A Comparison of Noninvasive Positive-Pressure Ventilation and Conventional Mechanical Ventilation in Patients with Acute Respiratory Failure

Study Design and Patient Selection

We enrolled consecutive adults with acute hypoxemic respiratory failure who were admitted to the intensive care unit of La Sapienza University Hospital in Rome. The patients were randomly assigned to receive either conventional mechanical ventilation with endotracheal intubation or noninvasive ventilation through a face mask. An ad hoc ethics committee approved the protocol, and all patients or their next of kin gave written informed consent.

The criteria for eligibility were acute respiratory distress that had deteriorated despite aggressive medical management, including severe dyspnea at rest as determined by a clinician who was not an investigator; a respiratory rate greater than 35 breaths per minute; a ratio of the partial pressure of arterial oxygen to the fraction of inspired oxygen (PaO₂ :FiO₂) of less than 200 while the patient was breathing oxygen through a Venturi mask; and active contraction of the accessory muscles of respiration or paradoxical abdominal motion. Patients with any of the following were excluded: chronic obstructive pulmonary disease according to previously defined criteria⁵; immunosuppressive therapy; a requirement of emergency intubation for cardiopulmonary resuscitation, respiratory arrest, severe hemodynamic instability, or encephalopathy; respiratory failure caused by neurologic disease or status asthmaticus; more than two new organ failures (e.g., the simultaneous presence of renal and cardiovascular failure)¹⁴; and tracheostomy, facial deformities, or recent oral, esophageal, or gastric surgery.

Medical management for the conditions causing acute respiratory failure was similar in the two groups. We used two types of mechanical ventilators: the Puritan Bennett 7200 (Puritan Bennett, Overland Park, Kans.) and the Servo 900 C (Siemens Elema, Uppsala, Sweden). The patients underwent continuous electrocardiography and monitoring of arterial oxygen saturation (Biox 3700, Ohmeda, Boulder, Colo.).

Conventional Ventilation

Patients assigned to the conventional-ventilation group underwent intubation with cuffed endotracheal tubes (internal diam-eter, 7.5 to 8.5 mm). The initial ventilator setting was in the assist–control mode, with a delivered tidal volume of 10 ml per kilogram of body weight and a respiratory rate of 14 to 18 breaths per minute, a positive end-expiratory pressure of 5 cm of water, and an FiO₂ of 0.8. Positive end-expiratory pressure was increased in increments of 2 to 3 cm of water up to 10 cm of water, until the FiO₂ requirement was 0.6 or less. Intravenous diazepam (0.2 mg per kilogram) or propofol (2 mg per kilogram) was given for sedation at the time of intubation; none of the patients received a paralyzing drug. The head of the bed was kept elevated at an angle of 45 degrees to minimize the risk of aspiration. When spontaneous breathing reappeared, the ventilator settings were changed to intermittent mandatory ventilation (rate, 4 to 7 breaths per minute) with pressure support (14 to 20 cm of water), adjusted to achieve a spontaneous tidal volume of 8 to 10 ml per kilogram, a respiratory rate of fewer than 25 breaths per minute, and the disappearance of accessory-muscle activity.¹⁵ All patients were weaned from the ventilator by reducing the level of pressure support by 4 cm of water twice and then decreasing the ventilatory rate by two breaths per minute at two-hour intervals, as tolerated. Patients who tolerated an intermittent-mandatory-ventilation rate of 0.5 breath per minute, a pressure-support level of 8 cm of water, and an FiO₂ of 0.5 or less had a two-hour T-piece trial.¹⁶ These patients then underwent extubation if they maintained a respiratory rate lower than 30 breaths per minute and a PaO₂ greater than 75 mm Hg.

Noninvasive Ventilation

For patients assigned to noninvasive ventilation, the ventilator was connected with conventional tubing to a clear, full-face mask with an inflatable soft-cushion seal and a disposable foam spacer to reduce dead space (Gibeck, Upplands, Sweden). The mask was secured with head straps to avoid an excessively tight fit, and the head of the bed was kept elevated at a 45-degree angle. In most cases, a hydrocolloid sheet was applied over the nasal bridge. For patients with a nasogastric tube, a seal connector in the dome of the mask was used to minimize air leakage. After the mask had been secured, pressure support was increased to achieve an exhaled tidal volume of 8 to 10 ml per kilogram, a respiratory rate of fewer than 25 breaths per minute, the disappearance of accessory-muscle activity (as evaluated by palpation of the sternocleidomastoid muscle),¹⁵ and patient comfort. Continuous positive airway pressure was increased by 2 to 3 cm of water repeatedly, up to 10 cm of water, until the FiO₂ requirement was 0.6 or less. Ventilator settings were adjusted on the basis of continuous oximetry and measurements of arterial-blood gases. The patients were not sedated.

The duration of ventilation was standardized according to the protocol of Wysocki et al.¹³ During the first 24 hours, ventilation was continuously maintained until oxygenation and clinical status improved. Subsequently, each patient was evaluated daily after breathing supplemental oxygen without ventilatory support for 15 minutes. Noninvasive ventilation was reduced progressively in accordance with the degree of clinical improvement and was discontinued if the patient maintained a respiratory rate lower than 30 breaths per minute and a PaO₂ greater than 75 mm Hg with an FiO₂ of 0.5, without ventilatory support.

In the patients randomly assigned to receive noninvasive ventilation, the criteria for switching them to endotracheal intubation and conventional ventilation were the failure to maintain a PaO₂ above 65 mm Hg with an FiO₂ of at least 0.6; the development of conditions necessitating endotracheal intubation to protect the airways (coma or seizure disorder) or to manage copious tracheal secretions; hemodynamic or electrocardiographic instability; or an inability on the part of the patient to tolerate the face mask because of discomfort. An attending physician who was not an investigator evaluated these criteria.

End Points and Definitions

The primary end points were the values for gas exchange and the frequency of complications of mechanical ventilation, including pneumonia, sepsis, and sinusitis. Arterial-blood gas values were determined at base line, at one hour, at four-hour intervals during mechanical ventilation, and before discontinuation of ventilatory support. Improvement in gas exchange was defined as the ability to increase the PaO₂ :FiO₂ ratio to more than 200 or an increase in this ratio of more than 100 from base line.¹² Improvement in gas exchange was evaluated within one hour after study entry (initial improvement) and over time (sustained improvement). Sustained improvement in gas exchange was defined as the ability to maintain the defined improvement in PaO₂ :FiO₂ until mechanical ventilation was discontinued, as confirmed by serial blood gas measurements.

Patients were monitored for the development of infections or other complications. Sepsis was defined as a systemic inflammatory response to an infectious process, with manifestations including tachycardia, tachypnea, hyperthermia, and hypothermia, and a high white-cell count; positive blood cultures were not required. Severe sepsis was diagnosed when sepsis was associated with evidence of organ dysfunction or hypoperfusion, such as altered mental status, metabolic acidosis, or oliguria. Severe sepsis associated with hypotension that was unresponsive to fluid therapy was referred to as septic shock.¹⁷

Patients in whom clinical manifestations of pneumonia developed, including radiographic evidence of persistent pulmonary infiltrates, hyperthermia or hypothermia, purulent tracheobronchial secretions, a high white-cell count, and worsening of pulmonary-gas exchange,^18,19 underwent bronchoscopy with bronchoalveolar lavage. The methods and laboratory procedures followed consensus guidelines.^20,21 Pneumonia was diagnosed when at least 100,000 colony-forming units of bacteria per milliliter were measured in bronchoalveolar-lavage fluids.²⁰ Since infections in patients receiving mechanical ventilation are frequently associated with the presence of an invasive device,¹⁸ an index of invasiveness was established by counting the number of devices (central venous, arterial, pulmonary arterial, and urinary catheters; drainage, endotracheal, and nasogastric tubes) per patient at entry to the study.

The secondary end points were survival, the duration of mechanical ventilation, and the duration of the stay in the intensive care unit. The criteria for the diagnosis of the acute respiratory distress syndrome were those of the American–European Consensus Conference.²² The simplified acute physiologic score was calculated 24 hours after admission to the intensive care unit.²³ This score takes into account 14 variables (age, heart rate, systolic blood pressure, body temperature, respiratory rate or need for ventilatory support, urinary output, white-cell count, hematocrit, Glasgow coma score, and serum glucose, potassium, sodium, bicarbonate, and urea nitrogen concentrations). A range of 0 to 4 is assigned for each variable (range of possible scores, 0 to 56). Higher scores indicate a higher risk of death; for instance, a score of 15 or 16 is associated with a mortality rate of approximately 32 percent, and for all scores of 21 or higher, mortality exceeds 80 percent.

Statistical Analysis

Results are given as means ±SD. Demographic and physiologic characteristics of the two groups were compared with use of Student's t-test for continuous data (separate estimates of variance were used when variances differed significantly) and with the Mantel–Haenszel extended chi-square test for categorical data. Fisher's exact test (two-tailed) was used when appropriate (when the expected number of cases per cell was below five). The SPSS package (SPSS, Chicago) was used for all analyses. The odds ratios, relative risks, and 95 percent confidence intervals are given with chi-square and P values to illustrate the amount of risk associated with some of the effects.²⁴

Table 1. Table 1. Base-Line Characteristics of the Patients and Causes of Acute Respiratory Failure.

Between April 1995 and March 1996, 486 patients were admitted to the intensive care unit; 295 had already undergone intubation, 19 underwent tracheostomy, and 95 had chronic obstructive pulmonary disease or were receiving immunosuppressive therapy. Of the 77 patients who met the entry criteria, 13 chose not to participate; thus, 64 were enrolled. Thirty-two patients were assigned to each group. The base-line characteristics of the two groups were similar (Table 1), with the exception that in the conventional-ventilation group the mean arterial pH was significantly lower (P=0.002) and more patients had a partial pressure of arterial carbon dioxide (PaCO₂) greater than 45 mm Hg (12 vs. 5 patients, P=0.05). The treatments for the conditions precipitating respiratory failure and the ventilatory requirements were similar. The mean level of applied positive end-expiratory pressure was similar (5.1±1.4 cm of water in the noninvasive-ventilation group and 5.3±1.2 cm of water in the conventional-ventilation group).

Figure 1. Figure 1. The Ratio of the Partial Pressure of Arterial Oxygen to the Fraction of Inspired Oxygen (PaO₂:FiO₂) at Base Line and after One Hour of Mechanical Ventilation in Patients with Acute Respiratory Failure in the Noninvasive-Ventilation and Conventional-Ventilation Groups.

A paired t-test was used for the statistical comparison. The degree of improvement in gas exchange after the start of mechanical ventilation was similar in the two groups. The values shown within the panels are means ±SD.

The patients in the two groups had a similar initial change in PaO₂ :FiO₂ (Figure 1). Within the first hour of mechanical ventilation, 20 patients (62 percent) in the noninvasive-ventilation group and 15 (47 percent) in the conventional-ventilation group had an improvement in PaO₂ :FiO₂ (P=0.21). The PaO₂ :FiO₂ improved over time in the 22 patients in the noninvasive-ventilation group who did not need intubation (116±25 at base line vs. 250±60 at the end of treatment, P=0.02) and in the 17 patients in the conventional-ventilation group who survived (126±25 at base line vs. 241±98 at the end of treatment, P=0.03). The change in PaCO₂ was similar in the two groups. Ten patients in the noninvasive-ventilation group (31 percent) required endotracheal intubation an average of 15±7 hours after entry into the study, but none required emergency intubation. The reasons for intubation were the failure of noninvasive ventilation to maintain the PaO₂ above 65 mm Hg (four patients), its inability to correct dyspnea (one patient), its inability to manage copious secretions (one patient), intolerance of noninvasive ventilation (two patients), and hemodynamic instability (two patients).

Overall, 42 patients, including 10 in the noninvasive-ventilation group, underwent intubation; 14 had orotracheal intubation, and 28 had nasotracheal intubation. The length of stay in the intensive care unit was shorter for patients in the noninvasive-ventilation group (9±7 days, vs. 16±17 days in the conventional-ventilation group; P=0.04). Nine patients in the noninvasive-ventilation group — all of whom required endotracheal intubation — and 15 in the conventional-ventilation group died in the intensive care unit. Thus, the rate of survival to discharge from the intensive care unit was 53 percent (17 patients) in the conventional-ventilation group and 72 percent (23 patients) in the noninvasive-ventilation group (odds ratio, 0.4; 95 percent confidence interval, 0.1 to 1.4; P=0.19). Two patients (one in each group) died in the hospital after discharge from intensive care. One, in the noninvasive-ventilation group, died of ventricular fibrillation. The other, in the conventional-ventilation group, died of cardiogenic shock due to a new myocardial infarction. The other patients were successfully discharged from the hospital without further complications.

Table 2. Table 2. Serious Complications and Complications Resulting in Death.

The complications and events leading to death are shown in Table 2. More patients in the conventional-ventilation group than in the noninvasive-ventilation group had serious complications (66 percent vs. 38 percent, P=0.02) and had pneumonia or sinusitis related to the endotracheal tube (31 percent vs. 3 percent, P=0.003). The rate of serious complications was higher in patients in the conventional-ventilation group. Among the patients in the noninvasive-ventilation group, 12 (38 percent) had serious complications after undergoing endotracheal intubation. One of the 12 patients in whom noninvasive ventilation failed had pneumonia diagnosed six days after undergoing endotracheal intubation.

Table 3. Table 3. Characteristics of Patients According to the Success or Failure of Noninvasive Ventilation and Survival or Death in the Conventional-Ventilation Group.

Further details of patient outcomes are shown in Table 3. In the noninvasive-ventilation group, there was a sustained improvement in gas exchange over time in 17 of the 22 patients who did not undergo intubation but in only 2 of the 10 patients who required intubation (P=0.003). Avoiding intubation was associated with a lower incidence of septic complications (P=0.006). Among the 40 patients who survived to be discharged from the intensive care unit, the patients in the noninvasive-ventilation group had a shorter duration of mechanical ventilation (3±3 vs. 6±5 days, P=0.006) and a shorter stay in the intensive care unit (6.6±5 vs. 14±13 days, P= 0.002) than those in the conventional-ventilation group.

Among the patients in the conventional-ventilation group, those who died had a higher simplified acute physiologic score (P=0.02). A post hoc subgroup analysis was performed for patients with simplified acute physiologic scores lower than 16 and for those with scores of at least 16. The 19 patients with simplified acute physiologic scores of at least 16 had similar outcomes regardless of the type of ventilation, whereas in the 45 patients with simplified acute physiologic scores lower than 16, noninvasive ventilation was superior to conventional ventilation (data not shown).

We found that, with similar ventilator settings, noninvasive ventilation was as effective as conventional ventilation in improving gas exchange in patients with acute hypoxemic respiratory failure. Furthermore, the rate of serious complications, in particular those related to intubation (pneumonia and sinusitis), was significantly lower in patients receiving noninvasive ventilation. Successful noninvasive ventilation was associated with shorter stays in the intensive care unit. Ten patients in whom noninvasive ventilation failed (31 percent of the group) required intubation despite an improvement in gas exchange; this finding agrees with a prior report.⁹

Noninvasive ventilation can relieve hypoxemic respiratory failure in patients with pneumonia, cardiogenic pulmonary edema, or postoperative complications.¹ Twenty-nine studies enrolling 748 patients described the successful application of noninvasive ventilation in patients with hypoxemic respiratory failure of various causes.¹ In one study, gas exchange improved when patients with hypoxemia had their endotracheal tubes replaced by noninvasive face masks with similar settings.²⁵

In our study, the noninvasive-ventilation and conventional-ventilation groups had received a similar number of invasive devices, not counting the endotracheal tubes in the patients in the conventional-ventilation group. Endotracheal intubation is the single most important predisposing factor for ventilator-associated pneumonia.²⁶ We identified pneumonia using bronchoscopic criteria in eight patients in the conventional-ventilation group (25 percent). The presence or absence of nosocomial pneumonia helps determine the outcome of respiratory failure.²⁶ Two patients in the conventional-ventilation group died of ventilator-associated pneumonia. The low rate of ventilator-associated pneumonia among patients in the noninvasive-ventilation group is in agreement with other findings.^1,5

As in other studies,^5,13 the outcome in patients with higher simplified acute physiologic scores was poor, irrespective of their randomization group. Among survivors, those in the noninvasive-ventilation group had a shorter duration of mechanical ventilation and a shorter stay in the intensive care unit. Factors that may have been involved in shortening the duration of mechanical ventilation include the avoidance of sedation, elimination of the extra work of breathing imposed by the endotracheal tube, the lower rate of ventilator-associated pneumonia, and earlier removal from ventilation. There were more patients with respiratory acidosis in the conventional-ventilation group, and this may have affected the duration of mechanical ventilation and the outcome in this group. However, in an earlier randomized study of patients without chronic obstructive pulmonary disease, patients with a PaCO₂ greater than 45 mm Hg who were treated with noninvasive ventilation had shorter durations of stay in the intensive care unit and a lower mortality rate than those treated with conventional ventilation.¹³

In conclusion, we found that noninvasive ventilation was as effective as conventional ventilation in improving gas exchange in patients with acute hypoxemic respiratory failure, and that when endotracheal intubation was avoided, the development of ventilator-associated pneumonia was unlikely.