OBJECTIVES: Primarily, to determine if respiratory variables, assessed on a daily basis on days 1-6 after ICU admission, were associated with mortality in non-ARDS and ARDS patients with respiratory failure requiring mechanical ventilation. Secondarily, to determine non-respiratory factors associated with mortality in ARDS and non-ARDS patients. DESIGN: Prospective multicentre clinical study. SETTING: Seventy-eight intensive care units in Sweden and Iceland. PATIENTS: Five hundred twenty non-ARDS and 95 ARDS patients. MEASUREMENTS AND RESULTS: Potentially prognostic factors present at inclusion were tested against 90-day mortality using a Cox regression model. Respiratory variables (PaO2/FIO2, PEEP, mean airway pressure (MAP) and base excess (BE)) were tested against mortality using the model. Primary aim: in non-ARDS a low PaO2/FIO2 on day 1, RR (risk ratio) = 1.17, CI (95% confidence interval) (1.00; 1.36), day 4, 1.24 (1.02; 1.50), day 5, 1.25 (1.02; 1.53) and a low MAP at baseline, 1.18 (1.00; 1.39), day 2, 1.24 (1.02; 1.52), day 3, 1.33 (1.06; 1.67), day 6, 2.38 (1.11; 5.73) were significantly associated with 90-day death. Secondary aim: in non-ARDS a low age, RR = 0.77 (0.67; 0.89), female gender, 0.85 (0.74; 0.98), and low APS (acute physiologic score), 0.85 (0.73; 0.99), were associated with survival; chronic disease, 1.31 (1.12; 1.52), and non-pulmonary origin to the respiratory failure, 1.27 (1.10; 1.47), with death. In ARDS low age, RR = 0.65 CI (0.46; 0.91), and low APS, 0.65 (0.46; 0.90), were associated with survival. CONCLUSIONS: No independent significant association was seen between 90-day mortality and degree of hypoxaemia, PEEP, MAP or BE for the first full week of ICU care in either ARDS or non-ARDS. In a sub-group of non-ARDS a lower PaO2/FIO2 and MAP tended to influence mortality where a significant association was seen for 3 of 7 study days. Age, gender, APS, presence of a chronic disease and a pulmonary/non-pulmonary reason for the respiratory failure were associated with mortality in non-ARDS, while only age and APS showed a similar association in ARDS.