Professional Documents
Culture Documents
ONLINE FIRST
Table 2.
Mild Moderate Severe
terion was removed from the defini- ARDS included the requirement of either definition and the draft Berlin
tion. Patients may qualify as having a low respiratory system compliance Definition.
ARDS as long as they have respiratory ("40 mL/cm H2O), a high VECORR (&10 Compared with patients from the
failure not fully explained by cardiac L/min), or both. These variables were population-based cohorts, patients from
failure or fluid overload as judged by identified for further study during the clinical trials and the academic cen-
the treating physician using all avail- evaluation phase. ters cohorts were younger, had more se-
able data. If no ARDS risk factor (eTable The panel considered a number of ad- vere hypoxemia, and had more opaci-
1) is apparent, some objective evalua- ditional measures to improve specific- ties on chest radiographs. The cohort
tion (eg, with echocardiography) is re- ity and face validity for the increased pul- of patients from the clinical trials had
quired to help eliminate the possibil- monary vascular permeability and loss the lowest mortality, likely reflecting the
ity of hydrostatic edema. of aerated lung tissue that are the hall- inclusion and exclusion criteria of the
Oxygenation. The term acute lung in- marks of ARDS, including CT scan- trials.31 The cohort of patients from aca-
jury as defined by the AECC was re- ning, and inflammatory or genetic mark- demic centers had the highest mortal-
moved, due to the perception that clini- ers (eTable 2). The most common ity and the lowest percentage of trauma
cians were misusing this term to refer to reasons for exclusion of these mea- patients, reflecting the referral popu-
a subset of patients with less severe hy- sures were lack of routine availability, lation (eTable 3).
poxemia rather than its intended use as lack of safety of the measure in criti- There were 269 patients in the physi-
an inclusive term for all patients with the cally ill patients, or a lack of demon- ological database with sufficient data to
syndrome. Positive end-expiratory pres- strated sensitivity, specificity, or both for classify ARDS by the AECC defini-
sure can markedly affect PaO2/FIO25,6; use as a defining characteristic for ARDS. tion, although the numbers of pa-
therefore, a minimum level of PEEP (5 Empirical Evaluation of the Draft tients in each cohort were small. Pa-
cm H2O), which can be delivered non- Definition. tients in the Turin cohort had worse
invasively in mild ARDS, was included Patients. A total of 4188 patients in the PaO2/FIO2 ratios and had higher mor-
in the draft definition of ARDS. A mini- clinical database had sufficient data to tality than the other studies (eTable 4).
mum PEEP level of 10 cm H2O was pro- classify as having ARDS by the AECC Evaluation of Ancillary Variables. The
posed and empirically evaluated for the definition. Of these patients, 518 (12%) draft Berlin Definition for severe ARDS
severe ARDS category. could not be classified by the draft Ber- that included a PaO2/FIO2 of 100 mm Hg
Additional Physiologic Measurements. lin Definition because PEEP was miss- or less, chest radiograph with 3 or 4
Compliance of the respiratory system ing or was less than 5 cm H2O. Pa- quadrants with opacities, PEEP of at least
largely reflects the degree of lung vol- tients who could not be classified by the 10 cm H2O, and either a CRS of 40 mL/cm
ume loss.2 Increased dead space is com- draft Berlin Definition had a mortality H2O or less or a VECORR of at least 10
mon in patients with ARDS and is asso- rate of 35% (95% CI, 31%-39%), a me- L/min identified a smaller set of pa-
ciated with increased mortality. 2 4 dian (interquartile range [IQR]) of 19 tients with identical mortality to the sim-
However, because the measurement of (1-25) ventilator-free days, and a me- pler severe ARDS category of PaO2/FIO2
dead space is challenging, the panel chose dian (IQR) duration of mechanical ven- of 100 mm Hg or less (TABLE 2). To ad-
minute ventilation standardized at a tilation in survivors of 4 (2-8) days. dress the possibility that the CRS and
PaCO2 of 40 mm Hg (VECORR = minute These patients were excluded from VECORR thresholds might be different in
ventilation % Pa CO 2 /40) as a surro- analyses of the draft Berlin Definition patients with higher body weight, we
gate.22 The draft definition of severe and comparisons between the AECC evaluated weight-adjusted cutoffs for
E4 Published online May 21, 2012 2012 American Medical Association. All rights reserved.
these variables in one of the cohorts. ate ARDS and 4% (95% CI, 3%-6%) pro- the results, the regression analysis was
There was no significant difference in the gressed to severe ARDS within 7 days; repeated without this cohort and
predictive validity of the weight- and 13% (95% CI, 11%-14%) of pa- yielded similar results.
adjusted criteria. The consensus panel re- tients with moderate ARDS at baseline The Berlin Definition performed simi-
viewed these results and considered the progressed to severe ARDS within 7 days. larly in the physiological database as in
lack of evidence for predictive validity of All differences between outcome vari- the clinical database (TABLE 5, eFigure
these ancillary variables and their po- ables across categories of modified AECC 1, and eFigure 2). Twenty-five percent
tential contribution to face validity and (ALI non-ARDS and ARDS alone) and (95% CI, 20%-30%) of patients met cri-
construct validity and decided to use the across categories of Berlin Definition teria for mild ARDS, 59% (95% CI, 54%-
simpler definition for severe ARDS that (mild, moderate, and severe) were sta- 66%) of patients met criteria for moder-
relied on oxygenation alone. tistically significant (P".001). ate ARDS, and 16% (95% CI, 11%-
The Berlin Definition. The final Ber- Compared with the AECC defini- 21%) of patients met criteria for severe
lin Definition of ARDS is shown in tion, the final Berlin Definition had bet- ARDS. Mortality increased with stages of
TABLE 3. Twenty-two percent (95% CI, ter predictive validity for mortality with ARDS from mild (20%; 95% CI, 11%-
21%-24%) of patients met criteria for an AUROC of 0.577 (95% CI, 0.561- 31%) to moderate (41%; 95% CI, 33%-
mild ARDS (which is comparable with 0.593) vs 0.536 (95% CI, 0.520- 49%) to severe (52%; 95% CI, 36%-
the ALI non-ARDS category of the 0.553; P".001), with the difference in 68%), with P = .001 for differences in
AECC definition; TABLE 4), 50% (95% AUROC of 0.041 (95% CI, 0.030- mortality across stages of ARDS. Me-
CI, 48%-51%) of patients met criteria 0.050). To ensure that missing PEEP dian (IQR) ventilator-free days de-
for moderate ARDS, and 28% (95% CI, data in one of the cohorts did not bias clined with stages of ARDS from mild
27%-30%) of patients met criteria for
severe ARDS. Mortality increased with
Table 3.
stages of ARDS from mild (27%; 95%
Acute Respiratory Distress Syndrome
CI, 24%-30%) to moderate (32%; 95%
Timing Within 1 week of a known clinical insult or new or worsening respiratory
CI, 29%-34%) to severe (45%; 95% CI, symptoms
42%-48%). Median (IQR) ventilator- Chest imaging a Bilateral opacitiesnot fully explained by effusions, lobar/lung collapse, or
free days declined with stages of ARDS nodules
from mild (20 [1-25] days) to moder- Origin of edema Respiratory failure not fully explained by cardiac failure or fluid overload
ate (16 [0-23] days) to severe (1 [0- Need objective assessment (eg, echocardiography) to exclude hydrostatic
edema if no risk factor present
20] day). Median (IQR) duration of me- Oxygenation b
chanical ventilation in survivors Mild 200 mm Hg " PaO2/FIO2 ! 300 mm Hg with PEEP or CPAP #5 cm H2O c
increased with stages of ARDS from Moderate 100 mm Hg " PaO2/FIO2 ! 200 mm Hg with PEEP #5 cm H2O
mild (5 [2-11] days) to moderate (7 [4- Severe PaO2/FIO2 ! 100 mm Hg with PEEP #5 cm H2O
14] days) to severe (9 [5-17] days). Abbreviations: CPAP, continuous positive airway pressure; FIO2, fraction of inspired oxygen; PaO2, partial pressure of
arterial oxygen; PEEP, positive end-expiratory pressure.
Using the Berlin Definition, 29% (95% a Chest radiograph or computed tomography scan.
b If altitude is higher than 1000 m, the correction factor should be calculated as follows: [PaO /FIO %(barometric pressure/
CI, 26%-32%) of patients with mild 760)].
2 2
ARDS at baseline progressed to moder- c This may be delivered noninvasively in the mild acute respiratory distress syndrome group.
Table 4.
Modified AECC Definition a Berlin Definition ARDS a
2012 American Medical Association. All rights reserved. Published online May 21, 2012 E5
Table 5.
Modified AECC Definition a Berlin Definition ARDS a
(8.5 [0-23.5] days) to moderate (0 [0- tients with severe ARDS that included were subjected to evaluation, these
16.5] days) to severe (0 [0-6.5] days), 15% of the entire ARDS population and parameters did not identify a group of
with P=.003 for differences in ventilator- had a mortality of 52% (95% CI, 48%- patients with higher mortality and were
free days across stages of ARDS. Me- 56%). Patients with severe ARDS who excluded from the final Berlin Defini-
dian (IQR) duration of mechanical ven- did not meet the higher-risk subset cri- tion after further consensus discus-
tilation in survivors increased with stages teria included 13% of the entire ARDS sion. Without this evaluation, a need-
of ARDS from mild (6.0 [3.3-20.8] days) population and had a mortality rate of lessly complex ARDS definition would
to moderate (12.0 [5.0-19.3] days) to se- 37% (95% CI, 33%-41%). The differ- have been proposed. However, static re-
vere (19.0 [9.0-48.0] days), with P=.045 ence between the mortality of patients spiratory system compliance and an un-
for differences in duration of mechani- with higher-risk severe ARDS and pa- derstanding of minute ventilation are
cal ventilation in survivors across stages tients with severe ARDS who did not important variables for clinicians to
of ARDS. meet these criteria was statistically sig- consider in managing patients with
Using the Berlin Definition, stages of nificant (P" .001). ARDS, even though those variables were
mild, moderate, and severe ARDS had in- not included as part of the defini-
creased mean lung weight by CT scan Comment tion.32
(1371 mg; 95% CI, 1268-1473; 1556 mg; Developing and disseminating formal The Berlin Definition addresses some
95% CI, 1474-1638; and 1828 mg; 95% definitions for clinical syndromes in of the limitations of the AECC defini-
CI, 1573-2082; respectively) and in- critically ill patients are essential for re- tion, including clarification of the ex-
creased mean shunt (21%; 95% CI, 16%- search and clinical practice. Although clusion of hydrostatic edema and add-
26%; 29%; 95% CI, 26%-32%; and 40%; previous proposals have relied solely on ing minimum ventilator settings, and
95% CI, 31%-48%; respectively). Com- the consensus process, this is to our provides slight improvement in pre-
parisons of lung weight and shunt (from knowledge the first attempt in critical dictive validity. Our study presents data
the single site providing these data) care to link an international consen- on the outcomes of patients with ARDS
across categories of modified AECC (ALI sus panel endorsed by professional so- defined according to the Berlin Defini-
non-ARDS and ARDS alone) and across cieties with an empirical evaluation. tion in a large heterogeneous cohort of
categories of Berlin Definition (mild, The draft Berlin Definition classi- patients including patients managed
moderate, and severe) were statistically fied patients with ARDS into 3 inde- with modern approaches to lung pro-
significant (P".001) (Table 5, eFigure pendent categories but relied on ancil- tective ventilation. Estimates of the
3, and eFigure 4). lary variables (severity of chest prevalence and clinical outcomes of
In a post hoc analysis, combining a radiograph, PEEP #10 cm H2O, CRS mild, moderate, and severe ARDS can
PaO2/FIO2 of 100 mm Hg or less with !40 mL/cm H 2 O, and V E CORR #10 be assessed from this database for re-
either a Crs of 20 mL/cm H2O or less or L/min) in addition to oxygenation to de- search and health services planning.
a VECORR of at least 13 L/min identified fine the severe ARDS group. When the Acute respiratory distress syndrome is
a higher-risk subgroup among pa- ancillary variables selected by the panel a heterogeneous syndrome with com-
E6 Published online May 21, 2012 2012 American Medical Association. All rights reserved.
plex pathology and mechanisms. The missing data seem unlikely because our Berlin Definition for ARDS may serve as
proposed definition does not resolve this results were robust to sensitivity analy- an example for linking consensus defi-
problem. Investigators may choose to de- ses that excluded individual cohorts. nition activities with empirical research
sign future trials using 1 or more of the Fourth, it is possible that the ancillary to better inform clinical care, research,
ARDS subgroups as a base study popu- variables did not identify a higher-risk and health services planning.
lation, which may be further refined subset because the number of quadrants Published Online:
using criteria specific to the putative on the chest radiograph cannot be as-
Authors/Writing Committee:
mechanism of action of the interven- sessed reliably, PEEP was not used in a
tion (eg, IL-6 levels for an antiIL-6 trial predictable fashion, or CRS and VECORR
or more stringent hypoxemia criteria for were not accurately measured. However,
a study on extracorporeal membrane if this is true, it is likely also to be true in
oxygenation). Furthermore, some vari- future studies and in clinical practice be-
ables that were excluded from the Ber- cause the study database was constructed
lin Definition because of current feasi- from clinical trial, academic, and com-
bility and lack of data on operational munity sites reflecting practice in the real
characteristics may become more use- world of clinical research. In addition, we
ful in the future. We anticipate that clini- evaluated PEEP and CRS as used by cli-
cal research using our model of defini- nicians in practice and not as a test of pre-
tion development will be used to revise specified ventilator settings that may be
the definition in the future. betterthanthevariablesevaluatedherein,
There are limitations to our ap- but may not be practical, particularly in
proach. First, although the Berlin Defi- observational cohort studies.5,6
nition had statistically significantly su- Fifth, because our study was not an
perior predictive validity for mortality exercise in developing a prognostic
compared with the modified AECC defi- model for ARDS, we only considered
nition, the magnitude of this difference the variables and cutoffs proposed by Author Contributions:
and the absolute values of the AUROC the consensus panel. We could not
are small and would be clinically unim- compare this definition directly to the
Study concept and design:
portant if the Berlin Definition was de- AECC definition because the catego-
signed as a clinical prediction tool. How- ries of that definition overlap. It is pos- Acquisition of data:
Analysis and interpretation of data:
ever, predictive validity for outcome is sible that the outcomes as well as the
only one criterion for evaluating a syn- relative proportion of patients within Drafting of the manuscript:
drome definition and the purpose of the each category of ARDS will change if Critical revision of the manuscript for important in-
Berlin Definition is not a prognostica- the underlying epidemiology of the syn- tellectual content:
tion tool.33 Although the Berlin Defini- drome evolves due to changes in clini-
Statistical analysis:
tion was developed with a framework in- cal practice or risk factors.34 This is par- Obtained funding:
cluding these criteria, we did not ticularly true for the post hoc higher- Administrative, technical, or material support:
empirically evaluate face validity, con- risk subset reported, for which the cut Study supervision:
tent validity, reliability, feasibility, or suc- points were derived from the data sets.
Conflict of Interest Disclosures:
cess at identifying patients for clinical trial
enrollment. Conclusion
Second, it is possible that our results In conclusion, we developed a consen-
are not generalizable because of the data sus draft definition for ARDS with an in-
sets we studied. This seems unlikely be- ternational panel using a framework that
cause patients from a broad range of focused on feasibility, reliability, and va-
populations, including clinical trials, aca- lidity. We tested that definition using em-
demic centers, and community pa- pirical data on clinical outcome, radio-
tients, were included in the analyses. graphic findings, and physiological
Third, some variables (eg, CRS and measures from 2 large databases con-
PEEP) were missing in some patients in structed from 7 contributing sources to
the data sets we used, either due to the assess the predictive value of ancillary
mode of mechanical ventilation that pre- variables, refine the draft definition, and
cluded their measurement or the prac- compare the predictive validity of the
ticalities of population-based research. definition to the existing AECC defini-
However, bias due to cohort selection or tion. This approach for developing the Members of the ARDS Definition Task Force:
2012 American Medical Association. All rights reserved. Published online May 21, 2012 E7
Online-Only Material:
Additional Contributions:
Provided data for the
empiric evaluation of the definition but were not part
of the consensus development:
Funding/Support:
REFERENCES
1. Health Measurement Scales. 24.
Chest
2. 13. N Engl J Med
Lancet
25.
3. 14.
N Engl J Crit Care Med
Med
Am J Respir Crit Care Med 15. 26.
4.
Biometrics
Crit Care Med Am J Respir Crit Care Med
27.
5. 16.
E8 Published online May 21, 2012 2012 American Medical Association. All rights reserved.