A total of 15 RCTs were included (n= 1,419) of which 14 contributed to analysis of blood loss and 10 to the transfusion analysis. Quality assessment was reported in a table ranging from 0 to 7 points out of a possible 7. Overall quality score was not correlated with the observed effect sizes in the blood loss studies. Funnel plots for both outcomes indicated low levels of publication bias from a fairly symmetrical plot.
Blood loss (14 RCTs, n=1,219): pooled analysis found normothermia was associated with significantly lower levels of blood loss than hypothermia; ratio of geometric mean total blood loss = 0.84 (95% CI: 0.74, 0.96). Significant heterogeneity was noted in this analysis (Q: 62.3, p<0.001).
Transfusion requirement (10 RCTs, n=895): pooled analysis found normothermia was associated with significantly lower need for transfusion compared with hypothermia, RR 0.78 (95% CI: 0.63, 0.97, p=0.027). No significant heterogeneity was found in this analysis (Q: 11.4, P=0.25).