Wednesday, April 17, 2013

Out to Lunch: Nutrition and Supplementation in Critical Illness

A study in week's issue of the NEJM (Heyland et al, Glutamine in Critical Illness, April 18th, 2013) left me titillated in consideration of how new evidence demonstrates underlying misconceptions, shortcomings, and biases in our understanding of, and general approach to, disease and its pathophysiology.  Before you read on, try to predict:  Will supplemental glutamine and anti-oxidants influence the course of critical illness?

The Canadian Critical Care Trials group has continued the effort to determine the causal role of macro- and micronutrients and their deficiency and supplementation in critical (and other) illness.  The results are discouraging (glutamine and anti-oxidants don't work), but only if we consider RCTs to be a tool for the assessment of the therapeutic value of putative molecules and their manipulation in disease states.  RCTs are such a tool, but only if we happen to be fortunate enough to be pursuing a causal pathway.  In the absence of this good fortune, RCTs remain valuable but only to help us understand that the associations we have labored to delineate are not causal associations, and that we should direct our focus to other, potentially more fruitful, investigations.  As I articulated in the last post, this dual role of RCTs represents a paradox which can be the source of great cognitive dissonance (and misunderstanding).  The (properly conducted and adequately powered) RCT is a method for determining if observational associations are causal associations, but the promise of confirming causal associations in an RCT by manipulating dependent variables with a potential therapeutic agents carries with it the possibility of proving the efficacy of a disease treatment.   During this protracted scientific process, there is a tendency to get carried away, such that our hypothesis mutates into a premise that we are studying a causal factor and the RCT is the last hurdle to confirming that we have advanced not only the science of causation, but also clinical therapeutics.  Alas, the historical record shows that we are far better at advancing our understanding (if we are willing to accept the results for what they are) than we are at finding new treatments for disease, because most of the associations we are investigating turn out not to be causal.

In the background of the study of macro- and micronutrients are three biased ways of thinking about the world that permeate laypersons and investigators alike.  Firstly, people are seduced by the idea that things that are "natural" are good for them and better than things that are not natural.  Jonathan Baron and coworkers demonstrated this in the 1990s by showing that people preferred to drink "natural water" as opposed to chemically identical distilled water (Spranca, 1992; and Rozin et al, 2004) and physicians showed a preference for giving patients "natural" postmenopausal hormones rather than identical synthetic compounds (they shouldn't have given either! - see Baron Holzman and Schulkin,1998).

This "natural bias" as I will call it has received inadequate attention, especially given its obviousness - why has the natural supplement market burgeoned into a multi-billion dollar industry?  Why do so many people want to take multivitamins, minerals, fish oil, bee pollen, ginseng, etc.?  Why have we been studying, in specific disease states, fish oil, folate, homocysteine, vitamin D, calcium, vitamin E for the last ten plus years? (See Multivitamins and Cancer in Men in JAMA and my post on it here, among too many others to list.)   Why did the NIH establish, in 1998, a National Center for Complementary and Alternative Medicine and why is it funded with $128 million?  Because physicians, researchers, and laypersons alike have a strong natural bias.  If you ask a person who's taking these supplements why they are, they are likely to make some sort of comment about "anti-oxidant" properties or some other statement that is utterly unsupported by credible evidence.  But evidence is not needed when the belief in the natural (and in vitamins and minerals) is so entrenched.  (This anti-oxidant argument was also promulgated as a possible explanation why coffee drinking lowers mortality - but I doubt that it does.)

Secondly, we have a collective bias towards believing that people must be fed when ill.  "Eat some chicken soup, you'll feel better," mother used to admonish us.  Similarly, patient's families seem relieved to know that their loved ones are being fed.  But believing it does not make it so.  Have you ever had influenza?  I had Influenza A when I was an intern, January 1999.  I stayed in bed and did not eat (or drink as I recall) for three days.  While it is possible that anorexia is a maladaptive response to illness, it is also possible, and perhaps even more likely, that anorexia during acute illness is an evolved adaptive strategy.  There will be more on this in a subsequent post as I continue to integrate my own ideas with those of a book I'm reading called Antifragile by Nassim Nicholas Taleb, author of TheBlack Swan.  One basic premise of the book is that unlike physical systems which are easily characterized and which respond poorly to stress, complex biological systems have evolved over the eons in such a way that they are difficult to characterize but are often strengthened by stress - that is, they are beyond robust - they are antifragile.  Thus, the belief that feeding in critical illness at all in the acute period (say, one week), based on observational associations and pathophysiological reasoning is open for debate.  Anorexia and catabolism may be an evolutionary way of adapting to acute illness.

Thirdly, we have a bias, reinforced if not engendered by formula companies and the curricula of nutrition programs, that the exact composition of what we feed patients matters.  I am aware of no credible evidence that it does.  I have commented before that it is ironic that a patient with ARDS or renal failure or diabetes or some other diagnostic category is fed a formula supposedly specifically directed at their unique nutritional needs, and then the first day they're eating by mouth they are served the same cafeteria tray as almost every other patient in the hospital.  It's really kind of amusing.  When you're on the ventilator, we think we can and we should control every variable of your physiology, to your benefit no less.

So, we have the current study by Heyland et al (NEJM April 18th, 2013) that showed that neither glutamine nor anti-oxidants (selenium, beta carotene, zinc, vitamins C and E) administered to 1223 critically ill patients showed benefit and there was perhaps evidence of harm from glutamine.  Because the primary analysis did not meet statistical significance for harm, because my prior probability for a harmful effect is low, because the statistical significance for harm at 6 months was marginal, and because I fail to understand why the harm would take 6 months to accrue, I am inclined to dismiss this as a type 1 error.  (As an aside, the editorialist botched the statistical interpretation of the trial.  That might have to be a letter to the editor.  See:  GVDB Glutamine Editorial.)  But the results are sobering, especially in light of related evidence.  Other recent studies in this genre include:
  • The EPANIC trial (NEJM, August 11, 2011) showed that late (and less) nutrition was superior in critically ill adults  
  • The OMEGA study (JAMA, October 12, 2011) showed that provision of supplemental n-3 fatty acids, γ-linolenic acid, and antioxidants in patients with ARDS caused increased time on the ventilator and a strong trend towards increased mortality
  • The EDEN study (JAMA,February 22/29, 2012) showed that delaying full feeding for the first week of illness had basically no effect on outcomes.
  • A post-hoc analysis of the EPANIC data (AJRCCM, February 1, 2013) concluded that "no dose or type of macronutrient was found to be associated with improved outcome."  The editorialist opines that we are pandering to gluttony in the ICU.  I am inclined to agree with him.  If less isn't more, it sure looks like it's the same as more.
In sum, I'm not sure that I agree with the authors that their intervention was harmful and I think it's just as likely that they have a Type 1 error in their secondary analyses and a near miss Type 1 error in the primary analysis.  (If the antioxidant component had showed a strong trend towards harm like in OMEGA, I would be more concerned.)  Nonetheless, I think we should abandon this therapeutic path, that the idea is now dead in the water.  It's not that the dose, the timing, the choice of nutrients, the selenium levels in the soil, the method of administration was wrong - none of that.  It's Ockham's Razor: the simplest explanation is that none of it works and we're impelled to study it because we have a biased way of thinking about natural things and nutrition and we give inadequate emphasis to the null hypothesis as we invest more and more time in pursuit of the alternative, with its attendant hopes of finding  interventions that work.

Alas, evolution, in addition to being a brilliant chemist, is a brilliant physician and nutritionist.

[Addendum 6/9/2013:  Here's a great JAMA piece by Paul Offit about the silliness of alternative and complementary medicine.]


  1. As usual very interesting. I need a lot more understanding of statistic which I plan to work on. I do believe that the tendency to trusting "natural products" comes from the same root as the "trusting our own body's evolution" that you talk about. Eating whole foods that have been around for thousands of years or longer makes sense compared to eating recently engineered foods. I don't know that they are healthier but I do know that with every change we make we have unintended consequences. High fructose corn syrup seems to coincide with increased obesity. It seemed like a good idea at the time but now maybe not so much. As far as nutrition in the ICU setting you may be right. Fluid administration in sepsis is important but nutrition may be the opposite. Maybe you could do a study with that opposite hypothesis of anorexia being more beneficial than early supplementation and find out.

  2. Scott. Appreciate your long term blogging as a medical specialist and statistical posts. In reference to might your statement "Nonetheless, I think we should abandon this therapeutic path, that the idea is now dead in the water." be a little premature ?-)

    Carotenoids Up-Regulate Connexin43 Gene Expression Independent of Their Provitamin A or Antioxidant Properties1 Li-Xin Zhang, Robert V. Cooney, and John S. Bertram2 Molecular Oncology Unit, Cancer Research Center of Hawaii. University of Hawaii, Honolulu, Hawaii

    Epidemiological evidence and studies in whole animals and cell culture have indicated that carotenoids have cancer chemopreventive action. In mouse C3H10T1/2 cells, this activity is highly correlated with the ability of carotenoids to up-regulate gap junctional intercellular communication. Here, we report that in mouse cells, carotenoids increase the expression of connexM.ì,a gene that encodes a major gap junction protein. This effect appears unrelated to their provitamin A or antioxidant properties, since carotenoids with and without provitamin A activity increased levels of connexin43 mRNA and protein, whereas the antioxidants methyl-bixin and a-tocopherol were inactive. More over, the active carotenoid canthaxanthin did not induce the vitamin A-inducible gene retinoic acid receptor-0. Connexln4i is the first carotenoid-inducible gene described in mammals. By indicating an additional pathway through which carotenoids function, these data provide a mechanistic basis for cancer chemoprevention by carotenoids and may lead to a re-evaluation of carotenoid physiology.

  3. Well written Scott. What was demonstrated, yet again, is that "normalisation" of a "deficiency" is ineffective and possibly harmful( although like you I am struggling to find a mechanism to explain why glutamine would produce a 5% increase in absolute mortality). I've had influenza too and agree anorexia is very much an adaptive response