Crystalloids vs. colloids in fluid resuscitation: a systematic review
Choi PT, Yip G, Quinonez LG, Cook DJ
Critical Care Medicine. 1999;27((1):):200-10.
OBJECTIVE To systematically review the effects of isotonic crystalloids compared with colloids in fluid resuscitation. DATA SOURCES Computerized bibliographic search of published research and citation review of relevant articles. STUDY SELECTION All randomized clinical trials of adult patients requiring fluid resuscitation comparing isotonic crystalloids vs. colloids were included. Pulmonary edema, mortality, and length of stay were evaluated. Independent review of 105 articles identified 17 relevant primary studies of 814 patients. Weighted c about article inclusion was high (0.76). DATA EXTRACTION Data on population, interventions, outcomes, and methodologic quality of the studies were obtained by duplicate independent review with differences resolved by consensus. Weighted ic on the validity assessment was moderate (0.54). DATA SYNTHESIS No difference was observed overall between crystalloid and colloid resuscitation with respect to mortality and pulmonary edema; however, the power of the aggregated data was insufficient to detect small but potentially clinically important differences. Subgroup analysis suggested a statistically significant difference in mortality in trauma in favor of crystalloid resuscitation (relative risk 0.39, 95% confidence intervals: 0.17 to 0.89). Several methodologic issues are noteworthy regarding the primary studies, including lack of blinding (except in three studies). The type, dose, and duration of fluid administration and outcomes measured were different across these trials. CONCLUSIONS Overall, there is no apparent difference in pulmonary edema, mortality, or length of stay between isotonic crystalloid and colloid resuscitation. Crystalloid resuscitation is associated with a lower mortality in trauma patients. Methodologic limitations preclude any evidence-based clinical recommendations. Larger well-designed randomized trials are needed to achieve sufficient power to detect potentially small differences in treatment effects if they truly exist.