A paramount goal of this blog is to understand the evidence as it applies to the epistemology of medical knowledge, hypothesis testing, and overarching themes in the so-called evidence based medicine movement. Swedberg et al report the results of a large[Amgen funded] randomized controlled trial of darbepoetin [to normalize hemoglobin values] in congestive heart failure (published online ahead of print this weekend) which affords us the opportunity to explore these themes afresh in the context of new and prior data.
The normalization heuristic, simply restated, is the tendency for all healthcare providers including nurses, respiratory therapists, nutritionists, physicians, and pharmacists among others, to believe intuitively or explicitly that values and variables that can be measured should be normalized if interventions to this avail are at their disposal. As an extension, modifiable variables should be measured so that they can be normalized. This general heuristic is deeply flawed, and indeed practically useless as a guide for clinical care.
Hemoglobin (Hgb) values are often abnormal in both acutely and chronically ill patients such as those with congestive heart failure, and patients with low Hgb often have worse outcomes in observational studies. The easy availability of transfusion and erythropoetin stimulating agents (ESAs) to normalize or correct low Hgb values, combined with this known association, makes Hgb a popular target for normalization. The act of normalizing Hgb implies that there will be net benefit for the patient, and this is where things fall apart. Underlying assumptions are that the low values themselves are causally related to untoward downstream outcomes, rather than just associations/markers of the underlying disease (epiphenomena) and that the ways in which we pursue modification affect the causal pathway. These assumptions are incredibly naive and sophomoric. That we can measure two things (say, Hgb and death rates in an observational study) does not mean that we have more than an inkling of understanding of causal pathways in the body (but we can easily rationalize them). After all, we don't measure Hgb levels because we know that they have anything to do with causation of anything - we measure them simply because we can, and we assume (or hope for) causality between the things we measure, as if the physiology of the body is all laid bare before us because of the measurement tools we currently have. The entire approach here is predicated on a hope that the variables that we know about, that we have already discovered usually by happenstance, are causal variables for clinically important outcomes. I argue that the crucial missing pieces of our understanding are the unknown unknowns that Donnie Rumsfeld famously referred to, and without knowing what we don't know (actual causal factors), we have only one hope of understanding causation in any specific case: the randomized controlled trial (RCT). And helped us understand the causal role of Hgb it has:
- The TRICC study (Hebert et al, NEJM, 1999) showed that more versus less transfusion was of no benefit is critically ill patients
- The CHOIR trial (Singh et al, NEJM, 2006) showed that increasing Hgb with epoetin alfa in patients with chronic kidney disease worsened outcomes
- The CREATE trial (Drueke et al, NEJM, 2006) showed that early normalization of Hgb in patients with chronic kidney disease did not affect cardiovascular outcomes
- The FOCUS trial (Carson et al, NEJM, 2011) showed that more versus less transfusion in elderly patients at risk for or with known high cardiovascular risk and undergoing surgery for hip fracture was unhelpful
- Villanueva et al (NEJM, 2013) showed that more versus less transfusion was harmful in acute upper gastrointestinal hemorrhage
- Finally, the RED-HF (published online ahead of print March 10th, 2013, NEJM) investigators found that correction of anemia in heart failure with darbepoetin alfa did not improve mortality and cardiovascular outcomes.
These trials involve over 6000 patients with a spectrum of diseases using several methods to raise Hgb, and not one of them showed benefit, rather they point to overall harm. I don't know how much more evidence we need -I have enough already to place my bet: the causal role of hemoglobin (in the range of mild or moderate anemia) in a pathway to death or adverse cardiovascular events is either in pieces or extremely fragile. To convince me otherwise, we will need a very large trial free from bias showing a noteworthy and statistically robust benefit in a specific patient population. Because we now have a prior probability informed by 6000 patients that this intervention simply does not work. The prior probability is so stout that even a new study with evidence of benefit should be regarded with suspicion.
Here is the next research population that will have to withstand the blinding light of that scrutiny: patients with acute coronary syndrome. A phase 3 pilot feasibility study is already underway (see this clinicaltrials.gov registration). I will go on the record saying that my wager is that outcomes will not be improved in this patient population either. If such a trial does show benefit, we have some serious talking to do. Because we will have to explain why, given the frailty of the observational association when subjected to an RCT in all these other studies, we have finally found a population in which it survives.
Step back, and take the view of the forester instead of that of the arborist. We naively believed that hemoglobin had a causal role, because of the observational data. We tested this hypothesis, over and over, with large RCTs, and could not confirm it. The simplest explanation is that we were wrong. The hemoglobin hypothesis is hemolyzed.
Right Doc, lecture us about Forests and Trees while you ignore the big picture.ReplyDelete
Just love it when you Evidence Based Med types decide that it is okay to only worry about clinical outcomes as if patients were the only stakeholder in this game.
What about investors? What about the politicians with manufacturers in their district? Do you even know what the word J-O-B-S means? I won't ask if you care about the fate of all those C-suite types who have approved continued study of this drug for this new, blockbuster indication. Your narrow focus makes that seem highly unlikely.
You toss around NEJM as if we are supposed to cower at its authority. Ever heard of WSJ, or Bloomsberg's?
If we are going to have evidence based medicine, wouldn't it make sense to present all the evidence?
In fact the statisticians in Marketing already see several silver lined clouds. When you do real analysis of the trial data, you can always find a subgroup for which a drug is important, maybe even vital. After all, there is a clear record that you just need approval for anything, Sales can take care of ensuring that ALL patients who could benefit get the drug.
Regulatory Affairs is already talking to the FDA about developing more realistic trial end points based on some ideas they've gotten from the psychiatric pharmaceutical industry. Is mortality really more important than Happiness as measured by a psychiatric assessment instrument from Berkeley?
And speaking of clinically irrelevant, manufacturers of drugs for "auto-immune" diseases such as MS have this down to a science. You can get the FDA to accept sensitivity to the smell of blue cheese from the moon as being relevant to an MS Disease Modifying Drug and as long as you are careful to cut the trial off at 27 months before elevated liver enzymes become frank hepatic failure.
Fortunately, the grassroots United Patients for YOUng eRythrocytes Society (UP YOURS) group (1600 K Street, Wash DC) is now lobbying the FDA to get the agency to think more broadly, an effort that is fully supported by the Southern California Congressional delegation I should note.
There is ample precedent for approval of drugs that fail clinical trials as long as you can get three positive ones. Might not be your kind of medicine Doc, but that's the law and this is a democracy, like it or not. This is clearly why manufacturers are continuing trials of drugs to clear a-beta in patients with AD in spite of the repeated failures. Three's the number. The FDA is not allowed to consider failed trials when you have three positive ones.
Finally, all this negativity is depressing. At some point, even if these drugs repeatedly fail to show benefit in humans, you can pull a "tepoxalin" and sell it to dog owners since Muffy is unlikely to complain about the little hole you just blew in his tummy and the FDA will approve any veterinary drug application, worst case requiring some post-approval studies which are then never done. Just think of all the dogs with CHF to say nothing of the CKD kitties already being treated.
Patients may get sent to the cemetery, but there is no burial place for failed drugs or indications. Once a certain amount of money and prestige are sunk into a drug for a particular indication, it will eventually be approved.
Get with the program. There's a speaker's bureau being organized at this very minute and Marketing is looking for "authors" for a bunch of very positive and uplifting papers that their team of professional writers have crafted.
You obviously aren't getting enough sun. I hear there's a CME-approved conference in Cancun on Advances in Cardiology and Hematology coming up soon. A few days of ocean air and golf will change your depressingly narrow view of the world.
Here you go: http://www.regulations.gov/#!documentDetail;D=FDA-2013-D-0077-0001ReplyDelete
"Draft Guidance for Industry on Alzheimer’s Disease"
"The purpose of this guidance is to ASSIST SPONSORS . . ." by lowering the standards for approving drugs to treat AD.
No pretense here that the FDA is going to protect the public from high-priced, ineffective drugs.
They are stepping right up and "assisting sponsors."
You might think that the FDA feels their mission is to protect drug manufacturers from the public, not the other way around.
Ed, I think you're being facetious but I like your points, especially on the FDA and ADs. In fact, a post on that nonsense is in the works. Thanks for your interest. ScottReplyDelete
Larger trials are usually more definitiveReplyDelete