Prospective Observational Study of Exhaled Carbon MonoxideTrajectories and Outcomes in ICU Patients

Background: Endogenous carbon monoxide is generated in the human body and may increase during critical illness due to hemolysis, oxidative stress, and inflammation. Whether longitudinal patterns of exhaled carbon monoxide (eCO) provide prognostic information in critically ill patients remains unclear.

Design and setting: This is a single-center, prospective, observational cohort study conducted in an adult intensive care unit (ICU). The study is designed to characterize longitudinal eCO patterns and evaluate their clinical significance in critically ill patients without altering routine clinical management.

Study procedures and data collection: eCO will be measured once daily from Day 0 (baseline, within 24 hours of ICU admission/enrollment) through Day 7 when feasible, using a standardized breath analysis device. For patients receiving mechanical ventilation, breath sampling will be performed according to a standardized protocol for ventilated patients. In addition to serial eCO measurements, demographic characteristics, admission diagnosis, comorbidities, severity-of-illness scores, key treatments, and laboratory variables relevant to hemolysis, inflammation, organ function, anemia, and iron metabolism will be collected from the electronic medical record and protocol-defined assessments, as applicable.

Analytic approach: Group-based trajectory modeling (GBTM) will be used to identify distinct eCO trajectory groups from repeated measurements. Associations between eCO trajectories and prespecified clinical outcomes will be evaluated using Kaplan-Meier methods and multivariable Cox proportional hazards models with adjustment for clinically relevant covariates. Additional exploratory analyses will assess relationships between eCO trajectories and laboratory biomarker patterns, including anemia-related, iron metabolism-related, and inflammatory indices. Machine learning methods (e.g., XGBoost with SHAP-based interpretation) may also be used in exploratory analyses to develop risk prediction models incorporating eCO and other clinical variables.

Study conduct: No intervention is assigned as part of the study, and all treatment decisions remain at the discretion of the treating clinicians.

Inclusion criteria

• Age ≥ 18 years
• Admitted to the intensive care unit (ICU) at the study site
• Expected ICU length of stay ≥ 48 hours
• Able to complete exhaled carbon monoxide (eCO) measurement (spontaneously breathing or mechanically ventilated)
• Written informed consent provided by the patient or a legally authorized representative

Exclusion criteria

• Carbon monoxide poisoning
• Active smoking within 24 hours prior to ICU admission
• Unable to perform breath sampling (e.g., massive airway bleeding, severe airway obstruction, extracorporeal membrane oxygenation)
• Ineffective sampling due to poor mask seal or special ventilation modes
• Pregnant or breastfeeding
• Known hereditary hemoglobinopathies (e.g., sickle cell disease, thalassemia)
• Patient or family refusal
• Other conditions deemed unsuitable by investigators