There appears to be considerable variability in the approach physicians use to manage arterial carbon dioxide tensions, in patients in the early phases [first 48 hours] of ARDS (Acute hypoxemic respiratory failure and). A number of specific concerns exist, particularly the use of greater than needed inspired oxygen concentrations (potentially in 40% patients), and the proportion of hypocapnic patients in our cohort.

This study will address a number of specific concerns, including:

1. the incidence of hypocapnia and hypercapnia;
2. the factors (e.g. ventilatory mode) contributing to hypocapnia and hypercapnia;
3. the impact of hypocapnia and hypercapnia on progression of ARDS, duration of ventilation, ICU stay, hospital stay and patient outcome;
4. the impact of pH (respiratory versus metabolic cause) on outcome variables.

LUNG SAFE was a prospective observational study. There was not patient intervention. The current analysis is a secondary analysis of this existing database.

Data will be abstracted form the database on all patients meeting the inclusion criteria for ARDS (during the first 2 days from AHRF onset) for the following variables: Patient's demographic characteristics (age, sex) and chronic comorbidities, Arterial PCO2, pH in first 48 hours, Data on P/F ratio over course of illness, Duration of ventilation, ICU stay, hospital survival in patients identified as being exposed to hypocapnia (pCO2 < or = to 30mmHg) and hypercapnia (PCO2 > or = to 45mmHg); Main ventilatory variables: mode of ventilation, tidal volume, respiratory.

Descriptive statistics will include proportions for categorical and mean (standard deviation) or median (interquartile range) for continuous variables, according the data distribution. The amount of missing data in the LUNG SAFE database is low so no assumptions will be made for missing data.

In order to investigate the relationship between PaCO2 and management factors, the investigators will stratify study population according the presence or absence of hypercapnia and will evaluate differences in management factors, using appropriate statistical tests. In detail, proportions will be compared using chi-squared or Fisher exact tests and mean values were compared using T-test or Wilcoxon rank sum test, as appropriate. Shapiro-Wilks test will be used to assess normality in data distribution. Moreover, they will perform univariate and multivariable logistic regression models using hypercapnia as dependent variable. Stepwise regression approach with significance alpha levels of 0.05 (both for entry and retention) will be used to establish a set of independent predictors. Odds ratios (ORs) and 95% confidence intervals (CIs) will be reported as measure of association.

They will assess the association of hypercapnia with different outcomes (ARDS progression, duration of mechanical ventilation, ICU and hospital stay and mortality) using generalized linear regression models (Poisson or Logistic regression models, according outcome distribution) and adjusting the relationship with all possible confounders identified with stepwise approach. Results will be reported as ORs, incidence rate ratios and corresponding 95% CI.

Particular attention will be given to the relationship and interaction between pH and PaCO2.

Survival analysis (Kaplan-Meier approach) will be used to estimate the time of liberation from invasive mechanical ventilation, of ICU and hospital discharge, and of hospital mortality within 28 days of AHRF onset in patients with and without hypercapnia. Differences in survival time will be assess by Log-Rank test.

Inclusion Criteria:

• All patients meeting criteria for ARDS. Fulfillment of criteria for Acute Hypoxaemic respiratory failure (AHRF) and fulfillment of Berlin criteria for ARDS during the first two days after AHRF onset.

Exclusion Criteria:

• Age < 18

Patients over 18 years of age who fulfilled clinical criteria for ARDS and Acute Hypoxaemic respiratory failure were recruited into the LUNG SAFE trial (NCT02010073) that was published in JAMA, 2016 Jul 19;316(3):347).