In recent years, noninvasive mechanical ventilation (NIMV) has become an alternative to orotracheal intubation and invasive mechanical ventilation for the management of respiratory failure (RF), since it can reduce the complications and shorten patient stay in the Intensive Care Unit (ICU). The first acute disorders treated with NIMV were exacerbations of chronic obstructive pulmonary disease (COPD)1 or acute lung edema (ALE).2,3 Since these first studies, the cumulative supporting evidence has been very consistent in reference to patients of this kind,4 and over the last 20 years the use of NIMV has been extended to patients with other kinds of diseases,5 such as hypoxemic RF or postextubation RF. The usefulness of NIMV in patients of this kind is much less certain, however. The latest clinical guides6 suggest that NIMV should be the first management option in patients with exacerbated COPD or with ALE, and that NIMV may be considered in patients with postoperative RF or in immune depressed individuals. In turn, the use of NIMV should also be considered for allowing the early extubation of patients with COPD.7 These guides offer no recommendations on the use of NIMV in patients with asthma exacerbations, acute lung injury (ALI), severe community-acquired pneumonia or thoracic trauma. Nevertheless, NIMV is often used in patients of this kind.

Considering the above, it is important to know the use of NIMV outside protocolized studies, in patients with different types of RF.

The present study has two main objectives:

• -

  To know how NIMV is being used in clinical practice and its effectiveness.

• -

  To identify early predictors of NIMV failure that can be used for improved indication of the technique in future patients.

A total of 410 cases of NIMV were recorded, with a mean patient age of 66.7±13.4 years and a mean APACHE II score of 20.8±7.4. Males predominated (60.2%).

The total percentage failure rate was 50.7%, with a global mortality rate of 33.4%. Fig. 2 shows the distribution according to etiology, together with percentage failure, mortality, and mortality among the patients in which NIMV failed. The most frequent indication of the technique was acute hypoxemic RF (38%), followed at a distance by postextubation RF (21.2%) and the exacerbation of COPD (19%). The least numerous groups were hypercapnic RF without COPD (14.9%) and ALE (6.8%). On analyzing failure according to the type of RF involved, the poorest results corresponded to acute hypoxemic RF (73.7%), while the best results corresponded to ALE (21.4%), the exacerbation of COPD (26.9%) and hypercapnic RF without COPD (31.1%). The percentage failure rate in postextubation RF was 54%. The highest mortality corresponded to patients with hypoxemic RF. In the patients with NIMV failure, the mortality rate was very high (83.3%) among the individuals with ALE, though the small number of patients in this group means that the data must be viewed with caution. In contrast, the lowest mortality in patients with NIMV failure corresponded to the group of individuals with hypercapnic RF without COPD (31.6%).

Figure 2.

Distribution by etiologies (n), with percentage failure (light gray), mortality (dark gray) and mortality among the patients in which NIMV failed (black). aCOPD, chronic obstructive pulmonary disease. bRF, respiratory failure. cALE, acute lung edema.

(0.14MB).

The patients in which NIMV proved successful presented lesser mortality (14.4 vs 51.9%, p<0.001), with a shorter stay in the ICU (4.7±5.0 vs 16.3±19.2 days, p<0.01) and in hospital (20.8±17 vs 31.7±26.7 days, p<0.01) compared with the patients in which NIMV failed.

Regarding the differences in stay according to the type of RF, we observed significant differences in stay in the ICU among the patients with exacerbation of COPD (2.7±1.6 vs 9.9±16.7 days, p<0.005), hypercapnic RF without COPD (4.1±5.6 vs 14.4±12.4 days, p<0.001), acute hypoxemic RF (6±7.4 vs 16.5±21.4 days, p<0.001) and postextubation RF (7.6±4 vs 20.8±16.7 days, p<0.001)—the difference in the ALE group being nonsignificant (3.4±2.6 vs 5.2±3.4 days, p=0.13). Regarding the stay in hospital, we observed significant differences in the postextubation RF group (23.9±11.7 vs 37.1±25 days, p<0.05)—the rest of the differences being nonsignificant: exacerbation of COPD (14.5±8.6 vs 18.2±19.2 days, p=0.94); hypercapnic RF without COPD (19.9±12.9 vs 25.9±17.8 days, p=0.11); acute hypoxemic RF (29±28.9 vs 33.6±29.1 days, p=0.19) and ALE (18.1±11.7 vs 20.3±24.1 days, p=0.53).

We also observed differences in mortality in all the groups except in the hypercapnic RF without COPD group (31% vs 31.6%, p=0.96): exacerbation of COPD (17.5% vs 57.1%, p<0.001); acute hypoxemic RF (7.3% vs 54.8%, p<0.005); ALE (4.5% vs 83.3% p<0.005); postextubation RF (5% vs 46.8% p<0.005).

The most frequent cause of NIMV failure (Fig. 3) was excessively labored breathing (41.3%), followed by refractory hypoxemia (35.6%).

Figure 3.

Causes of NIMV failure (number of patients).

(0.13MB).

There were few complications associated to the technique. In effect, only 18 hemodynamic alterations (4.4%) and a single case of NIMV-associated pneumonia (0.2%) were documented.

Table 1 shows the differences in general characteristics of the patients in which NIMV proved successful and those in which NIMV failed. There were significant differences in the APACHE II score, the use of NIMV before admission to the ICU, the use of inotropic/vasopressor support, home oxygen therapy, the need for sedation in order to tolerate NIMV, and the presence of some type of immunosuppression (according to the definition of the APACHE II scale).

Table 2 in turn shows the clinical and laboratory test findings at the start and 2h after the start of NIMV. At the start, the patients in which NIMV failed showed lower values corresponding to pH, respiratory frequency, GCS, serum lactate and serum bilirubin, and greater PaCO2, PaO2/FiO2, bicarbonate and systolic blood pressure values.

Table 3 shows the variables affording the best fit in the logistic regression model. The goodness of fit of the model was 0.48. The percentage of correct classifications was 77%, with a sensitivity of 70% and a specificity of 84%. With the validation sample, the correct classifications rate was 74%, with a positive predictive value of 66% and a negative predictive value of 72%.

Although the change in heart rate and the existence of radiological findings were not significant, they were retained in the model due to their clinical importance and because their presence in the construct afforded better correct classification rates. The need for sedation, the type of RF (hypercapnic versus non-hypercapnic), serum bilirubin at the start, and the existence of radiological findings were the risk factors with the greatest magnitude of effect, while the change in heart rate and respiratory frequency were found to be protective factors, though with a small magnitude of effect. It is necessary to clarify the interpretation of the variables defining the change from the start to 2h after the start: in the case of the change in GCS in patients who are improving, negative values are obtained (as a result of change from a lower to a higher GCS). A more negative result reflects better evolution than a less negative result; consequently, an increase in this scale (i.e., a less negative result) identifies lesser improvement and therefore points to a poorer prognosis.

The NIMV failure rate in our series was 50.7%. On examining the results of previous studies,10,11 the failure rate was found to be between 24% and 58%—the latter being the rate observed in a randomized study involving different types of RF.12 There is also evidence consistent with our own results in favor of a decrease in stay both in the ICU13,14 and in hospital14 among those patients in which NIMV proves successful.

However, the broad range in percentage failure observed in the literature is notorious. The fundamental reason for such variability is probably the different proportions of types of RF in the different studies. It therefore might be more useful to compare results according to type of RF.

In this regard, the success rate was high in exacerbations of COPD, in ALE and in hypercapnic RF without COPD, ranging between 70 and 78%. In the first two groups, the scientific evidence is overwhelmingly in favor of the use of NIMV.1,15,16 In these patients, our results are similar to those of previous studies as regards the success rates15,16 and decreases in mortality and stay in the UCI16 when NIMV proves successful.

Likewise, in the patients with ALE the observed failure rate was 21.4%, which is similar to the data published elsewhere,17 and lesser mortality was also recorded in the patients in which the technique did not fail. Despite these observations, however, the small number of patients involved in this group means that the data must be viewed with caution.

Regarding the last group, i.e., hypercapnic RF without COPD, the existing evidence is much less clear, with few studies and moreover involving patients with highly diverse disease conditions. The reported success rate in this case ranges from 89%18 to 38%.11 In our study, and in contrast to the previous evidence, we recorded no increased mortality or greater NIMV failure among the patients without COPD versus those with COPD.

In our experience, NIMV can be used with a high probability of success in these three types of patients (exacerbation of COPD, ALE and hypercapnic RF without COPD).

In contrast, the failure rate in patients with hypoxemic ARF was very high (73.7%). In this type of disease there is much less available evidence in support of the use of the technique. The great variability of causes of hypoxemic ARF complicates the comparison of results, with failure rates of between 12 and 60%.11,19,20

In our study, the patients in which NIMV proved successful presented lesser mortality and a shorter stay in the ICU. In the literature, although there are studies reporting a decrease in the number of serious complications, a lesser need for mechanical ventilation, and a shorter stay in the ICU,19–21 the metaanalysis conducted by Keenan et al.22 concluded that the great variability of results in the different studies does not allow recommendation of the generalized use of NIMV in patients with hypoxemic ARF.

The data of our study should cause reconsideration of the use of NIMV in patients of this kind.

In the postextubation RF subgroup, success was only moderate, with failure being observed in 54% of the cases. While the use of NIMV in the prevention of postextubation RF during the weaning process is warranted by the scientific evidence,23 its effectiveness in postextubation RF is much more controversial. In the same way as in hypoxemic ARF, the great variety of causes of RF and the diversity of disease conditions for which the patient required orotracheal intubation in the first place all complicate the evaluation of NIMV in these patients. Two randomized studies24,25 found no differences in reintubation, and one of them24 demonstrated greater mortality in the NIMV group.

The complications of NIMV were found to be infrequent (one case of NIMV-associated pneumonia and 18 hemodynamic alterations), though the retrospective nature of the study made it impossible to adequately evidence the minor complications such as the appearance of skin lesions secondary to application of the interface. There were 6 cases of cardiorespiratory arrest (1.5%)—all in asystole or electrical activity without a pulse. The incidence of cardiorespiratory arrest in earlier studies varied between 0%15 and 7%.12 We identified no myocardial ischemia episode (defined as the presence of acute changes in the ECG tracing and/or the elevation of cardiac enzymes), in contrast to the study of Mehta et al.,17 who found an increased incidence of myocardial infarction in the group subjected to bilevel positive airway pressure (BPAP) ventilation. The 6 cases of cardiorespiratory arrest occurred in patients with multiple previous disease conditions justifying conservative management (in fact, three of them were patients with instructions against intubation).

The results of the logistic regression analysis are concordant with the previous evidence which identified the following predictors of failure: a diagnosis of pneumonia,12 high APACHE II score,12 respiratory frequency >35rpm,12 high SAPS II score,8,11,26 poor tolerance of NIMV,26 high IPAP,27 low pH,27,28 GCS before NIMV,11 PaO2/FiO2 before NIMV11 or PaO2/FiO2 after 1h of NIMV.28 The exacerbation of COPD has been identified as a predictor of success.29

In general, our data suggest that the more seriously ill patients, as well as those with a poorer initial response to NIMV, are more likely to suffer NIMV failure.

The identification of serum bilirubin before the start of NIMV as a predictor of failure of the technique has not been reported elsewhere, and could help to identify patients with multiorgan dysfunction (since there were no significant differences regarding the existence of previous liver disease between the NIMV success and failure groups). Previous evidence30 showed that patients with greater organ failure (as measured by the SOFA score) had poorer results.

The main limitation of our study is the fact that it is not a randomized controlled trial but a retrospective study. On the other hand, its single-center nature means that the results might not be extrapolatable. Furthermore, its long duration (6 years) might imply the existence of changes in technology and in the management of patients that must be taken into account in interpreting the data obtained.

Since one of the key factors in deciding the use of NIMV is the probability of failure, which can worsen the prognosis of patients with RF, our results could contribute to the decision making process, since we have identified several predictors of failure that might be useful. With these predictors it may be possible to decide not to start NIMV, or to suspend it early, if the probability of failure is seen to be high.

Noninvasive mechanical ventilation is useful for managing patients with RF, and can avoid the need for orotracheal intubation in approximately one-half of the cases. Its effectiveness is greater in patients with exacerbation of COPD, with ALE or with hypercapnic RF without COPD. The use of NIMV in patients with hypoxemic RF or postextubation RF should be assessed on an individualized basis.

There has been no financial support for the conduction of this study.

The authors declare that they have no conflicts of interest.