Online data supplement
Acute respiratory distress syndrome mimickers lacking common risk factors of the
Berlin definition
METHODS
Identification of patients
The medical records of patients hospitalized in the two participating centers between January
2003 and December 2012 for ARDS were reviewed. The following keywords were searched
on the hospital record databases: “Acute Respiratory Distress Syndrome”, “ARDS”, “Acute
Interstitial Pneumonia”, “Diffuse Alveolar Hemorrhage”, “Drug-induced Acute Lung Injury”.
“Common” risk factors from the Berlin definition
These include the following factors (Table 1):[1] pneumonia, gastric aspiration, pulmonary
contusion, drowning, inhalation of gases, extra-pulmonary sepsis, polytrauma, pancreatitis,
burns, non-cardiogenic shock, drug intoxication, and polytransfusion. Patients with
“vasculitides” were deemed to have ARDS with no CRF because vasculitides are not
pathologically characterized by diffuse alveolar damage.
Patients with no “common” risk factors
ARDS patients with no common risk factors (ARDSCRF-) were separated into four etiological
groups: immune ARDS, drug-induced ARDS, malignant ARDS, and idiopathic ARDS. The
immune ARDS group included vasculitides, as defined by the American College of
Rheumatology criteria, [2] and connective tissue diseases according to ACR and ARA
criteria, with systemic lupus erythematosus,[3] rheumatoid arthritis,[4] mixed connective
tissue,[5] inflammatory myopathy[6] and scleroderma. Undifferentiated connective tissue
diseases were classified separately, in case of clinical and serological evidence of
autoimmunity within the criteria of a conventional connective disease.[7, 8] Thresholds used
for positivity were as follows: > 15 IU/mL for rheumatoid factor; > 5 IU/mL for anti-CCP
antibodies; titer > 1/160 for antinuclear antibodies, and > 20 IU/mL for anti-neutrophil
cytoplasmic antibodies (ANCA).
Assessment of lung lesions reversibility
The presence of lung lesions potentially responding to therapy with corticosteroids was
assessed by three experts blinded to the final diagnosis and to outcome (AP, BM, MF), based
on BAL fluid cytology and chest CT scan patterns, and patients were considered as “potential
responders” when an agreement of all three experts was met. From this, threshold values for
BAL cell counts associated with potential responsiveness to therapy with corticosteroids were
obtained with Receiver Operator (ROC) curves (Figures E1 and E2). The composite variable
obtained was finally entered into a multivariable logistic regression model aimed at
determining factors associated with in-ICU mortality among ARDSCRF- patients.[9]-[10]
Data presentation and statistical analysis
Categorical variables are expressed as number and percentages (95% confidence interval),
and compared with the chi2 test or the Fischer Exact test, as appropriate. Continuous variables
are reported as median [interquartile range IQR 25-75] or mean (± standard deviation SD),
and compared using the Student t-test or the Mann-Whitney test, as appropriate. The
distribution of variables was tested using the Shapiro-Wilk normality test. Comparisons of
more than two groups were performed using ANOVA and the Bonferroni post-hoc test, when
overall p value was less than 0.05, or using the Kruskal-Wallis test, as appropriate. Two
distinct sets of statistical analyses were performed. First, a comparison of ARDSCRF+ and
ARDSCRF- patients was performed. Second, ARDSCRF- patients were studied. Factors
associated with intensive care unit mortality were determined by univariate and multivariable
analyses, both within the whole cohort of patients, and within the ARDSCRF- subgroup.
Interaction between variables were tested using the Mantel-Haenzel test, and interaction terms
introduced in the multivariable analysis when appropriate. Odds ratios (OR) and their 95%
confidence intervals (95% CI) were computed. Independent variables with p<0.10 in
univariate analysis were included in the multivariable analysis, with backward elimination of
variables displaying a p value >0.05. Analyses were carried out using the SPSS Base 21.0
statistical software package (SPSS Inc., Chicago, IL).
RESULTS
Based on the experts’ consensus, the variable most significantly associated with the presence
of potentially reversible lung lesions was BAL lymphocyte counts, which yielded an area
under the ROC curve of 0.88 (Figure 1D). A proportion of lymphocytes >20% in the BAL
fluid was associated with the best discrimination between potential responders and nonresponders to steroids, as defined by the experts’ consensus (Figure 1C). BAL neutrophils
yielded an area under the curve of 0.79 (Figure 1F), and the threshold value associated with
the best discrimination was a proportion of neutrophils <22% in the BAL fluid (Figure 1E).
The composite variable “presence of potentially reversible lung lesions” also tended to be
associated with the lack of signs consistent with pulmonary fibrosis (honeycombing and
traction bronchiectasis, Figure 2) on chest CT scan.
References
1.
ARDS Definition Task Force, Ranieri VM, Rubenfeld GD, et al. (2012) Acute
respiratory distress syndrome: the Berlin Definition. JAMA J Am Med Assoc 307:2526–
2533. doi: 10.1001/jama.2012.5669
2.
Jennette JC, Falk RJ, Andrassy K, et al. (1994) Nomenclature of systemic vasculitides.
Proposal of an international consensus conference. Arthritis Rheum 37:187–192.
3.
Hochberg MC (1997) Updating the American College of Rheumatology revised criteria
for the classification of systemic lupus erythematosus. Arthritis Rheum 40:1725. doi:
10.1002/1529-0131(199709)40:9&lt;1725::AID-ART29&gt;3.0.CO;2-Y
4.
Lee DM, Weinblatt ME (2001) Rheumatoid arthritis. Lancet 358:903–911. doi:
10.1016/S0140-6736(01)06075-5
5.
Sharp GC, Irvin WS, Tan EM, et al. (1972) Mixed connective tissue disease--an
apparently distinct rheumatic disease syndrome associated with a specific antibody to an
extractable nuclear antigen (ENA). Am J Med 52:148–159.
6.
Dalakas MC, Hohlfeld R (2003) Polymyositis and dermatomyositis. Lancet 362:971–
982. doi: 10.1016/S0140-6736(03)14368-1
7.
Kinder BW, Collard HR, Koth L, et al. (2007) Idiopathic nonspecific interstitial
pneumonia: lung manifestation of undifferentiated connective tissue disease? Am J
Respir Crit Care Med 176:691–697. doi: 10.1164/rccm.200702-220OC
8.
Doria A, Mosca M, Gambari PF, Bombardieri S (2005) Defining unclassifiable
connective tissue diseases: incomplete, undifferentiated, or both? J Rheumatol 32:213–
215.
9.
Gay SE, Kazerooni EA, Toews GB, et al. (1998) Idiopathic pulmonary fibrosis:
predicting response to therapy and survival. Am J Respir Crit Care Med 157:1063–1072.
doi: 10.1164/ajrccm.157.4.9703022
10. Terriff BA, Kwan SY, Chan-Yeung MM, Müller NL (1992) Fibrosing alveolitis: chest
radiography and CT as predictors of clinical and functional impairment at follow-up in
26 patients. Radiology 184:445–449. doi: 10.1148/radiology.184.2.1620845
Table 1. Common risk factors of the Berlin definition (adapted from [1]).
ARDS Risk factor
Pneumonia
Non-pulmonary sepsis
Aspiration of gastric contents
Major trauma
Pulmonary contusion
Pancreatitis
Inhalational injury
Severe burns
Non-cardiogenic shock
Drug overdose
Multiple transfusions or transfusion-associated acute lung injury (TRALI)
Pulmonary vasculitis a
Drowning
a
Pulmonary vasculitis was not considered as a “common risk factor” in the current study, as it
is not associated with the presence of diffuse alveolar damage.
Table 2. Factors recorded on admission and within 48h of mechanical ventilation and
associated with ICU mortality (n=665)
n
Death
n (%)
-
Univariate analysis
OR (95% CI)
P
1.03 [1.02-1.04]
<0.0001
1.14 [1.10-1.18]
<0.0001
Multivariable analysis
OR (95% CI)
P
1.03 [1.02-1.04]
<0.0001
1.09 [1.04-1.06]
<0.0001
SAPS II a
665
b
LODS
665
Shock
Yes
516
299 (58)
1.99 [1.37-2.88]
<0.0001
No
149
61 (41)
1
Absence of
common risk factors for
ARDS
Yes
50
33 (66)
1.69 [0.92-3.09]
0.091
2.06 [1.02-4.18]
0.044
No
615
327 (53)
1
Berlin classification with
shock c
mild
58
31 (53)
1
moderate
280
126 (45)
1.66 [0.49-5.56]
0.41
severe
327
203 (62)
4.65 [1.36-15.94]
0.014
2.04 [1.41-2.95]
<0.001
Age d
665
1.02 [1.01-1.03]
<0.001
1.02 [1.01-1.03]
0.004
e
NIV duration
665
1.17 [1.04-1.32]
0.009
1.46 [1.24-1.71]
<0.001
No bacterial infection
Yes
306
189 (62)
1.78 [1.30-2.42]
<0.001
1.47 [1.03-2.11]
0.033
No
359
171 (48)
1
a
per unit of simplified acute physiology score (SAPS II); b per unit of logistic organ dysfunction score
(LODS); c interaction variable created from the severity of ARDS and shock; d per year; e per day of
non-invasive ventilation (NIV). The Hosmer-Lemeshow goodness of fit test showed good calibration
of the model (p=0.37). The area under the ROC curve of the model was 0.76 (sensitivity=74%,
specificity=63%).
Table 3. Histopathological and cytological diagnoses of ARDS patients with no common risk
factors.
Cases
Lung tissue
specimen
Cytological/Histopathological
patterns
Idiopathic ARDS (n=4/13)
case 1
OLB
DAD, fibroproliferative stage
case 2
TB
DAD
case 3
TB
Organizing pneumonia
case 4
TB
Organizing pneumonia
Immune ARDS (n=2/24)
case 1
OLB
Organizing pneumonia, fibrosis
case 2
Autopsy
Hemophagocytosis, TL infiltration
Drug-induced ARDS (n=2/20)
case 1
Autopsy
DAD, fibroproliferative stage
case 2
Autopsy
DAD, fibroproliferative stage
Malignant ARDS (n=10/12)
cases 1-4
BAL
Bronchiolo-alveolar carcinoma
cases 5-6
BAL
Monoclonal lymphocytes
cases 7-10
BB
Carcinoma cells
OLB, open Lung Biopsy; TB, transbronchial biopsy; BAL, broncho-alveolar lavage; BB,
bronchial biopsy; DAD, diffuse alveolar damage; TL, T lymphocytes
Figure legends
Figure 1. Broncho-alveolar lavage (BAL) fluid cytology for patients with ARDS and no risk
factors from the Berlin definition categorized as responders or non-responders to steroids by
three experts blinded to management and outcome. BAL cellularity (A), macrophage (B) and
eosinophil (G) counts, and the proportion of patients with diffuse alveolar hemorrhage (H)
were not different between both groups. In contrast, BAL lymphocytes (C) were higher and
neutrophils (E) lower in responders than in non-responders to steroids. Dashed lines show the
best threshold values for BAL lymphocytes (alveolar lymphocytes > 20% (C); area under the
ROC curve (AUC) = 0.88 (D)) and neutrophils (alveolar neutrophils<22% (E); area under the
ROC curve = 0.79 (F)) for discriminating responders from non-responders to steroids.
Figure 2. Chest CT scan patterns for patients with ARDS and no risk factors from the Berlin
definition categorized as responders or non-responders to steroids by three experts blinded to
management and outcome. There was no difference in the proportion of patients with ground
glass opacities (A) and septal thickening (D). In contrast, there was a trend towards less
frequent honeycombing (B) and traction bronchiectasis (C) in responders than in nonresponders to steroids.
Figure 1
A
BAL cellularity
P=0.52
1.0×106
5.0×105
0.0
C
Responders
40
20
0
40
20
Responders
Responders
D
Non-responders
BAL lymphocytes
100
Sensitivity (%)
BAL lymphocytes (%)
60
0
60
P<0.0001
80
P=0.092
80
Non-responders
BAL lymphocytes
100
100
BAL macrophages (%)
BAL cells / mL
1.5×106
BAL macrophages
B
50
AUC=0.88
0
Non-responders
0
50
100
100 - Specificity (%)
F
E
BAL neutrophils
BAL neutrophils
P=0.002
Sensitivity (%)
BAL neutrophils (%)
100
100
80
60
40
50
AUC=0.79
20
0
Responders
0
Non-responders
0
50
100
100 - Specificity (%)
H
G
Diffuse alveolar hemorrhage
BAL eosinophils
25
P=0.34
20
30
Patients (n)
BAL eosinophils (%)
40
20
10
0
15
P=0.71
Non-responders
Responders
10
5
Responders
Non-responders
0
Present
Absent
Figure 2
A
Ground glass opacities
25
B
20
P=0.63
Non-responders
Responders
15
10
Patients (n)
Patients (n)
20
5
0
Present
15
10
5
0
Absent
D
C
Patients (n)
Patients (n)
15
P=0.096
10
5
0
Present
Present
Absent
Absent
Septal thickening
Traction bronchectasies
15
Honeycombing
P=0.065
P=0.34
10
5
0
Present
Absent