To test whether assessing pulmonary permeability by transpulmonary thermodilution enables to differentiate increased permeability pulmonary edema (ALI/ARDS) from hydrostatic pulmonary edema. Retrospective review of cases. A 24-bed medical intensive care unit of a university hospital. Forty-eight critically ill patients ventilated for acute respiratory failure with bilateral infiltrates on chest radiograph, a PaO(2)/FiO(2) ratio < 300 mmHg and extravascular lung water indexed for body weight >/= 12 ml/kg. We assessed pulmonary permeability by two indexes obtained from transpulmonary thermodilution: extravascular lung water/pulmonary blood volume (PVPI) and the ratio of extravascular lung water index over global end-diastolic volume index. The cause of pulmonary edema was determined a posteriori by three experts, taking into account medical history, clinical features, echocardiographic left ventricular function, chest radiography findings, B-type natriuretic peptide serum concentration and the time-course of these findings with therapy. Experts were blind for pulmonary permeability indexes and for global end-diastolic volume. ALI/ARDS was diagnosed in 36 cases. The PVPI was 4.7+/-1.8 and 2.1+/-0.5 in patients with ALI/ARDS and hydrostatic pulmonary edema, respectively (p<0.05). The extravascular lung water index/global end-diastolic volume index ratio was 3.0 x 10(-2)+/-1.2 x 10(-2) and 1.4 x 10(-2)+/-0.4 x 10(-2) in patients with ALI/ARDS and with hydrostatic pulmonary edema, respectively (p<0.05). A PVPI >/= 3 and an extravascular lung water index/global end-diastolic index ratio >/= 1.8 x 10(-2) allowed the diagnosis of ALI/ARDS with a sensitivity of 85% and specificity of 100%. These results suggest that indexes of pulmonary permeability provided by transpulmonary thermodilution may be useful for determining the mechanism of pulmonary edema in the critically ill.