Review Articles
Noninvasive Ventilation and Survival in Acute Care
Settings: A Comprehensive Systematic Review and
Metaanalysis of Randomized Controlled Trials*
Luca Cabrini, MD; Giovanni Landoni, MD; Alessandro Oriani, MD; Valentina P. Plumari, MD;
Leda Nobile, MD; Massimiliano Greco, MD; Laura Pasin, MD; Luigi Beretta, MD; Alberto Zangrillo, MD
Objective: Noninvasive ventilation is increasingly applied to
prevent or treat acute respiratory failure, but its benefit on sur-
vival is still controversial for many indications. We performed a
metaanalysis of randomized controlled trials focused on the effect
of noninvasive ventilation on mortality.
Data Sources: BioMedCentral, PubMed, Embase, and the
Cochrane Central Register of clinical trials (updated December
31, 2013) were searched.
Study Selection: We included all the randomized controlled tri-
als published in the last 20 years performed in adults, reporting
mortality, comparing noninvasive ventilation to any other treatment
for prevention or treatment of acute respiratory failure or as a tool
allowing an earlier extubation. Studies with unclear methodology,
comparing two noninvasive ventilation modalities, or in palliative
settings were excluded.
Data Extraction: We extracted data on mortality, study design,
population, clinical setting, comparator, and follow-up duration.
*See also p. 927.
All authors: Department of Anesthesia and Intensive Care, IRCCS San
Raffaele Hospital, Vita-Salute University, Milan, Italy.
All authors developed the concept of this study. Drs. Landoni, Pasin,
Beretta, and Zangrillo developed the search strategy. Drs. Cabrini, Ori-
ani, Plumari, and Nobile retrieved the studies and extracted the data. Drs.
Landoni and Greco did the statistical analyses. The first draft of the article
was written by Drs. Cabrini and Pasin and was thoroughly revised by Drs.
Beretta and Zangrillo.
Supplemental digital content is available for this article. Direct URL cita-
tions appear in the printed text and are provided in the HTML and PDF
versions of this article on the journal’s website (http://journals.lww.com/
ccmjournal).
Supported, in part, by departmental funds. The Department of Anesthesia
and Intensive Care had no role in study design, data collection, data analy-
sis, data interpretation, or writing of the report.
Dr. Cabrini consulted for Medival (received honoraria for consultancy) and
disclosed a patent (applied for patent in a noninvasive ventilation [NIV]
device). Dr. Landoni has disclosed a patent (applied for patent in a NIV
device). The remaining authors have disclosed that they do not have any
potential conflicts of interest.
For information regarding this article, E-mail: cabrini.luca@hsr.it
Copyright © 2015 by the Society of Critical Care Medicine and Wolters
Kluwer Health, Inc. All Rights Reserved.
DOI: 10.1097/CCM.0000000000000819
880 www.ccmjournal.org
Data Synthesis: Seventy-eight studies were analyzed. Noninvasive
ventilation was associated with a reduction in mortality (12.6% in
the noninvasive ventilation group vs 17.8% in the control arm; risk
ratio = 0.73 [0.66–0.81]; p < 0.001; number needed to treat =
19 with 7,365 patients included) at the longest available follow-
up. Mortality was reduced when noninvasive ventilation was used
to treat (14.2% vs 20.6%; risk ratio = 0.72; p < 0.001; number
needed to treat = 16, with survival improved in pulmonary edema,
chronic obstructive pulmonary disease exacerbation, acute respira-
tory failure of mixed etiologies, and postoperative acute respiratory
failure) or to prevent acute respiratory failure (5.3% vs 8.3%; risk
ratio = 0.64 [0.46–0.90]; number needed to treat = 34, with sur-
vival improved in postextubation ICU patients), but not when used
to facilitate an earlier extubation. Overall results were confirmed for
hospital mortality. Patients randomized to noninvasive ventilation
maintained the survival benefit even in studies allowing crossover
of controls to noninvasive ventilation as rescue treatment.
Conclusions: This comprehensive metaanalysis suggests that
noninvasive ventilation improves survival in acute care settings.
The benefit could be lost in some subgroups of patients if nonin-
vasive ventilation is applied late as a rescue treatment. Whenever
noninvasive ventilation is indicated, an early adoption should be
promoted. (Crit Care Med 2015; 43:880–888)
Key Words: critical care; in-hospital mortality; mechanical
ventilation; metaanalysis; noninvasive ventilation; respiratory
failure
N
oninvasive ventilation (NIV) refers to the delivery of
positive pressure to the lungs without the insertion
of an endotracheal tube (1). It includes the delivery
of continuous positive airway pressure (CPAP) and all the
modalities of noninvasive positive pressure support/pressure-
controlled mechanical ventilation (noninvasive positive pres-
sure ventilation [NPPV]). The efficacy of NIV is mainly related
to its ability to improve lung volumes and to unload respira-
tory muscles. Furthermore, NIV avoids tracheal intubation–
associated risks (1).
April 2015 • Volume 43 • Number 4

In the last decades, NIV has been increasingly applied in a
growing number of acute care settings and conditions (1–3)
for prevention or treatment of acute respiratory failure (ARF).
Its efficacy has been evaluated in several randomized trials,
and it is currently considered a first-line treatment in some
common conditions such as chronic obstructive pulmonary
disease (COPD) exacerbation, cardiogenic pulmonary edema,
and pneumonia in immunosuppressed patients (1). However,
so far its benefit on survival has not yet been demonstrated
for many indications (especially in the most recent studied
fields such as the perioperative arena), and no recommen-
dation could be done out of few established indications (3).
Furthermore, most published metaanalyses on NIV did not
focus on the most relevant patient-centered outcome, which
is survival. Finally, a comprehensive metaanalysis including all
the applications of NIV in acute care settings either as a thera-
peutic or a preventive tool has never been performed.
The aim of the present study was to evaluate the impact
of NIV on mortality in all acute care settings, performing a
metaanalysis of randomized controlled trials (RCTs). As mor-
tality data can be greatly biased if crossover of patients from
the control group to the intervention group is allowed after
reaching predefined failure criteria, we took into account, for
the first time, this potential bias performing also separate anal-
yses of studies that allowed or not the crossover.
MATERIALS AND METHODS
Search Strategy
Four investigators (L.C., L.N., V.P.P., A.O.) independently
searched BioMedCentral, PubMed, Embase, and the Cochrane
Central Register of clinical trials (updated December 31,
2013) for pertinent articles. The full PubMed search strategy
(4) is available in Appendix 1. Furthermore, the investigators
scanned references of retrieved articles and pertinent reviews
to detect further studies.
Study Selection
The same four investigators independently examined at title/
abstract level the references obtained from database and litera-
ture searches with divergences resolved by consensus, and then,
if potentially relevant, retrieved the full-text articles.
To identify potentially relevant studies, we used the follow-
ing inclusion criteria: random allocation to treatment; articles
published in a peer-reviewed journal in the last 20 years (1994–
2013); and comparing NIV to another treatment for preven-
tion or treatment of ARF in any human clinical setting.
The exclusion criteria were as follows: duplicate publica-
tions; lack of data on mortality (at least two separate attempts
to contact the original authors were made in case of miss-
ing data); absent or unclear methods description; compari-
son of two NIV modalities; studies including only palliative
use of NIV (patients with expected low short-term survival);
patients less than 18 years old. Compliance to selection criteria
was checked independently by two investigators (G.L., M.G.),
resolving divergences by consensus.
Critical Care Medicine
Review Articles
Data Abstraction and Study Characteristics
Four investigators (L.C., L.N., V.P.P., A.O.) independently
abstracted baseline, procedural, and outcome data, with diver-
gences resolved by consensus. Specifically, we extracted data on
mortality, study design, population, clinical setting, compara-
tor, and follow-up duration. The primary endpoint of the pres-
ent systematic review was mortality at the longest follow-up
available. Secondary endpoints were hospital mortality and the
effect of crossovers on mortality.
Risk of Bias Assessment
Publication bias was assessed visually inspecting funnel plots
and by analytical appraisal based on Egger linear regression
test and Begg adjusted rank correlation test. When publication
bias was detected, we used the Trim-and-Fill method to handle
publication bias (5, 6). A two-sided p value of 0.10 or less was
regarded as significant in Peters or Begg test (5, 7). Sensitivity
analyses were performed by sequentially removing each study
and reanalyzing the remaining dataset (producing a new analy-
sis for each study removed), by meta-regression, and by per-
forming subanalyses on specific subsets (defined by setting, by
presence of crossovers between cases and controls, by year of
publications, and by comparator).
Data Analysis and Synthesis
Computations were performed with Stata (StataCorp, 2013,
Stata Statistical Software: Release 13, College Station, TX).
Hypothesis of statistical heterogeneity was tested using Cochran
Q test, with statistical significance set at the two-tailed 0.10
level. The extent of statistical consistency was measured by I2 of
Higgins and Thompson (8). Outcomes from individual studies
were analyzed in order to compute individual and pooled risk
ratios (RR) with pertinent 95% CIs. We used inverse variance
method with fixed effects model if low statistical inconsistency
was detected (I2 ≤ 25%) or with random effects model (to
better handle clinical and statistical variations) when moder-
ate or high statistical inconsistency were detected (I2 > 25%).
In addition, we computed number needed to treat (NNT) in
case of statistical significance. Cumulative metaanalysis was
performed in order to evaluate the trend in RR and the per-
formance of updated metaanalysis over time (9). To explore
the influence of length of follow-up and year of publication on
mortality, we performed univariate meta-regression analyses
of log-RR against these variables.
Unadjusted p values are reported throughout. Statistical
significance was set at the two-tailed 0.05 level for hypothesis
testing and at 0.10 for heterogeneity testing. This study was
performed in compliance with The Cochrane Collaboration
and the Preferred Reporting Items for Systematic Reviews and
Meta-Analyses guidelines (10, 11).
RESULTS
Database searches and bibliography scanning yielded a total of
1,322 articles. Excluding 1,189 nonpertinent titles or abstracts,
we retrieved 133 studies in the full-text form for complete
www.ccmjournal.org 881
Cabrini et al
assessment according to selection criteria (Supplemental Fig.
1, Supplemental Digital Content 1, http://links.lww.com/
CCM/B159). Fifty-five studies were excluded (references avail-
able from authors at request) because NIV was applied in all
groups; included only patients with expected low short-term
survival (12, 13); mortality data were not available and could
not be obtained by the authors; and methodology was unclear
or not reported. Ultimately, we identified 78 eligible random-
ized clinical trials for inclusion in the analyses (Supplemental
references 1–78 are available in the supplemental data, Supple-
mental Digital Content 1, http://links.lww.com/CCM/B159).
Study Characteristics
The 78 included trials randomized 7,365 patients (3,840 to
NIV and 3,525 to control). In 57 trials, NIV was used to treat
ARF: in four trials, the comparator was endotracheal intuba-
tion, whereas in all the others, NIV was compared to standard
oxygen therapy. In 14 trials, NIV was applied to prevent ARF:
in these studies, established ARF was not present at the time of
randomization, and NIV was mainly applied in the postopera-
tive period or in ICU after extubation, in unselected patients or
in patients at high-risk for subsequent ARF. In seven trials, NIV
was used to facilitate an earlier extubation in intubated ICU
patients who had not yet met the standard extubation crite-
ria: in this setting, NIV was applied after extubation in patients
who had failed at least one T-piece trial or to accelerate the
extubation before the criteria to attempt a T-piece trial were
reached; in all these studies, the controls remained intubated
after randomization, until extubation criteria were eventually
met. The characteristics and findings of the included trials
are reported in Supplemental Table 1 (Supplemental Digital
Content 1, http://links.lww.com/CCM/B159). Notably, every
year from 1995 to 2013, there was at least one RCT published
on NIV reporting survival data. Most trials (60%) were per-
formed in the ICU.
Quantitative Data Synthesis
Mortality at the Longest Follow-Up Available. Overall, the
use of NIV was associated with a significant reduction in
mortality (12.6% in the NIV group vs 17.8% in the control
arm; RR = 0.73; NNT = 19). Mortality reduction was con-
firmed when NIV was used to treat (RR = 0.72) or to prevent
(RR = 0.64) ARF, but not when NIV was used as a tool to
allow an earlier extubation (Table 1; Supplemental Figs. 2–4,
Supplemental Digital Content 1, http://links.lww.com/CCM/
B159). Subanalyses for each of these categories are reported in
Table 1. Figure 1 presents the forest plot for the effect of NIV
on hospital mortality and on mortality at longest follow-up
available.
We performed subanalyses of the most represented and rel-
evant subgroups. When NIV was applied to treat ARF, the ben-
eficial effect on survival was confirmed in several subgroups:
acute cardiogenic pulmonary edema (APE), COPD exacerba-
tion, ARF of various etiologies, and postoperative ARF. With
regard to preventive use of NIV, the effect on survival was lim-
ited to patients at high risk of postextubation ARF.
882 www.ccmjournal.org
Notably, patients randomized to receive NIV maintained
the survival benefit in studies that allowed controls to crossover
to NIV after reaching predefined failure criteria (Table 1), sug-
gesting that an early use of NIV is of fundamental importance.
Hospital Mortality. Hospital follow-up was reported in
59 trials, and the use of NIV was associated with a significant
reduction in hospital mortality (10.9% in the NIV group vs
16.9% in the control arm; RR = 0.69; NNT = 17) (Fig. 1). Sub-
analyses on hospital mortality are reported in Table 2 and in
Supplemental Figures 5–7 (Supplemental Digital Content 1,
http://links.lww.com/CCM/B159).
Sensitivity Analysis and Risk of Bias Assessment. Visual
inspection of funnel plots (Fig. 2; Supplemental Figs. 8–12,
Supplemental Digital Content 1, http://links.lww.com/CCM/
B159) and quantitative evaluation did not identify a skewed or
asymmetrical shape for most of the performed analyses exclud-
ing small publication bias. When a moderate small study bias
was identified by quantitative evaluation with Begg and Egger
tests or funnel plot analysis, the Trim and Fill approach was
used and confirmed the absence of small study bias (Supple-
mental Table 2, Supplemental Digital Content 1, http://links.
lww.com/CCM/B159).
Sensitivity analysis performed by sequential removal of
each included study confirmed the overall results and are
reported in Supplemental Table 2 (Supplemental Digital
Content 1, http://links.lww.com/CCM/B159). Sensitivity anal-
ysis by publication year, setting, and comparator is reported in
Supplemental Table 2 (Supplemental Digital Content 1, http://
links.lww.com/CCM/B159) and confirms overall results.
Furthermore, univariate meta-regressions for log-risk
mortality at longest follow-up available did not show any sig-
nificant effect of length of follow-up and of publication year
on mortality (Supplemental Figs. 13 and 14, Supplemental
Digital Content 1, http://links.lww.com/CCM/B159). A cumu-
lative metaanalysis, showing a constant effect of NIV over time,
is reported in Supplemental Figure 15 (Supplemental Digital
Content 1, http://links.lww.com/CCM/B159).
DISCUSSION
The present metaanalysis of 78 randomized trials supports
the hypothesis that NIV improves survival in most acute set-
tings when applied to treat or to prevent ARF. Furthermore,
it suggests that the survival benefit could be lost when NIV is
applied late, as a rescue treatment. Nevertheless, a metaanalysis
can generate but not prove hypotheses, and further large RCTs
are required.
This is the first metaanalysis that focuses on the survival
effect of NIV, embracing all acute care settings and trying
to overcome the bias related to patients’ crossover between
groups. To obtain a reliable evaluation of NIV impact on
survival, we included only RCTs published in the last 20
years and in which NIV was compared to other treatments
different from NIV. An original part of our study was to
evaluate if NIV improved survival in studies according to
the presence of crossover between groups. The fact that the
April 2015 • Volume 43 • Number 4
 Review Articles

Table 1. Effects of Noninvasive Ventilation on Mortality at the Longest Follow-Up

Available With Subanalyses in Different Settings
Purpose of Number p for Crossovera
Treatment and Events/ Events/ Relative Risk p for Needed Heterogene- No Crossovera p p (No. of
Disease Cases Controls [95% CI] Effect to Treat ity, I2 (%) (No. of Patients) Patients)

Treatment 385/2,717 476/2,314 0.72 < 0.001 16 0.22 (13) < 0.001 (2,873) 0.002 (2,158)b
[0.63–0.81]
Chronic obstructive 50/529 97/533 0.56 < 0.001 11 0.96 (0) 0.06 (254) < 0.001 (808)
pulmonary [0.42–0.74]
disease
Asthma 0/71 0/54 — — —
Acute pulmonary 170/1,278 148/890 0.80 0.04 30 0.53 (0) 0.04 (1,872) 0.66 (296)
edema [0.65–0.99]
 Out of hospital 15/195 17/193 0.88 0.69 0.95 (0) 0.88 (181) —
[0.45–1.70]
ARF 115/537 180/540 0.66 < 0.001 8 0.22 (20) 0.001 (378) 0.01 (699)b
[0.54–0.80]
 NIV vs 20/88 37/92 0.59 0.02 6 0.46 (0) — 0.05 (128)
 endotracheal [0.37–0.92]
tube
Treatment of 40/153 28/149 1.35 0.28b 0.19 (39) — 0.28 (312)b
postextubation [0.78–2.35]
ARF
Treatment of 10/149 23/148 0.51 0.02 11 0.45 (0) 0.02 (297) —
postoperative [0.28–0.92]
ARF
Prevention 48/906 76/919 0.64 0.01 34 0.42 (1) 0.96 (952) 0.04 (873)
[0.46–0.90]
Prevention of 33/403 55/408 0.65 0.03 19 0.32 (14) 0.30 (137)b 0.19 (674)b
 postextubation [0.44–0.96]
ARF
 High-risk 30/181 49/184 0.64 0.03 10 0.31 (14) — 0.15 (268)b
patients [0.43–0.97]
 Unselected 3/222 6/224 0.53 0.61b 0.13 (56) — —
patients [0.04–6.50]
Prevention of 11/490 14/498 0.81 0.72b 0.25 (26) 0.66 (789)b 0.34 (199)
 postoperative [0.26–2.53]
ARF
 Lung surgery 5/221 9/223 0.55 0.29 0.26 (20) 0.56 (444) —
[0.19–1.64]
 Obese patients 0/73 0/78 — — —
NIV used to 52/217 76/292 0.81 0.39 b
0.04 (55) — —
 facilitate [0.51–1.30]
an earlier
extubation
 Failed T-piece 49/160 66/239 0.91 0.70b 0.22 (35) — —
trial [0.56–1.47]
Overall 485/3,840 628/3,525 0.73 < 0.001 19 0.102 (18) < 0.001 (4,334) < 0.001
[0.66–0.81] (3,031)
ARF = acute respiratory failure, NIV = noninvasive ventilation.
a
Subanalyses according to trials with or without crossover between study groups after reaching failure criteria.
b
Random effects method.
Dashes indicate not applicable due to total zero events or data deriving from one study only.

Critical Care Medicine www.ccmjournal.org 883

Cabrini et al

Figure 1. Forest plot for the risk of mortality reporting both mortality at longest follow-up available (black) and hospital mortality (gray) within summary
estimate of subgroup analyses, overall analyses for the type of noninvasive ventilation (NIV) therapy, and total overall analyses. ARF = acute respiratory
failure, COPD = chronic obstructive pulmonary disease, LFU mortality = mortality at longest follow-up available, RR = risk ratio.

survival advantage for NIV treatment was present even when
crossover to NIV was allowed for controls suggests that NIV
used as a rescue treatment is less effective than early NIV
treatment. In other words, the group using NIV at an ear-
lier stage maintained its survival advantage even if the stan-
dard group used NIV later. The presence of a “window of
opportunity” was previously hypothesized by other authors
(1, 14) and for different etiologies of ARF, such as COPD
(Supplemental reference 57 is available in the supplemental
data, Supplemental Digital Content 1, http://links.lww.com/
CCM/B159) or acute respiratory distress syndrome (ARDS)
(Supplemental reference 77 is available in the supplemental
data, Supplemental Digital Content 1, http://links.lww.com/
CCM/B159).
In COPD exacerbations, NIV almost halves mortality when
compared to standard treatment, and it is currently recom-
mended as a first-line therapy (2, 3, 15, 16). COPD patients
are good responders to NIV even when they suffer from other
kinds of ARF such as community-acquired pneumonia or
when NIV is used to prevent postoperative or postextubation
ARF (1). Our metaanalysis confirmed the survival benefit of
NIV for COPD exacerbations and suggested that a late NIV
application is not useful (Tables 1 and 2).
NIV effect on survival in asthma is difficult to evaluate since
mortality is uncommon in this setting, and no death occurred in
the three RCTs that were included in the present metaanalysis.
Larger trials on asthma are required.
The largest RCT ever performed, the Three Interventions
in Cardiogenic Pulmonary Oedema trial (17), did not find a
significant survival benefit for NIV. However, two subsequent
metaanalyses on NIV in APE (18, 19), both including the 3CPO
trial, concluded that NIV improved survival. The efficacy of
NIV in APE has been widely evaluated. In this setting, beside
avoiding the complications associated with tracheal intuba-
tion and improving lung volumes and work of breathing, NIV
offers other advantages, including reduced cardiac preload and

884 www.ccmjournal.org April 2015 • Volume 43 • Number 4

 Review Articles

Table 2. Effects of Noninvasive Ventilation on Hospital Mortality With Subanalyses

Performed in Different Settings
Purpose of Number p for No Crosso-
Treatment and Events/ Events/ Relative Risk p for Needed to Heterogeneity, vera p (No. of Crossovera p
Disease Cases Controls [95% CI] Effect Treat I2 (%) Patients) (No. of Patients)

Treatment 181/1,619 292/1,558 0.64 < 0.001 13 0.53 (0) < 0.001 0.007 (1,566)
[0.54–0.75] (1,611)
Chronic obstructive 35/520 68/525 0.56 0.003 16 0.74 (0) 0.06 (254) 0.01 (791)
pulmonary [0.38–0.82]
disease
Asthma 0/43 0/92 — — —
Acute pulmonary 48/501 70/446 0.64 0.01 16 0.67 (0) 0.003 (651) 0.66 (296)
edema [0.45–0.90]
 Out of hospital 9/135 10/131 0.87 0.69 0.75 (0) — —
[0.37–2.06]
ARF 76/367 118/368 0.63 0.005 9 0.15 (29) 0.002 (337) 0.11 (398)b
[0.46–0.87]
 NIV vs 20/88 37/92 0.59 0.02 6 0.46 (0) — 0.05 (128)
 endotracheal [0.37–0.92]
tube
Treatment of 12/42 13/42 — — —
postextubation
ARF
Treatment of 10/149 23/148 0.51 0.02 11 0.45 (0) 0.02 (297) —
postoperative ARF [0.28–0.92]
Prevention 31/403 44/419 0.70 0.10 0.43 (0) 0.90 (425)b 0.11 (397)
[0.45–1.08]
Prevention of 25/201 43/204 0.63 0.04 12 0.62 (0) 0.30 (137)b 0.10 (268)
postextubation [0.40–0.99]
ARF
 High-risk patients 25/181 39/184 0.66 0.07 0.86 (0) — 0.10 (268)
[0.42–1.04]
 Unselected 0/20 4/20 — — —
patients
Prevention of 6/202 1/215 2.98 0.20 0.56 (0) 0.11 (288) —
postoperative [0.55–16.12]
ARF
 Lung surgery 1/40 0/44 — — —
NIV used to 28/171 40/245 0.89 0.76 b
0.09 (51) — —
facilitate an earlier [0.41–1.92]
extubation
 Failed T-piece 25/114 30/192 1.45 0.12 0.28 (21) — —
trial [0.91–2.31]
Overall 240/2,193 376/2,222 0.69 < 0.001 17 0.2 (14) < 0.001 0.002 (1,963)
[0.60–0.80] (2,452)
ARF = acute respiratory failure, NIV = noninvasive ventilation.
a
Subanalyses according to trials with or without crossover between study groups after reaching failure criteria.
b
Random effects method.
Dashes indicate not applicable due to total zero events or data deriving from one study only.

afterload (20). The lack of efficacy on mortality in the 3CPO supplemental data, Supplemental Digital Content 1, http://
trial was attributed to a low severity of patients and to the high links.lww.com/CCM/B159). In the present metaanalysis, NIV
rate of crossover from oxygen therapy to NIV in the control was associated with an improved survival, but the benefit was
group (Supplemental references 69 and 70 are available in the not confirmed in trials allowing crossover, with NIV used as

Critical Care Medicine www.ccmjournal.org 885

Cabrini et al
Figure 2. Funnel plot for the risk of mortality at longest follow-up avail-
able, by type of noninvasive ventilation therapy. RR = risk ratio.
rescue treatment (Tables 1 and 2). This finding, together with a
lack of survival benefit in trials applying NIV at the prehospital
stage (Supplemental references 29, 59, and 72 are available in
the supplemental data, Supplemental Digital Content 1, http://
links.lww.com/CCM/B159), might indicate that in APE an
early application of NIV is not fundamental. Further research
focusing on best timing for NIV treatment in APE is required.
Eighteen trials included patients with ARF of various eti-
ologies (including but not limited to COPD, APE, pneumonia,
asthma, pulmonary embolism, and trauma). The settings were
heterogeneous, including also immunocompromised patients
and prehospital acute care. NIV resulted effective in reducing
mortality also in this mixed population and even in a subgroup
of three trials including patients with severe ARF in which NIV
was compared to tracheal intubation. Nonetheless, it is impor-
tant to underline that caution should be exercised when severe
ARF is present, as in severe ARDS where a failure rate up to
80% was reported (1, 21), with 100% of NIV failure in patients
with ARDS presenting with shock (2). Furthermore, the het-
erogeneity of the populations in these studies makes their find-
ings less informative. Our analysis on trials allowing crossover
confirms that an early application in this setting is warranted
(Tables 1 and 2).
Only two RCTs evaluated NIV to treat established postex-
tubation ARF in the ICU (Supplemental references 22 and 37
are available in the supplemental data, Supplemental Digital
Content 1, http://links.lww.com/CCM/B159). The present
metaanalysis confirms previous studies (22–24) concluding
that NIV does not improve survival in this setting. On the con-
trary, we observed a significant reduction of mortality when
NIV was used to treat postoperative ARF.
A significant reduction in mortality was found when NIV
was applied after extubation in ICU to prevent ARF. The ben-
efit was present in all the RCTs evaluating patients at high risk
for postextubation failure (Supplemental references 24, 27,
and 51 are available in the supplemental data, Supplemental
Digital Content 1, http://links.lww.com/CCM/B159) while
controversial results were observed in unselected populations
886 www.ccmjournal.org
(Supplemental references 52 and 65 are available in the supple-
mental data, Supplemental Digital Content 1, http://links.lww.
com/CCM/B159). NIV did not reduce mortality when applied
to prevent postoperative ARF; the analyzed RCTs were per-
formed in obese patients or after cardiac, thoracic, or vascular
surgeries (Supplemental references 1, 7, 9, 23, 45, 56, 74, and 78
are available in the supplemental data, Supplemental Digital
Content 1, http://links.lww.com/CCM/B159). Only one trial
evaluated NIV during upper endoscopy and reported mortal-
ity data (Supplemental reference 11 is available in the supple-
mental data, Supplemental Digital Content 1, http://links.lww.
com/CCM/B159); even if high-quality data on NIV efficacy
during upper endoscopies are limited, the technique appears
quite promising (25). The same might be true for intraopera-
tive NIV applications (26).
NIV was applied in the liberation process from mechani-
cal ventilation with different aims (23, 24, 27). The present
metaanalysis suggests that NIV applied to facilitate an earlier
extubation has no impact on survival: one previous metaanal-
ysis including few RCTs reported the same result (23), whereas
a recent one with less stringent exclusion criteria found NIV
to be associated with mortality reduction, mainly in patients
with COPD (27). It should be noted that in patients extubated
despite one or more failed T-piece trials or even before reach-
ing the stage in which a T-piece trial is commonly attempted,
an increased mortality rate could be expected in the NIV group.
The fact that survival was not different might indicate that NIV
could be safely applied to obtain other positive outcomes, such
as reduced ICU length of stay.
Our metaanalysis presents strengths and limitations. All
analyzed articles were randomized and published in peer-
reviewed journals, and the methodology was clearly reported
and NIV was always compared to other treatments different
from NIV. This is the largest metaanalysis on NIV so far. The
elevated number of included trials allowed confirming the
beneficial effect of NIV on mortality in different settings. We
also divided studies allowing or avoiding crossover between
cases and controls, documenting that the effect of NIV on
survival is present (or higher) when it is applied at an early
stage compared to its use as rescue treatment. However, we
did not evaluate and analyze other potential relevant ben-
efits such as avoidance of nosocomial infections, reduction
in ICU length of stay, and lower hospital costs. The major
limitation of our study is the heterogeneity of the studies
with respect to inclusion and exclusion criteria, NIV tech-
niques, and severity of patients. This is a common limitation
for large metaanalyses. Nonetheless, it is worth noticing that
the effect of NIV on survival was confirmed in several sub-
groups and sensitivity analyses. Since in most cases mortality
data were not reported separately for hypoxemic, hyper-
capnic, or mixed ARF, an analysis focused on this clinically
relevant aspect was not possible. Another limitation is repre-
sented by the controlled condition in which the studies were
performed in contrast to daily practice (28). Furthermore,
individual experience with NIV is one of the most impor-
tant determinants of NIV effectiveness and the benefit on
April 2015 • Volume 43 • Number 4

survival can vary among different centers. However, several
observational studies in the “real world” supported NIV sur-
vival benefit in most settings (2, 24, 29). We included as NIV
both CPAP and NPPV use, even if CPAP does not provide
ventilation.
The findings of our metaanalysis have at least two relevant
clinical and organizational implications. First of all, NIV is
one of the few treatments with robust evidence supporting
its beneficial effect on survival in acute care settings, includ-
ing the postoperative period (30, 31); since NIV is relatively
cheap, its appropriate use should be vigorously promoted
worldwide (32). In particular, NIV appears greatly underuti-
lized in the postoperative population (23). Second, our data
suggest that an early use of NIV is more beneficial than its
use as rescue treatment, after failure of standard therapy. We
should consider that if we adopt NIV to treat more and more
patients and at an early stage, then a marked increase in the
number of treatments must be anticipated. Hospitals should
create the logistical, organizational, and educational frames
to allow for an increase in NIV treatments outside the ICUs,
as ICU beds could be largely insufficient for all patients who
could benefit from NIV (33, 34) (Supplemental reference 57
is available in the supplemental data, Supplemental Digital
Content 1, http://links.lww.com/CCM/B159). Safety and
cost-effectiveness should also be promoted and carefully
monitored.
CONCLUSIONS
The present large, comprehensive metaanalysis of RCTs sug-
gests that NIV improves survival in a wide range of acute
care settings. The benefit could be lost in some subgroups of
patients (in particular when treating COPD exacerbations)
if NIV is applied late as a rescue treatment. Research should
individuate the best organizational and educational inter-
ventions to promote an early, safe, and widespread adoption
of NIV whenever indicated.
REFERENCES
1. Nava S, Hill N: Non-invasive ventilation in acute respiratory failure.
Lancet 2009; 374:250–259
2. Hess DR: Noninvasive ventilation for acute respiratory failure. Respir
Care 2013; 58:950–972
3. Keenan SP, Sinuff T, Burns KE, et al; Canadian Critical Care Trials Group/
Canadian Critical Care Society Noninvasive Ventilation Guidelines
Group: Clinical practice guidelines for the use of noninvasive positive-
pressure ventilation and noninvasive continuous positive airway pres-
sure in the acute care setting. CMAJ 2011; 183:E195–E214
4. Robinson KA, Dickersin K: Development of a highly sensitive search
strategy for the retrieval of reports of controlled trials using PubMed.
Int J Epidemiol 2002; 31:150–153
5. Egger M, Davey Smith G, Schneider M, et al: Bias in metaanalysis
detected by a simple, graphical test. BMJ 1997; 315:629–634
6. Duval S, Tweedie R: Trim and fill: A simple funnel-plot-based method of
testing and adjusting for publication bias in metaanalysis. Biometrics
2000; 56:455–463
7. Dear KBG, Begg CB: An approach for assessing publication bias
prior to performing a metaanalysis. Stat Sci 1992; 7:237–245
8. Higgins JP, Thompson SG: Quantifying heterogeneity in a metaanaly-
sis. Stat Med 2002; 21:1539–1558
Critical Care Medicine
Review Articles
9. Lau J, Antman EM, Jimenez-Silva J, et al: Cumulative metaanalysis
of therapeutic trials for myocardial infarction. N Engl J Med 1992;
327:248–254
10. Higgins J, Green S: Cochrane Handbook for Systematic Reviews of
Interventions (v 5.0.2) 2009. Available at: http://www.cochrane-hand-
book.org. Accessed May 1, 2014
11. Liberati A, Altman DG, Tetzlaff J, et al: The PRISMA statement for
reporting systematic reviews and meta-analyses of studies that evalu-
ate healthcare interventions: Explanation and elaboration. BMJ 2009;
339:b2700
12. Nava S, Ferrer M, Esquinas A, et al: Palliative use of non-invasive ven-
tilation in end-of-life patients with solid tumours: A randomised feasi-
bility trial. Lancet Oncol 2013; 14:219–227
13. Azoulay E, Kouatchet A, Jaber S, et al: Noninvasive mechanical venti-
lation in patients having declined tracheal intubation. Intensive Care
Med 2013; 39:292–301
14. Liesching T, Kwok H, Hill NS: Acute applications of noninvasive posi-
tive pressure ventilation. Chest 2003; 124:699–713
15. Ram FSF, Picot J, Lightowler J, et al: Non-invasive positive pressure
ventilation for treatment of respiratory failure due to exacerbations of
chronic obstructive pulmonary disease. Cochrane Database Syst Rev
2004; 3:CD004104
16. Quon BS, Gan WQ, Sin DD: Contemporary management of acute
exacerbations of COPD: A systematic review and metaanalysis.
Chest 2008; 133:756–766
17. Gray A, Goodacre S, Newby DE, et al; 3CPO Trialists: Noninvasive
ventilation in acute cardiogenic pulmonary edema. N Engl J Med
2008; 359:142–151
18. Potts JM: Noninvasive positive pressure ventilation: Effect on mortal-
ity in acute cardiogenic pulmonary edema: A pragmatic metaanalysis.
Pol Arch Med Wewn 2009; 119:349–353
19. Weng CL, Zhao YT, Liu QH, et al: Metaanalysis: Noninvasive ventila-
tion in acute cardiogenic pulmonary edema. Ann Intern Med 2010;
152:590–600
20. Mehta S, Al-Hashim AH, Keenan SP: Noninvasive ventilation in
patients with acute cardiogenic pulmonary edema. Respir Care
2009; 54:186–195; discussion 195–197
21. Agarwal R, Aggarwal AN, Gupta D: Role of noninvasive ventilation in
acute lung injury/acute respiratory distress syndrome: A proportion
metaanalysis. Respir Care 2010; 55:1653–1660
22. Agarwal R, Aggarwal AN, Gupta D, et al: Role of noninvasive positive-
pressure ventilation in postextubation respiratory failure: A metaanaly-
sis. Respir Care 2007; 52:1472–1479
23. Glossop AJ, Shephard N, Shepherd N, et al: Non-invasive venti-
lation for weaning, avoiding reintubation after extubation and in
the postoperative period: A metaanalysis. Br J Anaesth 2012;
109:305–314
24. Hess DR: The role of noninvasive ventilation in the ventilator discon-
tinuation process. Respir Care 2012; 57:1619–1625
25. Cabrini L, Nobile L, Cama E, et al: Non-invasive ventilation during upper
endoscopies in adult patients. A systematic review. Minerva Anestesiol
2013; 79:683–694
26. Cabrini L, Nobile L, Plumari VP, et al: Intraoperative prophylactic
and therapeutic non-invasive ventilation: A systematic review. Br J
Anaesth 2014; 112:638–647
27. Burns KE, Meade MO, Premji A, et al: Noninvasive ventilation
as a weaning strategy for mechanical ventilation in adults with
respiratory failure: A Cochrane systematic review. CMAJ 2014;
186:E112–E122
28. Carson SS, Goss CH, Patel SR, et al; American Thoracic Society
Comparative Effectiveness Research Working Group: An official
American Thoracic Society research statement: Comparative effec-
tiveness research in pulmonary, critical care, and sleep medicine. Am
J Respir Crit Care Med 2013; 188:1253–1261
29. Schnell D, Timsit JF, Darmon M, et al: Noninvasive mechanical ventila-
tion in acute respiratory failure: Trends in use and outcomes. Intensive
Care Med 2014; 40:582–591
30. Landoni G, Rodseth RN, Santini F, et al: Randomized evidence for
reduction of perioperative mortality. J Cardiothorac Vasc Anesth 2012;
26:764–772
www.ccmjournal.org 887
Cabrini et al
31. Landoni G, Augoustides JG, Guarracino F, et al: Mortality reduction in
cardiac anesthesia and intensive care: Results of the first International
Consensus Conference. Acta Anaesthesiol Scand 2011; 55:259–266
32. Hess DR: How to initiate a noninvasive ventilation program: Bringing
the evidence to the bedside. Respir Care 2009; 54:232–243
APPENDIX 1. PubMed Research Query
(NIV[tiab] OR “non-invasive ventilation”[tiab] OR “noninva-
sive ventilation”[tiab] OR “Noninvasive Ventilation”[Mesh])
AND (randomized controlled trial[pt] OR controlled clinical
trial[pt] OR randomized controlled trials[mh] OR random
allocation[mh] OR double-blind method[mh] OR single-
blind method[mh] OR clinical trial[pt] OR clinical trials[mh]
888 www.ccmjournal.org
33. Cabrini L, Idone C, Colombo S, et al: Medical emergency team and
non-invasive ventilation outside ICU for acute respiratory failure.
Intensive Care Med 2009; 35:339–343
34. British Thoracic Society Standards of Care Committee: Non-invasive
ventilation in acute respiratory failure. Thorax 2002; 57:192–211
OR (clinical trial[tw] OR ((singl*[tw] OR doubl*[tw] OR
trebl*[tw] OR tripl*[tw]) AND (mask*[tw] OR blind[tw]))
OR (latin square[tw]) OR placebos[mh] OR placebo*[tw]
OR random*[tw] OR research design[mh:noexp] OR fol-
low-up studies[mh] OR prospective studies[mh] OR cross-
over studies[mh] OR control*[tw] OR prospectiv*[tw] OR
volunteer*[tw]))) NOT (animal[mh] NOT human[mh])
April 2015 • Volume 43 • Number 4