RESUMEN
OBJECTIVES: Studies correlating the arterial partial pressure of oxygen to the fraction of nonaerated lung assessed by CT shunt yielded inconsistent results. We systematically analyzed this relationship and scrutinized key methodological factors that may compromise it. We hypothesized that both physiological shunt and the ratio between PaO2 and the fraction of inspired oxygen enable estimation of CT shunt at the bedside. DESIGN: : Prospective observational clinical and laboratory animal investigations. SETTING: ICUs (University Hospital Leipzig, Germany) and Experimental Pulmonology Laboratory (University of São Paulo, Brazil). PATIENTS, SUBJECTS AND INTERVENTIONS: Whole-lung CT and arterial blood gases were acquired simultaneously in 77 patients mechanically ventilated with pure oxygen. A subgroup of 28 patients was submitted to different Fio2. We also studied 19 patients who underwent repeat CT. Furthermore we studied ten pigs with acute lung injury at multiple airway pressures, as well as a theoretical model relating PaO2 and physiological shunt. We logarithmically transformed the PaO2/Fio2 to change this nonlinear relationship into a linear regression problem. MEASUREMENTS AND MAIN RESULTS: We observed strong linear correlations between Riley's approximation of physiological shunt and CT shunt (R = 0.84) and between logarithmically transformed PaO2/Fio2 and CT shunt (R = 0.86), allowing us to construct a look-up table with prediction intervals. Strong linear correlations were also demonstrated within-patients (R = 0.95). Correlations were significantly improved by the following methodological issues: measurement of PaO2/Fio2 during pure oxygen ventilation, use of logarithmically transformed PaO2/Fio2 instead of the "raw" PaO2/Fio2, quantification of nonaerated lung as percentage of total lung mass and definition of nonaerated lung by the [-200 to +100] Hounsfield Units interval, which includes shunting units within less opacified lung regions. CONCLUSION: During pure oxygen ventilation, logarithmically transformed PaO2/Fio2 allows estimation of CT shunt and its changes in patients during systemic inflammation. Relevant intrapulmonary shunting seems to occur in lung regions with CT numbers between [-200 and +100] Hounsfield Units.