There's a lot of bad data out there, much of it retrospective, heterogeneous populations, and so on. When I was a MICU resident, often the team would prefer to get the Hgb up above 10 to "help with the oxygen delivery." Marino, OTOH, in his The ICU Book, makes a persuasive case that the reduced viscosity in anemia facilitates oxygen delivery, and you mess with this compensatory mechanism (with pRBCs) at your peril. Incidently, Marino provides a review of some more recent transfusion-related papers from his website. Worth reading.
Let me review the single best, and most-oft cited, paper on this topic, which was done by Hêbert et al, published in the NEJM in 1999. The paper (pdf) can be downloaded free from NEJM, I think, but here's a copy.
Real interesting stuff.
They prospectively studied the effect of two different transfusion thresholds, either 7 g/dL (restrictive strategy) or 10 g/dL (liberal strategy), in ICU patients. The ICUs were mixed, both surgical and medical, and both community and academic. They excluded patients with, amongst other issues, active hemorrhage or s/p cardiac surgery.
They found that 30-day mortality was the same in the two groups, but the hospitalization mortality was significantly lower in the restrictive-strategy group (22.3 percent vs. 28.1 percent, P=0.05).
Let me emphasize that a bit more:
28.1 - 22.3 = 5.8, or an Absolute Rate Reduction of 5.8%
That gives us a Number Needed to Harm of a little over 17. In other words:
For every 17 patients that were transfused to stay above 10 g/dL, instead of above 7 g/dL, 1 died.
Closer look at the breakdown on adverse effects:
Okay, that was in 1999 - what's changed since then? Not much. In fact, the evidence has become stronger with regard to the risks of transfusion, and the benefits of a restrictive strategy, across a spectrum of medical and surgical contexts.
Let me give the latest transfusion guidelines, published in 2009, written jointly by the Society for Critical Care Medicine,and the Eastern Association for Surgery on Trauma. This document is kinda unique, addressing both the medical and surgical perspectives regarding the literature. The salient recommendations are copied here. I just want to highlight a few of them:
1. Transfuse patients with active, ongoing hemorrhage. Be proactive!
2. If they're not bleeding out, don't transfuse until the Hgb drops below 7 g/dL.
3. At that point, only transfuse single units.