.3] vs. 54.7 [45.4/56.0] mA, p < 0.001). Interestingly, no difference in electrical current was observed for lower extremities (responders vs. non-responders median [IQR] 44.6 [31.1/46.1] vs. 54.4 [33.0/58.3] mA, p = 0.140) (Fig. 2e–g) (Additional file 3: Table S1). Muscle strength at first awakening showed no statistically significant differences between responders and non-responders in the upper extremities (responders vs. non-responders median [IQR] 3.8 [3.4/4.2] vs. 3.0 [2.0/3.4] MRC score, p = 0.145), lower extremities (responders vs. non-responders median [IQR] 3.7 [3.3/3.7] vs. 2.9 [2.4/3.9] MRC score, p = 0.630) and overall (responders vs. non-responders median [IQR] 3.8 [3.4/4.0] vs. 3.1 [2.1/3.5] MRC score, p = 0.282) (Figs. 2h–j and 3 a). At ICU discharge, a statistically significant difference in muscle strength can be observed with higher muscle strength in the upper extremities of responders (responders vs. non-responders median [IQR] 4.4 [4.1/4.6] vs. 3.3 [2.8/3.8] MRC score, p = 0.036), while in lower extremities (responders vs. non-responders median [IQR] 4.3 [3.5/4.6] vs. 3.6 [3.1/4.0] MRC score, p = 0.376) and overall (responders vs. non-responders median [IQR] 4.3 [3.8/4.6] vs. 3.5 [2.8/3.9] MRC score, p = 0.145) statistical significance was not reached (Fig. 2h–j and 3b).

The proportion of contractile responses to NMES correlated inversely with the applied electrical current overall (k = − 0.746, r² = 0.556, p < 0.001) as well as for upper extremities (k = − 0.653, r² = 0.427, p = 0.002). No correlation between electrical current and contractile response was found for the lower extremities.

Contractile response furthermore correlates inversely with the SOFA score at admission (k = − 0.544, r² = 0.296, p = 0.011). A SOFA score ≤ 13.5 has a sensitivity of 75% and a specificity of 61.5% to identify responders (AUC 0.788; p = 0.03) (Fig. 4a, b).

This retrospective sub-analysis aimed to outline the characteristics of, as well as predictors for, a contractile response to NMES, and also potential clinical benefits resulting from an adequate response to NMES. We were able to show that at the initiation of NMES, only two thirds of stimulations led to a contractile response in our cohort and that this number declined to one third throughout the first 7 days of stimulation, which we think is due to progressing pathophysiological processes. Established mechanisms that would hypothetically contribute to this effect are channelopathy, decreasing metabolic flexibility, advanced muscle atrophy and edema. Interestingly, the upper extremities proved to respond to NMES more frequently than the lower extremities and the distal extremities more frequently than the proximal extremities. If contractile response was present, the required electrical current for a muscle contraction did not change within the first 7 days, while in the lower extremities higher currents were usually required to elicit a contractile response. Patients classified as responders to NMES were shown to have a lower SOFA score, to require a lower electrical current and to have a significantly improved upper extremity muscle strength at discharge. An electrical current of 50.1 mA was sufficient to adequately stimulate responders to NMES, which can be identified by a SOFA score ≤ 13.5.

NMES has been investigated in the context of critical illness and ICU-acquired Weakness for almost a decade without a clear conclusion regarding its effectiveness mainly due to controversial and incoherent study results [13, 14, 17]. As an example, an improvement in muscle strength using NMES was shown by Karatzanos and colleagues [22]. On the other hand, Kho et al. were not able to reproduce these findings [23]. An aspect that is critical when reporting interventional trials is whether the intervention was delivered as planned. Segers et al. first suggested that not every electrical impulse during NMES is converted into a muscle contraction [18]. Nevertheless, most trials on NMES published to date merely report the number or total duration of NMES sessions without mentioning the relative amount of stimulations successfully translated into a muscular contraction, a potentially useful surrogate indicating successful delivery of NMES. A lack of contractile response in the participants could be one cause for the incoherent and contradictory study results in NMES interventional trials throughout the past years. Contractile response was only reported in one outcome-focused interventional trial by Rodriguez and colleagues, who conducted an intraindividual comparison in 16 patients [17]. The muscular response rate of 77% was relatively high compared to our observations as well as to the numbers presented by Segers et al. [17, 18]. Interestingly, they also report an increase in muscle strength at ICU discharge, which is similar to the increased muscle strength we found in the upper extremities of our responder group at ICU discharge [17]. When looking at markers for muscle mass—extremity circumference or muscle thickness via ultrasound—Rodriguez et al. found no effect due to NMES. Interestingly, the dissociation between muscle mass and function has been repeatedly observed in critically ill patients shortly after ICU admission as well as 6 months after ICU discharge [24, 25]. Nevertheless could muscle strength only be assessed via a nonvolitional method based on electrical stimulation of a muscle contraction in the trial investigating the timeframe shortly after ICU admission. The effect of an impaired contractile response has on this form of muscle strength measurement in critically ill patients is unknown [24]. An impaired bioenergetic status has been shown to be of relevance for the development of muscle weakness during the ICU stay and could be a plausible culprit for the difference between muscle mass and muscle function [26]. Future trials are required to elucidate the exact mechanisms.

Our investigation into multiple muscle groups revealed that the M. vastus lateralis has the lowest response rate to NMES and that there is a significant difference between upper and lower extremities in the response to NMES. Based on these facts, we think another factor introducing significant bias into most investigations of NMES is the fact that almost all of them only investigated the lower extremities or solely the quadriceps femoris, which, according to our findings, would carry a high risk of yielding negative study results due to lack of contractile response [13, 14, 23, 27, 28]. A standardised NMES reporting sheet (Additional file 4: Standardised NMES Reporting Sheet) was developed by us in order to improve reporting of NMES protocols and responder rates in future trials and through that enable an adequate comparison of patient cohorts, protocols and muscle groups.

We have shown that a maximum electrical current of 50 mA is sufficient to stimulate all patients that respond adequately to NMES. This number was doubled and almost tripled in some studies, and while we did not observe any harm from NMES with electrical currents up to 70 mA in our cohort, further investigations are needed to rule out any harm from electrical currents above our defined threshold [29].

ICU-acquired weakness has been shown to develop early during critical illness, while the likelihood of developing it increases with the severity of critical illness [15, 16, 30]. In agreement with our results, Segers and colleagues have found that patients with higher disease severity, in their case differentiated by sepsis or no sepsis, are less likely to respond to NMES [18]. Furthermore, we were able to show that distal muscle groups respond better to NMES than proximal muscle groups, which is in accordance with DeJonghe et al., who showed that proximal muscle groups are more severely affected by ICU-acquired weakness than distal muscle groups [31]. Due to the overlap between characteristics of, and risk factors for, ICU-acquired weakness and insufficient response to NMES, it appears likely and plausible that pathophysiological mechanisms involved in the development of ICU-acquired weakness are also responsible for hampering contractile response to NMES [30]. Furthermore, could differences in impedance due to edema or fat mass, a different electrode to muscle ratio or differing adaptive reactions to therapy have an important impact on contractile response [18]. Segers et al. were not able to find an association between electrophysiological characteristics of critical illness myopathy and the lack of contractile response, within our data such an association could not be shown as well but as our analysis is of exploratory nature a large sample size in future trials might yield different results [18]. As such, future investigations specifically dedicated to this question are necessary, in order to further elucidate the connection between electrophysiologic abnormalities and lack of contractile response.

ICU-acquired weakness is a clinical syndrome defined by an average MRC score < 4 [32]. Multiple pathophysiological mechanisms, e.g. muscle wasting, bioenergetic failure and membrane inexcitability, have been described in the context of ICU-acquired weakness [15, 16, 21, 33]. We therefore hypothesise that depending on the aetiology, different pathophysiological mechanisms could be predominantly responsible for the development of ICU-acquired weakness and also be the reason patients respond differently to interventions such as NMES. Future studies should further evaluate these underlying mechanisms in the development of ICU-acquired weakness in order to more closely define different types of ICU-acquired weakness and possibly identify sub-cohorts that respond to and consequently benefit from the different interventions.

Our pilot analysis is limited by the number of patients as multivariate analysis to investigate independency of association between, for example, SOFA score and responder status could not be performed. Furthermore, as our ICU admissions were not planned, we were not able to obtain pre-admission muscle strength values. Our muscle strength values at first awakening will therefore already be affected by the intervention, and correction for potential baseline differences is not possible. We did not included patient with pre-existing neuromuscular disease in order to keep bias due to missing pre-admission values to a minimum. Since our protocol did not include an electrical current above 70 mA, we are not able to reach any conclusion—neither beneficial nor harmful effects—regarding NMES with higher electrical currents. We see that 50.1 mA is a good cut-off to differentiate responders and non-responders on this basis we do not believe increasing electrical current further will increase responder rates.

Different muscle groups as well as different patients show a varying response to NMES, which appears to be dependent on multiple factors that are linked to the respective pathomechanisms behind the development of the ICU-acquired weakness. Since we observed a varying degree of clinical improvement when comparing patients that responded adequately to NMES to those who did not, it is strongly suggested that response rates to NMES should be reported in future trials (Additional file 4: Standardised NMES Reporting Sheet) and evaluated in relation to clinical outcomes. The described difference in response rates to NMES may explain the controversial findings of previous trials and improve the quality of evidence in future critical care NMES trials.