Thursday, July 17, 2008

Even More Dissonance

A recently published study comparing various weight-loss approaches has been getting a lot of press and Internet buzz, probably because the results contradict mainstream thinking about diet and health. It's pleasantly surprising that this is getting some media coverage - usually such dissonance-inducing results are largely ignored. Regina Wilshire posted an especially amusing blog, showing how different people interpret these results. You can taste the cognitive dissonance, as each individual spins the results according to their own beliefs.

The essence of the study results is that those following a low-carbohydrate diet had greater weight-loss and improvements in blood lipids. The Mediterranean diet did well also. Both of these results are predictable from what we know about metabolic regulation, but for the mainstream, this result clearly induces significant dissonance. I particularly enjoyed Dean Ornish's attempt at reconciling this dissonance. Here's a choice quote:

I'm also very skeptical of the quality of data in this study. For example, the investigators reported that those on the "low-fat" diet consumed 200 fewer calories per day—or 10,000 fewer calories per year—than those on the Mediterranean diet, yet people lost more weight on the Mediterranean diet. That's physiologically impossible.

I think Dr. Ornish needs to bone up on his biochem. We'll hit this point later in the series on Energy Regulation, but the body very definitely has a mechanism to dump excess fat calories in the form of heat. And of course Ornish's calorie-centric focus completely ignores other regulatory effects, such as insulin's effect on fat storage. Ornish does spend plenty of time telling you all about himself, what he believes, why his particular diet flavor is superior, etc. The article reads more like an infomercial than scientific exposition. Comparison of different scientific hypotheses requires inclusion of ALL relevant evidence. Ornish heavily weighs evidence of his own creation, which (not surprisingly) supports his own preconceived notions. If you selectively weigh evidence in this way, you can come to any conclusion you want.

Here's another fun quote from Ornish: "Most people associate an Atkins diet with bacon, butter and brie, not a plant-based diet like the one I recommend." There's that "I" again. Shouldn't the diet be recommended by the evidence, not one individual? I guess Dr. Ornish is smarter than the rest of us. Maybe he would grace us with a more detailed explanation of why he's "right" given our knowledge of metabolic regulation at the molecular and cellular level?

I'm not holding my breath.

Ornish's comment also highlights one of the major origins of dissonance surrounding these recent results: the seemingly unshakable belief that saturated fat ("bacon, butter, and brie") plays a role in a wide range of disease processes. We saw in the original post on cognitive dissonance that there actually exists essentially no evidence of causality (I just confirmed this with an ex-official of the American Heart Association). For example, there may be some statistical association between saturated fat consumption and development of heart disease (particularly if you limit the observational data set), but there's no evidence at all of causality at the molecular and cellular level. Let's look at a some ways in which this association might arise:

  • Fast food is often high in saturated fat. It's also often high in total refined carbohydrates, particularly fructose. The damage wrought by increase carbohydrates (fructose is particularly good at this) and the hormonal derangement from repeated insulin spikes (and probably fructose as well) quite logically predicts an increase in heart disease. The likely high consumption of oxidized fats from deep-fried foods is the cherry on top of this sundae. Lipoprotein molecules are composed of a water soluble membrane including both proteins and fatty acids. White blood cells have a specific receptor for oxidized LDL (but not unoxidized LDL), so if your LDL includes some oxidized fat from your French fries, you should expect an increased immune response, which is known to be important in the development of atherosclerosis. So if a population has a high consumption of fast food, not only is their saturated fat consumption higher, so is the consumption of refined carbohydrates and oxidized fats. Which of these actually causes the observed increase in heart disease?
  • Grain-fed beef is known to have some nutritional issues. Grains are not the natural food of cattle, who prefer to eat leafy material, which tends to be rich in the omega-3 alpha-linolenic acid. When compared with grass-fed beef, grain-fed has a significantly higher ratio of omega-6/omega-3 fatty acids. There is a biochemical reason to believe this could increase heart disease, due to the pro-inflammatory effect of omega-6 fats. Grain-fed beef is also much higher in saturated fat, so there would be an association between saturated fat intake and increased omega-6/omega-3 ratio.
  • Grain-fed beef is also higher in total fat. Guess what - carbohydrates make cows fat too! But this fat is essentially "empty calories" in that the increased fat intake does not bring significant additional micronutrients, probably displacing calories from foods that are nutrient dense. Again, at the molecular/cellular level, there are good reasons to believe these micronutrients (like magnesium) are protective against the development of heart disease.
  • Eating a crappy diet like fast food makes people sick. Sick people tend to stay inside. If you don't go outside, in all likelihood you are deficient in Vitamin D. Vitamin D deficiency is implicated in a whole host of diseases, including heart disease. I'll bet saturated fat consumption is correlated with Vitamin D deficiency as well.

I'm sure with a little thought we could come up with several more. The point is this: associating causality with an individual statistical correlation is a very slippery slope. If you have no evidence for causality, making such an association implies that you are ignoring other possible causes WITHOUT EVIDENCE. Attempting to treat sick people based on this association could be expected to be ineffective at best, harmful at worst. And of course you wind up with the precise situation we observe today, which is that some bogus dogmatic belief blocks the advancement of science due to cognitive dissonance.


OhYeahBabe said...

Great post! It's going to be interesting to see where all this discussion leads. I'm looking forward to the next batch of studies that will attempt to discredit this one... and fail to show that high carbohydrates are healthy!
My blog: Kimorexia
See Kimkins on Good Morning America

Dave said...

Thank you.

The widespread media coverage and ensuing discussion give me hope that we've hit the "tipping point" on this topic.

JPB said...

Bravo - I, too hope that we have reached the tipping point. It's about time!

Joanna M said...

Dean Ornish: "...those on the "low-fat" diet consumed 200 fewer calories per day—or 10,000 fewer calories per year—than those on the Mediterranean diet..."

200 fewer cal/day x 365 days = fewer 73,000 cal in a year, not 10,000, as Ornish stated. By Ornish's argument, those 73,000 fewer cal/yr / 3,500 cal/lb = 20.9 fewer lb/yr!! the low-fatters should have lost over 40 lbs in the two years of the study. No wonder Ornish didn't carry through on this line of reasoning. ;)

Dave said...

Hah! While claiming a "physiological impossibility", Ornish manufactured a mathematical impossibility!

OhYeahBabe said...

Good catch, Joanna! WTG!

Steve Parker, M.D. said...

It's funny how Dr. Ornish always refers to his diet program as "plant-based" rather than vegetarian, which it is. That's how I got snookered into buying his "Eat More, Weigh Less" book years ago. Great title! You don't find out until half-way through that it's vegetarian. "Plant-based" is good marketing. Casual observers don't reject it as soon as they might "vegetarian."

Ornish makes a valid point that the low-fat group in the study was not particularly low-fat. The researchers guidelines for the low-fat cohort aimed for 30% of calories from fat, 10% of calories from saturated fat. If you look at the baseline diet compositions of all three diet groups, before they ever started the intervention, you see percentage of caloric intake from fat being 31.4 to 32.1%, and saturated fat percentage between 9.7 and 9.9%. So the "low-fat" intervention group didn't really make any significant change from baseline fat intake percentage. The main intervention for the "low-fat" group was to reduce caloric intake, which would reduce fat gram intake as well carb and protein gram intake.

[For those who haven't read the study: The low-fat group was also "counseled to consume low-fat grains, vegetables, fruits, and legumes and to limit their consumption of additional fats, sweets, and high-fat snacks."]

Ornish is also correct in pointing out the calorie deficit issue. The researchers, in Table 2, report that all three diet groups reduced their daily caloric intake (compared to baseline) by 254 to 591, for 24 months. Just eye-balling the nine numbers presented in Table 2, the average deficit is roughly 450 cals/day. If true, shouldn't they have been losing about 3/4 pound per week, on average, for 104 weeks? They could have avoided that weight loss if they drastically reduced their activity levels, but that's not likely. You can come up with other hypotheses. I agree with Ornish that the most likely explanation is a methodological flaw, and the data points are invalid. Just my gut feeling.

I just thought of something. Maybe the reported daily energy deficit refers to entire groups of dieters, not to individual deficits!

Sorry to be so long-winded.


Dave said...

Hi Steve. I agree that this study has some design flaws. They certainly should have attempted to have greater differentiation between the diets. The other problem is in the reporting. The errors on some of the quantities in Table 2, like Energy Change, are very large. I doubt the data are normally distributed, though. Outliers probably occur mostly on the high side due to those individuals who did not comply with the diet. The "+/-" numbers are standard deviation. If taken as implying normal distribution, we would interpret them to mean that something like 15% of the people on the low-fat and low-carb diets reduced their energy intake by over 2000 kCal, which seems unlikely, unless the authors included people who dropped dead. We need to push the authors to release their raw data, if they haven't already.

So I don't think we can draw much conclusion from a lot of the published numbers, because they're almost certainly derived from non-normal data. The mean is a crummy statistic in such cases, the median would have been more useful. Their P-values are also probably polluted by non-normality, and of course P-value answers the wrong scientific question, telling you about your data rather than your hypothesis.

It's dangerous to assume that caloric intake is the only variable. Energy output consists of three components: base metabolic rate, physical activity, and thermogenesis. These are clearly variable and change in an effort to maintain energy homeostasis. Consider the following from this paper: "In a study on lean and obese ob/ob mice, the lean animals increased their energy intake by approximately 70% on the cafeteria diet. Nevertheless, there was no significant difference in energy deposition, because of the hyperphagia-induced stimulation of energy expenditure in the form of diet-induced thermogenesis (Trayhurn et al. 1982). In this particular experiment the lean animals receiving the cafeteria diet serendipitously had the same energy intake as ob/ob mice consuming the normal diet; in other words, they were ‘pair-fed’ on an energy basis (isocaloric intakes). However, the ob/ob mutants deposited considerably more energy than their lean wild type counterparts, indicating that diet-induced thermogenesis was substantially reduced in the absence of functional leptin (Trayhurn et al. 1982)."

There's also evidence that physical activity is modulated by dietary intake, see e.g. this paper. From the standpoint of comprehensive energy regulation, it makes sense that energy output would be regulated along with energy input. Food availability and macronutrient content can vary considerably, so there should be a mechanism for "dumping" extra calories, just like there's one for dumping excess water, salt, etc. Except in disease states, there's no way to excrete this calories in an unused chemical form (e.g. you can't pee out excess fat), so it needs to be lost as energy, either heat or metabolic activity.

Anonymous said...

Dave, I really enjoy reading your blog. I recently stumbled accross this study, and wondered if you had a take on it: