IDENTIFICATION OF PATIENTS WITH ACUTE LUNG INJURY - PREDICTORS OF MORTALITY

A recent North American-European Consensus Conference proposed new, un iform criteria for the definition of acute lung injury, in part to fac ilitate earlier identification of patients for clinical trials. Howeve r, these criteria have not been evaluated prospectively. We designed a prospective cohort study of 123 consecutive patients with acute lung injury prospectively identified on admission to the adult intensive ca re units of a tertiary care university hospital. The objectives were t o determine if selection of patients using the new criteria for acute lung injury results in a significant change in the clinical characteri stics, risk factors, or predictors of mortality when compared with pri or studies of patients with adult respiratory distress syndrome (ARDS) ; and to determine if a quantitative index of the severity of acute lu ng injury has prognostic value in identifying nonsurvivors of acute lu ng injury. We used three methods: (1) prospective identification of pa tients with acute lung injury using a Pa-O2/Fl(O2) ratio < 300 and bil ateral infiltrates on chest radiograph in the absence of left heart fa ilure; (2) evaluation of the severity of lung injury using a four-poin t scoring system; and (3) stepwise logistic regression analysis to ide ntify variables significantly associated with hospital mortality. Over all hospital mortality was 58%. Sepsis was the most common clinical di sorder (50/123 or 41%) associated with the development of acute lung i njury. Using the new definition for acute lung injury, 66 of the 123 p atients were enrolled with a Pa-O2/Fl(O2) ratio between 150 and 299; 5 7 of the 123 patients had a Pa-O2/Fl(O2) < 150 at the time of entry in to the study. The mortality of the patients was similar in the two gro ups (59% for the patients identified with a Pa-O2/Fl(O2) between 150 a nd 299 and 57% for the patients with a Pa-O2/Fl(O2) < 150). The degree of lung injury on days 1, 2, or 3 was not predictive of survival. Ste pwise logistic regression analysis identified three factors that were independently associated with mortality: (1) nonpulmonary organ system dysfunction between hospital admission and admission to the intensive care unit (odds ratio (OR) = 8.1; p < 0.0001); (2) chronic liver dise ase (OR = 5.2; p < 0.05); and (3) sepsis (OR = 2.8; p < 0.05). The ide ntification of patients with acute lung injury (using the recent defin ition of a Pa-O2/Fl(O2) ratio < 300 and bilateral infiltrates on the c hest radiograph) does not alter the hospital mortality or the clinical characteristics associated with more traditional definitions of ARDS that have used stricter oxygenation criteria. Thus, the initial severi ty of the oxygenation defect does not appear to have major prognostic value. Since a detailed four-point lung injury score did not predict s urvival, this result also emphasizes that assessment of lung injury by physiologic criteria is of limited prognostic value. The overriding i nfluence of nonpulmonary factors in determining mortality was evident in the multivariate analysis in which the three major predictors of no nsurvival were nonpulmonary organ system dysfunction before admission to an intensive care unit, chronic liver disease, and sepsis. These re sults have important implications for the selection and stratification of patients for clinical trials of new therapeutic strategies for acu te lung injury.