Sunday, March 3, 2013

Does the Mediterranean Diet Reduce Cardiovascular Risk?

By now, most of you have probably heard about the recent study on the "Mediterranean diet" (1), a diet that was designed by diet-heart researchers and is based loosely on the traditional diet of Crete and certain other Mediterranean regions.  The popular press has been enthusiastically reporting this trial as long-awaited proof that the Mediterranean diet reduces the risk of cardiovascular events-- by a full 30 percent over a 4.8-year period.  I wish I could share their enthusiasm for the study.

Let me start off by saying that I think the Mediterranean diet is a relatively healthy diet pattern, certainly a major improvement over the typical diet.  In this particular study, participants were encouraged to eat more olive oil, nuts, fish, fruit, vegetables, beans, white meat and wine; and eat fewer baked goods, spread fats, red meats and soda.  The Mediterranean diet group was split into two, with one half receiving extra-virgin olive oil and the other half receiving nuts.  Here is how they were instructed and followed up:
For participants in the two Mediterranean diet groups, dietitians ran individual and group dietary-training sessions at the baseline visit and quarterly thereafter. In each session, a 14-item dietary screener was used to assess adherence to the Mediterranean diet (Table S1 in the Supplementary Appendix) so that personalized advice could be provided to the study participants in these groups.
The comparison group was a "low-fat diet" group.  Diet patterns scarcely changed in this group over time, with dietary fat only decreasing from 39 to 37% of calories over the course of the trial.  The reason becomes clear when you read the description of how they were instructed and followed up:
Participants in the control group also received dietary training at the baseline visit and completed the 14-item dietary screener used to assess baseline adherence to the Mediterranean diet. Thereafter, during the first 3 years of the trial, they received a leaflet explaining the lowfat diet (Table S2 in the Supplementary Appendix) on a yearly basis.
So basically, they got training at baseline and a pamphlet in the mail once a year, while the Mediterranean diet group got quarterly visits with a dietitian, diet screening, and "personalized advice".  Anyone see a problem here?  The investigators did, halfway through the study:
However, the realization that the more infrequent visit schedule and less intense support for the control group might be limitations of the trial prompted us to amend the protocol in October 2006 [3 years after the beginning of the trial -SG]. Thereafter, participants assigned to the control diet received personalized advice and were invited to group sessions with the same frequency and intensity as those in the Mediterranean-diet groups, with the use of a separate 9-item dietary screener (Table S3 in the Supplementary Appendix).
Why did they have a low-fat diet control group to begin with?  The point of a control group is to eliminate all variables from your comparison except the factor you're interested in, in this case, diet.  In this study, they presumably chose a low-fat diet for the control group because multiple trials have shown it to be ineffective at preventing cardiovascular events*.  If you want to see a treatment effect, you need to use a relatively ineffective comparison group, or else there will be no difference between groups.  Since people tend to become healthier when they adopt any sort of health-oriented diet change, and/or meet with dietitians and doctors regularly, the low-fat diet intervention should have controlled for this potentially important variable.

However, since the diet of the control group remained mostly unchanged, and it received a less intensive intervention than the Mediterranean diet group, it was not a proper control, and there's no way to know how much of the 30 percent reduction in events was due to the Mediterranean diet itself, and how much was due to increased diet vigilance and visiting regularly with dietitians.  Dr. Peter Attia recently made the same point (2).  So while we can say that an intensive Mediterranean diet intervention lowers cardiovascular event risk more than a low-intensity intervention that has little effect on diet patterns, strictly speaking we can't say that the Mediterranean diet itself lowers cardiovascular risk.  This point will be lost on most of the popular press and many researchers, because this study seemed to provide the answer everyone wanted to hear.

That being said, if I had to guess, I'd say the Mediterranean diet probably does lower risk.  That's why I find this study so frustrating-- it came so close to conclusively demonstrating what other evidence has been hinting at for some time.

* At least in studies lasting a few years done on people who already have cardiovascular disease.  That doesn't necessarily mean a low-fat diet wouldn't prevent atherosclerosis if administered before the development of disease-- but these types of trials are virtually impossible in humans.


Unknown said...

Excellent critique of one of the largest flaws in this study. On initial read through of the paper I had high hopes, but as you articulated, we must curb our enthusiasm. Unfortunately, as you stated, long term, large population studies are difficult, nigh impossible, Additionally, one variable that will always confound the data in studies like these is the fact that participants adopting other healthier lifestyle choices as a result of the intervention. Your concluding statement says exactly how I felt upon a closer look at the paper. Enormous amounts of time, effort, and money were spent to drive home a key point and make a large statement, but alas we must spend more time and money to provide irrefutable evidence in large scale populations.

Now where do we go from here? How can we ensure the next large scale, large population, prospective, study is methodologically sound enough to provide this evidence before we devote resources to a follow up study? I hope greater minds than mind way in on the next step in the process.

Thanks for the great analysis and thought provoking position.


Newbie said...

Was there a difference between the olive oil (more n3) versus the nuts (more n6)? Also, Mediterraneans do eat lots of grains, so telling them to eat less would mean the study group is not really following a "Med" diet... just saying.

Unknown said...

Good post Stephan, and that's just the beginning of the flaws in my opinion. Intent to treat analysis and such a large dropout, the dietary recommendations for the control was considerably different. For instance why would the low fat group recommendation for vegetables be >2 servings/wk and the Med group be >2 servings/day? Also in the end there was no difference in myocardial infraction or all cause mortality, something that isn't being mentioned, and the realistic risk compared to the control was quite frankly small.

In the end the study was not that impressive, their conclusions seem overly enthusiastic and of course the media are having a field day.

Also the Mediterranean diet has long since shown to be a healthy diet and it's a high fat diet, another point that isn't being discussed in the media, or at least that I've seen. It's also worth noting the saturated fat content of both diets were virtually identical, but again, the outcome differences were not that impressive.

Gretchen said...

I think one problem with all popular press reports of nutritional studies is that they trumpet "reduces X" without saying in comparison to what.

Almost any sensible diet of real food will improve health if you're comparing it with the "standard American diet."

I think all diet studies in "free-living" subjects are flawed, and the type of research you're doing, trying to find out *how* dietary changes affect the body, are much more interesting.

Ross said...

This is a minor point, but in my glance through the paper it seemed like the overall group effects on "events" were driven by reduced risk of stroke; there were not reliable effects on risk of heart attack or death. Am I reading it correctly?

glib said...

For someone like me this study is worthless. Besides the points you make, the Mediterranean diet apparently had unlimited eggs. That is not Mediterranean, at least not the way Mediterranean is usually described. The eggs and meat were, surely, of the corn-and-soybeans variety. The low fat diet may have used corn oil spray to saute' things, who knows, and maybe even low fat milk. Not a particularly enlightening study, no.

I will, however, point out that the places with particularly high longevity described in the book The Blue Zones, in all of those vegetable garden play a big role. That is one of the pillars of the Mediterranean diet no doubt. Those places are also much warmer than the Midwest, and while I can and do eat like an Okinawan in summer (near vegetarian, mostly from the garden), in winter I am much more into meat and fats. The low is going to be 12 tonight.

Laurence Chalem said...

Hi Stephan. I enjoyed your article, as usual and would like to add that the 30% difference between the groups is a gross exaggeration, utilizing relative risk, a ratio of a ratio. The real difference is 3.4% on the Mediterranean diet suffered a heart attack, stroke or death from cardiovascular disease versus 4.4% on the control diet, which is 1%. That's the absolute risk in a population of 7,447 people. 30% seems like a lot, but 1% doesn't seem very significant. While the Insituto de salud Carlos III is an accredited Universidad, they are linked to Spain's Dept. of Agriculture, which, you may or may not agree, has an interest in the export of olive oil, nuts, et al.

A more aligned reality is that our bodies have evolved to be robust enough to thrive in spite of whatever insults we subject them to, grains and the feces of sugar-fed bacteria (e.g., wine) included.

Best wishes... - lc

Unknown said...

I am sure the Mediterranean diet does reduce cardiovascular risks. And of course, so does exercise. If you can make it part of your regular routine, it’s much easier to be motivated. I started commuting by bike so I have to exercise to get to work!

Stephan Guyenet said...

Hi Brad,

I think it goes to show how difficult it is to know anything with a high degree of certainty.

Hi Newbie,

No difference.

Howdy Ross,

Good point, most of the effect came from a reduction in stroke. The control group had twice as many strokes per cap as myocardial infarctions in this study! Very unusual for a Western population, but it would be typical for Japan.

Hi Laurence,

Personally I am not usually swayed by the argument that relative risk is misleading. It would be essentially impossible to design a trial to detect differences of 30% in absolute risk, and exceedingly difficult even for 10%. I think we have to take it on a case-by case basis; I usually find relative risk to be adequate, but from time to time it is true that relative risk doesn't make sense.

Will Hui said...

Hi Stephan,

The authors wrote,

"In addition, to account for the protocol change in October 2006 whereby the intensity of dietary intervention in the control group was increased, we compared hazard ratios for the Mediterranean-diet groups (both groups merged vs. the control group) before and after this date. Adjusted hazard ratios were 0.77 (95% CI, 0.59 to 1.00) for participants recruited before October 2006 and 0.49 (95% CI, 0.26 to 0.92) for those recruited thereafter (P=0.21 for interaction)."

Wouldn't this suggest that increasing the intensity of the low-fat intervention actually exacerbated the differences between the Mediterranean and control diet groups?

Reijo said...

While I think you have a point in stating that the control group received less attention during the 1st half oth the study, I don't think it's reason enough to abandon the data. The effect of attention is often overdriven; this notion is supported by the results of WHI trial. Despite the huge attention in the low fat arm, only some patients seemed to be able to reduce fat content of their diet substantially.

Results are exceptionally well in line with Lyon Diet Heart (& Oslo Diet Heart) and prospective cohort studies (Mente et al. 2009).

In addition, this trial had very low drop out rates, patients were already on multiple drugs and the design of the study (RCT with disease and mortality end points) is at the top level in the hierarchy of evidence based nutrition. The controls also consumed 5 servings of fruits or veggies a day (already at the baseline), and had only moderate amounts of saturated fat in their diet and used about 25 grams of fiber at, ie. 10 grams more than an average person in US and just in line with US fiber recommendations (AHA or American Dietetic Association).

That being said, I don't think there is such a thing as universally perfect diet. Mediterranean diet is a feasible diet for masses in western society though.

One of the researchers Jordi Salas-Salvadós responded the critique by Dean Ornish and some more points at my blog.

Stephan Guyenet said...

Hi Will and Reijo,

Thanks for pointing that out, it does go some ways toward making me feel more confident about the conclusion.

The hazard ratios of 0.77 and 0.49 were not significantly different from one another, so we can't say that HR decreased after the low-fat intervention intensified. However, we can say that the reduced HR was significant even after the LF diet group intervention was intensified.

I still have reservations about the study because 1) the low-fat diet group scarcely adhered to the intervention, which makes me suspicious since a number of other LF diet trials have gotten reasonable adherence, and 2) post-hoc analysis is not a perfect way to fix a study that was not properly designed to begin with (though it can help somewhat).

Mike said...

There was another interesting trend in the supplementary index that got no mention in the original article when you look at the HR over time. Considering only events that occurred during the first two years, the HR were 0.57 and 0.46, between years two and four they were 0.62 and 0.80, after year four they were 1.11 and 1.08 - showing fewer events in the control group. It makes me wonder what the results would have been if the trial hadn't been stopped, also makes me question the motives for stopping it. Maybe it's insignificant and I'm just overanalysing, but that seems very odd to me.

Anonymous said...

It's also a little disingenuous to to purposefully pick a control group you know does not reduce heart disease. This makes it a lot easier for your experimental group to do better.

raphi said...

1st comment so I wanted to say great blog Stephan right off the bat! :)

Could you please explain (with an example preferably) why absolute risk is not always the most important ratio stemming from nutritional studies? What else could be more informative? and WHY? I ask since after reading this article by DR. RD Feinman it seems very clear that understanding the differences between different HZ's is important, but that recommendations should first of all come from absolute risks.

Florent Berthet said...

I agree. It seems we will never find a definitive study on the matter. Maybe if there were better incentives to find the truth...

On a related note, I'd like to know what you think about the following abstract which concludes that "diet has benefits that come directly from foods, as well as from the reduction in saturated fats, cholesterol, meats, and fatty dairy foods" :

Chris Wilson said...

All indices have strengths and weaknesses. The relative risk ratio gives you more information about the efficacy of a particular intervention. The absolute risk reduction provides context. However, to interpret the absolute risk reduction correctly, you need to understand some probability. Over time, small rates can integrate surprisingly.


Unknown said...

Great post! I've been searching a good diet for a long time and I think mediterranean diet will be a great help to me.

Unknown said...

Good points. Thanks.

One factor seems conspicuously absent from this study, which is the relative changes in weight among the test and control groups.

As Peter Attia notes, these subjects were mostly overweight, and had metabolic syndrome or diabetes. Any changes in weight could easily explain these outcomes.

Unknown said...

I wish all that were true.

Like others already pointed out though, there are lots of other factors to consider. Like myself, coming from a usuall SE Asia diet modified to be more paleo friendly, eating med-style screws up my body in all sorts of ways.

My canary in the coal mine is my eyes. While paleo, I can see 20/10 with a -2D prescription. If i stray, after two months I can barely get to 20/30 on a Snellen chart.

On the whole eyes and paleo subject, might want to Google Frauenfeld Clinic. Alex is paleo friendly and amazing about the subject of vision improvement.

Unknown said...

Thanks for writing such a great article on cardiovascular diseases.I will, however, point out that the places with particularly high longevity described in the book The Blue Zones, in all of those vegetable garden play a big role.

milan said...

I am sure the Mediterranean diet does reduce cardiovascular risks. And of course, so does exercise

Unknown said...

I certainly believe that mediterranean diet can reduce cardiovascular risk. If you observe countries whose dominant diet is similar or close to mediterranean diet their cardiovascular risk statistics is much lower compare. So, there must be some truth that this diet can somehow reduce cardiovascular risk.