Monday, November 5, 2012

Do we need to EVOLVE our views on EBM in dialysis

I have posted on the release of the EVOLVE data at Kidney Week 2012 at the eAJKD blog.

This is a big deal and when I read the @NEJM tweet it makes me mad.
That tweet provides only the first sentence in what should be an important and longer discussion of the findings of EVOLVE. To completely discount EVOLVE as @fish2phil and @Nephroboy do...
...leaves the nephrologist in a bit of a bind. If we ignore studies that don't meet their primary endpoint, what are we left with? Here is a list of ALL the randomized controlled trials in dialysis that use hard outcomes. This list ignores any study where the end-point is an improvement in a biochemical parameter, echographic finding, or vital sign:
  1. NCDS from 1981: Positive. We learned that small molecule clearance is better than large molecule clearance
  2. Normalization of hematocrit: Negative. No advantage and nearly significant disadvantage to normalizing the hemoglobin
  3. HEMO: Negative. Definitively negative trial showing that small molecule clearance can not get any better with thrice weekly hemodialysis
  4. ADAMEX Negative. Peritoneal dialysis version of HEMO, same result
  5. AURORA Negative. Statin trial with rosuvastatin
  6. 4D Negative. Statin trial with atorvastatin
  7. SHARP Positive, kind of. Statin plus ezetimibe was able to reduce CV endpoints however they lumped dialysis and CKD patients together. The dialysis patients alone were not powered to find a difference and they did not. I think most nephrologists, in the face of 4D and AURORA are pretty skeptical of this data approach.
  8. DCOR Negative. Trial of sevelamer versus calcium containing binders.
  9. IDEAL Negative. Trial of early versus late initiation of dialysis. No advantage for early start. Some may argue that this is a CKD rather than dialysis study.
  10. EVOLVE. Negative.
Is that it? I can't think of any others. Nephrology operates in an evidence desert. (Hat tip Dr. Dale)

So if you want to throw EVOLVE away because it didn't reach statistical significance what are you going to do? The questions that EVOLVE attempted to answer do not go away because the study was negative.

In a few weeks I will start going through the November labs for dozens of patients who trust me to give them the best dialysis care. Some of them will have a PTH over 600 and a calcium over 8.4. What will I do? If I am relying on randomized controlled data I will probably just sit in a corner and cry. I could rely on Block's, Floege's and Kalantar-Zedah's data that ties increased PTH with increased risk of death. All of that data is, of course, controlled for age the sin for which Homer Smith is pounding the EVOLVE crew for:
Of course if you accept the data that elevated PTH is bad, none of the observational data provides any guidance on what happens when you lower PTH.

To see what happens when you lower the PTH we need to look at the data from the mortality studies done with paricalcitol (Zemplar) and doxercalciferol (Hecterol).*

* The reason there is no link to those studies is that those studies don't exist. Thanks Abbott. Thanks Genzyme. I have heard rumors of a Zemplar trial conducted in the 90's but the data was never published and the study buried in the bad old days before
So what is the EBM dedicated nephrologist to do?

I bet if the skeptical nephrologist went to his dialysis patients and explaned to them:
USRDS 2012 Annual Data Report
  • Dialysis patients have a 3 year survival of 50%.
  • There is a drug that is already approved (i.e. not experimental) that recently has been shown to have an 89% chance of being able to reduce mortality by 17%
That every one of his patients would beg for the drug.

Final thoughts: The first time I heard about intention to treat was during the HIV epidemic. Some of drugs were so toxic, especially in the early days of HAART when patients were taking handfuls of pills, that significant number of patients weren't able to tolerate them. I remember an ID doc expressing frustration that he needed to know if the drug would save the lives of the patients he had that could tolerate the drug. If they can't tolerate it, he was going to stop the drug, but if they could tolerate it he needed to know if it slowed the virus. In the middle of the epidemic, there is no room for statistical purity.

Like that ID doctor, I want to know if the drug will work if the patient actually swallows the pill, and the answer to that is a definitive "Yes." Hazard ratio for death of 0.83 P=0.009.

Disclosures: I was a principle investigator in EVOLVE, that means my practice was paid to recruit, and administer a bitch of a study that lasted longer than it was supposed to. It was a lot of work, a lot of meetings, a lot of signatures, a lot of responsibility, and not so much money. My name will not be among the authors.
My name on page 53 on the supplement
Related Posts Plugin for WordPress, Blogger...