Arterial oxygen tension and mortality in mechanically ventilated patients

Early hyperoxia may be an independent risk factor for mortality in mechanically ventilated intensive care unit (ICU) patients. We examined the relationship between early arterial oxygen tension (PaO2) and in-hospital mortality. We retrospectively assessed arterial blood gases (ABG) with 'worst' alveolar-arterial (A-a) gradient during the first 24 h of ICU admission for all ventilated adult patients from 150 participating ICUs between 2000 and 2009. We used multivariate analysis in all patients and defined subgroups to determine the relationship between PaO2 and mortality. We also studied the relationship between worst PaO2, admission PaO2 and peak PaO2 in a random cohort of patients. We studied 152,680 patients. Their mean PaO2 was 20.3 kPa (SD 14.6) and mean inspired fraction of oxygen (FiO(2)) was 62% (SD 26). Worst A-a gradient ABG identified that 49.8% (76,110) had hyperoxia (PaO2 > 16 kPa). Nineteen per cent of patients died in ICU and 26% in hospital. After adjusting for site, Simplified Acute Physiology Score II (SAPS II), age, FiO(2), surgical type, Glasgow Coma Scale (GCS) below 15 and year of ICU admission, there was an association between progressively lower PaO2 and increasing in-hospital mortality, but not with increasing levels of hyperoxia. Similar findings were observed with a sensitivity analysis of PaO2 derived from high FiO(2) (a parts per thousand yen50%) versus low FiO(2) (< 50%) and in defined subgroups. Worst PaO2 showed a strong correlation with admission PaO2 (r = 0.98) and peak PaO2 within 24 h of admission (r = 0.86). We found there was an association between hypoxia and increased in-hospital mortality, but not with hyperoxia in the first 24 h in ICU and mortality in ventilated patients. Our findings differ from previous studies and suggest that the impact of early hyperoxia on mortality remains uncertain.