Erythropoiesis-stimulating agents in critically ill trauma patients: a systematic review and meta-analysis.
In this paper, French and co-authors address the ongoing debate on the use of erythropoietin (EPO) in critically ill patients. Besides its haematopoietic effect resulting in increased haemoglobin levels and reduced transfusion requirements, EPO is also know to have non-haematopoietic effects as a pleiotropic cytokine.
The authors performed a systematic review and meta-analysis to assess whether EPO is associated with clinical outcome (mortality, functional neurological outcome, the need for renal replacement therapy and adverse events) in critically ill trauma patients. Nine randomised controlled trials that enrolled a total of 2607 patients were included in the analysis. EPO administration was associated with a significant reduction in mortality (most short-term) (risk ratio: 0.63; 95% CI: 0.49-0.79; P = 0.001). No difference was observed with regard to neurological outcome or adverse events. Renal replacement therapy was not reported in any of the studies.
As always, the studies differed in quality: 4 had a high risk of bias, 7 were blinded, and time from injury to EPO administration was highly variable, ranging from 6 hours to 6 days. Furthermore, EPO dose, frequency and duration differed between studies. Also quite interesting was the finding that EPO did not result in a reduction of transfusion incidence or volume in the studies that reported transfusion data (6 of 9 studies). So, here, the non-haematopoietic effects of EPO may play an important role. To check for the reliability of the conclusions, trial sequential analysis was performed, which is an elegant method to perform a check on false positive outcome when the meta-analysis is underpowered. The reduction in mortality was confirmed, also when performing sensitivity analysis of the data.
Although the authors’ findings still need to be confirmed by more supportive data, if EPO truly has a protective effect on mortality, we can only start speculating about the underlying mechanism, since human data are lacking.
– Cynthia So-Osman
