Association Between Early Postoperative Fraction of Inspired Oxygen and Pulmonary Complications After Coronary Artery Bypass Grafting: A Retrospective Cohort Study Using Public Databases

Technical Summary: Association Between Early Postoperative Fraction of Inspired Oxygen and Pulmonary Complications After Coronary Artery Bypass Grafting

Problem Statement
Patients undergoing coronary artery bypass grafting (CABG) frequently require postoperative mechanical ventilation and oxygen therapy. While increasing the fraction of inspired oxygen (FiO₂) is standard practice to prevent hypoxemia, the safety of higher FiO₂ exposure remains uncertain. Hyperoxia is known to potentially induce oxidative stress, absorption atelectasis, and alveolar injury, which could impair pulmonary recovery. However, existing literature lacks robust real-world evidence specifically linking early postoperative ICU FiO₂ exposure to a composite of pulmonary complications (PPCs) in CABG patients. Furthermore, the dose-response relationship and the external reproducibility of this association across different clinical settings have not been fully evaluated.

Methodology
This study employed a retrospective cohort design utilizing two large, publicly available critical care databases:

• Derivation Cohort: MIMIC-IV version 3.1, comprising 5,457 postoperative CABG patients.
• External Validation Cohort: The eICU Collaborative Research Database, comprising 4,521 postoperative CABG patients.

Exposure Definition:
Early postoperative FiO₂ exposure was defined as the mean FiO₂ recorded during the first 6 hours following ICU admission. Patients were categorized into three groups:

1. FiO₂ ≤ 0.60 (Reference group)
2. FiO₂ 0.61–0.80
3. FiO₂ > 0.80

Additionally, FiO₂ was analyzed as a continuous variable (per 0.1 increase).

Outcome Definition:
The primary outcome was a composite of postoperative pulmonary complications (PPCs), defined as the occurrence of any of the following during the postoperative ICU stay:

• Prolonged mechanical ventilation (> 48 hours)
• Reintubation
• Pneumonia

Statistical Analysis:
The study utilized multivariable logistic regression with stepwise adjustments:

• Model 1: Unadjusted.
• Model 2: Adjusted for demographics and baseline comorbidities (age, sex, BMI, race, COPD, CHF, CKD, diabetes, hypertension).
• Model 3: Further adjusted for early physiological status, ventilatory parameters (PEEP, heart rate, MAP), and laboratory variables (lactate, hemoglobin, creatinine, WBC).
• Model 4: Further adjusted for peri-ICU interventions (vasoactive drugs, transfusion, renal replacement therapy), primarily estimated in the eICU cohort due to data availability in MIMIC-IV.

Advanced analytical techniques included:

• Restricted Cubic Spline (RCS) Analysis: To evaluate non-linear dose-response relationships.
• Propensity Score Weighting: Generalized overlap weighting and inverse probability of treatment weighting (IPTW) were used in MIMIC-IV to mitigate baseline imbalance and indication bias.
• Subgroup Analysis: Stratified by age, COPD status, and obesity.

Key Results

• Incidence Trends: In both cohorts, PPC incidence increased stepwise with higher FiO₂ categories.
  • MIMIC-IV: 5.51% (≤0.60), 8.10% (0.61–0.80), and 14.61% (>0.80).
  • eICU: 9.41% (≤0.60), 14.25% (0.61–0.80), and 26.52% (>0.80).
• Multivariable Associations:
  • In the fully adjusted MIMIC-IV model, FiO₂ > 0.80 remained significantly associated with increased PPC risk (OR 1.69, 95% CI 1.16–2.48; P = 0.007). The association for the 0.61–0.80 group became non-significant after full adjustment.
  • In the eICU cohort, after adjusting for peri-ICU interventions (Model 4), both intermediate (0.61–0.80) and high (>0.80) FiO₂ groups showed significant associations with PPCs (OR 1.51 and OR 2.39, respectively).
• Continuous Exposure: Each 0.1 increase in FiO₂ was associated with increased PPC risk in both cohorts (MIMIC-IV fully adjusted OR 1.13; eICU Model 4 OR 1.25).
• Dose-Response: RCS analyses confirmed a significant overall linear association between FiO₂ and PPC risk in both cohorts without significant nonlinearity.
• Robustness: Propensity score weighting and sensitivity analyses in MIMIC-IV confirmed the monotonic increase in PPC risk across FiO₂ groups, supporting the robustness of the findings against confounding.

Significance and Claims
The authors claim that this study provides the first systematic evaluation and external validation of the association between early postoperative FiO₂ exposure and PPCs in CABG patients using large-scale public databases.

• Clinical Implication: The findings suggest that higher early postoperative FiO₂ exposure is independently associated with an increased risk of pulmonary complications. This supports a shift in clinical practice toward individualized oxygen titration, aiming to avoid unnecessary high FiO₂ exposure once adequate oxygenation is maintained, rather than maintaining high FiO₂ levels indiscriminately.
• Methodological Contribution: The study demonstrates the utility of using multiple large databases (MIMIC-IV and eICU) to validate critical care findings across different patient populations and clinical practice patterns. The use of continuous exposure analysis and RCS modeling offers a more nuanced understanding of the dose-response relationship compared to categorical analysis alone.
• Limitations Acknowledged: The authors explicitly state that as a retrospective observational study, causality cannot be established. Residual confounding remains possible, and the study relies on database records which may be subject to documentation variations. Furthermore, specific intraoperative factors (e.g., bypass duration) were not fully available for adjustment.

In conclusion, the paper posits that while oxygen therapy is essential, the "more is better" approach regarding FiO₂ may be detrimental to pulmonary outcomes in post-CABG patients, warranting prospective studies to define optimal titration strategies.