Murphy et al. conducted a very large (2003 patients) randomised trial comparing a liberal (threshold < 9 g/dL) versus a restrictive (threshold < 7.5 g/dL) transfusion strategy in non-urgent cardiac surgery. The primary outcome was a serious infection or an ischaemic event within 3 months after randomisation. They also estimated health care costs from the day of surgery to 3 months after the operation.
The primary outcome occurred in 35.1% of patients in the restrictive group and in 33% of patients in the liberal group (P = 0.30). Other serious postoperative complications and total costs were not different between groups. Thus, all is well since the results are similar to those published previously (TRICC, FOCUS, TRACS, for example).
However, there is, in my view, a major difference with previous trials: mortality was increased in the restrictive transfusion group (4.2% vs. 2.6%; HR 1.64, CI 1.00 to 2.67; P = 0.045) (Figure S4 in the supplementary appendix).
As pointed out by Spertus in an accompanying editorial, the difference in mortality was significant at 90 days, but not at 30 days. This finding was a secondary endpoint, unrelated to any increase in AEs, but I find it worrying, although it is difficult to imagine by what mechanism a 1.5 g/dL difference in haemoglobin concentration can increase mortality 90 days later. Nevertheless, Hajjar et al.presented somewhat similar results in TRACS (JAMA 2010;304:1559-67).
So, what do we do now in everyday clinical practice? I am tempted to answer, until we obtain more evidence from similar trials in cardiac patients, that we should perhaps consider a more “liberal” transfusion trigger (9 g/dL) in these patients. A 1.6% increase in mortality is not trivial.
– Jean-François Hardy
Discuss this article on the Nataonline forum