The best type of fluid replacement and optimal volume of resuscitation in the setting of severe sepsis have been heavily debated but studies have provided guidance to the clinician. One trial comparing 4% albumin with normal saline for fluid resuscitation found no difference in mortality at 28 days. 24 A 2004 meta-analysis similarly found no mortality advantage with the use of colloids compared with the use of crystalloids. 25 The trial of EGDT revealed that patients in the treatment arm received far greater volumes of fluid in the first 6 hours of resuscitation than those in the control arm. In a large study of European ICUs, patients with a positive fluid balance at 72 hours had a poor outcome. 26 In a clinical trial of patients with acute lung injury, the use of a conservative fluid strategy targeting a CVP lower than 4 mm Hg and a pulmonary artery occlusion pressure (PAOP) lower than 8 mm Hg was associated with a fewer number of ICU and ventilators days. 27 The preponderance of data would suggest that aggressive fluid management be done in the acute phase of sepsis, followed by a more conservative phase in the following few days.
Results We identified 17 randomized trials (n = 2138) and 3 quasi-randomized trials (n = 246) that had acceptable methodological quality to pool in a meta-analysis. Twenty-eight-day mortality for treated vs control patients was 388/1099 (%) vs 400/1039 (%) in randomized trials (risk ratio [RR], ; 95% confidence interval [CI], -; P =.05; I 2 =53% by random-effects model) and 28/121 (%) vs 24/125 (%) in quasi-randomized trials (RR, , 95% CI, -; P = .83). In 12 trials investigating prolonged low-dose corticosteroid treatment, 28-day mortality for treated vs control patients was 236/629 (%) vs 264/599 (44%) (RR, ; 95% CI, -; P = .02). This treatment increased 28-day shock reversal (6 trials; 322/481 [%] vs 276/471 [%]; RR, ; 95% CI, -; P = .02; I 2 = 4%) and reduced intensive care unit length of stay by days (8 trials; 95% CI, – to –; P < .001; I 2 = 0%) without increasing the risk of gastroduodenal bleeding (13 trials; 65/800 [%] vs 56/764 [%]; P = .50; I 2 = 0%), superinfection (14 trials; 184/998 [%] vs 170/950 [%]; P = .92; I 2 = 8%), or neuromuscular weakness (3 trials; 4/407 [1%] vs 7/404 [%]; P = .58; I 2 = 30%). Corticosteroids increased the risk of hyperglycemia (9 trials; 363/703 [%] vs 308/670 [46%]; P < .001; I 2 = 0%) and hypernatremia (3 trials; 127/404 [%] vs 77/401 [%]; P < .001; I 2 = 0%).