The new england journal of medicine

original article

Noninvasive Positive-Pressure Ventilation

for Respiratory Failure after Extubation
Andrés Esteban, M.D., Ph.D., Fernando Frutos-Vivar, M.D.,
Niall D. Ferguson, M.D., Yaseen Arabi, M.D.,
Carlos Apezteguía, M.D., Marco González, M.D., Scott K. Epstein, M.D.,
Nicholas S. Hill, M.D., Stefano Nava, M.D., Marco-Antonio Soares, M.D.,
Gabriel D’Empaire, M.D., Inmaculada Alía, M.D., and Antonio Anzueto, M.D.

abstract

background
From the Hospital Universitario de Getafe, The need for reintubation after extubation and discontinuation of mechanical ventilation
Madrid (A.E., F.F.-V., N.D.F., I.A.); the De- is not uncommon and is associated with increased mortality. Noninvasive positive-
partment of Medicine, Division of Respirol-
ogy, and the Interdepartmental Division of pressure ventilation has been suggested as a promising therapy for patients with respi-
Critical Care Medicine, University Health ratory failure after extubation, but a single-center, randomized trial recently found no
Network, University of Toronto, Toronto benefit. We conducted a multicenter, randomized trial to evaluate the effect of noninva-
(N.D.F.); King Fahad National Guard Hos-
pital, Riyadh, Saudi Arabia (Y.A.); Hospital sive positive-pressure ventilation on mortality in this clinical setting.
Profesor Posadas, Buenos Aires (C.A.);
Hospital General de Medellín, Medellín, methods
Colombia (M.G.); the Pulmonary and Crit-
ical Care Division, Department of Medicine, Patients in 37 centers in eight countries who were electively extubated after at least 48
New England Medical Center, Tufts Univer- hours of mechanical ventilation and who had respiratory failure within the subsequent
sity School of Medicine, Boston (S.K.E., 48 hours were randomly assigned to either noninvasive positive-pressure ventilation by
N.S.H.); the Respiratory Intensive Care
Unit, Fondazione S. Maugeri, Istituto Scien- face mask or standard medical therapy.
tifico di Pavia, Pavia, Italy (S.N.); Hospital
Universitario São José, Belo Horizonte, Bra- results
zil (M.-A.S.); Hospital de Clínicas, Caracas,
Venezuela (G.D.); and the University of Tex- A total of 221 patients with similar baseline characteristics had been randomly assigned
as Health Science Center, San Antonio to either noninvasive ventilation (114 patients) or standard medical therapy (107 pa-
(A.A.). Address reprint requests to Dr. Es- tients) when the trial was stopped early, after an interim analysis. There was no differ-
teban at the Unidad de Cuidados Intensi-
vos, Hospital Universitario de Getafe, Car- ence between the noninvasive-ventilation group and the standard-therapy group in the
retera de Toledo Km. 12,500, Getafe 28905, need for reintubation (rate of reintubation, 48 percent in both groups; relative risk in
Madrid, Spain, or at aesteban@ucigetafe. the noninvasive-ventilation group, 0.99; 95 percent confidence interval, 0.76 to 1.30).
com.
The rate of death in the intensive care unit was higher in the noninvasive-ventilation
N Engl J Med 2004;350:2452-60. group than in the standard-therapy group (25 percent vs. 14 percent; relative risk, 1.78;
Copyright © 2004 Massachusetts Medical Society. 95 percent confidence interval, 1.03 to 3.20; P=0.048), and the median time from res-
piratory failure to reintubation was longer in the noninvasive-ventilation group (12
hours vs. 2 hours 30 minutes, P=0.02).

conclusions
Noninvasive positive-pressure ventilation does not prevent the need for reintubation or
reduce mortality in unselected patients who have respiratory failure after extubation.

2452 n engl j med 350;24 www.nejm.org june 10, 2004

The New England Journal of Medicine

Downloaded from nejm.org on October 28, 2020. For personal use only. No other uses without permission.
Copyright © 2004 Massachusetts Medical Society. All rights reserved.
 noninvasive ventilation for respiratory failure after extubation

among patients who have respiratory failure within

t he process of discontinuing me-

chanical ventilation must balance the risk
of complications due to unnecessary delays
in extubation with the risk of complications due to
premature discontinuation and the need for reintu-
48 hours after elective extubation. We hypothesized
that a reduction in mortality would be mediated
through a decrease in the need for reintubation.

bation.1,2 Evidence-based guidelines recommend a methods

trial of spontaneous breathing to determine, in any
given patient, whether mechanical ventilation can The study protocol was approved by the institution-
be successfully discontinued3; with this approach, al ethics review board at each of the participating
the documented need for reintubation ranges from centers (see the Appendix). Written informed con-
13 to 19 percent.4-7 sent was obtained from the patients or their surro-
Patients who require reintubation have been not- gates at the time of their enrollment in the at-risk
ed to have a significantly higher mortality rate than cohort.
those who are successfully extubated on the first at-
tempt.8,9 The reasons for their increased risk of selection of patients
death may include both difficulties encountered Between November 1999 and May 2002, consecu-
during the reintubation period and the development tive patients in the intensive care units of 37 centers
of additional ventilator-related complications, such in eight countries were enrolled in a cohort of pa-
as pneumonia.10 The need for reintubation may also tients at risk of requiring reintubation (the at-risk
be a marker of increased severity of illness, but after cohort). To be enrolled, the patients had to be older
adjustment for coexisting conditions and severity of than 18 years of age, had to have undergone me-
illness, extubation failure is still an independent pre- chanical ventilation for more than 48 hours, and had
dictor of death.8 This suggests that, at least to some to have been successfully extubated after the com-
extent, the increased mortality seen in these patients pletion of a trial of spontaneous breathing. Exclu-
may be reduced by treatments aimed at reducing ei- sion criteria were the presence of a tracheostomy
ther the need for reintubation or its subsequent and the absence of informed consent.
complications. In all the participating intensive care units, the
Noninvasive positive-pressure ventilation has following procedures were in place for the discon-
been deemed by a recent international consensus tinuation of mechanical ventilation. Discontinua-
conference to be a promising therapy after failure tion was considered appropriate when all of the fol-
of extubation.11 This conference, which considered lowing criteria were met: the underlying cause of
the findings of physiological and nonrandomized acute respiratory failure had improved; the patient
studies,12,13 called for randomized, controlled trials was alert and able to communicate; the patient’s
with clinical end points to examine the value of non- core temperature was less than 38°C; there was no
invasive ventilation as a means of averting the need need for vasoactive drugs, with the exception of do-
for reintubation.11 Subsequently, the results of one pamine at doses lower than 5 µg per kilogram of
randomized, controlled trial, conducted while our body weight per minute; and the partial pressure of
study was being planned, were published. Keenan oxygen was greater than 60 mm Hg while the pa-
et al.14 reported no difference in either the rate of tient was breathing an inspired fraction of oxygen
reintubation or mortality with the use of noninva- of 0.40 or less with a positive end-expiratory pres-
sive positive-pressure ventilation as compared with sure of 5 cm of water or less. Discontinuation of
standard medical therapy in patients who had res- mechanical ventilation was performed by means of
piratory failure within 48 hours after extubation. one of the following techniques: a trial of sponta-
Their study was a relatively small, single-center tri- neous breathing (a single daily trial or multiple daily
al that evaluated the rate of reintubation as a primary trials of spontaneous ventilation with the use of a
end point. The extent to which its results can be gen- T tube, continuous positive airway pressure, flow-
eralized has been questioned.15 by, or pressure support of 5 to 8 cm of water) for up
We performed a large, multicenter, international to 120 minutes or a gradual reduction of pressure
study to determine whether noninvasive ventilation, support in steps of 2 cm of water every 2 hours until
as compared with standard medical therapy, would a pressure of 7 cm of water was reached. Patients
reduce the rate of death in the intensive care unit who successfully tolerated a trial of spontaneous

n engl j med 350;24 www.nejm.org june 10, 2004 2453

The New England Journal of Medicine

Downloaded from nejm.org on October 28, 2020. For personal use only. No other uses without permission.
Copyright © 2004 Massachusetts Medical Society. All rights reserved.
 The new england journal of medicine

breathing (according to published criteria)2-4 were subsequently adjusted as needed for the patient’s
subsequently extubated. comfort (notably, if there was a decrease in the res-
After extubation, patients were observed for 48 piratory rate and heart rate), while an adequate ar-
hours for the onset of respiratory failure, as defined terial oxygen saturation and an arterial pH above
by the presence of two or more of the following: res- 7.35 were maintained. The facial skin was assessed
piratory acidosis (defined as an arterial pH below every four hours to prevent damage from the tightly
7.35 with a partial pressure of arterial carbon di- fitting face mask used to deliver the ventilation.
oxide greater than 45 mm Hg), clinical signs sug- The patients in this group were encouraged to
gestive of respiratory-muscle fatigue or increased use noninvasive ventilation continuously for four-
respiratory effort (i.e., use of accessory muscles, hour periods. The face mask could be removed,
intercostal indrawing, or paradoxical motion of the however, for 15-to-20-minute periods to allow the
abdomen), a respiratory rate greater than 25 breaths patient to drink fluids or receive nursing care. The
per minute for two consecutive hours, and hypox- decision regarding when to discontinue noninva-
emia (defined as an arterial oxygen saturation of less sive ventilation was left to the attending physician.
than 90 percent or a partial pressure of arterial oxy- If respiratory failure subsequently developed, how-
gen of less than 80 mm Hg with a fraction of in- ever, noninvasive ventilation was restarted.
spired oxygen greater than 0.50).
Patients who met at least two of these criteria criteria for reintubation
were randomly assigned to either standard medical In both study groups, patients were reintubated if
therapy or noninvasive ventilation. The assignments they met at least one of the following criteria, as
were made with the use of a random-number table judged after they had undergone the assigned treat-
and opaque, sealed, numbered envelopes. Random- ment for at least 1 hour: lack of improvement in the
ization was performed with variable block sizes and pH or in the partial pressure of carbon dioxide;
was stratified according to the study center and the changes in mental status, rendering the patient un-
presence or absence of chronic obstructive pulmo- able to tolerate noninvasive ventilation; a decrease
nary disease. in the oxygen saturation to less than 85 percent, de-
spite the use of a high fraction of inspired oxygen;
standard medical therapy lack of improvement in signs of respiratory-muscle
Patients assigned to the standard-therapy group fatigue; hypotension, with a systolic blood pressure
received supplemental oxygen, respiratory physio- below 90 mm Hg for more than 30 minutes despite
therapy, bronchodilators, and any other therapies as adequate volume challenge, the use of vasopressors,
directed by the attending physician. These patients or both; or copious secretions that could not be ade-
could be reintubated or crossed over to receive non- quately cleared or that were associated with acidosis,
invasive ventilation if they met the prespecified cri- hypoxemia, or changes in mental status. The final
teria for reintubation (described below). In this decision to reintubate was made by the treating phy-
group, the application of noninvasive ventilation sician, who recorded the single most relevant rea-
was considered to indicate that standard medical son for reintubation from the list of six possible
therapy had failed. reasons.

noninvasive ventilation data collection

Patients assigned to the noninvasive-ventilation At baseline, demographic data, the Simplified Acute
group received ventilation through a full facial mask Physiology Score II (which can range from 0 to 163
from a ventilator located in the intensive care unit. points, with higher scores indicating greater impair-
Before ventilation was begun, the head of the pa- ment), the reason for mechanical ventilation, the
tient’s bed was positioned at a 45-degree angle. The duration of mechanical ventilation before extuba-
initial ventilator mode was pressure support, set to tion, and the method of discontinuation of mechan-
achieve a tidal volume of more than 5 ml per kilo- ical ventilation were recorded. Before extubation, the
gram of body weight and a respiratory rate of less spontaneous tidal volume, the maximal negative
than 25 breaths per minute. The fraction of inspired inspiratory pressure, and the respiratory rate were
oxygen and the positive end-expiratory pressure measured, and the rapid shallow breathing index
were titrated to maintain the arterial oxygen satura- (the ratio of the respiratory rate [expressed in
tion above 90 percent. The ventilator settings were breaths per minute] to tidal volume [expressed in

2454 n engl j med 350;24 www.nejm.org june 10 , 2004

The New England Journal of Medicine

Downloaded from nejm.org on October 28, 2020. For personal use only. No other uses without permission.
Copyright © 2004 Massachusetts Medical Society. All rights reserved.
 noninvasive ventilation for respiratory failure after extubation

Enrollment in at-risk cohort

(N=980)

Respiratory failure within

48 hr after extubation
(N=244)

Exclusion because of immediate need for reintubation

(N=23)
Decreased level of consciousness (N=8)
Increased respiratory effort (N=5)
Shock (N=4)
Hypoxemia (N=4)
Upper-airway obstruction (N=2)

Randomization
(N=221)

Noninvasive ventilation Standard medical therapy Crossover to noninvasive ventilation

(N=114) (N=107) (N=28)

Reintubation No reintubation Reintubation No reintubation Reintubation No reintubation

(N=55) (N=59) (N=51) (N=56) (N=7) (N=21)

Death Death Death Death Death Death

(N=21) (N=7) (N=11) (N=4) (N=1) (N=2)

Figure 1. Randomization and Patient Outcomes.

The numbers of patients with various outcomes in the crossover group are included among the numbers of patients with those outcomes
in the standard-therapy group.

liters], where a value greater than 105 breaths per solute reduction in the risk of death with the use of
minute per liter is associated with an increased risknoninvasive ventilation (relative to a mortality rate
of reintubation) was calculated. At the time of ran- of 33 percent in controls5) with a type I error of
5 percent and a power of 80 percent. One interim
domization, the respiratory rate, heart rate, arterial
blood gas values, and oxygen saturation were re- analysis by an independent data-monitoring board
corded. was planned after 200 patients had been recruited.
If needed, the time and reason for reintubation The interim analysis was based on a comparison of
were recorded. Every patient was followed until dis- the mortality in the two treatment groups with the
charge from the intensive care unit. The length of use of a normal approximation for a two-sided test,
stay in the intensive care unit and the vital status at
according to the following rules: first, if noninvasive
the time of discharge were recorded. ventilation was associated with a reduction in mor-
tality, the P value needed to stop the trial should
statistical analysis be less than 0.029 (Pocock’s value for one interim
We calculated that 194 patients would be required analysis)16; and second, if noninvasive ventilation
in each group to allow detection of a 13 percent ab- was associated with an increase in mortality, the

n engl j med 350;24 www.nejm.org june 10, 2004 2455

The New England Journal of Medicine

Downloaded from nejm.org on October 28, 2020. For personal use only. No other uses without permission.
Copyright © 2004 Massachusetts Medical Society. All rights reserved.
 The new england journal of medicine

Table 1. Baseline Characteristics of the Patients, According to Study Group.* results

Non- Standard A total of 980 patients who had been electively ex-
invasive Medical tubated after receiving ventilation for more than 48
Ventilation Therapy P
Characteristic (N=114) (N=107) Value
hours were enrolled in the at-risk cohort. Respirato-
ry failure developed within 48 hours after extubation
Age — yr 61±17 58±19 0.25 in 244 of them (25 percent; 95 percent confidence
Female sex — no. (%) 47 (41) 47 (44) 0.68 interval, 22 to 28 percent). Twenty-three patients
Simplified Acute Physiology Score II on 37±13 36±10 0.77 (four of whom died in the intensive care unit) did
admission† not undergo randomization because their clinical
Reason for initiation of mechanical ventilation 0.65 condition necessitated urgent reintubation. There-
Acute respiratory failure — no. (%) fore, 221 patients were randomly assigned to a study
Pneumonia 28 (25) 20 (19)
group — 114 to receive noninvasive ventilation and
107 to receive standard medical therapy (Fig. 1).
Postoperative respiratory failure 20 (18) 23 (21)
The groups did not differ significantly at baseline,
Sepsis 13 (11) 11 (10) at the time of extubation, or at the time of random-
Trauma 11 (10) 7 (7) ization (Tables 1, 2, and 3).
Cardiac failure 8 (7) 12 (11)
Acute respiratory distress syndrome 4 (4) 8 (7) primary outcome
Other 12 (11) 10 (9)
Mortality from all causes in the intensive care unit
was higher in the noninvasive-ventilation group
Acute-on-chronic respiratory failure —
no. (%) than in the standard-therapy group: 28 of the pa-
tients assigned to noninvasive ventilation died (25
Chronic obstructive pulmonary disease 14 (12) 9 (8)
percent; 95 percent confidence interval, 17 to 34
Asthma 1 (1) 3 (3)
percent), as compared with 15 of those assigned to
Neuromuscular disease — no. (%) 3 (3) 4 (4) standard medical therapy (14 percent; 95 percent
confidence interval, 8 to 23 percent; P=0.048). The
* Plus–minus values are means ±SD.
† The Simplified Acute Physiology Score II can range from 0 to 163 points.
absolute difference in mortality was thus 11 per-
A higher score indicates greater impairment. centage points (95 percent confidence interval, 0 to
21 percentage points). This difference corresponds
to a relative risk of death from all causes of 1.78 (95
P value needed to stop the trial should be less than percent confidence interval, 1.03 to 3.20) and a
0.05 (Pocock’s recommended value).16 In the inter- number needed to harm of 9 (95 percent confidence
im analysis, the observed mortality rate in the non- interval, 4.5 to 100) associated with noninvasive
invasive-ventilation group was higher than that in ventilation.
the standard-therapy group, and the P value was less This difference appeared to be due to differences
than 0.05 (P=0.048); the trial was therefore stopped in the rate of death among the patients who required
according to the second rule. The analysis was per- reintubation. Of the 55 patients in the noninvasive-
formed on an intention-to-treat basis. ventilation group who required reintubation, 21
Continuous variables are presented as means died (38 percent; 95 percent confidence interval,
±SD or as medians and interquartile ranges. Com- 26 to 52 percent); in comparison, 11 of the 51 pa-
parison of continuous variables between the two tients in the standard-therapy group who required
groups was conducted with the use of Student’s reintubation died (22 percent; 95 percent confi-
t-test for variables with a normal distribution and dence interval, 12 to 36 percent; P=0.06).
with the use of the Mann–Whitney U test for varia-
bles with a non-normal distribution. Results with secondary outcomes
respect to categorical variables are presented as pro- There was no difference in the rate of reintubation
portions with 95 percent confidence intervals and between the two groups: reintubation was necessary
were analyzed with the use of chi-square tests, ex- in 55 of the patients assigned to noninvasive venti-
cept when small samples required the use of Fish- lation (48 percent; 95 percent confidence interval,
er’s exact test. All reported P values are two-sided. 38 to 57 percent) and 51 of those assigned to stan-

2456 n engl j med 350;24 www.nejm.org june 10 , 2004

The New England Journal of Medicine

Downloaded from nejm.org on October 28, 2020. For personal use only. No other uses without permission.
Copyright © 2004 Massachusetts Medical Society. All rights reserved.
 noninvasive ventilation for respiratory failure after extubation

dard medical therapy (48 percent; 95 percent confi-

Table 2. Characteristics of the Patients at the Time of Extubation, According to
dence interval, 39 to 58 percent). The reasons for Study Group.*
reintubation were similar in the two groups (Table
4). However, the interval between the development Non- Standard
invasive Medical
of respiratory failure and reintubation was signifi- Ventilation Therapy P
cantly longer in the noninvasive-ventilation group Characteristic (N=114) (N=107) Value
(median, 12 hours; interquartile range, 2 hours Duration of mechanical ventilation — days 0.67
10 minutes to 28 hours) than in the standard-therapy
Median 7 8
group (median, 2 hours 30 minutes; interquartile
Interquartile range 4–12 4–13
range, 45 minutes to 16 hours 30 minutes; P=0.02).
Tidal volume — ml/kg 6±2 6±2 0.76
In a post hoc analysis of the 23 patients with
chronic obstructive pulmonary disease who were Maximum inspiratory pressure — cm of water 33±15 32±15 0.72
included in the study, we observed that the rate of Respiratory rate — breaths/min 24±7 23±7 0.20
reintubation was lower among those who had been Rapid shallow breathing index — breaths/min/ 62±31 58±30 0.33
liter†
assigned to noninvasive ventilation (7 of 14 [50 per-
cent]) than among those who had been assigned to Method of withdrawal from mechanical 0.99
ventilation — no. (%)
standard therapy (6 of 9 [67 percent], P=0.67), but
Single daily trial of spontaneous breathing 61 (54) 54 (50)
the sample was too small to allow us to draw mean-
Multiple daily trials of spontaneous breathing 19 (17) 19 (18)
ingful conclusions about this subgroup. Similarly,
in a post hoc examination of the data from the pa- Gradual reduction of pressure support 29 (25) 28 (26)
tients who had respiratory acidosis after extubation, Other 5 (4) 6 (6)
there was no significant difference between the
* Plus–minus values are means ±SD.
study groups in the rate of reintubation, which was † The rapid shallow breathing index is the ratio of the respiratory rate (expressed
52 percent (12 of 23 patients) in the noninvasive- in breaths per minute) to the tidal volume (expressed in liters). A value greater
ventilation group and 35 percent (6 of 17 patients) than 105 breaths per minute per liter is associated with an increased risk of re-
intubation.
in the standard-therapy group (P=0.29).
The difference between the groups in the length
of stay in the intensive care unit was also not signif-
icant. The median time was 18 days (interquartile Table 3. Characteristics of the Patients at the Time of Randomization,
range, 11 to 30) in the noninvasive-ventilation group According to Study Group.*
and 18 days (interquartile range, 11 to 26) in the Non- Standard
standard-therapy group (P=0.59). invasive Medical
Of the 114 patients assigned to noninvasive ven- Ventilation Therapy P
Characteristic (N=114) (N=107) Value
tilation, 5 did not tolerate the procedure, and 2 of
these 5 patients were reintubated. Of the 107 pa- Interval between extubation and respiratory 0.90
failure (hr)
tients assigned to standard medical therapy, 28 (26
Median 9 9
percent) received noninvasive ventilation as rescue
therapy; of these 28 patients, 7 (25 percent; 95 per- Interquartile range 3–21 2–21
cent confidence interval, 11 to 45 percent) subse- Respiratory rate (breaths/min) 29±7 29±6 0.70
quently required reintubation, and 3 died (Fig. 1). Heart rate (beats/min) 98±24 95±24 0.36
pH 7.39±0.09 7.39±0.08 0.48

discussion Partial pressure of arterial carbon dioxide 47±18 45±16 0.37

(mm Hg)
The main finding of this study is that noninvasive Partial pressure of arterial oxygen (mm Hg) 73±29 79±29 0.11
ventilation did not reduce mortality or the need for Oxygen saturation (%) 93±6 93±5 0.88
reintubation among patients receiving mechanical
ventilation who had respiratory failure after extuba- * Plus–minus values are means ±SD.
tion. The mortality rate tended to be higher among
the patients assigned to noninvasive ventilation than er with noninvasive ventilation than with standard
among those assigned to standard medical thera- therapy.
py, and the interval from the development of respi- Since the first report, in the late 1980s, of the
ratory failure to reintubation was significantly long- use of noninvasive ventilation instead of intubation

n engl j med 350;24 www.nejm.org june 10, 2004 2457

The New England Journal of Medicine

Downloaded from nejm.org on October 28, 2020. For personal use only. No other uses without permission.
Copyright © 2004 Massachusetts Medical Society. All rights reserved.
 The new england journal of medicine

ventilation was ineffective in preventing the need for

Table 4. Reasons for Reintubation, as Defined in the Protocol Guidelines,
According to Study Group.
reintubation in our study. First, the success of non-
invasive ventilation could be dependent on the ex-
Non- Standard perience of the health care team using the tech-
invasive Medical
Ventilation Therapy P nique. All the centers participating in this study,
Reason (N=55) (N=51) Value however, had incorporated noninvasive ventilation
no. (%)
into their routine clinical practices at least one year
before the start of the study. In addition, the find-
Lack of improvement in signs of muscle fatigue 25 (45) 23 (45) 0.97
ings in the study by Keenan et al.,14 in which a one-
Hypoxemia 9 (16) 15 (29) 0.11
year training period was undertaken before the start
Copious secretions 5 (9) 6 (12) 0.65
of the trial, were similar to ours. Second, the tim-
Lack of improvement in pH or partial pressure 8 (15) 3 (6) 0.13 ing of the initiation of noninvasive ventilation could
of carbon dioxide
be important. We chose to start noninvasive posi-
Changes in mental status 4 (7) 2 (4) 0.45
tive-pressure ventilation when respiratory failure
Hypotension 4 (7) 2 (4) 0.45
first developed. The only trial of noninvasive posi-
tive-pressure ventilation after extubation in which
the procedure was begun early, before the develop-
in patients with acute respiratory failure,17 this tech- ment of signs of respiratory failure, also failed to
nique has been used in a number of clinical situa- show positive results.26 Finally, the composition of
tions. The majority of studies examining noninva- the study population may influence the results. It
sive ventilation have assessed its role in averting the has been consistently shown that patients with
need for primary endotracheal intubation in patients chronic obstructive pulmonary disease benefit from
with acute respiratory failure.18-22 When the results noninvasive ventilation.23 Only approximately 10
of these studies were examined together in a recent percent of our patients had chronic obstructive pul-
meta-analysis, they showed that noninvasive venti- monary disease (a proportion consistent with that
lation is effective in reducing morbidity and mor- reported in previous observational studies),27,28 and
tality in patients with acute-on-chronic respiratory thus there may be a beneficial effect in these patients
failure but that the benefit in patients with hypox- that was not detected in our study.
emic respiratory failure is less clear.23 The use of An important observation from our study was
noninvasive ventilation to avert the need for reintu- that the rate of death in the intensive care unit ap-
bation in patients with respiratory failure after ex- peared to be greater among patients assigned to
tubation has been less extensively studied. Until very noninvasive ventilation than among those assigned
recently, the published literature addressing this to standard therapy. In a recent observational study
question consisted entirely of physiological case se- conducted by our group, we found that patients with
ries and nonrandomized studies.12,13,24-26 One ran- acute respiratory failure who were intubated after
domized, controlled trial examining the use of non- having first received noninvasive ventilation had a
invasive positive-pressure ventilation in patients higher mortality rate than those who were intubat-
with respiratory failure after extubation was con- ed without having received noninvasive ventilation
ducted while we were planning this study and was (48 percent vs. 31 percent, P=0.01).28 In the cur-
recently published.14 In that trial, Keenan and col- rent trial, the interval between the onset of respira-
leagues enrolled 81 patients in a single-center study tory failure and reintubation was significantly long-
and, as in the current study, found that the use of er in the noninvasive-ventilation group than in the
noninvasive ventilation did not significantly alter standard-therapy group. It is possible that this de-
the need for reintubation. These authors did not find lay in reintubation was the reason for the significant
differences between patients assigned to noninva- increase in the risk of death in the former group,
sive ventilation and those assigned to standard ther- through a number of mechanisms such as cardiac
apy in the rate of death either in the intensive care ischemia, increased respiratory muscle fatigue, as-
unit or in the hospital overall, but their study was piration pneumonitis, and complications of emer-
powered only to detect large differences in the rates gency intubation.
of reintubation.14 We conclude that noninvasive ventilation is not
There may be several reasons why noninvasive effective in averting the need for reintubation in un-

2458 n engl j med 350;24 www.nejm.org june 10 , 2004

The New England Journal of Medicine

Downloaded from nejm.org on October 28, 2020. For personal use only. No other uses without permission.
Copyright © 2004 Massachusetts Medical Society. All rights reserved.
 noninvasive ventilation for respiratory failure after extubation

selected patients in whom respiratory failure devel- Supported by a grant (00/315) from the Fondo de Investigación
Sanitaria and by grants (Red Gira G03/063 and Red Respira C03/11)
ops after extubation. In addition, noninvasive pos- from the Instituto de Salud Carlos III — both in Spain; and by a Ca-
itive-pressure ventilation does not improve survival nadian Institutes of Health Research–Canadian Lung Association
and may in fact be harmful. Although selected pa- Post-Doctoral Fellowship (to Dr. Ferguson).
We are indebted to Victor Abraira (Unidad de Bioestadística of
tients in specialized centers may benefit from this Hospital Ramón y Cajal, Madrid) and Miguel Angel de la Cal (Un-
therapy, specific hypotheses need to be tested pro- idad de Cuidados Intensivos, Hospital Universitario de Getafe,
spectively. Madrid) for statistical advice.

ap p e n d i x
The following investigators participated in the study: Argentina — S. Bauque and S. Giannasi (Hospital Italiano de Buenos Aires), L. Bettini
and A.R. Diez (Hospital Provincial del Centenario de Rosario), H.S. Canales (Hospital Interzonal General de Agudos General San Martin de
la Plata de Mar del Plata), M.F. Costa and H. Solar (Hospital Profesor Posadas de Haedo), P.M. Desmery and A. Gómez (Sanatorio Mitre de
Buenos Aires), P. Gómez and O. Yunk (Hospital Español de Buenos Aires), M. Grill and W. Vázquez (Hospital Español de Mendoza), E. Tur-
chetto (Hospital Privado de la Comunidad de Mar del Plata), and R. Valentín (Centro de Estudios Médicos e Investigaciones Clínicas de Bue-
nos Aires); Brazil — I.M. de Oliveira Rezende (Hospital Universitario São José, Belo Horizonte); Colombia — B. Gil (Clínica Medellín de
Medellín), M. Granados (Fundación Valle de Lily de Cali); A. Guerra (Hospital General de Medellín and Clínica Soma de Medellín), and F.
Molina (Clínica Universitaria Bolivariana de Medellín); Saudi Arabia — S. Haddad (King Fahad National Guard Hospital of Riyadh); Spain —
A. Abella and M. Prieto (Hospital Universitario de Getafe, Madrid), J.M. Allegue and S. Rodríguez Fernández (Hospital Santa Maria del
Rosell, Cartagena), S. Alonso and C. Boqué (Hospital Universitario Joan XXIII, Tarragona), A. Belenguer and T. Mut (Hospital General de
Castellón), S. Benito and A. Claramunt (Hospital de Santa Creu i Sant Pau, Barcelona), J. Blanco (Hospital del Río Hortega, Valladolid), J.L.
Buendía and J.A. Gómez Rubí (Hospital Virgen de la Arrixaca, El Palmar), R. Fernández Fernández and M.M. Fernández Fernández (Com-
plejo Hospitalari de Parc Taulí, Sabadell), J. Gener and R. Tomás (Hospital Germans Trias i Pujol, Badalona), S. Macias and F. Martínez Soba
(Hospital General de Segovia), and F. Esteban and I. Vallverdú (Hospital Universitari de Reus, Tarragona); United States — J. Houtchens and
T. Liesching (Rhode Island Hospital, Brown University Medical School, Providence, R.I.), A. Pelaez and D. Vines (University of Texas Health
Science Center, San Antonio), and N. Singh (New England Medical Center, Tufts University School of Medicine, Boston); Venezuela — M.
Capdevielle (Hospital Universitario de Caracas), J.M. España (Hospital Universitario de Caracas), A. Medina (Hospital Militar), F. Pérez
(Hospital de Clínicas de Caracas), and R.A. Zerpa (Hospital Militar).

references
1. Manthous CA. The anarchy of weaning fect of failed extubation on the outcome of 15. Girault C, Auriant I. Differences in suc-
techniques. Chest 2002;121:1738-40. mechanical ventilation. Chest 1997;112: cess rates of noninvasive ventilation. JAMA
2. Gil B, Frutos-Vivar F, Esteban A. Delete- 186-92. 2002;288:2540.
rious effects of reintubation of mechanically 9. Epstein SK, Ciubotaru RL. Independent 16. Pocock SJ. Clinical trials: a practical ap-
ventilated patients. Clin Pulm Med 2003;10: effects of etiology of failure and time to re- proach. Chichester, England: John Wiley,
226-30. intubation on outcome for patients failing 1983.
3. MacIntyre NR, Cook DJ, Ely EW Jr, et al. extubation. Am J Respir Crit Care Med 1998; 17. Meduri GU, Conoscenti CC, Menashe P,
Evidence-based guidelines for weaning and 158:489-93. Nair S. Noninvasive face mask ventilation in
discontinuing ventilatory support: a collec- 10. Torres A, Gatell JM, Aznar E, et al. Re- patients with acute respiratory failure. Chest
tive task force facilitated by the American intubation increases the risk of nosocomial 1989;95:865-70.
College of Chest Physicians, the American pneumonia in patients needing mechanical 18. Brochard L, Mancebo J, Wysocki M, et
Assocation for Respiratory Care, and the ventilation. Am J Respir Crit Care Med 1995; al. Noninvasive ventilation for acute exacer-
American College of Critical Care Medicine. 152:137-41. bations of chronic obstructive pulmonary
Chest 2001;120:Suppl:375S-395S. 11. International Consensus Conferences disease. N Engl J Med 1995;333:817-22.
4. Esteban A, Frutos F, Tobin MJ, et al. in Intensive Care Medicine: noninvasive 19. Kramer N, Meyer TJ, Meharg J, Cece RD,
A comparison of four methods of weaning positive pressure ventilation in acute respi- Hill NS. Randomized, prospective trial of
patients from mechanical ventilation. N Engl ratory failure. Am J Respir Crit Care Med noninvasive positive pressure ventilation in
J Med 1995;332:345-50. 2001;163:283-91. acute respiratory failure. Am J Respir Crit
5. Esteban A, Alia I, Gordo F, et al. Extuba- 12. Kilger E, Briegel J, Haller M, et al. Ef- Care Med 1995;151:1799-806.
tion outcome after spontaneous breathing fects of noninvasive positive pressure venti- 20. Wysocki M, Tric L, Wolff MA, Millet H,
trials with T-tube or pressure support venti- latory support in non-COPD patients with Herman B. Noninvasive pressure support
lation. Am J Respir Crit Care Med 1997;156: acute respiratory insufficiency after early ex- ventilation in patients with acute respiratory
459-65. [Erratum, Am J Respir Crit Care tubation. Intensive Care Med 1999;25:1374- failure: a randomized comparison with con-
Med 1997;156:2028.] 80. ventional therapy. Chest 1995;107:761-8.
6. Esteban A, Alia I, Tobin MJ, et al. Effect 13. Hilbert G, Gruson D, Portel L, Gbikpi- 21. Celikel T, Sungur M, Ceyhan B, Kara-
of spontaneous breathing trial duration on Benissan G, Cardinaud JP. Noninvasive kurt S. Comparison of noninvasive positive
outcome of attempts to discontinue me- pressure support ventilation in COPD pa- pressure ventilation with standard medical
chanical ventilation. Am J Respir Crit Care tients with postextubation hypercapnic res- therapy in hypercapnic acute respiratory
Med 1999;159:512-8. piratory insufficiency. Eur Respir J 1998;11: failure. Chest 1998;114:1636-42.
7. Ely EW, Baker AM, Dunagan DP, et al. 1349-53. 22. Antonelli M, Conti G, Rocco M, et al.
Effect on the duration of mechanical venti- 14. Keenan SP, Powers C, McCormack DG, A comparison of noninvasive positive-pres-
lation of identifying patients capable of Block G. Noninvasive positive-pressure ven- sure ventilation and conventional mechani-
breathing spontaneously. N Engl J Med tilation for postextubation respiratory dis- cal ventilation in patients with acute respira-
1996;335:1864-9. tress: a randomized controlled trial. JAMA tory failure. N Engl J Med 1998;339:429-35.
8. Epstein SK, Ciubotaru RL, Wong JB. Ef- 2002;287:3238-44. 23. Peters JV, Moran JL, Phillips-Hughes J,

n engl j med 350;24 www.nejm.org june 10, 2004 2459

The New England Journal of Medicine

Downloaded from nejm.org on October 28, 2020. For personal use only. No other uses without permission.
Copyright © 2004 Massachusetts Medical Society. All rights reserved.
 noninvasive ventilation for respiratory failure after extubation

Warn D. Noninvasive ventilation in acute tory support following extubation: avoid- tion review. Am J Respir Crit Care Med 2000;
respiratory failure — a meta-analysis up- ance of critical care unit costs. Chest 1999; 161:1450-8.
date. Crit Care Med 2002;30:555-62. 116:1025-8. 28. Esteban A, Anzueto A, Frutos F, et al.
24. Putensen C, Hörmann C, Baum M, 26. Jiang JS, Kao SJ, Wang SN. Effect of early Characteristics and outcomes in adult pa-
Lingnau W. Comparison of mask and nasal application of biphasic airway pressure on tients receiving mechanical ventilation: a
continuous positive airway pressure after the outcome of extubation in ventilator 28-day international study. JAMA 2002;287:
extubation and mechanical ventilation. Crit weaning. Respirology 1999;4:161-5. 345-55.
Care Med 1993;21:357-62. 27. Esteban A, Anzueto A, Alia I, et al. How Copyright © 2004 Massachusetts Medical Society.
25. Munshi IA, DeHaven B, Kirton O, Slee- is mechanical ventilation employed in the
man D, Navarro M. Reengineering respira- intensive care unit? An international utiliza-

2460 n engl j med 350;24 www.nejm.org june 10 , 2004

The New England Journal of Medicine

Downloaded from nejm.org on October 28, 2020. For personal use only. No other uses without permission.
Copyright © 2004 Massachusetts Medical Society. All rights reserved.