Friday, January 11, 2008

Jumping the Gun with Intensive Insulin Therapy (Leuven Protocol):How ICUs across the nation rushed to adopt a therapy which is probably not beneficial

In this week's NEJM is an anxiously awaited article about intensive insulin therapy in severely septic patients in the ICU: http://content.nejm.org/cgi/content/short/358/2/125
This business of intensive insulin therapy began with publication in the NEJM in 2001 an article by Van den Berghe et al showing a remarkable reduction in mortality in surgical (mostly post-cardiac surgery) patients in a surgical ICU. Thereafter ensued a veritable rush to adopt this therapy, and ICUs around the country began developing and adopting protocols for "tight glucose control" in spite of concerns about the study and its generalization to non-surgical patients who were not being fed concentrated intravenous dextrose solutions....

I vividly remember one of the ICU attendings at Johns Hopkins Hospital, Dr. Jimmy Sylvester, telling us on the morning after the study was published that "this is either the largest break-through in intensive care therapeutics ever, or these data are faked". In essence what he was saying was that the prior expectation of a result as dramatic as demonstrated by Van den Berghe was very low (see also: http://jama.ama-assn.org/cgi/content/full/294/17/2203 ). That lower prior probability should have reduced our confidence in the results, and made us more skeptical of the population studied and the dextrose solutions and the applicability to non-surgical patients. Well then, why didn't it?

My colleague James M. O'Brien, Jr, MD, MSc and I have one possible explanation for the rush to adopt "intensive insulin therapy" which we have dubbed the "normalization heuristic." Physicians, for all of our training, remain quite simple-minded. We like simple, feel-good fixes. Normalizing lab values is one of those things. "Make it normal and all will be fine," goes the mantra. We like to make the potassium normal. We like to make the hematocrit normal. We love it when the magnesium increases after we order 4 grams. It's satisfying. And it feels like we're doing some measurable, that is, easily measurable good in the world. Normalizing blood sugars fits that paradigm and makes us feel like we are doing good. But are we?

We have learned the hard way over the years that many of the things we do to "normalize" some surface value causes an undercurrent of harm for patients. Think suppression of PVCs (the CAST trial: http://content.nejm.org/cgi/content/abstract/321/6/406 ) or transfusion thresholds (the TRICC study and others: http://content.nejm.org/cgi/content/abstract/340/6/409 ). Oftentimes, it seems, our efforts to "normalize" some value cause more harm than good. It is quite possible that this is also the case with intensive insulin, and that the "feel-good" appeal of making the blood sugars normal in the short term in acutely ill patients propelled us to early adoption of this probably useless and possibly harmful therapy.

(For an analogous contemporaneous story about biology's complexity and defiance of simple explanations and logic such as the normalization heuristic, see: http://www.nytimes.com/2008/01/11/science/11ants.html?scp=1&sq=aiding+trees+can+kill+them.)

The interesting thing regarding the "adoption" of Van den Berghe's "Leuven protocol" is that no ICU I have worked in really adopted that protocol. They softened it up, making the target blood sugar not 80-120, but rather 120-150 or some similar range. So what was adopted was "moderate insulin therapy" rather than intensive insulin therapy. Nobody has any idea whether such an approach is beneficial. It's certainly safer. But it has substantial costs in terms of nursing care that might be better spent on other interventions (think sedation interruption).

(I have been highly critical of Van den Berghe's medical insulin article, and my criticisms were published in the NEJM. I was delighted that she did not even address me/them in "the authors reply" - apparently I left her speechless: http://content.nejm.org/cgi/content/extract/354/19/2069.)

So this wonderful article in the current issue by Brunkhorst et al is music to my ears. Rather than hiding the high rate of severe hypoglycemia in supplementary material, Brunkhorst et al come right out and say that not only was the Leuven protocol NOT associated with reduced mortality, but also that it had a very high incidence of severe side effects and that their DSMB had the wherewithal to stop the study early for safety reasons. Bravo!

We await the results of several other ongoing studies of intensive insulin therapy before we nail shut the coffin on the Leuven protocol. Meanwhile, I hope that someone somewhere will design a protocol to test the "moderate insulin therapy" that we rushed to adopt after the first Van den Berghe article as a half-hearted hedge/compromise between our "normalization heuristic", our tempered enthusiasm for the Leuven protocol, our desire to "do something" for critically ill patients, and our fear of causing side effects that result directly from our interventions (omission bias: http://mdm.sagepub.com/cgi/content/abstract/26/6/575 ).

Thank you, Brunkhorst et al, for testing the Leuven protocol in an even-handed and scientifically unbiased manner and for reporting your results candidly.

2 comments: