Monday, October 20, 2008

DART: Many Lessons Learned

The Diet and Reinfarction Trial (DART), published in 1989, is one of the most interesting clinical trials I've had the pleasure to read about recently. It included 2,033 British men who had already suffered from an acute myocardial infarction (MI; heart attack), and tested three different strategies to prevent further MIs. Subjects were divided into six groups:
  • One group was instructed to reduce total fat to 30% of calories (from about 35%) and replace saturated fat (SFA) with polyunsaturated fat (PUFA).
  • The second group was told to double grain fiber intake.
  • The third group was instructed to eat more fatty fish or take fish oil if they didn't like fish.
  • The remaining three were control groups that were not advised to change diet; one for each of the first three.
Researchers followed the six groups for two years, recording deaths and MIs. The fat group reduced their total fat intake from 35.0 to 32.3% of calories, while doubling the ratio of PUFA to SFA (to 0.78). After two years, there was no change in all-cause or cardiac mortality. This is totally consistent with the numerous other controlled trials that have been done on the subject. Here's the mortality curve:

Here's what the authors have to say about it:
Five randomised trials have been published in which a diet low in fat or with a high P/S [polyunsaturated/saturated fat] ratio was given to subjects who had recovered from MI. All these trials contained less than 500 subjects and none showed any reduction in deaths; indeed, one showed an increase in total mortality in the subjects who took the diet.
So... why do we keep banging our heads against the wall if clinical trials have already shown repeatedly that total fat and saturated fat consumption are irrelevant to heart disease and overall risk of dying? Are we going to keep doing these trials until we get a statistical fluke that confirms our favorite theory? This DART paper was published in 1989, and we have not stopped banging our heads against the wall since. The fact is, there has never been a properly controlled clinical trial that has shown an all-cause mortality benefit for reducing total or saturated fat in the diet (without changing other variables at the same time). More than a dozen have been conducted to date.

On to fish. The fish group tripled their omega-3 intake, going from 0.6 grams per week of EPA to 2.4 g (EPA was their proxy for fish intake). This group saw a significant reduction in MI and all-cause deaths, 9.3% vs 12.8% total deaths over two years (a 27% relative risk reduction). Here's the survival chart:

Balancing omega-6 intake with omega-3 has consistently improved cardiac risk in clinical trials. I've discussed that here.

The thing that makes the DART trial really unique is it's the only controlled trial I'm aware of that examined the effect of grain fiber on mortality (without simultaneously changing other factors). The fiber group doubled their grain fiber intake, going from 9 to 17 grams by eating more whole grains. This group saw a non-significant trend toward increased mortality and MI compared to its control group. Deaths went up from 9.9% to 12.1%, a relative risk increase of 18%. I suspect this result was right on the cusp of statistical significance, judging by the numbers and the look of the survival curve:

You can see that the effect is consistent and increases over time. At this rate, it probably would have been statistically significant at 2.5 years.

I think the problem with whole grains is that the bran and germ contain a disproportionate amount of toxins, such as the mineral-binding phytic acid.  The bran and germ also contain a disproportionate amount of nutrients. To have your cake and eat it too, soak, sprout or ferment grains. This reduces the toxin load but preserves or enhances nutritional value. Wheat may be a problem whether it's treated this way or not.

Subjects in the studies above were eating grain fiber that was not treated properly, and so they were increasing their intake of some pretty nasty toxins while decreasing their nutrient absorption. Healthy non-industrial cultures would never have made this mistake. Grains must be treated with respect, and whole grains in particular.


Pasi said...

Interesting post again!

Here is a sequal for DART:

Proceedings of the Nutrition Society (2007), 66, 9–15; Secondary prevention of CHD in UK men: the Diet and Reinfarction Trial and its sequel

Alexandra said...

Thanks - very interesting.

I saw a story this morning that "junk food" raises risk for heart attack. I haven't looked it up, but the news identified junk food as "fried food, eggs, and meat." How much sense does that make? I'm sure it had more to do with sugar and "bad" least I'll bet it did, as those studies often do. They just never link it to such.

Dave said...

"Are we going to keep doing these trials until we get a statistical fluke that confirms our favorite theory?"

Probably. If we're lucky, this won't happen while the current crop of dogmatists fades away into retirement.

The term "non-significant trend" always gives me a chuckle. How can a trend be identified if it's not significant? The fiber results are just screaming for the sort of Probability Theory analysis I discussed in my last post, where you could directly compare hypotheses.

Peter N said...

I'm curious about Ornish. I know that he changes more than one variable at a time, but he's claiming an actual reversal of plaque in the arteries, and meditation and slight increase in exercise just don't seem likely to cause that effect (in my opinion- unscientific, I know). He always claims that 30% fat is still much too high to be called low-fat. Are his results relevant, and if so, could he be....gulp......right about fat?

Stephan Guyenet said...

Hi Westie,

I've read a bit about DART-2. It's confusing without a doubt. I'd have to look at it closely before making any statements about it.


Those studies are a dime a dozen. I haven't seen it but let me guess: it's either an epidemiological study that can only show associations, a short-term clinical trial that only measures surrogate outcomes, or it was done in rodents.

Stephan Guyenet said...


True. I guess it would have been better to simply state the p-value. Unfortunately, they didn't give p-values in the paper. It makes me think the fiber difference may have been uncomfortably close to p=0.05.

Peter N,

Ornish did do a small controlled trial that showed plaque regression in CHD patients. Unfortunately for him, his intervention didn't show a mortality benefit: to the contrary, more people died in his intervention group. He doesn't like to talk about that. To be fair, it could easily have been a statistical fluke given the small size of his study.

Ornish then conducted a larger trial that also failed to show any difference in CHD or all-cause mortality. That is despite his patients also quitting smoking, exercising and doing relaxation techniques. So his interventions have fallen totally flat, but he likes to ignore the little glitch of "mortality data".

He continues to claim that his diet is the only one capable of reversing arterial plaque. Well that's wonderful, until you die.

Dave said...

Hi Stephan. Just to be clear, I was not casting aspersions on your reporting of the results, just on statistics in general.

A p-value is still data-centric, rather than hypothesis centric. The idea that some arbitrarily chosen p-value represents "significance" is weird, if you think about it, both due to the arbitrariness and the data-centricity.

Stephan Guyenet said...


I agree that p-values don't tell the whole story, but don't you think they're useful? At a glance, they give you a sense of how likely the result could have been a statistical fluke.

Dave said...

Hi Stephan. If the question you're trying to answer is whether or not the data indicates some deviation from the null hypothesis, then the p-value is useful. But I feel these sorts of statistics are over-interpreted to mean that they indicate signficant (or not) evidence for some hypothesis, get erroneously used for making treatment decisions, etc.

And as you probably gathered, I get annoyed with arbitrary "significance" cutoffs. Significance as such only matters if that is particularly the question you're asking. There may be reasons to ask that question. When I was in high energy astrophysics, X-ray and gamma-ray telescopes would often find sources that were not identified in other wavelengths of light. A significance cutoff was sort of useful there, if only as a back-of-the-envelope cut to limit the places you'd look for counterparts. But once you started posing hypotheses about the physical nature of the source, significance ceased to be a useful concept. And the cutoff was just made up so that there were enough sources to investigate to keep people funded, but not so many as to be ridiculous. Sensitive instruments would set high cutoffs, less sensitive lower, etc.

Maybe I'm missing something because I didn't read the paper (yet), but I have a hard time believing the difference between those curves would not be significant when taken over the entire time period. Did they only compare the endpoint values?

Stephan Guyenet said...


Thanks for the interpretation. I'm not very knowledgeable about statistics at this point. I'm hoping to change that. I've been considering getting myself a textbook.

About the significance of the fiber group, you make a good point. If I remember correctly, they only analyzed the data statistically at specific timepoints, once at two years and once at either 6 mo or one year (I forget which).

With such fine-grained mortality data as the survival curves, you'd think they'd be able to get more power by analyzing the whole curve.

Maybe they didn't want to do an analysis that they hadn't set out to do from the beginning, or maybe they didn't want their statistics to be too sensitive because they didn't like the fiber result.

Dave said...

Hi Stephan. 3rd possibility: they didn't know any better in terms of what statistical analysis to apply. It happens a lot.

The Sivia book I linked from my last blog is a good one to get you started. Jaynes is really diving in to the deep end. Sivia is compact and reasonably accessible, and hits the major conceptual high points.

Stephan Guyenet said...

Thanks Dave, I will check out the Sivia statistics text.

Sadraki said...

Things are changing. All the patients I've ever cared for in the hospital post-MI have been on fish oil and lectured much harder core about their carbs and sugars then other diet factors. They also usually get a pretty good diabetes talk and then of course the standard stop smoking advice.

Stephan Guyenet said...


You're right, things are changing. People are starting to get it, at least in some ways.

Peter N said...

Belated thanks for the reply. No prob if you don't have time to explain this, but if you do, could you clarify: did arterial plaque regress in the second, larger trial as it did in the smaller one? And, in both cases, did heart-attack incidence decrease, stay equal, or increase? In other words, did the studies basically show that arterial plaque isn't the cause of heart attacks? That would change people's thinking a bit.

Stephan Guyenet said...

Hi Peter,

The study was called the Multicenter Lifestyle Demonstration Project. It was not randomized or controlled. 194 people who completed his intervention were compared to 139 who didn't. The intervention included exercise, relaxation techniques, and a whole foods, low-fat, plant-based diet.

I don't think plaque regressed in this study, because I can't find that in any abstracts. He would have emphasized the hell out of it if it had. His intervention did improve blood pressure, exercise tolerance and overweight.

At the end of 3 years, there was no differences in cardiac events, cardiac deaths or overall mortality between the two groups.

The fact that he couldn't show decreased mortality, even though his study was biased toward his intervention due to a lack of a control group, is telling. It's also telling that exercise, relaxation, and a low-fat vegetarian/vegan diet together failed to show any benefit. That says to me that the poor diet was negating most of the benefits of the exercise and relaxation.

Anonymous said...
This comment has been removed by the author.
Anonymous said...

"I think the problem with whole grains is that the bran and germ contain a disproportionate amount of toxins, among which are the lectins."

PUFAs are another factor in the toxicity of whole grains, IMO. Tubers are better, as they contain very little PUFAs: below 1% of calories, whereas whole grains are 2.5-6.0% PUFAs by calories and about 60% PUFAs as a percentage of fat. The safest grain is un-enriched white rice (organic white rice usually isn't enriched, but I have seen exceptions.) If you eat wheat, unbleached unenriched white flour is the safest. It should also be fermented 6 or more hours to break down the gluten.

Are You Gluten Intolerant? (or intolerant to modern quick-rise breads)

I feel best eating low-residue. No fruit unless peeled, seeded, juiced, or cooked (pref all of the above). No raw veggies, nuts, seeds, whole grains, beans or dark chocolate. Carbs are unheated honey, 1-3 glasses of strained juice, peeled potato (white), organic white rice, or natural sourdough white bread (not quick-rising junk). I don't pay any attention to lame studies claiming fruits, vegetables, and fiber are healthy, saturated fat is bad, or unsaturated fats are good. I only eat tropical oils: coconut, macadamia, cocoa butter, and so forth. Most grains should be avoided, because they are filled with junk and improperly prepared.

The notion that whole grains are healthy is a myth, like the notion that fiber is essential or healthy. I feel better with little or no fiber. Lack of fiber causes elimination of food cravings, increased digestive speed, and less waste. How can anybody think that fiber is anything but an anti-nutrient that should be avoided, along with PUFA oils, nuts, seeds, quick rise breads, and other garbage? All this propaganda for fiber came from the boxed cereal manufacturers.

BTW, Peter N, replacing PUFAs with SFAs seems to cause plaque regression, as was noted in Chris Masterjohn's analysis of the China Study. Dean Ornish has refused to isolate the reduction of fats in his study from the reduction of PUFAs. I say that reduction of PUFAs is the only way low-fat diets give benefits. And Chris M has expressed the same view.

Stephan Guyenet said...

Hi Bruce,

Thanks for the information. I didn't realize whole grains contained a non-negligible amount of n-6. I'm not too worried about it though, I don't think it's necessary to avoid all n-6 in the context of a healthy diet. I think refined n-6 vegetable oils are the main problem.

Peter N said...

Thanks for the run-down. I'm amazed that cardiac events and deaths didn't reduce in number in the second study. Doesn't make low-fat vegetarianism look all that healthy.

Thank for the info.

Unknown said...

What should be included into the discussion is the DART follow up:

"The first of these papers presented long-term, post-trial follow-up data from the DART study (6). Surprisingly, the all-cause mortality in the 2–5-y follow-up period of the original group who received advice to consume fish was significantly higher (31%) than that of the control group. Thus, the all-cause mortality of the 2 groups after a mean follow-up of 15 y became almost identical [adjusted hazard ratio of the fish group: 0.95 (95% CI: 0.85, 1.07)]. The second recent paper presented data on the effect of advice to consume fatty fish or of 3 daily fish oil capsules on CAD mortality in a group of 3114 male patients with stable angina (7). Another surprising result was found: advice to consume fish and the intake of fish oil capsules were associated with increases in CAD mortality of 20% (P = 0.16) and 45% (P = 0.02), respectively! Furthermore, the increase in risk was even more prominent for sudden death. Such results force us to rethink our view on fish and fish oil in the primary and secondary prevention of CAD."

This would be conistent with Bruce´s arguments. Short term (up to 2 years or so) fish oil seems to beneficial. Long term effects are everyones guess - maybe not so beneficial.

Stephan Guyenet said...


Thanks for pointing that out. There are a few things to keep in mind. First of all, in the DART follow-up, the patients weren't necessarily following the intervention anymore. And you can imagine what the control group did once they learned that the intervention was successful: they started eating fish. So the narrowing of the difference between control group and fish oil group is not surprising.

The other thing to keep in mind is that the increase in mortality in DART-2 was confined largely to the fish oil group, not the group eating fish. They speculate that it could be because the fish oil group was taking it on an empty stomach and receiving a bolus of PUFA-rich lipoproteins in the bloodstream. I recommend taking fish oil with a meal, preferably one that contains saturated fat.

I don't believe that fish fat can contribute to ischemic heart disease in any major way, because the traditionally-living Inuit didn't get heart disease. Even the modern-living Inuit, who still have a high consumption of omega-3 fat, have a very low rate of heart attacks.

I'd be willing to believe that too much n-3, on top of excessive n-6 vegetable oil, could be a problem. That may explain the different result between DART and DART-2. The background n-6 consumption may have been higher in the second case.

Unknown said...

"First of all, in the DART follow-up, the patients weren't necessarily following the intervention anymore. And you can imagine what the control group did once they learned that the intervention was successful: they started eating fish. So the narrowing of the difference between control group and fish oil group is not surprising."

The relative fish-intake was nevertheless higher in the intervention group. Because it went down you are right that a narrowing was to be expected. But that was somewhat more than narrowing.
The increased stroke risk should be mentioned too. That would be in line with more n3-intake. Stroke risk is quite high in Japan - most likely because of high n3 intake.
There are certainly valid arguments pro n3 but it don´t think it´s an "all clear"-situation. In my eyes those long term follow ups shouldn´t be swept under the carpet. Of course it would be nice to have a long term trial to confirm that positve short term results are still present after 10 or more years.

Stephan Guyenet said...


Definitely worth keeping in mind. And you are right about stroke risk. I'm not sure the Inuit had much stroke, but they did have a tendency to bleed, even on their traditional diet.

luke said...

Is there any significant difference between a total fat intake of 35.0% calories vs 32.3% calories? Also, assuming subjects were self reporting fat intake levels, the error in there reporting probably dwarfs the 2.7% difference in fat intake described in the study.