Tuesday, April 29, 2008

Blood Substitutes Doomed by Natanson's Meta-Analysis in JAMA

When the ARMY gives up on something, you should be on the lookout for red flags. (Pentagon types beholden to powerful contractors and highly susceptible to sunk cost bias still haven't given up on that whirligig of death called the Osprey, have they?) But the ARMY's abandonment of a blood substitute that it found was killing animals in tests was apparently no deterrent to Northfield Laboratories, Inc., makers of "Polyheme", as well as Wall Street investors in this an other companies working on products with a similar goal - to cook up an extracellular hemoglobin-based molecule that can be used in lieu of red blood cell transfusions in trauma patients and others.

Charles Natanson, an intramural researcher at the NIH and co-workers performed a meta-analysis of trials of blood substitutes which was published on-line today at the JAMA website: http://jama.ama-assn.org/cgi/content/full/299.19.jrv80007 . They found that these trials, which were powered for outcomes such as number of transfusions provided or other "surrogate-sounding" endpoints, when combined demonstrate that these products were killing subjects in these studies. The relative risk of death for study subjects receiving one of these products was 1.3 and the risk of myocardial infarction increased more than threefold. The robustness of these findings is enhanced by the biological plausibility of the result - cell-free hemoglobin is known to eat up nitric oxide from the endothelium of the vasculature leading to substantial vasoconstriction and other untoward downstream outcomes.

In addition to my penchant for cautionary tales, my interest in this study has to do with study design. We are beholden to "conventional" study design expectations where a p-value is a p-value, they're all 0.05, and an outcome is an outcome, whether it be bleeding, or pain or death, we don't differentially value them. But if you're studying a novel agent, looking for some crumby surrogate endpoint like number of transfusions, and your alpha threshold for that is 0.05, then the alpha threshold for death should be higher (say 0.25 or so), especially if you're underpowered to detect excess deaths. That kind of arrangement would imply that we value death at least 5 times higher than transfusion (I for one would rather have 500 or more transfusions that be dead, but that's a topic for another discussion).

Fortunately for any patients that may have been recruited to participate in such studies, Natanson et al undertook this perspicacious meta-analysis, and the editiorialists extended their recommendations for more transparency in data dissemination to argue, almost, that future trials of blood substitutes should be banned or boycotted. Even if the medical community does not have the gumption to go that far, prospective participants in such studies and their surrogates can at least perform a simple google search, and from now on the Natanson article is liable to be on the first page.


  1. As a patient advocate for those with cancer, I have been pushing for integrative treatments.

    Amazingly, Traditional Chinese Medicine -herbs and acupuncture can really help with anemia, low white blood counts and much more.

    NO Unwanted/adverse Effects - except an occasional teeny tiny needle stick.

    Isn't it time we (government, advocates, researchers), took a more serious look at this?

    Ann F.
    The Annie Appleseed Project

  2. Everyone knows that polyheme is bad. Did you really fall that short of finding a topic that you actually had to be a blow-hard and talk about it? I'm assuming it's only for you to get on your soapbox so you could use the words "boycott" and "medicine" once again. Yea, we get it, you think you are the authority on what the medical community should and should not be using to treat patients. Annoying.

  3. Thanks for your comment, that is a good point.

    Wait a minute, what was your point?

    Was there a point at all?

    If you have something meaningful to say, feel free to post it here, so that others may be enlightened by another's perspective, which hopefully is based upon information, evidence, etc.

    On the other hand, if you would only like to issue ad hominem attacks against me, feel free to e-mail me and say whatever you wish, but please do not bother readers of the blog with unenlightening nonsense, especially anonymous nonsense.


  4. Scott,

    Can you explain the concept of "alpha threshold"?



  5. Marilyn, alpha threhsold is simply the value of alpha that will be considered statistically significant, traditionally a two-sided alpha of 0.05. But it doesn't have to be that way. Any other value could be chosen. Alpha threshold is needed to perform power calculations.

  6. Um, what is "alpha"?

    I think I understand your basic point, though. You're saying that even if there is a 25% chance that the difference in deaths or MIs is due to chance, that you would view that as concerning, right?