To our knowledge the present multicenter observational study is the first to document the prevalence and type of asynchronies in patients receiving NIV for acute respiratory failure of mixed etiology. The results show that each specific type of asynchrony was present in 12–23% of patients, while 43% of patients presented with severe asynchrony (AI > 10%). The two factors predictive of the latter were the level of pressure support and the magnitude of leaks.
Before discussing these results, some limitations of the study should be pointed out. First, the tracings were analyzed by only one investigator (L.V.), rather than one from each center, which could lead to systematic bias in their interpretation. However, the methodology had been strictly defined beforehand and tracings on which doubt persisted were discussed among investigators during data session meetings. Furthermore, the random analysis of samples comparing clinical and tracing analysis for ineffective breaths and auto-triggering showed no difference. Still, one cannot exclude that some events were missed or erroneously reported as present. Second, the three participating centers are considered as experienced in the use of NIV. This suggests that if the recordings had been made in centers with less experience with the technique, a higher prevalence of asynchrony might have been documented. Therefore, our figures probably reflect the lower end of the range of prevalence of asynchrony in patients receiving NIV for acute respiratory failure. Third, most patients had already received a few NIV sessions before being included in the study. Had the recordings been made in all patients during the first NIV session, the prevalence and severity of asynchrony would likely have been higher. Fourth, our patient population was fairly modest in number and was heterogeneous both in terms of chronic disease and cause of acute respiratory failure, which might tend to not reflect the specific problems experienced with patients in whom NIV is difficult, e.g., severe COPD or pneumonia. However, the purpose of the study was to “take a picture” of a real-life population of patients in need of NIV, and to that end the various acute and chronic diagnoses seem adequately representative of this situation and are in line with published data [14, 15]. Finally, an ICU ventilator was used in all patients. This reflects the usual practice of the participating centers, and is in line with other published results [14], mainly because of the higher capacity of these machines to deal with a high inspiratory demand [16]. Nonetheless, our results might not be applicable to the use of a bilevel device, given that these machines deal better with leaks than ICU ventilators when the latter’s NIV mode is not activated [17]. In this same line of thought, due to the observational nature of the study, eight different ventilator types were used, which might have influenced the results. However, had only one type of machine been used, asynchrony might have been attributed to this particular type of ventilator, whereas we found no obvious differences in asynchronies between machines.
Let us now discuss the various asynchronies observed:
Ineffective triggering
Ineffective triggering has been shown to increase the work of breathing, most often due to the added inspiratory threshold load associated with dynamic hyperinflation [18, 19]. Ineffective breaths were present in eight (13%) of patients. During NIV, one would intuitively expect leaks to increase the occurrence of ineffective efforts, although no clinical study has directly addressed this point. However, in two bench model studies mimicking NIV conditions, leaks were shown to increase the trigger delay [17] and the number of ineffective breaths [20]. This hypothesis is further supported by the fact that in our patients presenting ineffective breaths the magnitude of leak was higher and the VTe lower than those of patients without such wasted efforts (Table 4). Nonetheless, the weak correlation between the magnitude of leaks and the number of ineffective breaths suggests that leaks are not the only cause of such breaths, or that there is no direct quantitative relationship between these two parameters. Intrinsic PEEP has been shown to lead to ineffective efforts in COPD patients [21], but this was not assessed in the present study. In any case, ineffective efforts were not more prevalent or severe in patients with COPD.
Auto-triggering
Leaks have been shown to be a major factor leading to auto-triggering [17, 22]. In our eight (13%) patients in whom auto-triggering was present, the magnitude of leak was higher (Table 4). However, no correlation could be found between leak volume and the severity of auto-triggering. This suggests that once a critical threshold has been reached, above which auto-triggering occurs, the frequency of its occurrence depends on several factors such as trigger setting, type of trigger, and design of the ventilator, as shown in a recent bench study [17].
Double-triggering
Double triggering is often associated with an insufficient level of pressure support, and results from the same pronounced inspiratory effort re-triggering the ventilator after it has discontinued pressurization. In the 9 (15%) patients exhibiting double triggering, the tip was higher and the level of pressure support lower (Table 4). Even though tip was not measured in the study by Thille et al., analysis of the results shows that double triggering was associated with the same imbalance whereby the ventilator’s pressurization time is too short in the face of an increased inspiratory demand [10]. This was also the case in our patients, to which an insufficient level of pressure support probably contributed.
Short cycle (premature cycling)
Premature cycling occurred in only seven (12%) of the patients, and was associated with a longer tip, an expected finding. One of the main causes of premature cycling is the presence of restrictive respiratory mechanics [23, 24]. Interestingly, in our study, acute community-acquired pneumonia was more prevalent in patients in whom premature cycling was present.
Prolonged cycle (delayed cycling)
Delayed cycling was present in 14 (23%) of our patients. In intubated patients the main cause of delayed cycling is the presence of obstructive mechanics [23, 25]. On the other hand, during NIV, while obstructive mechanics are likely to be a contributing factor, leaks have been shown to be a major determinant of delayed cycling [7, 8]. This was probably also the case in our patients, as suggested by the higher magnitude of leak and lower VTe and minute volume in patients in whom delayed cycling was present (Table 4). Furthermore, among the 14 patients exhibiting delayed cycling, only 5 had documented COPD. The 19 other patients with COPD had no delayed cycling.
In a recent study testing the predictive value of an automatic algorithm in detecting asynchrony, Mulqueeny et al. found that 40% of their patients recovering from an episode of acute respiratory failure, and receiving NIV with a bilevel device exhibited an AI > 10% [26], which is in line with our documented 43%. It is of interest to note that a comparable prevalence of severe asynchrony was observed even though the patients were studied at a later stage in the course of treatment with NIV and the type of ventilator used was different [26]. This might be coincidental or point to one or several factors that are inherent to NIV, such as leaks, or the difficulty of adapting to PSV. In the study by Thille et al., only 25% of patients had an AI > 10% [10]. However, the patients were intubated, a situation with very little, if any, leaks. Therefore, it seems reasonable to assume that leaks play a key role in causing many of the asynchrony events documented in our study and that of Mulqueeny et al. [26]. This assumption is in line with the results of the multivariate analysis showing that both the level of pressure support and the magnitude of leaks were associated with an AI > 10%. A higher level of pressure support has been shown to increase the occurrence of leaks [9]. Regarding the level of leaks in our study (27% of MV), it was in line with the range (25–39%) found in two other studies in patients receiving NIV [27, 28].
In the six patients in whom the “NIV mode” was activated, no asynchrony event was observed. The primary goal of the “NIV mode” is to compensate for the various interferences of leaks with ventilator function, thereby diminishing the incidence of asynchrony events [17]. Given that one of the two main determinants of an AI > 10% was the magnitude of leaks, it is reasonable to assume that the “NIV mode” proved effective in those six patients. As the study was not designed to test this hypothesis, however, such a conclusion remains speculative.
The absence of difference in outcome between patients with and without an AI of 10% might suggest that asynchrony during NIV is of little clinical importance. In our opinion, however, such a conclusion is not warranted at this stage for several reasons. First, the study was not designed to document such an impact. It only looked at the prevalence of asynchrony during one NIV session. Compared to mechanical ventilation in intubated patients which is a continuous process, NIV is applied intermittently. Hence, asynchrony occurring during one NIV session might not be present during the next session, thereby having only a limited impact on NIV failure. Second, the cutoff level of 10% to define severe asyncrhony, associated with adverse outcome in intubated patients [10], might not be appropriate in the setting of NIV. Third, the cause of asynchrony might vary, e.g., respiratory muscle fatigue or weakness, poor lung mechanics, and agitation in intubated patients versus leaks during NIV, the prognostic significance being quite different. Supporting this view, it has been shown that the nature of the underlying lung disease and absence of arterial blood gas improvement are predictors of NIV failure [29]. In line with these findings, admission PaO2 was lower and respiratory rate at admission and during NIV was higher in patients in whom NIV failed. Most of the time NIV is delivered over a relatively short period of the time and the influence of the early sessions of NIV is probably critical. Indeed, most failures occur within the first 48 hours of treatment. Finally, it is possible that some patients did not require NIV anymore. Indeed, although the patients had received only 4 ± 2 prior sessions of NIV, one cannot exclude that some patients could have already been weaned off NIV.
Not surprisingly, patients with an AI < 10% had a higher VAS comfort score, which suggests that, irrespective of improving or not improving synchrony impacts outcome, it would at least make the NIV experience more comfortable for patients. Beyond the obvious merit of such an improvement, one must bear in mind that in large epidemiologic study on NIV, patient intolerance to the technique was one of the key factors predicting its failure [4].