Normocapnia Versus Mild Hypercapnia During Lung Protective Ventilation in Acute Respiratory Distress Syndrome: A Target Trial Emulation

OBJECTIVES: This trial emulation aimed to answer the question: Among patients with acute respiratory distress syndrome (ARDS) who receive lung protective ventilation with low tidal volume, is normocapnia through an increase in respiratory rate associated with lower mortality than mild hypercapnia? DESIGN: Target trial emulation. SETTING: Hospitals of the ARDS Network and Prevention and Early Treatment of Acute Lung Injury (PETAL) Network. PATIENTS: Patients with ARDS enrolled in seven ARDS Network and PETAL Network randomized controlled trials. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: A target trial was emulated, assigning participants in a 1:1 ratio to normocapnia (target Paco2 35-45 mm Hg) or mild hypercapnia (target Paco2 50-55 mm Hg) for 4 days following the initiation of lung protective ventilation. The primary outcome was 90-day in-hospital mortality. Of 5367 potentially eligible subjects, 292 patients were assigned to normocapnia and 292 to mild hypercapnia. Separation in Paco2 values was preserved over the 4-day period of emulated intervention. Both groups had median pH greater than or equal to 7.33. Bicarbonate levels, driving pressure, tidal volume, and dead-space fraction did not differ clinically significantly between groups, while median respiratory rate was 33.0 breaths/min in the normocapnia vs. 28.0 breaths/min in the mild hypercapnia group (p < 0.001). Death within 90 days occurred in 84 of 292 patients (28.8%) in the normocapnia group and in 84 of 292 patients (28.8%) in the mild hypercapnia group (log-rank p = 0.842). This result persisted in subgroups such as older (≥ 65 yr) and obese (body mass index ≥ 30 kg/m2) patients. CONCLUSIONS: Among patients with ARDS who received lung protective ventilation, normocapnia through an increase in respiratory rate was not associated with lower mortality than mild hypercapnia. As this finding is derived from a trial emulation, it should be interpreted cautiously given the potential for significant residual bias.