For the next few weeks, possibly months, I'll be working on a new series titled "Did I (need to) do that?" What I want to do is look into interventions that we regularly perform in the hospital or clinic but don't really have a good reason for doing them (or do we?). We often get bogged down in the "standard of care" in our busy lives and can tend to just accept certain routine things as status quo. Hopefully with the next few blogs I can question some of these practices and find out once and for all what the data says.
If you have ever played blackjack, you know that the closer you get to 21, the more timid you get about forcing the action. The opposite can be said about life in the hospital, specifically with anemia. Once the hematocrit starts to trend towards 21, we start throwing in the type & crosses and anxiously trending the hematocrit until we bleed the patient to that threshold. The number 21 has long been considered the transfusion threshold of choice without much understanding. It’s like many a tradition that has no known origin or reason but one that we continue anyways. I’m back on our Family Medicine service this month after a long hiatus in the outpatient world so I’ve already had my fair shares of run ins with the number 21 or even higher (yeah, looking at you Ortho). So I have two quick questions today.
Looking back quickly, the first blood transfusion was done in 1818 by stud physician James Blundell. It was performed on a patient who had postpartum hemorrhage and it worked! From that point on, vampires and humans alike have enjoyed a better quality of life. The origins of the 21 threshold are a bit more nebulous to say the least. I couldn't find a source that cited any landmark study or finding that created this phenomena.
Being in Family Medicine, by far the leading source of bleeding we see is GI bleeds so it's no coincidence this was the topic I was most interested in. Now let's just throw out hemodynamically unstable patients right off the bat since that's a different ballgame. What I found here was a randomized trial (n=921) looking at hemodynamically stable patients who had severe, acute, upper GI bleeds, didn't have acute coronary syndrome and who received endoscopy within 6 hours of presentation (yeah, I know it's very specific).  This isn't a study you could necessarily apply to a small, rural hospital. They looked at a restrictive (threshold 21) vs a liberal (< 27) strategy and found the restrictive strategy to be significantly superior in complications, subsequent bleeding, overall deaths (at 45 days). Great, so at least some evidence pointing towards 21 vs. 27.
Let's look at pre-op/post-op patients.
In pre-op patients, we have a lovely Cochrane systematic review (n=6264) of 19 RCTs comparing restrictive and liberal transfusion thresholds in adults and in children.  There was a wide range of definitions of the two strategies with restrictive being in the hematocrit 21-27 range and liberal anywhere from 27-40 range. Seriously? Who's going to transfuse with a hematocrit of 40?! Anyways, overall the restrictive strategy was superior or equal to the liberal strategy finding the following:
- A 39 percent decrease in the probability of receiving a transfusion (46 versus 84 percent; relative risk 0.61; 95% CI 0.52-0.72). This seems obvious but we have to remember that limiting transfusions limits the risks that come with transfusions (infection, fever, allergic reactions, TRALI etc)
- No difference in functional recovery or hospital/intensive care length of stay.
- No increased risk of myocardial infarction when all trials were included (relative risk = 0.88; 95% CI, 0.38-2.04).
In post-op patients, it's even more 21 with a retrospective cohort by Carson et al. of 2083 patients who declined transfusions for religious regions that found zero deaths from the 99 patients with hemoglobin levels between 7.1-8, which was significant. Zero. Mortality then trended up with each point of hemoglobin that went down. (p < 0.01). They controlled for cardiovascular disease and age, which they found to be the two largest variables in the study. 
- Hgb 7.1 to 8.0 (n = 99) — zero percent
- Hgb 5.1 to 7.0 (n = 110) — 9 percent
- Hgb 3.1 to 5.0 (n = 60) — 30 percent
- Hgb ≤3.0 (n = 31) — 64 percent
So it looks like 21 is still the magic number which isn't a surprise. But what about cardiac patients? We always debate whether it's a hematocrit of 21, 25, or 30.
Well in short, it looks like 24.
The two main trials that have evaluated this specific question are the FOCUS  and TRICC trials . The FOCUS trial looked at post-op patients with cardiovascular disease or risk factors (HLD, DM, HTN, Smoking, creatinine > 2) and found no difference in mortality, hospital complications, CHF, or stroke between the restrictive strategy (hct 24 threshold) and liberal strategy (hct 30 threshold). The TRICC trial (n=838) was a randomized study in ICU patients again looking at restrictive (hct 21-27) vs. liberal (hct 30-36) strategies. They found no difference in 30 days mortality and actually found a significantly lower mortality rate in those deemed "less acutely ill" (APACHE II score less than or equal to 20) or patients < 55 years old. In patients with ischemic heart disease, they found a non-significant trend towards greater 30-day mortality in the restrictive arm (26 vs. 21%).
So what did I learn? For some reason I am referencing the OJ Simpson trial quite a bit. In GI bleed, pre-op, and post-op patients I can be comfortable with a restrictive strategy which for my purposes will continue to be 21. For cardiac patients (including CHF patients), my threshold will be 24 and next time I hear a cardiologist tell me to transfuse < 30 I'll ask for some evidence. Nicely, of course.
So next time the transfusion steamroller tries to roll you over. Just say:
(no guarantee your refusal to transfuse will end up any different than above though)
TL;DR - Transfuse at 21 for GI Bleed, pre-op, and post-op patients that are hemodynamically stable. Transfuse at 24 for cardiac patients that are stable.