Blood transfusion in the ICU (part 3): Transfusion thresholds for PRBCs

Read part 1 (Overview) and part 2 (Cross-matching) here

Now that we’ve learned the general lay of the transfusion landscape, and discussed the process of cross-matching, let’s talk about when to transfuse.

The target: numeric thresholds or clinical indications?

The most iron-clad approach is to transfuse a product when the patient demonstrates a need for that product.

Red blood cells carry oxygen; transfuse them when the patient needs more oxygen delivery. This might be indicated by symptomatic anemia, such as tachycardia, tachypnea, and shortness of breath, or by outright shock, as indicated by a rising lactate, low SvO2, or organ dysfunction.

FFP and platelets, on the other hand, help form platelet plugs and strengthen clots, so transfuse those products when the patient is bleeding.

While these are all valid indications, and probably constitute 75% of the clear, well-supported reasons to give blood, we also transfuse blood to a lot of patients who don’t meet those criteria. We worry that an anemic patient may have subclinical coronary ischemia, or that a patient may not be bleeding but may be at risk of bleeding. So instead of targeting symptoms alone, we pick a number and transfuse to maintain that threshold.

Although appealingly simple, the threshold method has real limitations. The main problem is that arbitrary, non-individualized cutoffs are very unlikely—neither sensitive nor specific—to accurately match the physiology of any particular patient. Since we like to err on the side of caution, we therefore tend to pick conservative goals, and probably transfuse many patients who don’t need it. In fact, because transfusion has its own risks, this isn’t necessarily “safer,” it’s just electing one risk over another.

Still, we do it. As the saying goes, you require some blood to sustain life. Let’s look at the common thresholds for transfusing red cells.

Packed red blood cells

In general, a good hemoglobin target or transfusion threshold is >7.0 g/dl. This is equivalent to a hematocrit of 21%, but since hemoglobin is for winners and hematocrit is for thieves and degenerates, we’ll stick with the former.

This is far from a “normal” hemoglobin, which is usually >12 or 13 g/dl, so in years past we targeted higher goals such as >10 g/dl. Over time, however, and with the general trend toward more conservative medical intervention, we’ve pushed that number lower through a series of studies demonstrating equivalent safety with increasingly restrictive transfusion practices. A brief summary of the evidence:

A couple of patterns emerge from all of this. The first is that in the vast majority of studies, a hemoglobin goal of >7 was just as safe as a higher threshold; the main exceptions mostly seem to be cardiac patients, and even in that group, the contrary signal is fleeting and ephemeral.

The second is that nearly all of these are dichotomized equivalence studies showing that a lower threshold is just as good as a higher one. But proving 7 is as good as 10 doesn’t mean that 6 isn’t also as good as 7, or 5 isn’t as good as 6, or even less isn’t fine too. It merely means that, taken as an averaged, amorphous blob of statistics, pretty much all patients tolerate 7 as a routine goal. Many patients would also tolerate being lower still (patients have survived—sometimes with minimal symptoms—despite a hemoglobin of 3, 2, or even less), while others may experience symptoms whilst above this range.

That being said, for housekeeping purposes we do need some routine goal, and for most patients >7 g/dl is fine. The main exceptions are patients with coronary artery disease; synthesizing the aforementioned studies, many clinicians transfuse this group to a higher goal of >8. But this is tricky, because the characteristics of the patients with “coronary artery disease” differed between studies, raising the question of when to use that higher goal. In all patients with stable CAD? Symptomatic ischemia? Only outright unstable ACS, such as active STEMI? Who knows.

A reasonable exception to these cutoffs might be made for “expectant” transfusions, i.e. patients above 7 but expected to bleed more—such as a patient currently oozing, or about to undergo major surgery. However, this justification is much more flimsy, and in reality, many patients you expect to need lots of blood ultimately never do. This is why the old practice of transfusing two units at a time (anyone who needs one unit needs two) is now deprecated in favor of single transfusions followed by repeated labs (why give two when one will do?).

Continue to part 4 to discuss transfusion of plasma and platelets.

2 Replies to “Blood transfusion in the ICU (part 3): Transfusion thresholds for PRBCs”

Leave a Reply

Your email address will not be published. Required fields are marked *