Abstract
Purpose
To describe the prevalence of hypocapnia and hypercapnia during the earliest period of mechanical ventilation, and determine the association between PaCO2 and mortality.
Materials and Methods
A cohort study using an emergency department registry of mechanically ventilated patients. PaCO2 was categorized: hypocapnia (< 35 mmHg), normocapnia (35 - 45 mmHg), and hypercapnia (> 45 mmHg). The primary outcome was survival to hospital discharge.
Results
A total of 1,491 patients were included. Hypocapnia occurred in 375 (25%) patients and hypercapnia in 569 (38%). Hypercapnia (85%) had higher survival rate compared to normocapnia (74%) and hypocapnia (66%), P < 0.001. PaCO2 was an independent predictor of survival to hospital discharge [hypocapnia (aOR 0.65 (95% confidence interval [CI] 0.48 - 0.89), normocapnia (reference category), hypercapnia (aOR 1.83 (95% CI 1.32 - 2.54)]. Over ascending ranges of PaCO2, there was a linear trend of increasing survival up to a PaCO2 range of 66 - 75 mmHg, which had the strongest survival association, aOR 3.18 (95% CI 1.35 - 7.50).
Conclusions
Hypocapnia and hypercapnia occurred frequently after initiation of mechanical ventilation. Higher PaCO2 levels were associated with increased survival. These data provide rationale for a trial examining the optimal PaCO2 in the critically ill.
Keywords: mechanical ventilation, hypercapnia, clinical outcomes
Introduction
Mechanical ventilation is a common indication for critical care services, both in the intensive care unit (ICU) and emergency department (ED)[1, 2]. As the population ages, the need is increasing[3]. Several best practices, including low tidal volume for prevention of ventilator-induced lung injury (VILI), protocols for sedation and weaning, and early detection and treatment of sepsis have improved outcomes in mechanically ventilated patients. Even with these approaches, mortality for mechanically ventilated ICU patients remains high, at over 30%[4, 5]. Therefore finding new approaches to reduce mortality is crucial.
The management of the partial pressure of arterial carbon dioxide (PaCO2) is a fundamental aspect of care in mechanically ventilated patients. In the critically ill, derangements in PaCO2 occur in up to 70% of ventilated patients[6, 7]. The prevailing paradigm is that hypercapnia has either a deleterious effect on outcome or is a simple by-product of low tidal volume ventilation (i.e. permissive hypercapnia). A recent secondary analysis found severe hypercapnia to be associated with mortality among patients with acute respiratory distress syndrome (ARDS)[8]. Hence, the normalization of PaCO2 levels can be an intuitive therapeutic goal. However, increasing data supports the notion that hypercapnia has biologically important beneficial effects through various mechanisms, including anti-inflammation, mitigation of VILI, and modulation of gene expression [9-12]. Among post-cardiac arrest patients, hypercapnia has been suggested to attenuate injury through vasodilation and increased cerebral blood flow [13-15]. These data suggest that hypercapnia could confer additional advantages beyond low stretch ventilation and could be a therapeutic target to improve outcome. Conversely, hypercapnia also has potentially negative consequences, including increased pulmonary vascular tone, elevated intracranial pressure, and negative inotropy [16]. Despite the equipoise with respect to the optimal management of PaCO2, this has not been tested rigorously across a diverse cohort of mechanically ventilated patients, and it is currently unclear if hypercapnia has the same association with mortality among mechanically ventilated patients without ARDS. Given the fact that mechanical ventilation is delivered globally to hundreds of thousands of patients annually, and is a cornerstone of therapy in acute respiratory failure, investigating the impact of PaCO2 on outcome could have large-scale implications for many critically ill patients.
The objective of this study was to describe the prevalence of hypocapnia and hypercapnia during the earliest period of mechanical ventilation, and to determine the association between PaCO2 and mortality among mechanically ventilated non-ARDS arrest patients. We hypothesized that alterations in PaCO2 would be common, and given the pre-clinical data in this domain, hypercapnia would be associated with reduction in mortality.
Materials and Methods
Study design and participants
We conducted a cohort study, using an ED registry of patients with acute initiation of mechanical ventilation, at a tertiary academic center in St. Louis, Missouri, USA (September 2009 to March 2016)[17, 18]. Patients with initiation of mechanical ventilation in the ED were assessed for inclusion. Inclusion criteria: 1) age ≥ 18 years; and 2) mechanical ventilation via an endotracheal tube. Exclusion criteria: 1) death or discontinuation of mechanical ventilation within 24 hours; 2) chronic mechanical ventilation; 3) presence of a tracheostomy; 4) transfer to another hospital; 5) presence of acute respiratory distress syndrome (ARDS) during ED presentation[19]; and 6) cardiac arrest or drug overdose as the reason for mechanical ventilation. This study was approved by the institutional review board under waiver of informed consent, and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement (Additional File 1, Table S1) [20].
Procedures
Baseline demographics, comorbid conditions, indication for mechanical ventilation, illness severity, and blood pressure were collected. Pertinent treatment variables in the ED included the receipt of antibiotics and vasopressors. Arterial blood gas (ABG) analyses were performed after initiation of mechanical ventilation while subjects were in the ED. This initial blood gas was used in the analysis of PaCO2 and outcome.
All ED mechanical ventilator variables, airway pressures, pulmonary mechanics, and gas exchange variables were collected. ICU ventilator settings, airway pressures, and pulmonary mechanics were followed for up to two weeks and collected twice daily. For purposes of the analyses, the mean values for the ICU ventilator variables over the two-week period were used.
Data was retrieved from the electronic medical record and organized and maintained in an electronic database. Data accuracy was verified with the aid of a research assistant, trained and blinded to study objectives.
Sepsis was defined as previously described[21]. Lung-protective tidal volume was defined as the use of tidal volume of ≤ 8 mL/kg predicted body weight (PBW)[22]. Static compliance (mL/ cm H2O) of the respiratory system (CRS) was calculated as: tidal volume/[plateau pressure - positive end expiratory pressure (PEEP)]. Driving pressure (cm H2O) was calculated as: tidal volume/ CRS.
The primary outcome was survival to hospital discharge.
Statistical analysis
For descriptive statistics the categorical data was displayed as counts and proportions, and continuous data as mean values and standard deviation (SD) or median values and interquartile range (IQR). Categorical variables were compared using the chi-square test. Continuous variables were compared using one-way analysis of variance (ANOVA) or Kruskal-Wallis test, based on the distribution of the data. To correct for multiple comparisons we used the Bonferroni correction. Differences in PaCO2 categories were considered statistically significant if P < 0.017. Spearman's correlation coefficient (r) was used to assess the relationship between PaCO2, tidal volume, respiratory rate, and pH.
To test the relationship between PaCO2 and survival, a multivariable logistic regression model was constructed with survival to hospital discharge as the dependent variable. PaCO2 in the ED was a priori categorized into the following groups: hypocapnia (< 35 mmHg), normocapnia (35 - 45 mmHg), and hypercapnia (> 45 mmHg), and treated as a categorical variable, with normocapnia as the reference. A priori, candidate variables for inclusion in the model included: 1) baseline characteristics with known prognostic significance for outcome in mechanically ventilated patients; 2) clinically relevant ED treatment variables; 3) ED ventilator variables; 4) ICU ventilator variables; and 5) pulmonary mechanic variables. Candidate variables were entered into the model if statistically different at P < 0.20 between PaCO2 groups and are displayed in Additional File 2, Table S2. Because of the high correlation between the variables related to pulmonary mechanics (plateau pressure, static compliance, driving pressure), only one of the variables was included in the model. Therefore, given the increasing evidence demonstrating the importance of driving pressure on outcome, it was decided to enter driving pressure into the model [23]. Backward elimination with a criterion of P < 0.05 for retention in the model was used. Statistical interactions and collinearity were assessed. Goodness of fit was evaluated with the Hosmer-Lemeshow test.
To further examine if an association between PaCO2 and outcome existed, survival to hospital discharge was graphed across ascending ranges of PaCO2 (< 35, 35-45, 46-55, 56-65, 66-75, >75 mmHg). This graph was inspected to assess if there was a threshold signal for survival over PaCO2 ranges. The presence of significant linear trends in the odds of survival to hospital discharge was assessed using chi-square test for linear trend. To further test a relationship between ranges of PaCO2 and survival, PaCO2 was entered into a logistic regression model as a categorical variable using ascending ranges (i.e. < 35, 35-45, 46-55, 56-65, 66-75, >75 mmHg), with 35-45 mmHg as the reference. The model was adjusted for variables that were retained in the original model and used variables that were statistically independent of the other variables (i.e. the model was tested for collinearity).
To better control for the influence of diagnosis and etiology of respiratory failure, a priori subgroup analyses focused on patients with: a) sepsis; and b) trauma. As chronic obstructive pulmonary disease (COPD) is associated not only with hypercapnic respiratory failure but some data also suggests lower mortality in this cohort, a final a priori subgroup analysis that excluded patients with COPD was performed [4]. Since 374 patients could not be grouped into a specific indication for mechanical ventilation, a post hoc subgroup analysis which excluded patients mechanically ventilated for the indication of “other” was conducted. To further examine the influence of PaCO2 on outcomes in those patients with progressive pulmonary dysfunction, a second post hoc subgroup analysis was conducted on patients that developed ARDS after admission to the ICU. Finally, as there was a statistical difference in lactate levels between the groups, a subgroup analysis was conducted on those patients with lactate measured in the ED. For the subgroup analyses, we used variables that were retained in the original model. For the subgroups based on indication for mechanical ventilation (i.e. sepsis and trauma), indication for mechanical ventilation was removed as a covariable secondary to collinearity. For all models, robust standard errors to reduce the risk of type I error was used.
Assuming an approximate 1:1:1 ratio of patients in the hypocapnia, normocapnia, and hypercapnia groups, the sample size that was analyzed allowed > 80% power to detect a 10% absolute difference in survival between groups (assuming an alpha level of 0.017 when adjusted for multiple comparisons).
Results
Study population
A total of 3,525 subjects were assessed for inclusion; 1,491 were included in the final population (Figure 1).
Figure 1. Study flow diagram.
Baseline characteristics of the study population are shown in Table 1. The median (IQR) APACHE II score for the entire cohort was 15 (11-19). The most common reason for initiation of mechanical ventilation was sepsis, followed by trauma. Post-intubation data
Table 1.
Baseline characteristics at the time of intubation.
| All Subjects n = 1491 | Hypocapniaa n = 375 | Normocapniab n = 547 | Hypercapniac n = 569 | P - value | |
|---|---|---|---|---|---|
|
| |||||
| Age (years) | 59 (17) | 58 (17) | 60 (18) | 59 (16) | 0.305 |
|
| |||||
| Female gender, n (%) | 688 (46) | 178 (47) | 240 (44) | 270 (47) | 0.409 |
|
| |||||
| Race, n (%) | 0.003 | ||||
| Black | 859 (58) | 211 (56) | 308 (56) | 340 (60) | |
| White | 615 (41) | 153 (41) | 235 (43) | 227 (39) | |
| Other | 17 (1) | 11(3) | 4 (1) | 2 (1) | |
|
| |||||
| Comorbidities, n (%) | |||||
| Chronic obstructive pulmonary disease | 375 (25) | 41 (11) | 83 (15) | 251 (44) | <0.001 |
| Malignancy | 262 (18) | 83 (22) | 83 (15) | 96 (17) | <0.001 |
| Congestive heart failure | 350 (23) | 60 (16) | 109 (20) | 181 (32) | <0.001 |
| Diabetes mellitus | 526 (35) | 118 (31) | 190 (35) | 218 (38) | 0.132 |
| End stage renal disease | 116 (8) | 24 (6) | 46 (8) | 46 (8) | 0.885 |
| Cirrhosis | 117 (8) | 53 (14) | 36 (7) | 28 (5) | 0.148 |
|
| |||||
| Indication for mechanical ventilation, n (%) | |||||
| Asthma | 39 (3) | 0 | 11 (2) | 28 (5) | <0.001 |
| Chronic obstructive pulmonary disease | 122 (8) | 4 (1) | 9 (2) | 109 (19) | <0.001 |
| CHF/pulmonary edema | 100 (7) | 12 (3) | 29 (5) | 59 (10) | <0.001 |
| Sepsis | 471 (31) | 142 (38) | 159 (29) | 170 (30) | 0.010 |
| Trauma | 385 (26) | 91 (24) | 198 (36) | 96 (17) | <0.001 |
| Other | 374 (25) | 126 (34) | 141 (26) | 107 (19) | <0.001 |
|
| |||||
| APACHE II score* | 15 (11 - 19) | 14 (10 - 19) | 13 (9 -18) | 16 (12 - 21) | <0.001 |
|
| |||||
| Mean arterial pressure (mmHg) | 87 (69 - 107) | 85 (69 - 101) | 89 (71 - 112) | 86 (68 - 107) | 0.045 |
|
| |||||
| Hypotension (MAP < 70 mmHg), n (%) | 380 (25) | 96 (26) | 126 (23) | 158 (28) | 0.193 |
|
| |||||
| Vasopressor infusion, n (%) | 320 (21) | 100 (27) | 102 (19) | 118 (21) | 0.012 |
|
| |||||
| Antibiotic administration, n (%) | 697 (47) | 205 (55) | 222 (41) | 270 (47) | <0.001 |
|
| |||||
| Lactate, mmol/L (n= 1,070) | 2.3 (1.4 - 4.0) | 2.8 (1.7 - 4.9) | 2.4 (1.5 - 4.2) | 2.0 (1.2 - 3.4) | <0.001 |
PaCO2 < 35 mmHg;
PaCO2 35-45 mmHg;
PaCO2 > 45 mmHg.
Continuous variables are reported as mean (standard deviation) and median (interquartile range).
CHF: congestive heart failure; APACHE: acute physiology and chronic health evaluation; MAP: mean arterial pressure
modified score, which excludes Glasgow Coma Scale
P values are from the chi-square test for categorical variables, the one-way analysis of variance (ANOVA) for continuous variables, and the Kruskal-Wallis test (lactate).
Table 2 displays post-intubation ventilator variables. After initial initiation of mechanical ventilation in the ED (n= 2, 854 ventilator settings), the minority of subjects had changes to tidal volume (11%), respiratory rate (18%), PEEP (7%), and FiO2 (32%). Tidal volume also remained fairly stable after admission to the ICU (n= 20, 364 ventilator settings) with 55% of subjects having no change to tidal volume, 23% with one to two changes, and 22% having three or more changes during the first two weeks of their ICU stay. The median (IQR) PaCO2 for the entire cohort was 41 (34 - 53) mmHg. Three hundred and seventy-five (25%) subjects had exposure to hypocapnia, 547 (37%) had exposure to normocapnia, and 569 (38%) had exposure to hypercapnia. PaCO2 had a poor correlation with prescribed tidal volume (r = -0.06 P = 0.014) and respiratory rate (r = 0.14, P = <0.001). There was a modest correlation between PaCO2 and pH in the ED (r = -0.54, P < 0.001). Hypoxia in the ED (PaO2 < 60 mmHg) was more common among subjects with hypercapnia compared to normocapnia and hypocapnia, 8% vs. 2% and 1% respectively (P < 0.001).
Table 2. Ventilator variables in the emergency department and intensive care unit.
| All Subjects n = 1491 | Hypocapniaa n = 375 | Normocapniab n = 547 | Hypercapniac n = 569 | p - values | |
|---|---|---|---|---|---|
| Emergency Department | |||||
| Tidal volume (mL/kg PBW) | 7.5 (6.4 - 8.8) | 7.7 (6.8 - 8.8) | 7.3 (6.4 - 8.5) | 7.3 (6.4 - 8.7) | 0.002 |
| Tidal volume 8 ≤ mL/kg PBW | 937 (63) | 212 (57) | 364 (67) | 361 (63) | 0.008 |
| Respiratory rate | 16 (14 - 20) | 14 (14 - 20) | 16 (14 - 20) | 16 (14 - 20) | <0.001 |
| Minute Ventilation (mL/kg PBW*min) | 123 (105 - 146) | 122 (103 - 144) | 121 (103 - 138) | 128 (108 - 153) | <0.001 |
| FiO2 | 67 (40 - 100) | 70 (40 - 100) | 60 (40 - 100) | 70 (50 - 100) | 0.012 |
| PEEP | 5 (5 - 5) | 5 (5 - 5) | 5 (5 - 5) | 5 (5 - 7) | <0.001 |
| pH | 7.33 (7.23-7.41) | 7.41 (7.33-7.47) | 7.37 (7.31-7.41) | 7.24 (7.17-7.30) | <0.001 |
| PaCO2 (mmHg) | 41 (34 - 53) | 30 (26 - 33) | 39 (37 - 42) | 58 (50 - 75) | <0.001 |
| PaO2 (mmHg) | 142 (93 - 216) | 169 (113 - 239) | 146 (103 - 212) | 117 (80 - 201) | <0.001 |
| Hypoxia (PaO2 < 60 mmHg), n (%) | 66 (4) | 5 (1) | 12 (2) | 49 (8) | <0.001 |
| PaO2/FiO2 | 233 (142 - 345) | 278 (166 - 390) | 250 (166 - 355) | 195 (117 - 294) | <0.001 |
| Intensive care unit | |||||
| Tidal volume (mL/kg PBW) | 8.0 (7.0 - 9.0) | 8.0 (7.0 - 9.0) | 8.0 (7.0 - 9.0) | 8.0 (7.0 - 9.0) | 0.312 |
| FiO2 | 43 (40 - 51) | 43 (40 - 51) | 43 (40 - 50) | 45 (40 - 53) | <0.001 |
| PEEP | 5 (5 - 5) | 5 (5 - 5) | 5 (5 - 5) | 5 (5 - 6) | <0.001 |
| pH | 7.40 (7.35-7.43) | 7.41 (7.37-7.45) | 7.40 (7.36-7.43) | 7.38 (7.36-7.42) | <0.001 |
| Hypoxia (PaO2 < 60 mmHg), n (%) | 217 (15) | 51 (14) | 75 (14) | 91 (16) | 0.383 |
| Pulmonary Mechanics | |||||
| Plateau pressure (mmHg) | 20 (17 - 23) | 19 (17 - 22) | 19 (17 - 22) | 22 (19 - 25) | <0.001 |
| Plateau pressure > 30 mmHg, n (%) | 64 (4) | 8 (2) | 11 (2) | 45 (8) | <0.001 |
| Static compliance (mL/ cm H20) | 36 (28 - 46) | 39 (31 - 50) | 39 (29 - 50) | 32 (25 - 42) | <0.001 |
| Driving pressure (cm H2O) | 14 (11 - 18) | 12 (10 - 16) | 13 (10 - 17) | 15 (12 - 20) | <0.001 |
| Duration of mechanical ventilation (hours) | 77 (41 - 171) | 71 (42 - 177) | 83 (39 - 173) | 74 (42 - 166) | 0.937 |
Continuous variables are reported as median (interquartile range).
PaCO2 < 35 mmHg;
PaCO2 35-45 mmHg;
PaCO2 > 45 mmHg.
Pulmonary mechanics are the mean values during the first two weeks of mechanical ventilation, combining values from emergency department and intensive care unit.
PBW: predicted body weight; FiO2: fraction of inspired oxygen; PEEP: positive end-expiratory pressure; PaCO2: partial pressure of arterial carbon dioxide; PaO2: partial pressure of arterial oxygen: FiO2: fraction of inspired oxygen
P values are from the one-way analysis of variance (ANOVA) or Kruskal-Wallis test, based on data distribution.
Outcome analysis
The primary outcome of survival to hospital discharge occurred in 76% of subjects. Those with exposure to hypercapnia had a higher proportion of survival to hospital discharge when compared to subjects with normocapnia and hypocapnia, 85% vs. 74% and 66%, respectively (P < 0.001) (Figure 2).Table 3 displays the multivariable logistic regression model with PaCO2 treated as a categorical variable and survival to hospital discharge as the dependent variable. After adjusting for all identified significant confounders, including tidal volume, PaCO2 was an independent predictor of survival to hospital discharge [hypocapnia (aOR 0.65 (95% confidence interval [CI] 0.48 - 0.89), normocapnia (reference category), hypercapnia (aOR 1.83 (95% CI 1.32 - 2.54)].
Figure 2. Partial pressure of arterial carbon dioxide (PaCO2) and survival to hospital discharge among patients with hypocapnia, normocapnia, and hypercapnia.
Hypercapnia was associated with a higher proportion of survival to hospital discharge when compared to subjects with normocapnia and hypocapnia, 85% vs. 74% and 66%, respectively (P < 0.001).
Table 3.
Multivariable logistic regression model with partial pressure of arterial carbon dioxide (PaCO2) entered as a categorical variable and survival to hospital discharge as the dependent variable.
| Variables | aOR | 95% LCI | 95% UCI | Standard Error | p-value |
|---|---|---|---|---|---|
| Hypocapniaa | 0.64 | 0.46 | 0.88 | 0.10 | 0.006 |
| Normocapniab | Reference | ||||
| Hypercapniac | 2.00 | 1.43 | 2.79 | 0.34 | <0.001 |
| Age | 0.76 | 0.70 | 0.83 | 0.03 | <0.001 |
| ED Apache II | 0.98 | 0.96 | 1.00 | 0.01 | 0.034 |
| Malignancy | 0.51 | 0.37 | 0.71 | 0.09 | <0.001 |
| Reason for Mechanical Ventilation | |||||
| Respiratory failure | 2.45 | 1.52 | 3.95 | 0.60 | <0.001 |
| Trauma | 0.52 | 0.37 | 0.74 | 0.09 | <0.001 |
| ICU FiO2 | 0.97 | 0.96 | 0.98 | 0.01 | <0.001 |
| ICU ph | 1.41 | 1.19 | 1.67 | 0.12 | <0.001 |
| Vasopressor infusion | 0.68 | 0.48 | 0.95 | 0.12 | 0.025 |
| Antibiotics | 1.53 | 1.14 | 2.05 | 0.23 | 0.005 |
| Driving pressure | 0.97 | 0.95 | 1.00 | 0.01 | 0.022 |
PaCO2 < 35 mmHg;
PaCO2 35-45 mmHg;
PaCO2 > 45 mmHg.
Removed from model for non-significance: emergency department (ED) respiratory rate (p = 0.98), race (p = 0.66), ED tidal volume (p = 0.58), congestive heart failure (p = 0.87), partial pressure of arterial oxygen/fraction of inspired oxygen ratio (p = 0.62), ED pH (p = 0.54), ED hypotension (p = 0.55), ED positive end expiratory pressure (p = 0.33), ICU positive end expiratory pressure (p = 0.36), diabetes mellitus (p = 0.39), liver cirrhosis (p = 0.18), reason for mechanical ventilation: sepsis (p = 0.16), chronic obstructive pulmonary disease (p = 0.13), obesity (p = 0.07).
APACHE: acute physiology and chronic health evaluation; ED: emergency department; FiO2: fraction of inspired oxygen; aOR: adjusted odds ratio; LCI: lower confidence interval; UCI: upper confidence interval; ICU: intensive care unit
Figure 3 displays survival to hospital discharge rates across ascending ranges of PaCO2. There was a significant linear trend of increasing survival (chi-square for linear trend, P < 0.0001) across the range. There was an increasing odds of survival to hospital discharge with incremental increases in PaCO2 up to a range of 66 - 75 mmHg, which had the strongest association with survival to hospital discharge, [aOR 3.18 (95% CI 1.35 - 7.50)] (Table 4). There was a decrease in the strength of the relationship between PaCO2 and survival when PaCO2 was greater than 75 mmHg, [aOR 1.72 (95% CI 0.95 - 3.09)].
Figure 3. Partial pressure of arterial carbon dioxide (PaCO2) and survival to hospital discharge.
There was a significant linear trend of increasing survival (chi-square for linear trend, P < 0.0001) across the range.
Table 4.
Multivariable logistic regression model with partial pressure of arterial carbon dioxide (PaCO2), entered as a categorical variable using ascending ranges of PaCO2, and survival to hospital discharge as the dependent variable.
| Variables | aOR | 95% LCI | 95% UCI | Standard Error | p-value |
|---|---|---|---|---|---|
| PaCO2 (mmHg) | |||||
| < 35 | 0.64 | 0.46 | 0.88 | 0.10 | 0.006 |
| 35-45 | Reference | ||||
| 46-55 | 1.68 | 1.13 | 2.52 | 0.34 | 0.011 |
| 56-65 | 2.69 | 1.36 | 5.33 | 0.94 | 0.004 |
| 66-75 | 3.53 | 1.47 | 8.48 | 1.58 | 0.005 |
| >75 | 1.74 | 0.98 | 3.10 | 0.51 | 0.058 |
| Age | 0.76 | 0.70 | 0.83 | 0.03 | 0.000 |
| ED Apache II | 0.98 | 0.96 | 1.00 | 0.01 | 0.031 |
| Malignancy | 0.51 | 0.36 | 0.71 | 0.09 | 0.000 |
| Reason for Mechanical Ventilation | |||||
| Respiratory failure | 2.46 | 1.51 | 4.01 | 0.61 | 0.000 |
| Trauma | 0.52 | 0.37 | 0.73 | 0.09 | 0.000 |
| ICU FiO2 | 0.97 | 0.96 | 0.98 | 0.01 | 0.000 |
| ICU ph | 1.42 | 1.20 | 1.68 | 0.12 | 0.000 |
| Vasopressor infusion | 0.68 | 0.49 | 0.95 | 0.12 | 0.024 |
| Antibiotics | 1.52 | 1.13 | 2.04 | 0.23 | 0.005 |
| Driving pressure | 0.97 | 0.95 | 0.99 | 0.01 | 0.016 |
APACHE: acute physiology and chronic health evaluation; ED: emergency department; FiO2: fraction of inspired oxygen; aOR: adjusted odds ratio; LCI: lower confidence interval; UCI: upper confidence interval; ICU: intensive care unit
Subgroup analyses
When the analysis was restricted to patients mechanically ventilated for sepsis [aOR 2.47 (95% CI 1.40 - 4.33)], and trauma [aOR 2.94 (95% CI 1.46 - 5.90)], elevated PaCO2 remained an independent predictor of survival (Additional File 3, Table S3). After exclusion of patients with COPD (Additional File 4, Table S4), elevated PaCO2 was an independent predictor of survival [aOR 1.84 (95% CI 1.24 - 2.73)]. In the subgroup analysis that excluded patients mechanically ventilated for the indication of “Other” (Additional File 5, Table S5) elevated PaCO2 was an independent predictor of survival [aOR 2.74 (95% CI 1.90 - 3.96)]. In the subgroup analysis of patients that developed ARDS after admission (Additional File 6, Table S6), elevated PaCO2 was again an independent predictor of survival [aOR 2.49 (95% CI 1.04 - 5.94)]. Finally, in patients with lactate measured in the ED (Additional File 7, Table S7) elevated PaCO2 remained an independent predictor of survival [aOR 1.96 (95% CI 1.33 - 2.89)].
Discussion
The original goal of mechanical ventilation was the normalization of arterial oxygen and carbon dioxide tension[24]. With increasing knowledge of VILI, the dangers of hyperoxia, and the tolerance of hypercapnia, this therapeutic goal has appropriately shifted to protecting the patient from the mechanical power and oxygen delivered by the ventilator[22, 25-27]. While significant clinical data exists regarding VILI and hyperoxia, comparatively little has been devoted to the influence that PaCO2 management may have on outcome, and these analyses have focused primarily on the cardiac arrest population[6, 7, 14]. A previous secondary analysis found an association between hypercapnia (i.e. PaCO2 > 50 mmHg) and mortality among patients with ARDS [8]. It is possible this association between severe hypercapnia and mortality was driven by increased dead-space, which is common in patients with ARDS, and associated with increased mortality [28]. The focus of this investigation was to test the association between PaCO2 and survival among mechanically ventilated patients without ARDS, and the results have several implications.
First, hypocapnia and hypercapnia were common, occurring in 25% and 38% of patients respectively. This is consistent with previous data, demonstrating that deviations in PaCO2 away from normocapnia occur frequently during the earliest time period of mechanical ventilation[6, 7, 14]. There was also a poor correlation between PaCO2 and prescribed minute ventilation (tidal volume, respiratory rate). We are unable to explain factors potentially associated with this poor correlation, which could be related to factors such as increased dead space, hyperinflation, spontaneous respiratory efforts, or different sedation depth between groups. But given our observed association between PaCO2 and survival, this finding suggests that: 1) early, frequent arterial blood gas analysis should be used to titrate ventilator settings quickly; and/or 2) non-ventilator parameters that affect arterial CO2 should be addressed (i.e. temperature, neuromuscular blockade, and shock resuscitation).
The main finding of this study was an association between higher PaCO2 levels and survival. These findings were stable across models and subgroup analyses. Furthermore, PaCO2 was adjusted for lung-protective tidal volume, driving pressure, and arterial pH, which suggests that hypercapnia may have exerted a beneficial effect independent of pulmonary mechanics or acidosis. When placed in context of the frequency of abnormalities in arterial CO2 tensions during early mechanical ventilation, our findings suggest that early hypercapnia (or avoiding hypocapnia) could be a therapeutic target going forward, and provide scientific rationale for future studies to determine the optimal PaCO2 range for patients undergoing acute mechanical ventilation.
There are several strengths to the study. The sample size is large and the population of ventilated patients is diverse, enhancing external validity. Our results also remained consistent across several analyses, suggesting internal consistency. These data also have biological plausibility in that pre-clinical data shows hypercapnia can modulate inflammation, VILI, and immunity [9, 11, 12, 29, 30]. Hypercapnia can decrease some key components of inflammatory pathways, such as tumor necrosis factor-α and interleukin (IL)-1, which contribute to tissue injury [9, 31]. Hypercapnia has also been found to attenuate neutrophil activity through lowered intracellular pH [12, 32]. In pulmonary endothelial cells, hypercapnia decreases the DNA-binding activity of nuclear factor (NF)-κB, a regulator of pro-inflammatory pathways. This decreased binding attenuates IL-8 production, with a commensurate decrease in cell injury[33].
There are also several limitations to the study. First, as an observational cohort study, we can only describe associations and not causal inference. The dose-response association between increasing PaCO2 levels (up to a point) and survival seen in this study is consistent with previous pre-clinical data, and could suggest causality [34]. While we hypothesize that the association between hypercapnia and survival is secondary to hypercapnia-induced gene modification and anti-inflammation, without a biomarker assessment, we are unable to determine the direct effects of hypercapnia on end-organ damage or survival. Second, although we used multivariable logistic regression analyses to adjust for multiple potential confounders, there still exists the possibility that a CO2-associated confounder (i.e. low tidal volume, pH) drove the observed association between hypercapnia and survival. There was no observed difference in ICU tidal volume between the groups, and our statistical models adjusted for these confounders and were consistent in their results. It is also possible that unmeasured confounders which may affect both PaCO2 levels as well as outcome (e.g. pulmonary embolism, cardiac output, treatment/resuscitation variables) were responsible for the observed associations. We also did not formally study the time from intubation until the blood gas analyses were obtained; it is possible that the incidence of PaCO2 derangements, as well as the strength of the association with outcome, could have been different had this been standardized. Going forward, this level of granular data will be critical to examine in a prospective fashion in order to draw stronger conclusion regarding PaCO2 and outcome. Third, hypercapnia may also reflect a patient population that is less ill and therefore has a higher likelihood to survive. A higher APACHE II score and higher incidence of hypotension in the hypercapnic group suggests this was not the case. Finally, this is a very heterogeneous sample of mechanically ventilated patients. This, along with the study design, make it impossible to draw any conclusive results, and this study should be viewed as exploratory in nature and hypothesis-generating for future clinical studies. Although we adjusted the multivariable logistic regression model for the indication for mechanical ventilation, and performed multiple subgroup analyses, further studies are needed to test the association between PaCO2 and outcomes among specific, more homogeneous, patient populations. Other patient-centered clinical outcomes, such as lengths of stay and ventilator duration, should also be examined with these future studies.
Conclusions
In this observational study, derangements in arterial CO2 tension were common during the earliest phases of mechanical ventilation, and higher PaCO2 levels are associated with increased survival to hospital discharge. Given the continued high mortality seen in mechanically ventilated patients, and the heterogeneity seen within this cohort, future research should focus on interventions which are relatively simple and scalable to real world conditions. Our current findings, placed in context of previous data, suggest that future investigation into early manipulation of arterial CO2 tension is warranted.
Supplementary Material
Additional file 1: Table S1. STROBE checklist for reporting items in cohort studies.
Additional file 2: Table S2. Candidate covariables for the multivariable model.
Additional file 3: Table S3. Subgroup analysis of sepsis and trauma patients.
Additional file 4: Table S4. Subgroup analysis excluding patients with COPD.
Additional file 5: Table S5. Subgroup analysis excluding patients intubated for “Other”.
Additional file 6: Table S6. Subgroup analysis of patients who developed ARDS.
Additional file 7: Table S7. Subgroup analysis of patients with lactate values measured.
Highlights.
Hypocapnia and hypercapnia occur frequently after initiation of mechanical ventilation.
Higher PaCO2 levels are associated with increased survival.
Over ascending ranges of PaCO2, there is a linear trend of increasing survival up to a PaCO2 range of 66 - 75 mmHg, which has the strongest survival association
Acknowledgments
Not applicable.
Declarations: Funding: BMF and AMD were funded by the KL2 Career Development Award, and this research was supported by the Washington University Institute of Clinical and Translational Sciences (Grants UL1 TR000448 and KL2 TR000450) from the National Center for Advancing Translational Sciences (NCATS). BMF was also funded by the Foundation for Barnes-Jewish Hospital Clinical and Translational Sciences Research Program (Grant # 8041-88). AMD was also funded by the Foundation for Anesthesia Education and Research. NMM was supported by grant funds from the Health Resources and Services Administration. MHK was supported by the Barnes-Jewish Hospital Foundation. BWR was supported by a grant from the National Institutes of Health/National Heart, Lung, and Blood Institute (K23HL126979). Funders played no role in the design and conduct of the study, nor the collection, management, analysis, and interpretation of the data, nor the preparation, review, or approval of the manuscript.
Abbreviations
- ABG
arterial blood gas
- ANOVA
one-way analysis of variance
- aOR
adjusted odds ratio
- APACHE
Acute Physiology and Chronic Health Evaluation
- ARDS
acute respiratory distress syndrome
- CI
confidence interval
- COPD
chronic obstructive pulmonary disease
- CRS
static compliance of the respiratory system
- ED
emergency department
- FiO2
fraction of inspired oxygen
- ICU
intensive care unit
- IL
interleukin
- IQR
interquartile range
- NF
nuclear factor
- PaCO2
partial pressure of arterial carbon dioxide
- PBW
predicted body weight
- PEEP
positive end-expiratory pressure
- SD
standard deviation
- STROBE
Strengthening the Reporting of Observational Studies in Epidemiology
- VILI
ventilator-induced ung injury
Footnotes
The authors declare that they have no conflicts of interest or financial disclosures to declare.
Ethics approval and consent to participate: This study was approved by the Human Research Protection Office of Washington University in St. Louis under waiver of informed consent.
Consent for publication: Not applicable.
Author Contributions: BMF had full access to all of the data and takes responsibility for the integrity of the data and content of the manuscript, including the data and analysis. BMF, MHK, and BWR contributed to the conception and design of the study. ITF, BMF, and AMD contributed to the acquisition of the data. BWR did the analysis. BMF and BWR wrote the first and final draft. All authors were involved in interpretation of the data, and for critical writing and revisions of the manuscript for important intellectual content. All authors provided approval of the final version to be published.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Contributor Information
Brian M. Fuller, Departments of Emergency Medicine and Anesthesiology, Division of Critical Care Medicine, Washington University School of Medicine in St. Louis, St. Louis, MO 63110, Telephone: 314.7475368, Fax: 3143620419.
Nicholas M. Mohr, Departments of Emergency Medicine and Anesthesiology, Division of Critical Care Medicine, Roy J. and Lucille A. Carver College of Medicine, University of Iowa, 200 Hawkins Drive, 1008 RCP, Iowa City, IA 52242.
Anne M. Drewry, Department of Anesthesiology, Division of Critical Care Medicine, Washington University School of Medicine in St. Louis, St. Louis, MO 63110.
Ian T. Ferguson, School of Medicine and Medical Science, University College Dublin, Dublin 4, Ireland.
Stephen Trzeciak, Departments of Medicine and Emergency Medicine, Division of Critical Care Medicine, Cooper University Hospital, One Cooper Plaza, K152, Camden, NJ 08103.
Marin H. Kollef, Department of Medicine, Division of Pulmonary and Critical Care Medicine, Washington University School of Medicine in St. Louis, St. Louis, MO 63110.
Brian W. Roberts, Department of Emergency Medicine, Cooper University Hospital, One Cooper Plaza, K152, Camden, NJ 08103.
References
- 1.Easter BD, Fischer C, Fisher J. The use of mechanical ventilation in the ED. Am J Emerg Med. 2012;30(7):1183–1188. doi: 10.1016/j.ajem.2011.09.008. [DOI] [PubMed] [Google Scholar]
- 2.Wunsch H, Linde-Zwirble WT, Angus DC, Hartman ME, Milbrandt EB, Kahn JM. The epidemiology of mechanical ventilation use in the United States. Critical care medicine. 2010;38(10):1947–1953. doi: 10.1097/CCM.0b013e3181ef4460. [DOI] [PubMed] [Google Scholar]
- 3.Needham DM, Bronskill SE, Calinawan JR, Sibbald WJ, Pronovost PJ, Laupacis A. Projected incidence of mechanical ventilation in Ontario to 2026: Preparing for the aging baby boomers. Critical care medicine. 2005;33(3):574–579. doi: 10.1097/01.ccm.0000155992.21174.31. [DOI] [PubMed] [Google Scholar]
- 4.Esteban A, Anzueto A, Frutos F, Alía I, Brochard L, Stewart TE, et al. Characteristics and outcomes in adult patients receiving mechanical ventilation: a 28-day international study. Jama. 2002;287(3):345–355. doi: 10.1001/jama.287.3.345. [DOI] [PubMed] [Google Scholar]
- 5.Kahn JM, Goss CH, Heagerty PJ, Kramer AA, O'Brien CR, Rubenfeld GD. Hospital volume and the outcomes of mechanical ventilation. New England Journal of Medicine. 2006;355(1):41–50. doi: 10.1056/NEJMsa053993. [DOI] [PubMed] [Google Scholar]
- 6.Roberts BW, Karagiannis P, Coletta M, Kilgannon JH, Chansky ME, Trzeciak S. Effects of PaCO 2 derangements on clinical outcomes after cerebral injury: A systematic review. Resuscitation. 2015;91:32–41. doi: 10.1016/j.resuscitation.2015.03.015. [DOI] [PubMed] [Google Scholar]
- 7.Roberts BW, Kilgannon JH, Chansky ME, Mittal N, Wooden J, Trzeciak S. Association between Post-Resuscitation Partial Pressure of Arterial Carbon Dioxide and Neurological Outcome in Patients with Post-Cardiac Arrest Syndrome. Circulation. 2013 doi: 10.1161/CIRCULATIONAHA.112.000168. CIRCULATIONAHA. 112.000168. [DOI] [PubMed] [Google Scholar]
- 8.Nin N, Muriel A, Peñuelas O, Brochard L, Lorente JA, Ferguson ND, et al. Severe hypercapnia and outcome of mechanically ventilated patients with moderate or severe acute respiratory distress syndrome. Intensive Care Medicine. 2017;43(2):200–208. doi: 10.1007/s00134-016-4611-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Laffey JG, Tanaka M, Engelberts D, Luo X, Yuan S, Keith Tanswell A, et al. Therapeutic hypercapnia reduces pulmonary and systemic injury following in vivo lung reperfusion. American journal of respiratory and critical care medicine. 2000;162(6):2287–2294. doi: 10.1164/ajrccm.162.6.2003066. [DOI] [PubMed] [Google Scholar]
- 10.O'Toole D, Hassett P, Contreras M, Higgins B, McKeown ST, McAuley DF, et al. Hypercapnic acidosis attenuates pulmonary epithelial wound repair by an NF-κB dependent mechanism. Thorax. 2009;64(11):976–982. doi: 10.1136/thx.2008.110304. [DOI] [PubMed] [Google Scholar]
- 11.Shibata K, Cregg N, Engelberts D, Takeuchi A, Fedorko L, Kavanagh BP. Hypercapnic acidosis may attenuate acute lung injury by inhibition of endogenous xanthine oxidase. American journal of respiratory and critical care medicine. 1998;158(5):1578–1584. doi: 10.1164/ajrccm.158.5.9804039. [DOI] [PubMed] [Google Scholar]
- 12.Sinclair SE, Kregenow DA, Lamm WJ, Starr IR, Chi EY, Hlastala MP. Hypercapnic acidosis is protective in an in vivo model of ventilator-induced lung injury. American journal of respiratory and critical care medicine. 2002;166(3):403–408. doi: 10.1164/rccm.200112-117OC. [DOI] [PubMed] [Google Scholar]
- 13.Eastwood GM, Tanaka A, Bellomo R. Cerebral oxygenation in mechanically ventilated early cardiac arrest survivors: The impact of hypercapnia. Resuscitation. 2016;102:11–16. doi: 10.1016/j.resuscitation.2016.02.009. [DOI] [PubMed] [Google Scholar]
- 14.Schneider AG, Eastwood GM, Bellomo R, Bailey M, Lipcsey M, Pilcher D, et al. Arterial carbon dioxide tension and outcome in patients admitted to the intensive care unit after cardiac arrest. Resuscitation. 2013;84(7):927–934. doi: 10.1016/j.resuscitation.2013.02.014. [DOI] [PubMed] [Google Scholar]
- 15.Vaahersalo J, Bendel S, Reinikainen M, Kurola J, Tiainen M, Raj R, et al. Arterial blood gas tensions after resuscitation from out-of-hospital cardiac arrest: associations with long-term neurologic outcome. Critical care medicine. 2014;42(6):1463–1470. doi: 10.1097/CCM.0000000000000228. [DOI] [PubMed] [Google Scholar]
- 16.Curley G, Laffey JG, Kavanagh BP. Bench-to-bedside review: carbon dioxide. Critical Care. 2010;14(2):220. doi: 10.1186/cc8926. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Fuller BM, Ferguson I, Mohr NM, Stephens RJ, Briscoe CC, Kolomiets AA, et al. Lung-protective ventilation initiated in the emergency department (LOV-ED): a study protocol for a quasi-experimental, before-after trial aimed at reducing pulmonary complications. BMJ Open. 2016;6(4):e010991. doi: 10.1136/bmjopen-2015-010991. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Fuller BM, Ferguson IT, Mohr NM, Drewry AM, Palmer C, Wessman BT, et al. Lung-Protective Ventilation Initiated in the Emergency Department (LOV-ED): A Quasi-Experimental, Before-After Trial. Annals of Emergency Medicine. 2017 doi: 10.1016/j.annemergmed.2017.01.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Acute respiratory distress syndrome. The ARDS definition task force. JAMA. 2012;307(23):2526–2533. doi: 10.1001/jama.2012.5669. [DOI] [PubMed] [Google Scholar]
- 20.von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: Guidelines for Reporting Observational Studies. Annals of internal medicine. 2007;147(8):573–577. doi: 10.7326/0003-4819-147-8-200710160-00010. [DOI] [PubMed] [Google Scholar]
- 21.Bone RC, Balk RA, Cerra FB, Dellinger RP, Fein AM, Knaus WA, et al. Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee American College of Chest Physicians/Society of Critical Care Medicine. Chest. 1992;101(6):1644–1655. doi: 10.1378/chest.101.6.1644. [DOI] [PubMed] [Google Scholar]
- 22.Ventilation with lower tidal volumes as compared with traditional tidal volumes for acute lung injury and the acute respiratory distress syndrome. The Acute Respiratory Distress Syndrome Network. N Engl J Med. 2000;342(18):1301–1308. doi: 10.1056/NEJM200005043421801. [DOI] [PubMed] [Google Scholar]
- 23.Neto AS, Barbas CSV, Simonis FD, Artigas-Raventós A, Canet J, Determann RM, et al. Epidemiological characteristics, practice of ventilation, and clinical outcome in patients at risk of acute respiratory distress syndrome in intensive care units from 16 countries (PRoVENT): an international, multicentre, prospective study. The Lancet Respiratory Medicine. 2016;4(11):882–893. doi: 10.1016/S2213-2600(16)30305-8. [DOI] [PubMed] [Google Scholar]
- 24.Pontoppidan H, Geffin B, Lowenstein E. Acute respiratory failure in the adult. New England Journal of Medicine. 1972;287(14):690–698. doi: 10.1056/NEJM197210052871404. [DOI] [PubMed] [Google Scholar]
- 25.Girardis M, Busani S, Damiani E, et al. Effect of conservative vs conventional oxygen therapy on mortality among patients in an intensive care unit: The oxygen-icu randomized clinical trial. JAMA. 2016:E1–E7. doi: 10.1001/jama.2016.11993. [DOI] [PubMed] [Google Scholar]
- 26.Stub D, Smith K, Bernard S, Nehme Z, Stephenson M, Bray JE, et al. Air versus oxygen in ST-segment elevation myocardial infarction. Circulation. 2015 doi: 10.1161/CIRCULATIONAHA.114.014494. CIRCULATIONAHA. 114.014494. [DOI] [PubMed] [Google Scholar]
- 27.Gattinoni L, Tonetti T, Cressoni M, Cadringher P, Herrmann P, Moerer O, et al. Ventilator-related causes of lung injury: the mechanical power. Intensive care medicine. 2016;42(10):1567–1575. doi: 10.1007/s00134-016-4505-2. [DOI] [PubMed] [Google Scholar]
- 28.Nuckton TJ, Alonso JA, Kallet RH, Daniel BM, Pittet JF, Eisner MD, et al. Pulmonary dead-space fraction as a risk factor for death in the acute respiratory distress syndrome. New England Journal of Medicine. 2002;346(17):1281–1286. doi: 10.1056/NEJMoa012835. [DOI] [PubMed] [Google Scholar]
- 29.Cummins EP, Oliver KM, Lenihan CR, Fitzpatrick SF, Bruning U, Scholz CC, et al. NF-κB links CO2 sensing to innate immunity and inflammation in mammalian cells. The Journal of Immunology. 2010;185(7):4439–4445. doi: 10.4049/jimmunol.1000701. [DOI] [PubMed] [Google Scholar]
- 30.Ni Chonghaile M, Higgins B, Costello J, Laffey J. Hypercapnic acidosis attenuates severe acute bacterial pneumonia-induced lung injury by a neutrophil-independent mechanism. Critical care medicine. 2008;36(12):3135. doi: 10.1097/CCM.0b013e31818f0d13. [DOI] [PubMed] [Google Scholar]
- 31.West MA, Baker J, Bellingham J. Kinetics of decreased LPS-stimulated cytokine release by macrophages exposed to CO 2. Journal of Surgical Research. 1996;63(1):269–274. doi: 10.1006/jsre.1996.0259. [DOI] [PubMed] [Google Scholar]
- 32.Coakley R, Taggart C, Greene C, McElvaney N, O'Neill S. Ambient pCO2 modulates intracellular pH, intracellular oxidant generation, and interleukin-8 secretion in human neutrophils. Journal of leukocyte biology. 2002;71(4):603–610. [PubMed] [Google Scholar]
- 33.Takeshita K, Suzuki Y, Nishio K, Takeuchi O, Toda K, Kudo H, et al. Hypercapnic acidosis attenuates endotoxin-induced nuclear factor-κB activation. American journal of respiratory cell and molecular biology. 2003;29(1):124–132. doi: 10.1165/rcmb.2002-0126OC. [DOI] [PubMed] [Google Scholar]
- 34.Vannucci RC, Towfighi J, Brucklacher RM, Vannucci SJ. Effect of extreme hypercapnia on hypoxic-ischemic brain damage in the immature rat. Pediatric research. 2001;49(6):799–803. doi: 10.1203/00006450-200106000-00015. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Additional file 1: Table S1. STROBE checklist for reporting items in cohort studies.
Additional file 2: Table S2. Candidate covariables for the multivariable model.
Additional file 3: Table S3. Subgroup analysis of sepsis and trauma patients.
Additional file 4: Table S4. Subgroup analysis excluding patients with COPD.
Additional file 5: Table S5. Subgroup analysis excluding patients intubated for “Other”.
Additional file 6: Table S6. Subgroup analysis of patients who developed ARDS.
Additional file 7: Table S7. Subgroup analysis of patients with lactate values measured.