Discussion
Studies performed in the past 10 years indicate that many mechanically ventilated MICU patients have severe diaphragm weakness [
1‐
5]. Moreover, diaphragm weakness is thought to be associated with poor outcomes in this patient population, with the weakest patients requiring more prolonged mechanical ventilation and having a significantly higher mortality than patients with better diaphragm strength [
1,
4]. Because of this evidence, it is speculated that mechanically ventilated patients may benefit from treatment with anabolic agents which increase skeletal muscle strength. In theory, such agents could improve patient outcomes, reducing duration of mechanical ventilation and patient mortality. Several pharmacological agents have been shown to improve skeletal muscle strength in other patient populations (e.g., the elderly, patients with cancer, patients with chronic obstructive pulmonary disease, COPD) and it is reasonable to believe that one or more of these agents may also be capable of increasing muscle strength in the mechanically ventilated MICU population [
13‐
15]. There are, however, methodological issues that will need to be addressed to optimize performance of therapeutic trials of anabolic agents in critically ill mechanically ventilated patients. Most importantly, it will be useful if therapeutic trials incorporate an objective technique to accurately assess respiratory muscle strength.
There are two commonly used techniques that have been employed in the past to assess respiratory muscle strength in mechanically ventilated patients, i.e., the Pimax and the PdiTw in response to bilateral anterolateral magnetic phrenic nerve stimulation (BAMPS) [
1‐
7,
16‐
18]. The Pimax has been severely criticized in the past as being unreliable in mechanically ventilated patients [
7] and as a measure that does not discriminate between weaning success and weaning failure [
3,
19]. In addition, there is no report, of which we are aware, that has shown that Pimax levels correlate with MICU patient survival. As a result, it could be argued that the Pimax should not be used to assess respiratory muscle strength in mechanically ventilated patients for clinical trials. If so, one could argue, future trials will be forced to employ techniques such as the PdiTw in response to BAMPS, to assess responses to therapies. Unfortunately, while the Pimax technique is simple, the PdiTw technique is complex and is only available in a handful of MICUs in the world. The purpose of the present study, therefore, was to compare measurements of Pimax and PdiTw in mechanically ventilated MICU patients to see how these two indices correlate with each other, and to ascertain the degree to which the Pimax, like the PdiTw, could predict outcomes in this patient population. To our knowledge, this is the largest MICU patient population in which these two measurements were made and were correlated with outcomes.
When comparing Pimax and PdiTw, it is important to first consider that there are several fundamental differences between these techniques. For one thing, the Pimax assesses the cooperative generation of inspiratory pressure by all the inspiratory muscles, while the PdiTw elicited by magnetic stimulation of the phrenic nerves assesses mainly the diaphragm. As a result, variation in the degree of rib cage activation during the Pimax maneuver can alter the level of Pimax for a given level of diaphragm strength and, by inference, a given level of PdiTw. In addition, the Pimax measures maximal voluntary inspiratory muscle pressure generation while the PdiTw assesses submaximal performance (i.e., pressure generation in response to a single neural impulse). Third, PdiTw levels are influenced by the previous contraction history of the diaphragm (i.e., a phenomenon termed twitch potentiation) while the Pimax is not [
5,
6]. The findings in the present report underscore the potential for differences between Pimax and PdiTw measurements. In keeping with these factors, we found that there was substantial scatter when plotting Pimax against PdiTw, indicating that it is not possible to reliably predict PdiTw values from determination of Pimax alone.
The present data also provide the novel finding that infections systematically change the relationship between Pimax and PdiTw measurements; the present manuscript is the first to suggest this same phenomenon occurs in both animals and patients. Specifically, we found that infection selectively reduced skeletal muscle low-frequency force generation more than high-frequency force generation by comparing diaphragm force frequency curves for muscles obtained from control and mice treated with endotoxin. Similarly, we found that infected patients demonstrated a proportionately greater reduction of the PdiTw, an index of low-frequency force generation, than the Pimax, an index of high-frequency force generation, when compared to values for noninfected mechanically ventilated MICU patients. This greater reduction in low-frequency force generation may reflect the reported effect of infection to reduce sarcoplasmic reticulum calcium release and reuptake [
20]. Another factor that may contribute to this phenomenon is the reported effect of infection to reduce the calcium sensitivity of skeletal muscle contractile proteins [
21]. Regardless of the mechanism, this phenomenon results in a higher Pimax/PdiTw ratio for infected patients when compared to noninfected mechanically ventilated MICU patients.
It is also clear patients on mechanical ventilators have varying levels of respiratory drive due to differences in levels of lung disease, reflex sensitivity, ventilator settings, level of sedation, and other factors. In theory, lower levels of respiratory drive should depress the level of Pimax generated for a given level of intrinsic diaphragm force generating capacity, e.g., for a given level of PdiTw. As an extreme, Pimax could hypothetically be reduced to zero when extremely high levels of sedation are provided, even when PdiTw is very good. Consistent with this possibility, several patients had very low Pimax/PdiTw ratios in the present study (i.e., ratios as low as 1.54). Also in keeping with this concept, we found that patients with high levels of respiratory drive, i.e., that initiated (triggered) all breaths, had significantly higher Pimax/PdiTw ratios than patients with submaximal levels of ventilator triggering. Thus the dual influences of varying levels of infection and respiratory drive from patient to patient may well account for much of the variability in the Pimax to PdiTw relationship, with infection raising and low levels of respiratory drive lowering the Pimax for a given level of PdiTw.
Another factor that, in theory, may have contributed to the differences in PdiTw and Pimax measurements in the present study is related to the fact that we assessed PdiTw while patients continued on mechanical ventilation (to minimize twitch potentiation) while we assessed Pimax while patients were temporarily removed from mechanical ventilation (to facilitate recruitment of respiratory muscles during the maneuver). It is theoretically possible then that inspiratory muscle length may have changed, in some patients, when patients were transiently removed from mechanical ventilation in order to assess Pimax, further artifactually altering the PdiTw and Pimax relationship.
A final consideration when comparing these two indices, i.e., Pimax and PdiTw, are technical issues that may limit the accuracy of these assessments. As indicated above, inadequate levels of respiratory drive may result in artifactually low Pimax levels. There are several approaches to minimizing this artifact, including employment of prolonged airway occlusion (e.g., more than 20 seconds) for intubated patients, adding dead space to the respiratory circuit to increase respiratory drive, and ascertainment that the negative pressure generated at 100 milliseconds after initiation of an inspiratory effort (P0.1) is more negative than 2 cm H
2O before determination of the Pimax [
22]. For the PdiTw, attainment of supramaximal phrenic nerve stimulation during magnetic stimulation is necessary to ensure that all motor nerve fibers to diaphragm are maximally activated during this maneuver. Under laboratory conditions, proof of supramaximal phrenic nerve stimulation is best ascertained by repeatedly stimulating the phrenic nerves over a broad range of magnetic stimulation field strengths (from 60–100 % for the MagStim 200 unit) with multiple stimuli applied (e.g., more than five) at each level of field strength. The magnetic stimulation paradigm employed to obtain PdiTw data in the present manuscript represents a compromise between the need to attain perfect data and ethical constraints to minimize patient discomfort. While it is therefore possible that our phrenic stimulation technique may have underestimated PdiTw in some of our patients, previous studies suggest that supramaximal phrenic nerve stimulation can be achieved in the large majority of critically ill patients using the procedures employed in the present study [
4,
5].
While the above discussion evaluates factors and artifactual issues that differentially influence Pimax and PdiTw determinations, a larger issue is which of these two indices is a better index of physiological function and which one is a better predictor of patient outcomes. Which is the more relevant physiologically may well depend upon the specific respiratory task that a patient may encounter. Even in the presence of lung disease, the diaphragm does not normally generate “ballistic” repetitive contractions at near maximal levels of force generation (and extremely high levels of stimulation frequencies) during breathing efforts but is driven in response to mean stimulation frequencies in the 5–15 Hz range. In theory, therefore, diaphragm force generation in response to relatively low stimulation frequencies (closer to twitch forces) may be a more relevant index of diaphragm performance during normal breathing than diaphragm force output in response to maximal levels of excitation (i.e., Pimax of MIP maneuvers). On the other hand, ballistic efforts, such as coughing or sneezing, involve high-level activation of both inspiratory and expiratory muscles. During such maneuvers, the Pimax may well be a better index of predictor of task performance.
An even more important issue, however, is which parameter (Pimax or PdiTw) is a better predictor of patient outcomes. Arguably, the two most important outcomes for critically ill mechanically ventilated ICU patients are whether they survive, and if they survive, how long it takes to wean them from mechanical ventilation. Our data indicate that PdiTw is a good predictor of both mortality and duration of mechanical ventilation. Surprisingly, we also found that Pimax also was a fair predictor of these two outcomes, with patients with lower Pimax levels having a statistically greater mortality and, if they survived, a longer duration of mechanical ventilation than patients with high Pimax levels. While both indices were statistically predictive of mortality and mechanical ventilation duration, PdiTw was the better predictor, since this index had higher F scores than the Pimax for correlation with both outcomes.
This analysis also raises the question as to why PdiTw level was the better predictor of clinical outcomes. This may result because there is more biological variability in Pimax than the PdiTw due to the factors listed above (infection, alterations in sedation, variation in respiratory drive) so that it is more difficult to distinguish the true level of weakness using Pimax than using PdiTw measurements. PdiTw is also a better measure of diaphragm function at low frequencies of neural activation. As suggested above, since the mean endogenous phrenic neuron firing frequency is relatively low [
23], it is possible that the PdiTw may be a more relevant index of the ability of the diaphragm to function close to the range of firing frequencies that are physiologically achieved. In addition, while the other inspiratory muscles clearly contribute to breathing in patients with respiratory failure, the diaphragm is the major muscle of respiration and diaphragm strength per se may be a better determinant of outcomes.
An even more fundamental issue raised by our data is whether or not the observed correlation between respiratory muscle strength and mortality might be influenced by other factors that may have contributed to death in our patient population. To further analyze this issue, we performed a series of multivariate analyses examining the impact of several potentially confounding variables (namely, sepsis, age, gender, steroid usage, respiratory system static compliance, inspiratory airway resistance, SOFA score, and comorbidity score) on the relationship between weakness and mortality. As indicated by our results (Tables
2 and
3), age, gender, steroid usage, respiratory system static compliance, and inspiratory airway resistance were not significantly correlated with mortality in any of our statistical analyses, while our two indices of respiratory muscle strength (PdiTw and Pimax) were highly significant correlates of survival in all analyses. When included in this multivariate analysis, the presence and absence of sepsis was also not significantly correlated with mortality. Finally, when we statistically analyzed only the subgroup of patients in our study that were septic, respiratory muscle weakness remained, by far, the best predictor of mortality (Tables
4 and
5).
We should consider the potential mechanism(s) by which diaphragm weakness may have influenced mortality. One of the patients that died in our study had a large stroke, never regained consciousness, and died when his family withdrew care. An additional five patients were in shock at the time of death. It is therefore difficult to implicate weakness as a direct contributor to death in these six patients. In the remaining 12 patients who died in our study, however, none were in shock, all had good Glasgow coma scores, all were weak, and in all 12 it was not possible to successfully wean these patients from mechanical ventilatory support. In all 12, death occurred once the patients’ families decided to withdraw life support, and the only form of life support that these 12 patients were receiving and that was withdrawn was mechanical ventilation. Importantly, the lung function of these 12 patients was not different, on average, from patients that survived (also, lung function was not statistically associated with death in our multivariate analysis). Based on this analysis, we believe that it is reasonable to speculate that it is possible that diaphragm weakness may contribute to death by preventing weaning of patients from mechanical ventilation, thereby contributing to a decision to withdraw life support. The only way to conclusively prove this point, however, is to perform prospective studies to determine if therapies that increase diaphragm strength reduce mortality in mechanically ventilated ICU patients.
Authors’ contributions
GSS drafted the protocol, performed the measurements, analyzed the pressure tracings, obtained patient data, interpreted the data, drafted and revised the final manuscript. PW performed a statistical analysis of the data, interpreted data, and contributed to revision of the final manuscript. LAC assisted in performing the measurements, obtaining patient data, had a major impact on the interpretation of the data and revision of the manuscript. All authors read and approved the final manuscript.