STUDY OBJECTIVE: To evaluate the importance of fluid balance and changes in extravascular lung water (EVLW) on survival in the ICU and short-term outcome in patients with pulmonary edema. DESIGN: Retrospective analysis of data (sorting by survival and "treatment received") from a recent randomized controlled trial of fluid restriction in this population. SETTING: Medical ICU of a university-affiliated, tertiary-care medical center. PATIENTS: Eighty-nine patients (from the previously mentioned study) requiring pulmonary artery catheterization with abnormally high EVLW (greater than 7 ml/kg). MEASUREMENTS AND RESULTS: When analyzed by survival, the survivors had no significant fluid gain or change in EVLW but decreased wedge pressure and body weight, compared to nonsurvivors. When analyzed by fluid balance, patients who gained less than 1 L of fluid by 36 hours into the study had a better rate of survival (74 percent) than the rest (50 percent; p less than 0.05). Also, the median duration of days on the ventilator, ICU days, and days of hospitalization was approximately half as long for each variable in the group with less than 1 L of fluid gain. Even accounting for baseline differences in the severity of illness, fluid balance was an independent predictor of survival (p less than 0.05). When analyzed by whether or not EVLW decreased by more than 15 percent between the first and last measurement, only patients with ARDS or sepsis had decreased days on the ventilator and ICU days. CONCLUSIONS: These data support the concept that positive fluid balance per se is at least partially responsible for poor outcome in patients with pulmonary edema and defend the strategy of attempting to achieve a negative fluid balance if tolerated hemodynamically.