INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Noninvasive ventilation (NIV) has been shown in randomized controlled trials to be efficient for reducing the need for endotracheal intubation (ETI) in several groups of patients with acute respiratory failure (ARF) (1). Patients with acute-on-chronic respiratory failure are the most likely to benefit from NIV in terms of morbidity and mortality (1, 6). Selected patients with various forms of hypoxemic respiratory failure may also benefit from NIV in terms of intubation rate and complications (2). In this category of patients, however, the results are more variable (7, 8), which may be explained by the type of patients, the location at which treatment is provided, or the skill of the care team. Careful studies of the nursing and therapist workload required to deliver NIV suggest that this technique requires specific training, and that its efficacy probably follows a learning curve (9). When the investigators who conducted these studies computed the specific time required to treat patients with NIV, they found that in the first hours of treatment, more time was required of the respiratory therapists than with a conventional approach (10, 11).

The careful selection of patients, as well as the nonblinded nature of the clinical trials of NIV, are among the factors that could explain why the results of randomized controlled trials may differ slightly from those of everyday practice. For this reason, it appears important to assess the use and efficacy of NIV not only through clinical trials but also in everyday practice. We thus conducted an observational survey in France among 42 intensive care units (ICUs) to evaluate the use of NIV and to assess its efficacy in everyday clinical practice by comparing patients with similar etiologies treated with either NIV or ETI, after adjusting for the severity of their illness.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A prospective survey was performed between September 15 and October 15, 1997. A total of 52 ICUs were contacted through the collaborative group on mechanical ventilation belonging to the French-speaking medical intensive care society (Société de Réanimation de Langue Française [SRLF]). Recruitment was based on the voluntary participation of centers, and took place through meetings of this working group. Although all 52 ICUs were interested, 42 participated actively and sent a complete set of data; of these, 16 belonged to non-university hospitals and 26 were within university hospitals.

Most of the participating centers were in France; other countries included Switzerland, Belgium, Spain, and Tunisia. All participating centers were run by physicians belonging to the SRLF, which involves the French-speaking countries (France, Switzerland, part of Belgium, and Tunisia, as well as other centers who have a strong collaboration with French ICUs and with the SRLF, in Spain, Italy, Portugal, and elsewhere). Every patient admitted during the period of the study and who needed invasive mechanical ventilation or NIV for ARF (i.e., through an endotracheal tube or through a face mask, at the beginning of or during hospitalization) was included in this observational study. The total number of admissions during this period was also recorded.

A questionnaire was completed for each patient at admission and during the hospital stay until ICU discharge. All questionnaires were sent back by mail at the end of the 3-wk survey period of the study, or when the follow-up of patients admitted in this period was completed.

The following data were collected for every patient in the survey:

1. Demographic data and clinical history. This included the patient's age, institutional or other origin before hospitalization and ICU admission, and severity of illness as assessed with the Simplified Acute Physiologic Score (SAPS) II (12). The patient's prognosis before the acute event requiring ICU admission was scored from 0 to 2 as likely to be not fatal, ultimately fatal, or rapidly fatal, respectively, according to the criteria of McCabe and Jackson (13). Previous ICU stay and previous treatment with invasive mechanical ventilation or NIV, or mechanical ventilation at home, were recorded.

2. Data relative to the institution of mechanical ventilation. The condition precipitating respiratory failure; arterial blood gas values before ventilation, when available; respiratory rate; hemodynamic data; ability of the patient to efficiently mobilize secretions (yes = 1, or no = 0); and degree of encephalopathy scored from 0 to 4 (1) (grade 0 = no abnormality detected; grade 1 = hypertonia or cogwheel rigidity without an abnormality in consciousness; grade 2 = tremor, asterixis, or sleepiness; grade 3 = confusion or agitation; and grade 4 = prostration or vigil coma.

3. Follow-up during mechanical ventilation (NIV and/or conventional ventilation). This included time to the beginning and end of ventilation, mode of ventilation, and development of nosocomial pneumonia. No criteria were imposed for the diagnosis of nosocomial pneumonia. All centers, however, reported using clinical, biologic, and radiologic criteria together with quantitative cultures of protected brush specimens for endotracheally intubated patients. For NIV patients, the clinicians did not require quantitative cultures of protected brush specimens in every case, and the diagnosis was based on clinical (fever, sputum, lung crackles), radiologic (new infiltrate), and biologic (increase in white blood cells, bacteria in the sputum or in bronchial aspirates) criteria.

In addition, patients treated with NIV were followed with a special questionnaire during the first week of ventilation or until ETI. The duration of NIV in hours per day, the mode of ventilation, the kind of mask used, and the results of arterial blood gas analysis were recorded (arterial blood gases were obtained on Day 1, which was the first day of noninvasive ventilation, usually at the end of the first session of NIV or at the time NIV was stopped in the case of failure). A semiquantitative score was used to quantify the clinical tolerance of the patient and the severity of air leaks during NIV sessions (from 1 = very good to 5 = bad). Reasons for discontinuing NIV and the need for and timing of secondary ETI were also recorded.

Time to discharge from the ICU or to death was recorded (until Day 28) for all patients.

Statistical Analysis

Data are expressed as frequency for nominal variables and as mean ± SD for continuous variables. Nominal variables were compared by using the chi-square test or Fisher's exact test, as appropriate. Continuous variables were compared through the unpaired Student's t test. If the distribution of the continuous variables was not normal or if the test for the homogeneity of variance (Bartlett's test) gave a significant result, the Mann-Whitney U test was used.

Forward stepwise logistic regression was used for risk factor assessment. To assess the possible impact of NIV on nosocomial pneumonia and mortality rates, we compared the two study groups of patients, with attempts to adjust for differences at baseline. It was observed that coma (i.e., in which the need to protect the airway justifies intubation) was considered in all centers as a contraindication to NIV. Patients with coma were therefore not included in the analysis. First, the data for patients who developed nosocomial pneumonia and those who did not were compared, and those for survivors and nonsurvivors, through use of the chi-square test or Fisher's exact test, when necessary, for categorical variables, and through Student's t test for continuous variables. Second, variables found to be associated with pneumonia or death (p value  0.10 in the univariate analysis described earlier) were entered into a logistic regression model, which allows simultaneous control of multiple factors. Continuous variables were introduced as uncorrected data. The adjusted odds ratio (OR) of acquiring pneumonia and of dying, and the 95% confidence interval (CI), were calculated for all independent significant predictors of mortality (p < 0.05). The statistical analysis was performed using the BMDP statistical software package (BMDP, Berkeley, CA). A similar analysis was conducted in which failure of NIV was considered as an independent variable. All tests were two-tailed. A value of p < 0.05 was considered significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Demographics

Data were completely collected in 42 of the 52 ICUs contacted for the study. A total of 1,337 patients were admitted over the 3-wk-period of the study; 689 patients required invasive or noninvasive mechanical ventilation and were included in the study.

NIV was attempted in 108 of 689 patients (16%) as a first-line treatment. Among the 581 (84%) patients receiving conventional treatment through ETI, 382 received tracheal intubation before admission to the ICU (66%); the incidence of NIV for patients admitted to the ICU without intubation was therefore 35%. The percentage of patients receiving NIV ranged from 0 (in eight ICUs) to 67% (in one ICU).

The conditions precipitating respiratory failure and the need for ventilation were classified into four groups, whose relative distributions in the study population was as follows: 48% of patients had hypoxemic ARF, including pneumonia (10%) and postoperative failure (15%); 15% had hypercapnic ARF; 30% had coma; and 7% had cardiogenic pulmonary edema.

Main demographic and physiologic characteristics at the start of mechanical ventilation for patients treated either with ETI or NIV are shown in Table 1. NIV was never used for patients in coma, and these patients (n = 201) were therefore not retained in the analysis, which was directed at comparing patients with similar etiologies treated with either NIV or ETI. SAPS II and McCabe/Jackson score, previous treatment with NIV, and the origin of the patient before hospitalization with ICU admission differed significantly between the two study groups. Concerning arterial blood gases, a significant difference in carbon dioxide tension (Pa_CO2) was found between the two groups. NIV was used in 14% of the hypoxemic ARF patients, in 27% of those with pulmonary edema, and in 50% of those with hypercapnic ARF (Figure 1).

View larger version (28K): [in this window] [in a new window]   Figure 1.   Distribution of the number of patients mechanically ventilated with ETI or NIV in the groups of patients with coma, hypercapnic respiratory failure, hypoxemic respiratory failure, and pulmonary edema.

Response to NIV

In 52 of 108 patients (48%) NIV was discontinued early (i.e., while the physician wished to continue it). Reasons for discontinuing NIV were inefficacy of the method in 84% of cases (following a lack of clinical [54%] or arterial blood gas improvement [46%]), inability to manage copious secretions in 32%, the patient's refusal to continue NIV in 22%, and full dependence on ventilatory support in 11%. Forty-three patients (40% of all NIV patients, 77% of the patients with early termination of NIV) eventually required endotracheal intubation at 6.3 ± 6.8 h (mean ± SD) after entry into the study. The other nine patients with early termination of NIV, but who did not require secondary ETI, had received NIV for a mean of 8.1 ± 6.4 h (range: 1 to 22 h) before ventilation was stopped. Figure 2 illustrates the distribution of patients according to treatment.

View larger version (25K): [in this window] [in a new window]   Figure 2.   Distribution of patients according to treatment and response to treatment. ETI = endotracheal intubation; NIV = noninvasive ventilation.

Predictors of Success of NIV

The principal characteristics of the NIV population according to the success or failure of this ventilatory mode are shown in Table 2. Arterial blood gas values before ventilation differed only for pH, but changes in arterial blood gases after NIV (at Day 1) did not differentiate responders and nonresponders. Pressure-support ventilation was the most frequent ventilatory mode used during NIV (67%), often in association (74%) with external positive end-expiratory pressure; assist-control ventilation was used in 15% of patients, and spontaneous breathing with continuous positive airway pressure in 18%. Neither the mode of ventilation nor the kind of ventilator used (76% of ICU ventilators and 24% of "home" devices) were found to influence the outcome of NIV. The semiquantitative scores relative to the toleration of NIV and the severity of air leaks were significantly different between the two groups.

A multiple regression analysis for the risk of intubation, including the factors used in the univariate analysis, found that the SAPS II score (OR = 1.05 per point; 95% CI: 1.02 to 1.08; p < 0.005) and poor toleration of NIV, included as a binary variable (OR = 1.68; 95% CI: 1.14 to 2.49; p < 0.01), were the two independent predictive factors of the need for mechanical ventilation.

The mean length of NIV according to the reason for mechanical ventilation was 6.3 ± 6.8 d (range: 1 to 29 d) in hypoxemic ARF, 5.6 ± 5.4 d (range: 1 to 24 d) in hypercapnic ARF, and 2.4 ± 2 d (range: 1 to 6 d) in pulmonary edema. The mean daily duration of NIV during the first week of ICU stay is shown in Figure 3.

View larger version (12K): [in this window] [in a new window]   Figure 3.   Mean daily duration of NIV during the first week of ICU stay for patients with hypercapnic respiratory failure, hypoxemic respiratory failure, and pulmonary edema.

Intubated Patients

A total of 624 patients received ETI, and 31 (5%) were eventually tracheostomized within 18.5 ± 12.5 d. Among patients who were extubated, the rate of reintubation was 47 of 374 (13%). Distribution of the reintubation rate, durations of ETI, and mortality rates according to the reasons for ventilation are shown in Table 3.

In order to determine whether the failure of NIV was a risk factor for death, nosocomial pneumonia, or prolonged length of mechanical ventilation or stay, we performed several analyses. First, survivors and deceased patients were compared among intubated patients, and failure of NIV was found in univariate analysis to have a similar incidence in the two groups, indicating that it was not a risk factor for death (21 of 243 cases, or 9%, among survivors, versus 21 of 173 cases, or 12%, among deceased patients, p = 0.3). The same analysis was repeated for patients with or without pneumonia, and again no difference was found in the incidence of failure of NIV (10 of 82 or 12% in patients with pneumonia, versus 31 of 331 or 9% in patients without pneumonia, p = 0.57). For the latter analysis (length of mechanical ventilation or ICU stay), the patients were separated into two groups based on the median value of length of ventilation (< 5 d versus  5 d) or the median value of length of stay (< 9 d versus  9 d). In this analysis, failure of NIV was significantly more frequent in patients with a long duration of mechanical ventilation or stay. In a stepwise multiple logistic regression analysis, prolonged mechanical ventilation was found to be independently associated with SAPS II (OR per SAPS II point = 1.02; CI: 1.01 to 1.04; p < 0.01) and failure of NIV (OR = 5.17; CI: 1.41 to 19.01; p < 0.02).

Complications and Outcome: Comparison of NIV and ETI

For the purpose of comparing patients without coma who received NIV or ETI, the incidence of nosocomial pneumonia is presented in Table 4 for the two groups. The rate of pneumonia was significantly lower for NIV patients. Ten of the 43 patients intubated after NIV failure (23%) developed pneumonia at a mean of 3.7 d (range: 0 to 13 d) after endotracheal intubation. The overall incidence of pneumonia was 19% in all patients requiring ETI and 2% in patients successfully treated with NIV, respectively (p < 0.002). The mortality rate was also significantly lower in the NIV group (22% versus 41%, respectively; p < 0.001) (Table 5).

The two study populations differed in their SAPS II score, which was higher in the ETI group; McCabe/Jackson score, which was higher in the NIV group; and in their origins before admission. After adjusting for these differences, a logistic regression analysis performed on this population found that the success of NIV was the main protective factor against nosocomial pneumonia (OR = 0.06; CI: 0.01 to 0.45; p < 0.0001), and also found that SAPS II (OR = 1.05 per point; CI: 1.04 to 1.06; p < 0.001), McCabe/Jackson score (OR = 2.18 per point; CI: 1.57 to 3.03; p < 0.001), and hypoxemic ARF (OR: = 2.30; 1.33 to 4.01) were the main risk factors for mortality, whereas success of NIV was an independent predictor of survival (OR: = 0.16; CI: 0.05 to 0.54; p < 0.004).

Duration of Ventilation and of ICU Stay

The duration of ventilation and the length of ICU stay were significantly lower in the NIV group (13.9 ± 14.5 d in the ETI versus 8 ± 6.3 d in the NIV group, p < 0.002; and 7.8 ± 9.8 d in the ETI versus 5.1 ± 5.7 d in the NIV group, p < 0.04, respectively). In the hypercapnic respiratory failure subgroup, the lengths of stay and of ventilation were also significantly higher in the ETI than in the NIV group (p = 0.03 and p = 0.0004, respectively).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Evidence now exists to support the efficacy of noninvasive ventilation (NIV). Several prospective controlled studies have demonstrated that NIV can reduce the need for intubation, decrease the duration of endotracheal ventilation, and reduce mortality during ARF in patients with chronic obstructive pulmonary disease (COPD) (1, 3, 4). It has also been suggested that NIV could be used in hypoxemic respiratory failure and be potentially helpful in a wide range of clinical situations including asthma, pneumonia, cardiogenic pulmonary edema, weaning, and respiratory failure after extubation (4, 5, 14). However, some negative results and drawbacks related to the human cost and excessive workload induced by NIV have also been emphasized (8, 9, 21). These factors may limit the clinical application of this ventilatory strategy in routine practice. Moreover, positive results of NIV obtained in concealed but nonblinded prospective trials, in which greater attention was focused on assuring success, may differ from results when routine implementation of the technique is performed. Few data are available about the practical use of NIV on a large scale. For these reasons, the current extent of use and effectiveness of NIV in clinical practice needed to be assessed.

Pennock and colleagues tested the hypothesis that the effective results of NIV obtained with a research team could be transferred to usual care providers (18). The research team (special care) initially administered NIV to 31 patients, and both the research team and the usual care personnel managed the next 45 patients; only the usual care providers managed a further 34 patients (usual care). The proportions of successful outcomes were similar during usual and special care, suggesting that NIV was efficient for current use. Meduri and colleagues have addressed the same issue in a series of 158 patients noninvasively ventilated in a university medical center (14). A simple management protocol designed for routine NIV application was implemented in the framework of an educational program by a dedicated respiratory therapist. NIV was found to be effective in avoiding ETI in 65% of the patients. In this series, improvement in pH and Pa_CO2 predicted success. These authors concluded that application of NIV in clinical practice is safe and effective. Alsous and associates described their experience with NIV in a retrospective observational study of 80 patients conducted at a community teaching hospital over a 3-yr-period (22). In this series, less than half of the patients were treated for hypercapnic ARF, and 63% of these patients did not require ETI. The success of NIV was greater in patients with hypercapnic ARF than in patients with hypoxemic ARF. A recent study performed in the United Kingdom also showed that NIV could be performed successfully in respiratory or general wards for patients with COPD (23). Guérin and coworkers reported a prospective observational study of 320 patients requiring mechanical ventilation, conducted over a 2-yr period in a single ICU at a university teaching hospital. Of the 98 patients who initially received NIV, 39% required ETI, a percentage similar to the results of our study (24). The incidence of ventilator-associated pneumonia was lower in the NIV group. Nourdine and associates reported similar results, again collected in a single ICU (25). Using a stepwise multiple regression analysis, they found that NIV was associated with a decrease in nosocomial pneumonia and in all infection as well (25).

In our survey, all the patients admitted to 42 ICUs during a 3-wk period of study and requiring ventilatory support were followed. NIV represented the first-line intervention for ARF in 16% of patients. However, many of the patients admitted to the ICU had already been intubated in the emergency department or out of the hospital, or were transferred from other ICUs. This may explain why the incidence of NIV in hypercapnic respiratory failure did not exceed 50%, despite the favorable results reported in the literature for this group of patients (1, 3, 11). The incidence of use of NIV reported in the present study may have been influenced by the selection of the centers participating in the survey. This selection was based on a voluntary participation of centers belonging to a working group of the SRLF. The selected centers were larger, and a greater number belonged to university hospitals than is the case for all ICUs in France (26). This may overestimate the frequency of use of NIV among smaller, nonuniversity hospitals. No correlation, however, was found between the frequency of NIV per center and the failure rate. Differences in demographic and physiologic characteristics between the NIV and ETI groups may explain the choice between the two modes of ventilation: patients with the highest SAPS II were more frequently intubated, whereas older patients and those with the poorest life expectancy based on the McCabe/Jackson score were more likely to be treated with NIV. This is well explained by the more frequent proposal of NIV to patients in whom invasive mechanical ventilation is not desirable, or who refuse it (27, 28). In addition, patients treated with NIV more frequently came from medical wards and less frequently from surgical wards.

In our survey, NIV was followed by ETI in 40% of cases. Comparison of successes and failures showed a significant difference in some variables that could explain the failure of NIV, such as encephalopathy score, pH at admission, SAPS II, clinical toleration of NIV, and the presence of air leaks. Both a high severity score and the presence of leaks have previously been shown to predict failure (2, 29). The mode of ventilation, type of mask and ventilator used (home versus ICU ventilators), and improvement in gas exchange on the first day of ventilation did not differ between the two groups. A multiple regression analysis for prediction of NIV failure found that the severity score, SAPS II, and poor toleration of NIV were independent risk factors for secondary intubation. The higher SAPS II score indicates that the most severely ill patients, especially in terms of associated organ dysfunctions, are poor candidates for NIV, as already shown by others (2). Severity scores are usually calculated at the end of the first 24 h, however, and cannot be used as such on an individual basis. A semiquantitative score for the patient's clinical toleration of the treatment was an important predictive factor of success. Such a gross evaluation does not allow differentiation of the patient component from the component having to do with quality of the equipment or adequacy of settings.

Causes precipitating respiratory failure were differently distributed in the two study groups of patients. Patients with coma represented 35% of the ETI population, and none of these patients received NIV, since inability to protect the upper airway is a contraindication to NIV. Patients with coma were therefore not included in the comparisons of the effect of ETI and NIV on incidence of pneumonia or outcome.

Pneumonia has been reported as a rare complication of NIV (2, 14, 24, 25). A recent case-control study (30) comparing 56 consecutive cases treated with NIV and 56 matched intubated cases found a pneumonia rate significantly lower in the NIV group (4% versus 23%, p < 0.01). In our survey we found a significant difference in the rate of nosocomial pneumonia in noncomatose patients. In this study, patients with COPD represented only 48% of the NIV group, and a separate analysis in this subgroup also found a significantly lower incidence of pneumonia in the NIV-treated patients. The diagnosis of nosocomial pneumonia was usually not made with similar techniques in the NIV and ETI patients. The rationale for obtaining protected brush specimens with quantitative cultures is based on the presence of colonization of the large airways in intubated patients (31, 32). This rationale therefore does not apply to nonintubated patients. The majority of cases of pneumonia in our NIV group, however, occurred in patients in whom NIV failed and who required ETI. In these patients, the diagnosis of pneumonia was therefore based on similar criteria to those for the ETI group.

The mortality rate was significantly different in our two study populations. We found SAPS II, McCabe/Jackson score, and hypoxemic respiratory failure to be risk factors for mortality, and NIV, when successful, to be an independent factor associated with survival. No significant difference was observed in the hypoxemic subgroup (38% NIV versus 44% ETI), which is in agreement with findings reported in literature (2, 7, 8). Success of NIV was also found to be a protective factor against pneumonia. When NIV alone was introduced in an intention-to-treat analysis, it did not influence the results. Therefore, the interesting positive results observed with successful NIV are based on the assumption that every patient treated with NIV had a real need for ventilatory support, and that for respiratory failure of the same degree of severity and type, sucessful treatment with NIV instead of ETI improves survival. This is a limitation of this analysis.

Failure of NIV could potentially represent a risk factor for complications. The analysis performed suggests that it did not influence nosocomial pneumonia or death. It was, however, significantly associated with a longer duration of both ventilation and ICU stay. This should make the clinician cautious in prolonging the use of NIV over periods of days when criteria for improvement are not rapidly met. Delaying intubation in such cases may not benefit the patient.

In conclusion, in the 42 ICUs participating in the study, NIV represented the first attempted ventilatory technique in 16% of ARF patients and in 35% of patients admitted without ETI. NIV failure (ETI) occurred in 40% of patients. Efforts at promoting information and education to avoid ETI as a first-line intervention whenever possible, and to improve the clinical toleration of NIV in the future, seem necessary.