These retrospective or uncontrolled analyses leave open the question as to whether positive fluid balance contributed to death or was merely a marker of severity of illness. We believe that further controlled study is warranted. In a prospective perioperative trial21(elective colorectal surgery), fluid restriction improved outcome. Subjects were randomized to typical intraoperative fluids, including preloading, maintenance fluids, and replacement of blood and third-space losses, or to restrictive fluids (no preload or third-space replacement, less blood replacement). The restrictive approach reduced minor, major, cardiopulmonary, and tissue-healing complications. Two prospective trials22–23 in subjects with ARDS have shown that diuresis improves outcome, including time on the ventilator and ICU length of stay. The second of these trials, the Fluid and Catheter Treatment Trial (FACTT),23randomized 1,001 subjects with acute lung injury24 or ARDS to conservative (CVP < 4 or pulmonary artery occlusion pressure [PAOP] < 8 mm Hg) vs liberal (CVP, 10 to 14 mm Hg; or PAOP, 14 to 18 mm Hg) fluid management. Care was taken to ensure validity of the hemodynamic measurements, while ventilator management and weaning were controlled by protocol. Although there was no difference in 60-day mortality (the primary outcome), the conservative fluid strategy improved lung function, increased ventilator-free days, and reduced ICU length of stay.23 Of course, all of these subjects had pulmonary edema, a condition expected to respond to diuresis, and only some of them had severe sepsis. Moreover, active fluid management was only performed when subjects were hemodynamically stable. Nevertheless, since many patients with sepsis have acute lung injury, it is likely that more fluid leads to more lung edema.