ARDS Berlin Definition - JAMA
ARDS Berlin Definition - JAMA
ARDS Berlin Definition - JAMA
ONLINE FIRST
The acute respiratory distress syndrome (ARDS) was defined in 1994 by the American-European Consensus Conference (AECC); since then, issues regarding the reliability and validity of this definition have emerged. Using a consensus process, a panel of experts convened in 2011 (an initiative of the European Society of Intensive Care Medicine endorsed by the American Thoracic Society and the Society of Critical Care Medicine) developed the Berlin Definition, focusing on feasibility, reliability, validity, and objective evaluation of its performance. A draft definition proposed 3 mutually exclusive categories of ARDS based on degree of hypoxemia: mild (200 mm Hg PaO2/FIO2 300 mm Hg), moderate (100 mm Hg PaO2/FIO2 200 mm Hg), and severe (PaO2/ FIO2 100 mm Hg) and 4 ancillary variables for severe ARDS: radiographic severity, respiratory system compliance (40 mL/cm H2O), positive endexpiratory pressure (10 cm H2O), and corrected expired volume per minute (10 L/min). The draft Berlin Definition was empirically evaluated using patientlevel meta-analysis of 4188 patients with ARDS from 4 multicenter clinical data sets and 269 patients with ARDS from 3 single-center data sets containing physiologic information. The 4 ancillary variables did not contribute to the predictive validity of severe ARDS for mortality and were removed from the definition. Using the Berlin Definition, stages of mild, moderate, and severe ARDS were associated with increased mortality (27%; 95% CI, 24%-30%; 32%; 95% CI, 29%-34%; and 45%; 95% CI, 42%-48%, respectively; P .001) and increased median duration of mechanical ventilation in survivors (5 days; interquartile [IQR], 2-11; 7 days; IQR, 4-14; and 9 days; IQR, 5-17, respectively; P .001). Compared with the AECC definition, the final Berlin Definition had better predictive validity for mortality, with an area under the receiver operating curve of 0.577 (95% CI, 0.561-0.593) vs 0.536 (95% CI, 0.520-0.553; P .001). This updated and revised Berlin Definition for ARDS addresses a number of the limitations of the AECC definition. The approach of combining consensus discussions with empirical evaluation may serve as a model to create more accurate, evidence-based, critical illness syndrome definitions and to better inform clinical care, research, and health services planning.
JAMA. 2012;307(23):2526-2533 Published online May 21, 2012. doi:10.1001/jama.2012.5669 www.jama.com
criteria for defining acute, sensitivity of PaO2/FIO2 to different ventilator settings, poor reliability of the chest radiograph criterion, and difficulties distinguishing hydrostatic edema (TABLE 1).4
*Authors/Writing Committee and the Members of the ARDS Definition Task Force are listed at the end of this article. Corresponding Author: Gordon D. Rubenfeld, MD, MSc, Program in Trauma, Emergency, and Critical Care, Sunnybrook Health Sciences Center, 2075 Bayview Ave, Toronto, ON M4N 3M5, Canada (gordon .rubenfeld@sunnybrook.ca).
For these reasons, and because all disease definitions should be reviewed periodically, the European Society of Intensive Care Medicine convened an international expert panel to revise the ARDS definition, with endorsement from the American Thoracic Society and the Society of Critical Care Medicine. The objectives were to update the definition using new data (epidemiological, physiological, and clinical trials) to address the current limitations of the AECC definition and explore other defining variables. Methods Consensus Process. Three co-chairs were appointed by the European Society of Intensive Care Medicine, who in turn selected panelists based on their work in the area of ARDS and to ensure geographic representation from both Europe and North America. An overview of the consensus process used by the panel is outlined in the FIGURE. In revising the definition of ARDS, the panel emphasized feasibility, reliability, face validity (ie, how clinicians recognize ARDS), and predictive validity (ie, ability to predict response to therapy, outcomes, or both). In addition, the panel determined that any revision of the definition should be compatible with the AECC definition to facilitate interpretation of previous studies. After initial preparations and an in-person consensus discussion, a draft definition was proposed,13 which underwent empirical evaluation. The definition was further refined through consensus discussion informed by these empirical data. Empirical Evaluation of Draft Definition. Cohort Assembly. Through the review of the literature presented at the consensus meeting, discussions with other experts, and review of personal files, the panel identified studies that met the following eligibility criteria: (1) large, multicenter prospective cohorts, including consecutive patients or randomized trials, or smaller, single-center prospective studies with unique radiological or physiological data that enrolled adult patients with ALI as defined by AECC;
Table 1. The AECC Definition3Limitations and Methods to Address These in the Berlin Definition
AECC Definition Acute onset All patients with PaO2/ FIO2 300 mm Hg PaO2/FIO2 300 mm Hg (regardless of PEEP) AECC Limitations No definition of acute4 Misinterpreted as PaO2/FIO2 = 201-300, leading to confusing ALI/ARDS term Inconsistency of PaO2/ FIO2 ratio due to the effect of PEEP and/or FIO25-7 Addressed in Berlin Definition Acute time frame specified 3 Mutually exclusive subgroups of ARDS by severity ALI term removed Minimal PEEP level added across subgroups FIO2 effect less relevant in severe ARDS group Chest radiograph criteria clarified Example radiographs created a PAWP requirement removed Hydrostatic edema not the primary cause of respiratory failure Clinical vignettes created a to help exclude hydrostatic edema Included When none identified, need to objectively rule out hydrostatic edema
Oxygenation
Chest radiograph
Bilateral infiltrates observed on frontal chest radiograph PAWP 18 mm Hg when measured or no clinical evidence of left atrial hypertension
PAWP
Poor interobserver reliability of chest radiograph interpretation8,9 High PAWP and ARDS may coexist10,11 Poor interobserver reliability of PAWP and clinical assesments of left atrial hypertension12
Risk factor
None
Abbreviations: AECC, American-European Consensus Conference; ALI, acute lung injury; ARDS, acute respiratory distress syndrome; FIO2, fraction of inspired oxygen; PaO2, arterial partial pressure of oxygen; PAWP, pulmonary artery wedge pressure; PEEP, positive end-expiratory pressure. aAvailable on request.
(2) studies collected granular data necessary to apply the individual criteria of both the draft Berlin Definition and the AECC definition; and (3) authors of these original studies were willing to share data and collaborate. The panel identified 7 distinct data sets (4 multicenter clinical studies for the clinical database14-17 and 3 single-center physiological studies for the physiological database18-20) that met these criteria. Further details of these studies are included in the eMethods (http://www.jama .com). Variables. Studies provided data on hospital or 90-day mortality. Ventilatorfree days at 28 days after the diagnosis of ALI were calculated as a composite measure of mortality and duration of mechanical ventilation. Duration of mechanical ventilation in survivors was selected as an indirect marker of severity of lung injury because this outcome is not biased by mortality or decisions
related to the withdrawal of lifesustaining treatments.21 Progression of severity of ARDS within 7 days was assessed using the longitudinal data collected within each cohort. We distinguished patients with more extensive involvement on the frontal chest radiograph (3 or 4 quadrants) from those with the minimal criterion of bilateral opacities (2 quadrants). Static compliance of the respiratory system (CRS) was calculated as tidal volume (mL) divided by plateau pressure (cm H2O) minus positive endexpiratory pressure (PEEP) (cm H2O). The corrected expired volume per min ECORR) was calculated as the meaute (V sured minute ventilation multiplied by the arterial partial pressure of carbon dioxide (PaCO2) divided by 40 mm Hg.22 Total lung weight was estimated from quantitative computed tomography (CT) images.23 Shunt was calculated at one site as previously reported.24
JAMA, June 20, 2012Vol 307, No. 23 2527
In-person discussions (September 30 to October 2, 2011, Berlin, Germany) Presentations of key background material Development of the conceptual model of ARDS Draft of Berlin Definition based on informal consensus discussions
Empirical evaluation of draft definition (October 2011 to January 2012) Assembling clinical and physiologic cohorts Demonstration of patient characteristics and distribution according to definition categories Evaluation of impact of ancillary variables for severe ARDS subgroup
Follow-up of consensus discussions and analysis (February 2012 by multiple teleconferences) Presentation of empirical evaluation Final definition created based on further informal consensus discussions Decision to present the results of a post hoc higher-risk subset Testing of predictive validity
Analytic Framework and Statistical Methods. The analytic framework for evaluating the draft Berlin ARDS Definition was to (1) determine the distribution of patient characteristics across the defined severity categories; (2) evaluate the value of proposed ancillary variables (more severe radiographic criterion, higher PEEP levels, static respiratory compliance, and ECORR) in defining the severe ARDS V subgroup in the draft definition; (3) determine the predictive validity for mortality of the final Berlin Definition; and (4) compare the final Berlin Defini2528 JAMA, June 20, 2012Vol 307, No. 23
tion to the AECC definition. In addition, in a post hoc analysis, we sought E CORR that thresholds for C RS and V would identify a severe group of patients with ARDS who had more than 50% mortality and include more than 10% of the study population. We did not evaluate other PaO2/FIO2 cutoffs or the requirement of a minimum PEEP level (5 cm H2O) as they were selected by the panel using face validity criteria and to ensure compatibility with prior definitions. Similarly, we did not explore other variables that might improve predictive validity, such as age and severity of nonpulmonary organ failure, because they were not specific to the definition of ARDS.25 To compare the predictive validity of the AECC definition and the Berlin Definition, we used the area under the receiver operating curve (AUROC or C statistic) in logistic regression models of mortality with a dummy variable for the ARDS definition categories.26 Because this technique requires independent categories to create the dummy variable and the AECC definition for ARDS is a subset of ALI, we could not compare the AECC definition as specified. Therefore, we modified the AECC definition and divided ALI into the independent categories of ALI nonARDS (200 mm Hg PaO2/FIO2 300 mm Hg) and ARDS alone (Pa O 2 / FIO2 200 mm Hg). Although the category of ALI non-ARDS is not explicitly described by the AECC, it has been used by many investigators.27,28 P values for categorical variables were calculated with the 2 test; P values for continuous variables were estimated with the t test, Mann-Whitney, analysis of variance, or Kruskal-Wallis, depending on the distribution and number of variables. The receiver operating curve statistical analyses were performed by using MedCalc for Windows version 12.1.4.0 (MedCalc Software) and other statistical tests were performed with SAS/STAT for Windows version 9.2 (SAS Institute Inc). Statistical significance was assessed at the 2-sided P .05 level.
Results Draft Consensus Definition. The ARDS Conceptual Model. The panel agreed that ARDS is a type of acute diffuse, inflammatory lung injury, leading to increased pulmonary vascular permeability, increased lung weight, and loss of aerated lung tissue. The clinical hallmarks are hypoxemia and bilateral radiographic opacities, associated with increased venous admixture, increased physiological dead space, and decreased lung compliance. The morphological hallmark of the acute phase is diffuse alveolar damage (ie, edema, inflammation, hyaline membrane, or hemorrhage).29 Draft Definition Criteria. Following 2 days of consensus discussions, the panel proposed a draft definition with 3 mutually exclusive severity categories (mild, moderate, and severe) of ARDS. A set of ancillary variables was proposed to further characterize severe ARDS and these were explicitly specified for further empirical evaluation.13 Timing. Most patients with ARDS are identified within 72 hours of recognition of the underlying risk factor, with nearly all patients with ARDS identified within 7 days.30 Accordingly, for a patient to be defined as having ARDS, the onset must be within 1 week of a known clinical insult or new or worsening respiratory symptoms. Chest Imaging. The panel retained bilateral opacities consistent with pulmonary edema on the chest radiograph as defining criteria for ARDS, but also explicitly recognized that these findings could be demonstrated on CT scan instead of chest radiograph. More extensive opacities (ie, 3 or 4 quadrants on chest radiograph) were proposed as part of the severe ARDS category and identified for further evaluation. Origin of Edema. Given the declining use of pulmonary artery catheters and because hydrostatic edema in the form of cardiac failure or fluid overload may coexist with ARDS,10,11 the pulmonary artery wedge pressure criterion was removed from the defini-
220 (22)
27 (24-30)
1820 (50)
32 (29-34)
1031 (28)
45 (42-48) b,c
Abbreviations: ARDS, acute respiratory distress syndrome; CRS, compliance of the respiratory system; FIO2, fraction of inspired oxygen; PaO2, arterial partial pressure of oxygen; ECORR, corrected expired volume per minute. PEEP, positive end-expiratory pressure; V a The moderate group includes patients with PaO /FIO 200 mm Hg and patients with PaO /FIO 100 mm Hg who do not meet the additional criteria for severe ARDS in the draft 2 2 2 2 definition. All patients are receiving at least 5 cm H2O PEEP and have bilateral infiltrates on chest radiograph. b P .001 comparing mortality across stages of ARDS (mild, moderate, severe) for draft and final definitions. c P =.97 comparing mortality in consensus draft severe ARDS to consensus final severe ARDS definitions.
tion. Patients may qualify as having ARDS as long as they have respiratory failure not fully explained by cardiac failure or fluid overload as judged by the treating physician using all available data. If no ARDS risk factor (eTable 1) is apparent, some objective evaluation (eg, with echocardiography) is required to help eliminate the possibility of hydrostatic edema. Oxygenation. The term acute lung injury as defined by the AECC was removed, due to the perception that clinicians were misusing this term to refer to a subset of patients with less severe hypoxemia rather than its intended use as an inclusive term for all patients with the syndrome. Positive end-expiratory pressure can markedly affect PaO2/FIO25,6; therefore, a minimum level of PEEP (5 cm H2O), which can be delivered noninvasively in mild ARDS, was included in the draft definition of ARDS. A minimum PEEP level of 10 cm H2O was proposed and empirically evaluated for the severe ARDS category. Additional Physiologic Measurements. Compliance of the respiratory system largely reflects the degree of lung volume loss.2 Increased dead space is common in patients with ARDS and is associated with increased mortality. 2 4 However, because the measurement of dead space is challenging, the panel chose minute ventilation standardized at a ECORR = minute PaCO2 of 40 mm Hg (V ventilation Pa CO 2 /40) as a surrogate.22 The draft definition of severe ARDS included the requirement of either
a low respiratory system compliance ECORR (10 (40 mL/cm H2O), a high V L/min), or both. These variables were identified for further study during the evaluation phase. The panel considered a number of additional measures to improve specificity and face validity for the increased pulmonary vascular permeability and loss of aerated lung tissue that are the hallmarks of ARDS, including CT scanning, and inflammatory or genetic markers (eTable 2). The most common reasons for exclusion of these measures were lack of routine availability, lack of safety of the measure in critically ill patients, or a lack of demonstrated sensitivity, specificity, or both for use as a defining characteristic for ARDS. Empirical Evaluation of the Draft Definition. Patients. A total of 4188 patients in the clinical database had sufficient data to classify as having ARDS by the AECC definition. Of these patients, 518 (12%) could not be classified by the draft Berlin Definition because PEEP was missing or was less than 5 cm H2O. Patients who could not be classified by the draft Berlin Definition had a mortality rate of 35% (95% CI, 31%-39%), a median (interquartile range [IQR]) of 19 (1-25) ventilator-free days, and a median (IQR) duration of mechanical ventilation in survivors of 4 (2-8) days. These patients were excluded from analyses of the draft Berlin Definition and comparisons between the AECC
definition and the draft Berlin Definition. Compared with patients from the population-based cohorts, patients from clinical trials and the academic centers cohorts were younger, had more severe hypoxemia, and had more opacities on chest radiographs. The cohort of patients from the clinical trials had the lowest mortality, likely reflecting the inclusion and exclusion criteria of the trials.31 The cohort of patients from academic centers had the highest mortality and the lowest percentage of trauma patients, reflecting the referral population (eTable 3). There were 269 patients in the physiological database with sufficient data to classify ARDS by the AECC definition, although the numbers of patients in each cohort were small. Patients in the Turin cohort had worse PaO2/FIO2 ratios and had higher mortality than the other studies (eTable 4). Evaluation of Ancillary Variables. The draft Berlin Definition for severe ARDS that included a PaO2/FIO2 of 100 mm Hg or less, chest radiograph with 3 or 4 quadrants with opacities, PEEP of at least 10 cm H2O, and either a CRS of 40 mL/cm ECORR of at least 10 H2O or less or a V L/min identified a smaller set of patients with identical mortality to the simpler severe ARDS category of PaO2/FIO2 of 100 mm Hg or less (TABLE 2). To address the possibility that the CRS and ECORR thresholds might be different in V patients with higher body weight, we evaluated weight-adjusted cutoffs for
JAMA, June 20, 2012Vol 307, No. 23 2529
these variables in one of the cohorts. There was no significant difference in the predictive validity of the weightadjusted criteria. The consensus panel reviewed these results and considered the lack of evidence for predictive validity of these ancillary variables and their potential contribution to face validity and construct validity and decided to use the simpler definition for severe ARDS that relied on oxygenation alone. The Berlin Definition. The final Berlin Definition of ARDS is shown in TABLE 3. Twenty-two percent (95% CI, 21%-24%) of patients met criteria for mild ARDS (which is comparable with the ALI non-ARDS category of the AECC definition; TABLE 4), 50% (95% CI, 48%-51%) of patients met criteria for moderate ARDS, and 28% (95% CI,
27%-30%) of patients met criteria for severe ARDS. Mortality increased with stages of ARDS from mild (27%; 95% CI, 24%-30%) to moderate (32%; 95% CI, 29%-34%) to severe (45%; 95% CI, 42%-48%). Median (IQR) ventilatorfree days declined with stages of ARDS from mild (20 [1-25] days) to moderate (16 [0-23] days) to severe (1 [020] day). Median (IQR) duration of mechanical ventilation in survivors increased with stages of ARDS from mild (5 [2-11] days) to moderate (7 [414] days) to severe (9 [5-17] days). Using the Berlin Definition, 29% (95% CI, 26%-32%) of patients with mild ARDS at baseline progressed to moderate ARDS and 4% (95% CI, 3%-6%) progressed to severe ARDS within 7 days; and 13% (95% CI, 11%-14%) of pa-
Abbreviations: CPAP, continuous positive airway pressure; FIO2, fraction of inspired oxygen; PaO2, partial pressure of arterial oxygen; PEEP, positive end-expiratory pressure. 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/ 2 2 760)]. c This may be delivered noninvasively in the mild acute respiratory distress syndrome group.
tients with moderate ARDS at baseline progressed to severe ARDS within 7 days. All differences between outcome variables across categories of modified AECC (ALI non-ARDS and ARDS alone) and across categories of Berlin Definition (mild, moderate, and severe) were statistically significant (P .001). Compared with the AECC definition, the final Berlin Definition had better predictive validity for mortality with an AUROC of 0.577 (95% CI, 0.5610.593) vs 0.536 (95% CI, 0.5200.553; P .001), with the difference in AUROC of 0.041 (95% CI, 0.0300.050). To ensure that missing PEEP data in one of the cohorts did not bias the results, the regression analysis was repeated without this cohort and yielded similar results. The Berlin Definition performed similarly in the physiological database as in the clinical database (TABLE 5, eFigure 1, and eFigure 2). Twenty-five percent (95% CI, 20%-30%) of patients met criteria for mild ARDS, 59% (95% CI, 54%-66%) of patients met criteria for moderate ARDS, and 16% (95% CI, 11%21%) of patients met criteria for severe ARDS. Mortality increased with stages of ARDS from mild (20%; 95% CI, 11%31%) to moderate (41%; 95% CI, 33%49%) to severe (52%; 95% CI, 36%68%), with P = .001 for differences in mortality across stages of ARDS. Median (IQR) ventilator-free days declined with stages of ARDS from mild
No. (%) [95% CI] of patients Progression in 7 d from mild, No. (%) [95% CI] Progression in 7 d from moderate, No. (%) [95% CI] Mortality, No. (%) [95% CI] b Ventilator-free days, median (IQR) b Duration of mechanical ventilation in survivors, median (IQR), d b
Abbreviations: AECC, American-European Consensus Conference; ALI, acute lung injury; ARDS, acute respiratory distress syndrome; FIO2, fraction of inspired oxygen; IQR, interquartile range; PaO2, arterial partial pressure of oxygen; PEEP, positive end-expiratory pressure. a The definitions are the following for ALI non-ARDS (200 mm Hg PaO /FIO 300 mm Hg, regardless of PEEP), ARDS (PaO /FIO 200 mm Hg, regardless of PEEP), mild Ber2 2 2 2 lin Definition (200 mm Hg PaO2 /FIO2 300 mm Hg with PEEP 5 cm H2O), moderate Berlin Definition (100 mm Hg PaO2 /FIO2 200 mm Hg with PEEP 5 cm H2O), and severe Berlin Definition (PaO2 /FIO2 100 mm Hg with PEEP 5 cm H2O). b Comparisons of mortality, ventilator-free days, and duration of mechanical ventilation in survivors across categories of modified AECC (ALI non-ARDS and ARDS) and across categories of Berlin Definition (mild, moderate, and severe) are all statistically significant (P .001).
2530
No. (%) [95% CI] of patients Mortality, No. (%) [95% CI] b Ventilator-free days Median (IQR) Missing, No. Duration of mechanical ventilation in survivors, median (IQR), d Lung weight, mg c Mean (SD) Missing, No. Shunt, mean (SD), % c,d
Abbreviations: AECC, American-European Consensus Conference; ALI, acute lung injury; ARDS, acute respiratory distress syndrome; FIO2, fraction of inspired oxygen; IQR, interquartile range; PaO2, arterial partial pressure of oxygen; PEEP, positive end-expiratory pressure. a The definitions are the following for ALI non-ARDS (200 mm Hg PaO /FIO 300 mm Hg, regardless of PEEP), ARDS (PaO /FIO 200 mm Hg, regardless of PEEP), mild Ber2 2 2 2 lin Definition (200 mm Hg PaO2 /FIO2 300 mm Hg with PEEP 5 cm H2O), moderate Berlin Definition (100 mm Hg PaO2 /FIO2 200 mm Hg with PEEP 5 cm H2O), and severe Berlin Definition (PaO2 /FIO2 100 mm Hg with PEEP 5 cm H2O). b Eight patients are missing in the moderate Berlin Definition ARDS group. P =.001 for difference in mortality across Berlin stages of ARDS. c Comparisons of lung weight and shunt across categories of modified AECC (ALI non-ARDS and ARDS) and across categories of Berlin Definition (mild, moderate, and severe) are statistically significant (P .001). d Only available at 1 site.
(8.5 [0-23.5] days) to moderate (0 [016.5] days) to severe (0 [0-6.5] days), with P =.003 for differences in ventilatorfree days across stages of ARDS. Median (IQR) duration of mechanical ventilation in survivors increased with stages of ARDS from mild (6.0 [3.3-20.8] days) to moderate (12.0 [5.0-19.3] days) to severe (19.0 [9.0-48.0] days), with P =.045 for differences in duration of mechanical ventilation in survivors across stages of ARDS. Using the Berlin Definition, stages of mild, moderate, and severe ARDS had increased mean lung weight by CT scan (1371 mg; 95% CI, 1268-1473; 1556 mg; 95% CI, 1474-1638; and 1828 mg; 95% CI, 1573-2082; respectively) and increased mean shunt (21%; 95% CI, 16%26%; 29%; 95% CI, 26%-32%; and 40%; 95% CI, 31%-48%; respectively). Comparisons of lung weight and shunt (from the single site providing these data) across categories of modified AECC (ALI non-ARDS and ARDS alone) and across categories of Berlin Definition (mild, moderate, and severe) were statistically significant (P .001) (Table 5, eFigure 3, and eFigure 4). In a post hoc analysis, combining a PaO2/FIO2 of 100 mm Hg or less with either a Crs of 20 mL/cm H2O or less or ECORR of at least 13 L/min identified aV a higher-risk subgroup among pa-
tients with severe ARDS that included 15% of the entire ARDS population and had a mortality of 52% (95% CI, 48%56%). Patients with severe ARDS who did not meet the higher-risk subset criteria included 13% of the entire ARDS population and had a mortality rate of 37% (95% CI, 33%-41%). The difference between the mortality of patients with higher-risk severe ARDS and patients with severe ARDS who did not meet these criteria was statistically significant (P .001). Comment Developing and disseminating formal definitions for clinical syndromes in critically ill patients are essential for research and clinical practice. Although previous proposals have relied solely on the consensus process, this is to our knowledge the first attempt in critical care to link an international consensus panel endorsed by professional societies with an empirical evaluation. The draft Berlin Definition classified patients with ARDS into 3 independent categories but relied on ancillary variables (severity of chest radiograph, PEEP 10 cm H2O, CRS E CORR 10 40 mL/cm H 2 O, and V L/min) in addition to oxygenation to define the severe ARDS group. When the ancillary variables selected by the panel
were subjected to evaluation, these parameters did not identify a group of patients with higher mortality and were excluded from the final Berlin Definition after further consensus discussion. Without this evaluation, a needlessly complex ARDS definition would have been proposed. However, static respiratory system compliance and an understanding of minute ventilation are important variables for clinicians to consider in managing patients with ARDS, even though those variables were not included as part of the definition.32 The Berlin Definition addresses some of the limitations of the AECC definition, including clarification of the exclusion of hydrostatic edema and adding minimum ventilator settings, and provides slight improvement in predictive validity. Our study presents data on the outcomes of patients with ARDS defined according to the Berlin Definition in a large heterogeneous cohort of patients including patients managed with modern approaches to lung protective ventilation. Estimates of the prevalence and clinical outcomes of mild, moderate, and severe ARDS can be assessed from this database for research and health services planning. Acute respiratory distress syndrome is a heterogeneous syndrome with comJAMA, June 20, 2012Vol 307, No. 23 2531
plex pathology and mechanisms. The proposed definition does not resolve this problem. Investigators may choose to design future trials using 1 or more of the ARDS subgroups as a base study population, which may be further refined using criteria specific to the putative mechanism of action of the intervention (eg, IL-6 levels for an antiIL-6 trial or more stringent hypoxemia criteria for a study on extracorporeal membrane oxygenation). Furthermore, some variables that were excluded from the Berlin Definition because of current feasibility and lack of data on operational characteristics may become more useful in the future. We anticipate that clinical research using our model of definition development will be used to revise the definition in the future. There are limitations to our approach. First, although the Berlin Definition had statistically significantly superior predictive validity for mortality compared with the modified AECC definition, the magnitude of this difference and the absolute values of the AUROC are small and would be clinically unimportant if the Berlin Definition was designed as a clinical prediction tool. However, predictive validity for outcome is only one criterion for evaluating a syndrome definition and the purpose of the Berlin Definition is not a prognostication tool.33 Although the Berlin Definition was developed with a framework including these criteria, we did not empirically evaluate face validity, content validity, reliability, feasibility, or success at identifying patients for clinical trial enrollment. Second, it is possible that our results are not generalizable because of the data sets we studied. This seems unlikely because patients from a broad range of populations, including clinical trials, academic centers, and community patients, were included in the analyses. Third, some variables (eg, CRS and PEEP) were missing in some patients in the data sets we used, either due to the mode of mechanical ventilation that precluded their measurement or the practicalities of population-based research. However, bias due to cohort selection or
2532 JAMA, June 20, 2012Vol 307, No. 23
missing data seem unlikely because our results were robust to sensitivity analyses that excluded individual cohorts. Fourth, it is possible that the ancillary variables did not identify a higher-risk subset because the number of quadrants on the chest radiograph cannot be assessed reliably, PEEP was not used in a ECORR predictable fashion, or CRS and V were not accurately measured. However, if this is true, it is likely also to be true in future studies and in clinical practice because the study database was constructed from clinical trial, academic, and community sites reflecting practice in the real world of clinical research. In addition, we evaluated PEEP and CRS as used by clinicians in practice and not as a test of prespecified ventilator settings that may be betterthanthevariablesevaluatedherein, but may not be practical, particularly in observational cohort studies.5,6 Fifth, because our study was not an exercise in developing a prognostic model for ARDS, we only considered the variables and cutoffs proposed by the consensus panel. We could not compare this definition directly to the AECC definition because the categories of that definition overlap. It is possible that the outcomes as well as the relative proportion of patients within each category of ARDS will change if the underlying epidemiology of the syndrome evolves due to changes in clinical practice or risk factors.34 This is particularly true for the post hoc higherrisk subset reported, for which the cut points were derived from the data sets. Conclusion In conclusion, we developed a consensus draft definition for ARDS with an international panel using a framework that focused on feasibility, reliability, and validity. We tested that definition using empirical data on clinical outcome, radiographic findings, and physiological measures from 2 large databases constructed from 7 contributing sources to assess the predictive value of ancillary variables, refine the draft definition, and compare the predictive validity of the definition to the existing AECC definition. This approach for developing the
Berlin Definition for ARDS may serve as an example for linking consensus definition activities with empirical research to better inform clinical care, research, and health services planning.
Published Online: May 21, 2012. doi:10.1001 /jama.2012.5669 Authors/Writing Committee: V. Marco Ranieri, MD (Department of Anesthesia and Intensive Care Medicine, University of Turin, Turin, Italy); Gordon D. Rubenfeld, MD, MSc (Program in Trauma, Emergency, and Critical Care, Sunnybrook Health Sciences Center, and Interdepartmental Division of Critical Care Medicine, University of Toronto, Toronto, Ontario, Canada); B. Taylor Thompson, MD (Department of Medicine, Massachusetts General Hospital and Harvard Medical School, Boston); Niall D. Ferguson, MD, MSc (Department of Medicine, University Health Network and Mount Sinai Hospital, and Interdepartmental Division of Critical Care Medicine, University of Toronto, Toronto, Ontario, Canada); Ellen Caldwell, MS (Division of Pulmonary and Critical Care Medicine, University of Washington, Seattle); Eddy Fan, MD (Department of Medicine, University Health Network and Mount Sinai Hospital, University of Toronto, Toronto, Ontario, Canada); Luigi Camporota, MD (Department of Critical Care, Guys and St. Thomas NHS Foundation Trust, Kings Health Partners, London, England); and Arthur S. Slutsky, MD (Keenan Research Center of the Li Ka Shing Knowledge Institute of St. Michaels Hospital; Interdepartmental Division of Critical Care Medicine, University of Toronto, Toronto, Ontario, Canada). Author Contributions: Dr Rubenfeld and Ms Caldwell had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Ranieri, Rubenfeld, Thompson, Ferguson, Caldwell, Camporota. Acquisition of data: Ranieri, Rubenfeld, Thompson. Analysis and interpretation of data: Rubenfeld, Thompson, Ferguson, Caldwell, Fan, Slutsky. Drafting of the manuscript: Rubenfeld, Ferguson, Caldwell, Slutsky. Critical revision of the manuscript for important intellectual content: Ranieri, Rubenfeld, Thompson, Ferguson, Caldwell, Fan, Camporota, Slutsky. Statistical analysis: Rubenfeld, Caldwell, Slutsky. Obtained funding: Ranieri. Administrative, technical, or material support: Rubenfeld, Thompson, Fan, Camporota. Study supervision: Ranieri, Rubenfeld, Thompson, Slutsky. Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Ranieri reported receiving consulting fees or honoraria from Maquet and Hemodec and board membership from Faron. Dr Rubenfeld reported receiving consulting fees or honoraria from Ikaria, Faron, and Cerus. Dr Thompson reported receiving support for travel from European Society of Intensive Care Medicine; being an advisory board member of Hemodec and AstraZeneca; receiving consultancy fees from US Biotest, Sirius Genetics, sanofi-aventis, Immunetrics, Abbott, and Eli Lilly; and receiving grants from the National Heart, Lung, and Blood Institute. Dr Slutsky reported receiving support for travel expenses from European Society of Intensive Care Medicine; board membership from Ikaria; receiving consultancy fees from GlaxoSmithKline and Tarix; having stock/stock options with Apeiron and Tarix; and sitting on advisory boards for Maquet Medical and NovaLung and steering committees for HemoDec and Eli Lilly. No other authors reported any financial disclosures. Members of the ARDS Definition Task Force: V. Marco