The main finding of the current study is that surface EMG of extradiaphragmatic inspiratory muscles does not reliably reflect activity of the diaphragm under different levels of inspiratory support. There is a moderate to low correlation and low agreement between changes in diaphragm and extradiaphragmatic inspiratory muscle activity in response to unloading of the respiratory muscles. Furthermore, there are notable differences in timing of activation between the diaphragm and extradiaphragmatic inspiratory muscles.
As expected, activity of the extradiaphragmatic inspiratory muscles increased in response to reducing level of assist on a group level. Our results are largely in accordance with previous studies in intubated patients. Schmidt et al. reported that parasternal intercostal, scalene and alae nasi activity increases when a low inspiratory PS level is applied as compared to a high PS level [
26]. Cecchini et al. [
28] showed that both NAVA and PS ventilation reduced alae nasi and scalene activity in proportion to the level of assistance. In addition, we found that this also holds for the genioglossus, despite that the endotracheal tube bypasses the upper airways. Remarkably, in most patients extradiaphragmatic inspiratory muscles remain active up to a PS level of 15 cmH
2O. Brochard et al. [
39] also demonstrated that the sternocleodomastoid muscle remains active at high inspiratory PS levels. These findings indicate high respiratory drive even at high levels of pressure support. High respiratory drive despite high levels of inspiratory assist may be explained by persistent abnormal arterial blood gas, feedback from afferents from the lung and chest wall or systemic inflammation (for review see [
40]).
Relationship between diaphragm and extradiaphragmatic inspiratory muscle activity
In the current study, we showed with repeated measures observation analysis that there are only moderate correlations between the changes in diaphragm and extradiaphragmatic inspiratory muscle activity (Fig.
2). Moreover, we demonstrate that there are large limits of agreement for all PS levels when comparing the changes in diaphragm and extradiaphragmatic inspiratory muscle activity (Additional file
1: Figure S4). For example, for the scalene and diaphragm, the 95% limits of agreement are between – 50 and + 50% PS 3 (normalized to muscle activity at PS level 3 cmH
2O) for changes in surface EMG and EAdi
peak (Additional file
1: Figure S4). In clinical practice, such a measurement error is unacceptable, because this means, for example at an average EAdi
peak of 50% PS 3 (normalized to muscle activity at PS level 3 cmH
2O) at PS 15, that there could be either no scalene activity or scalene EMG
peak could be doubled.
The relationship between diaphragm and extradiaphragmatic inspiratory muscle activity has been addressed previously. These studies reported that the recruitment pattern of extradiaphragmatic inspiratory muscles is comparable to the diaphragm in response to lower inspiratory support levels during noninvasive ventilation in healthy subjects [
18,
37,
41], patients with chronic obstructive pulmonary disease (COPD) [
18] and ventilated ICU patients [
26,
28]. In contrast to our study, no correlation or agreement analysis were reported in most of these studies. In the study by Lin et al., there were also large limits of agreement between diaphragm and scalene muscle activity during noninvasive ventilation in COPD patients, whereas the parasternal intercostal muscles performed better [
18]. COPD patients often have high levels of neural respiratory drive (for review see [
42]) and thereby extradiaphragmatic inspiratory muscle activity is easier to detect with surface EMG.
Taken together, reducing inspiratory assist does not have a uniform effect on the diaphragm and extradiaphragmatic inspiratory muscles. Differences in responses among muscles may be partly explained by the fact that extradiaphragmatic inspiratory muscles are involved in other functions, such as patency of the upper airways, rotation of the head, flexion of the neck and stabilization of the trunk [
20‐
25]. For the parasternal intercostal muscles the same motoneurons are depolarized during postural and inspiratory tasks; their output during inspiration is depending on the direction of the rotation of the trunk [
23]. Furthermore, it has been shown that neural respiratory drive is not uniform in healthy subjects, and respiratory muscles recruit according to their mechanical advantage. In other words, respiratory muscles (or portions of muscles) with the greatest mechanical advantage for a specific task will be recruited earlier and to a larger extent [
43,
44]. It seems plausible that the same is true in disease, which could result in differences in recruitment of extradiaphragmatic inspiratory muscles and the diaphragm with changes in ventilator support. Parthasarathy et al. [
34] suggested such a hierarchy of respiratory muscle recruitment in patients failing a T-piece trial. In addition to the diaphragm and intercostal muscles, they demonstrated an immediate increase in sternocleidomastoid muscle activity with little change thereafter. The expiratory muscles are recruited relatively late during the T-piece trial: the largest increase in activity occurred only after 17–20 min. Finally, drive to the diaphragm may underestimate the true respiratory drive due to the contribution of the extradiaphragmatic inspiratory muscles, especially in critically ill patients there may be a discrepancy. Respiratory drive can be higher in critically ill patients not only due to the load on the respiratory muscles, but also due to metabolic acidosis and hypoxemia, brain, lung or chest wall pathologies (for review see [
40]).
In healthy subjects there is a clear hierarchy with respect to respiratory muscle recruitment [
45,
46]. For example, upper airway muscles recruit ± 100 ms before the diaphragm recruits in healthy subjects [
21]. We studied more different muscles and applied different PS levels as compared to previous studies [
26,
27]. The alae nasi recruited earlier as compared to other extradiaphragmatic inspiratory muscles (132–172 ms) and the diaphragm (122 ms). We found no differences in timing of the extradiaphragmatic inspiratory muscles between ventilator settings. These results were expected based on previous studies [
26,
27]. Schmidt et al. [
27] observed that recruitment onset times were similar among the scalene, sternocleidomastoid and genioglossus in mechanically ventilated patients.
Practical limitations of surface EMG
In addition to the poor correlations and low agreement between changes in activity between the diaphragm and extradiaphragmatic inspiratory muscles in response to unloading, there are practical issues that limit the applicability of surface EMG to monitor drive to the diaphragm. First, we found that in several patients no muscle activity could be detected from the genioglossus, alae nasi, parasternal intercostals and scalene during the whole study protocol. This could be the result from real inactivity of the muscles or low signal-to-noise ratio. Second, surface EMG is vulnerable to noise (e.g. electromagnetic noise) and artifacts (e.g. due to movement), these cannot be avoided, but the effects can be minimized in the preprocessing and analyzing process [
47]. Third, the technique is technically challenging in obese patients, restless patients, or patients with diaphoresis. Note that data used in the current study were highly selected. Large periods of data were not useful to study breathing activity because patients were moving their head or body resulting in non-breathing-related muscle activity.
Study limitations
The current study has some limitations. First, we did not measure force, only EMG as a measure for respiratory drive. Respiratory drive can be evaluated at the bedside by several methods (for recent review see [
40]). The only method to measure the contribution of extradiaphragmatic inspiratory muscles to respiration is by surface EMG. Therefore, we wanted to evaluate the recruitment pattern of extradiaphragmatic inspiratory muscles with respect to the diaphragm. Second, we did not measure surface EAdi. Bellani et al. [
17] demonstrated that surface EAdi correlated well with EAdi, although there was a high variability in the slopes between patients. They showed that respiratory effort could be calculated from surface EAdi, but when comparing surface EAdi with esophageal pressure to compute muscular pressure, this resulted in low bias but large limits of agreement. Calculation of effort from both the diaphragm and extradiaphragmatic inspiratory muscles did not result in an improved estimation of respiratory effort as compared to EAdi or surface EAdi. Third, the study was not blinded. The signals were analyzed offline, only periods to be analyzed were selected manually, while the rest of the analysis was performed automatically using a custom-written script. Therefore, the unblinded nature is unlikely to affect the results. Fourth, accuracy of calculating recruitment times depends on the manner in which the threshold for muscle activity is determined, and also on the noise level. Therefore, not only relative onset times were computed, but also peak and termination times. For all three parameters the same trends were found and recruitment times were in the same range.