Indices of pulmonary oxygenation in pathological lung states: an investigation using high-fidelity, computational modelling.

BACKGROUND Existing indices of pulmonary oxygenation vary misleadingly with external factors such as inspired oxygen fraction (FI(O2)), arterial carbon dioxide tension (PaCO2), and haemoglobin (Hb). Previous work suggested that some indices may be acceptably useful in particular scenarios such as acute respiratory distress syndrome (ARDS) or where FI(O2)>60%. However, it is not possible to identify such scenarios in most clinical contexts; therefore we aimed to examine the induced variability of existing indices in a population of patients with a variety of lung defects. METHODS We configured nine virtual patients within the Nottingham Physiology Simulator, each with a unique pulmonary configuration but identical arterial blood gases at FI(O2) 30%, PaCO2 6.0 kPa and Hb 8.0 g dl(-1). Factors (FI(O2), P(CO2), Hb) were varied independently and indices of oxygenation including calculated venous admixture (Qs/Qt), arterial oxygen tension (PaO2/FI(O2)), arterio-alveolar gas tension gradient (PA-aO2), and respiratory index (PA-aO2/PaO2) were recorded. RESULTS All indices varied with FI(O2), with greatest variation with lung defects having least true (absolute) shunt. Calculated Qs/Qt resisted induced variation best of all the indices, but varied by 30% of its mean value during FI(O2) variation. PaO2/FI(O2) varied greatly, especially during variation in FI(O2) (up to 74% of its average value), and most markedly in defects with little true (absolute) shunt. PaCO2 and Hb variation caused small, consistent changes in all indices that were similar between lung-states. CONCLUSIONS No existing index of oxygenation adequately describes the severity of gas exchange defect. Existing indices of oxygenation vary with disease severity, disease type, and external factors such as FI(O2). A novel and robust index is needed.