Background: The incidence of late postoperative hypoxaemia and complications after upper abdominal surgery is 20-50% among cardiopulmonary healthy patients. Atelectasis development during anaesthesia and surgery is the main hypothesis to explain postoperative hypoxaemia. This study tested the predictive value of PaO₂<19 kPa during combined general and thoracic epidural anaesthesia and the preoperative functional residual capacity (FRC) reduction in the 30°head tilt-down position for the development of late prolonged postoperative hypoxaemia, PaO₂<8.5 kPa for a minimum of 3 out of 4 days, and other complications. Methods: Forty patients without cardiopulmonary morbidity, assessed by ECG, spirometry, FRC and diffusion capacity preoperatively, underwent upper abdominal surgery. PaO₂ during anaesthesia and preoperative FRC reduction were compared to known risk factors for the development of hypoxaemia and complications: age, pack-years of smoking and duration of operation. The effect of optimizing pulmonary compliance with preoperative positive end-expiratory pressure (PEEP) on postoperative hypoxaemia and complications was evaluated in a blinded and randomized manner. Results: Late prolonged postoperative hypoxaemia and other complications were found in 37% and 38% of the patients, respectively. Patients with PAO₂>19 kPa during anaesthesia with F₁O₂=0.33 exhibited a risk, irrespective of PEEP status, of suffering late prolonged hypoxaemia of 0% (0;23) and patients with PaO₂<19 kPa a risk of 52% (32;71), P<0.005. Having smoked more than 20 pack-years was associated with a 47% (19;75) higher incidence of postoperative complications than having smoked less than 20 pack-years, P<0.006. Conclusions: PaO₂ during anaesthesia and smoked pack-years provide new tools evaluating patients undergoing upper abdominal surgery in order to predict the patients who develop late postoperative hypoxaemia and complications.