BACKGROUND: As the clinical application of non-invasive shunt estimation to operation under one-lung ventilation has not been reported, this study was carried out to evaluate the validity and accuracy of the non-invasive shunt estimations in one-lung ventilation with hemodilution. METHODS: Following general anesthesia with enflurane 0.5 1 vol.% and 100% oxygen in ten Mongrel dogs (B.W. around 16 kg), tracheostomy and insertion of left-side endobronchial tube and one-lung ventilation were performed. Acute normovolemic hemodilution was produced by sequential hemodilution with hydroxyethyl starch. The intrapulmonary shunt (QS/QT) was calculated by the classic shunt equation, by the oxygen contents-based estimated shunt equation, and by oxygen tension-based estimations such as alveolar to arterial oxygen difference (P(A-a)O2), respiratory index (RI, P(A-a)O2/PaO2), arterial oxygen tension to alveolar oxygen ratio (PaO2/PAO2), and PaO2 to FiO2 ratio. To assess the quantitative accuracy of the estimated shunt, the data were divided arbitrarily into two groups on the basis of the mean arteriovenous oxygen content difference (C(a-v)O2) being 3.6 ml/dl or greater (group 1) and less than 3.6 ml/dl (group 2). Relationships to QS/QT were analyzed by simple linear regression. RESULTS: In 104 measurements, the correlation between QS/QT and non-invasive shunt were poor (r = 0.66 - 0.76). However, in group 1 (n = 45), the correlation between QS/QT and the estimated shunt were very good (r = 0.93) and good for P(A-a)O2 (r = 0.83), RI (r = 0.87), PaO2/PAO2 (r = - 0.84), and PaO2/FiO2 (r = - 0.85). In group 2 (n = 58), the correlation between QS/QT and non-invasive shunt were worse than in group 1. Group 2 had lower hematocrit (20.6% vs 26.7 %, P < 0.001), higher cardiac output, and lower pulmonary and systemic vascular resistance than group 1 (P < 0.05). The difference between the estimated shunt and the classic shunt in group 1 remained constant when the classic shunt was increased further. However, the difference in group 2 was enhanced by the increment of the classic shunt. CONCLUSIONS: We conclude that even if the non-invasive shunt estimation might be affected by hemoglobin and cardiac output, it is a viable method in mild hemodiluted patients with good cardiovascular reserve.