Supplementary Table 1: Multiple logistic regression analyses of variables associated with mortality
Variable / Adjusted odds ratio / 95% confidence interval / p ValueModel 1a:
PRISM III
Non-pulmonary organ dysfunctions at PARDS onset
Immunocompromised (yes)
Worst OI in the first 24 hours / 1.05
2.05
3.08
1.02 / 1.01 to 1.10
1.33 to 3.17
1.21 to 7.85
1.00 to 1.05 / 0.022
0.001
0.019
0.100
Model 2a:
PRISM III
Non-pulmonary organ dysfunctions at PARDS onset
Immunocompromised (yes)
Worst OI in the first 24 hours
Cumulative dose of corticosteroids (mg/kg MPE) / 1.06
1.94
3.19
1.03
1.02 / 1.01 to 1.11
1.26 to 3.00
1.25 to 8.13
1.00 to 1.05
1.00 to 1.04 / 0.010
0.003
0.015
0.062
0.056
a In both models, variables modeled but not significant (p > 0.100) were: age, the maximum vasopressor score in the first 72 hours of PARDS, and fluid balance in the first 72 hours of PARDS. Corticosteroid exposure > 24 hours was modeled but not significant in the first model.
Supplementary Table 2: Multiple linear regression analyses of variables associated with VFD at 28 days
Variable / Coefficient / 95% confidence interval / p ValueModel 1a:
Non-pulmonary organ dysfunctions at PARDS onset
Worst OI in the first 24 hours
Maximum vasopressor score in first 72 hours
Corticosteroids > 24 hours (yes) / -2.56
-0.10
0.04
-2.98 / -3.61 to -1.50
-0.18 to -0.02
0.00 to 0.09
-5.09 to -0.88 / <0.001
0.010
0.077
0.006
Model 2a:
Non-pulmonary organ dysfunctions at ARDS onset
Worst OI in the first 24 hours
Maximum vasopressor score in first 72 hours
Cumulative dose of corticosteroids (mg/kg MPE) / -2.34
-0.12
0.04
-0.13 / -3.37 to -1.31
-0.19 to -0.04
0.00 to 0.08
-0.18 to -0.07 / <0.001
0.002
0.078
<0.001
a In both models, variables modeled but not significant (p > 0.100) were: age, PRISM III, immunocompromised condition, reactive airway disease, and fluid balance in the first 72 hours of PARDS.
Supplementary Table 3: Multiple linear regression analyses of variables associated with (log-transformed) total ventilator days in survivors
Variable / Coefficient / 95% confidence interval / p ValueModel 1a:
Age (years)
Reactive airway disease (yes)
Non-pulmonary organ dysfunctions at PARDS onset
Worst OI in the first 24 hours
Corticosteroids > 24 hours (yes) / -0.02
-0.28
0.19
0.01
0.30 / -0.04 to 0.00
-0.58 to 0.01
0.08 to 0.30
0.00 to 0.02
0.09 to 0.51 / 0.020
0.057
0.001
0.062
0.005
Model 2a:
Age (years)
Reactive airway disease (yes)
Non-pulmonary organ dysfunctions at PARDS onset
Worst OI in the first 24 hours
Cumulative dose of corticosteroids (mg/kg MPE) / -0.02
-0.48
0.18
0.01
0.02 / -0.02 to 0.01
-0.76 to -0.20
0.07 to 0.28
0.00 to 0.02
0.01 to 0.03 / 0.051
0.001
0.001
0.025
<0.001
a In both models, the response variable “total ventilator days” was log-transformed to achieve a normal distribution. In both models, variables modeled but not significant (p > 0.100) were: PRISM III, immunocompromised condition, the maximum vasopressor score in the first 72 hours of PARDS, and fluid balance in the first 72 hours of PARDS.
Supplementary Table 4: Additional subgroups stratified by exposure to ≤ 24 hours or > 24 hours of corticosteroids
All patients / No corticosteroids or exposure ≤ 24 hours / Corticosteroid exposure > 24 hours / p valueImmunocompetent
Sample size (n)
PRISM III at 12 hours
Non-pulmonary organ dysfunctions
At corticosteroid initiation
Pao2/Fio2
OI
Serum cortisol (µg/dL, n = 134)
Vasopressor score
Outcomes
Ventilator days (survivors, n = 211)
VFD at 28 days
PICU deaths (%) / 229
9 [5, 16]
1 [1, 2]
156 [109, 205]
10.3 [7.1, 16.3]
21.4 [9.8, 34.9]
9 [3, 18]
10 [7, 17]
17 [7, 21]
18 (8) / 101
8 [4, 15]
1 [1, 2]
161 [113, 219]
9.7 [6.1, 15.2]
22.5 [16.5, 31.5]
8 [2, 16]
9 [6, 15]
19 [10.5, 22]
7 (7) / 128
10 [5, 16]
1 [1, 2]
151 [103, 200]
10.4 [8, 17.4]
17.8 [6.5, 37.7]
10 [3, 20]
11 [7, 19]
16 [5, 20]
11 (9) / 0.130
0.541
0.079
0.020
0.136
0.188
0.040
0.039
0.642
Septic shock
Sample size (n)
PRISM III at 12 hours
Immunocompromised (%)
Non-pulmonary organ dysfunctions
At corticosteroid initiation
Serum cortisol (µg/dL, n = 115)
Vasopressor score
Outcomes
Ventilator days (survivors, n = 145)
VFD at 28 days
PICU deaths (%) / 166
11 [6, 16]
30 (18)
2 [1, 3]
21 [10.3, 37.4]
13.5 [7, 20]
10 [7, 17]
16 [4, 20]
21 (13) / 60
9 [5, 15]
8 (13)
2 [1, 3]
21.4 [14.2, 37.4]
11 [6, 16]
10 [7, 17]
18 [8, 21]
4 (7) / 106
11.5 [7, 18]
22 (21)
2 [1, 3]
20 [8.6, 36.3]
15 [8, 24]
11 [7, 17]
14.5 [0, 20]
17 (16) / 0.139
0.233
0.473
0.288
0.051
0.587
0.157
0.081
Prematurity or reactive airway disease
Sample size (n)
PRISM III at 12 hours
At corticosteroid initiation
Pao2/Fio2
OI
Outcomes
Ventilator days (survivors, n = 63)
VFD at 28 days
PICU deaths (%) / 68
9 [5, 13]
150 [105, 199]
10.5 [8.1, 16.4]
10 [7, 17.5]
18 [8.5, 21]
5 (7) / 18
8 [6, 15]
135 [109, 183]
13.2 [7.1, 16.4]
10 [6, 16]
18 [12, 22]
0 (0) / 50
9 [5, 12]
153 [100, 200]
10.3 [8.2, 16.5]
9 [7, 18]
18 [6, 21]
5 (10) / 0.862
0.739
0.808
0.709
0.802
0.315
Supplementary Table 5: High-risk subgroups stratified by exposure to ≤ 24 hours or > 24 hours of corticosteroids
All patients / No corticosteroids or exposure ≤ 24 hours / Corticosteroid exposure > 24 hours / pvalue≥ 3 nonpulmonary organ failures
Sample size (n)
Ventilator days (survivors, n = 50)
VFD at 28 days
PICU deaths (%) / 76
13 [8, 24]
4 [0, 18]
26 (34) / 27
10 [7, 22]
13 [0, 21]
6 (22) / 49
15 [9, 32]
0 [0, 15]
20 (41) / 0.072
0.013
0.132
Immunocompromised
Sample size (n)
Ventilator days (survivors, n = 35)
VFD at 28 days
PICU deaths (%) / 54
10 [5, 18]
10 [0, 20]
19 (35) / 13
7 [5, 12]
20 [10, 23]
2 (15) / 41
14 [7, 22]
0 [0, 18]
17 (41) / 0.121
0.012
0.107
Severe PARDS (worst OI ≥ 16)
Sample size (n)
Ventilator days (survivors, n = 79)
VFD at 28 days
PICU deaths (%) / 101
11 [8, 22]
13 [0, 19]
22 (22) / 31
9.5 [6, 20]
17 [1, 20]
5 (16) / 70
12 [9, 24]
9.5 [0, 19]
17 (24) / 0.157
0.093
0.588