Don't Switch to the Mediterranean Diet Just Yet

The New York Times made big news with reports that the New England Journal of Medicine study on the beneficial effects of the Mediterranean diet showed it could dramatically reduce the rates of heart attack and stroke. But this study has major issues that bear directly on whether or not physicians should make new recommendations about dietary intake of fats like olive oil, or whether patients should adopt the diet as a whole. Let’s talk about the trial.
First of all, this is a randomized, controlled trial, in which 7447 men and women between 55 and 80 years of age who had major risk factors for cardiovascular disease such as diabetes, obesity, smoking, hyperlipidemia etc., were divided evenly between 3 groups, one which received recommendations on a “low fat” diet, and two in which there was extensive counseling on the Mediterranean combined with either a ready free supply of extra-virgin olive oil, or alternatively a variety of nuts.
The primary end points being studied was the combined number of heart attacks, strokes, and death, and over the course of about 5 years of study about 288 such events occurred. If you combine all three of these end points together, and evaluate their frequency between the groups you find 96 of these end points occurred in the “Mediterranean diet with extra-virgin olive oil” or 3.8% of the group, 83 occurred in the Mediterranean diet with nuts for 3.4% of that population, and 109 in the control group for 4.4% of the controls.
But before anyone takes these results to heart, we have to recognize major flaws with the study design, and the populations that comprised these three groups. First, the rate of primary events was surprisingly low for such a high risk group, and because the study was stopped early, absurdly for “ethical reasons”, the number of events is quite low. For the life of me I can’t think of what that ethics committee was thinking. These results are not that dramatic. Further, the “low fat” diet was very ineffectually enforced or counseled, to the point that midway through the study the authors revised the protocol to include more counseling sessions. Evaluating the supplementary data, specifically table S7, you see this control group was in no way on a low fat diet. They still were consuming 37-39% of their calories from fat! “Low fat” should have 10-15% of calories from fats, so basically, everyone ignored the diet. Further, all of the groups consumed a similar amount of total fat, mono and poly-unsaturated fats, and even a used olive oil as their main culinary fat. All groups consumed (see table S5) a similar amount of red meat (forbidden from all diets), butter, soda, baked goods, etc. The places where there seemed to be more dramatic differences were in olive oil consumption (about 50% of controls had > 4 tbsp a day, vs 80% of the “nuts” group and about 90% of the “extra-virgin olive oil group), wine consumption (modest at about 30% in diet groups vs 25% in “low fat” control), nuts (crazy high at 90% in nuts group, vs 40% and 20% in “olive oil” and “low fat”, as well as modest elevation of the amount of fish, fruits and vegetables in the Mediterranean groups. Further, some of these differences, such as the consumption of alcohol, fruits and vegetables, was higher in the Mediterranean groups at baseline (notice no mean change in table S6) so the groups may have started out in a different place.
What does this mean? First of all, we have to reject the notion that this study compared Mediterranean diet to “low fat” diet. This was a study of basically no diet intervention versus increasing your intake of fish, nuts and/or olive oil. Otherwise, there didn’t appear to be compliance with the negative suggestions of the Mediterranean diet, to decrease red mean intake, baked goods, dairy, etc. The participants basically took the recommended items and increased them in their diets, but didn’t exclude any of the “discouraged” items. This is very interesting, but to call it the “Mediterranean diet” is misleading. In reality, it’s diet supplementation with olive oil, nuts and fish.
Second, the final results, while they sound impressive (30% reduction in combined primary end points!) are actually not as important as some of the less-emphasized findings. For this we have to evaluate the secondary endpoint, which happens to be the one we really care about – all cause mortality. They could not show a difference in mortality! So while you might be less likely to have a heart attack or stroke, you’re no less likely to die. This is why I’m so confused they ended the study early. This is really the only end point that matters, and it was unchanged at the interval at which the ethics committee decided this study had to be stopped for efficacy. Why did they do this? The evidence is suggestive that with more participants, the Mediterranean diet + olive oil might have diverged a bit and shown a benefit compared to the do nothing “low fat” control, but this didn’t reach significance.
What have we learned? Compared to other Spanish folks between the ages of 55 and 80, all with cardiovascular risk factors, those that added olive oil, nuts, and fish to their diet had fewer cardiovascular events, but no difference in their mortality compared with people that did nothing to change their diet.
Why did this make the front page of the New York Times? Let’s show a little bit more critical analysis of findings, and not just swallow the PR.

20 thoughts on “Don't Switch to the Mediterranean Diet Just Yet”

  1. “They could not show a difference in mortality! So while you might be less likely to have a heart attack or stroke, you’re no less likely to die. This is why I’m so confused they ended the study early.”
    Sorry, but I disagree with your interpretation. In safety studies, it is common for mortality to have lower incidence than the indicator endpoint (in this case, CV events of stroke, MI, or death). Use of the combined CV endpoint as surrogate for death is standard in the field of CV intervention; without doing this, studies would have to be many-fold larger and would never be practical to complete. In other words, the study was underpowered for the secondary endpoint but correctly powered for the primary endpoint, which is appropriate.
    Because it is so widely accepted that MI and stroke are leading causes of death, it is unavoidable to infer that substantive decreases in MI and stroke would, if the study had been larger or continued longer, have been associated with a difference in CV mortality. Hence it was unethical to continue the study once a statistically significant difference was observed (with proper correction for early testing and multiplicity, which was done).

  2. David, I’d generally agree, but the problem is only one of their arms had a difference on mortality, which I think further confounds this result. While med diet + olive oil appeared as though it could have become significant with more numbers, med diet + nuts appears no different at all in mortality compared to the do nothing control. The mixture of results on this end point leaves me feeling queasy about whether or not something real is happening here.
    It makes me happy to see the dominant fat I use in my cooking doesn’t appear to increase cardiac events. But these data wouldn’t make me start adding an additional 4 tbsp a day of olive oil, and certainly not a pile of nuts, for the hell of it. I don’t think based on the patient characteristics that they implemented a realistic low fat diet, so it doesn’t represent a “superiority” study, and I still think they ended it too soon given the divergence between the two study arms.

  3. So people can’t stick to a low-fat diet, even in a highly controlled setting?
    Maybe that in itself is an important result.
    Obviously there are two solutions when research hits a wall like that:
    First of all, more of the same! The researchers can say they didn’t hit the wall hard enough with their heads. Or their patients heads. Maybe a forced partial starvation regime? Clearly people are more gluttonous than we thought. Or are there any psychiatric medications with a positive influence on BMI? Or maybe try heavy metal poising – I’m sure one could find a dose of heavy metals with a positive influence on BMI. Or maybe a combination of forceful starvation with low-dose heavy metals. Surely a creative mind can find new approaches to hit one’s head harder against an unyielding wall. I call it the denialism approach – it’s always this unruly reality.
    Or if one hits an unyielding wall like that, one might need to consider that one is wrong.
    Maybe the christian image of the human being as gluttonous and lazy (you know, calories in, calories out, thermodynamics, yadayada), and the only animal that needs to be forced to eat less, and forced to exercise more, maybe that image is wrong? And have you considered that saturated fat – the most basic of the fatty acids – might actually be *not* unhealthy? Unlike say a consumption of supra-natural amounts of n-6 fatty acids, that gets metabolized among others in supra-natural amounts of series 2 prostaglandins? And I’m sure if you really wanted to know, you would find other evolutionary novel foods besides seed oils that have the potential to be involved in those pesky diseases of civilization.
    Whatever it is, I am highly optimistic that another couple of decades of more of the same kind of research (that tries so hard to frame fat and especially saturated fat) will find the answer. Can’t wait for medical science to come up with an answer that is based in reality! I’m holding my breath! Until then, try to half-starve yourself and ignore evolution!

  4. Mark, I really need to apologize that I am so lousy in hiding my contempt for idiots. It’s just that nutritional science reminds me so much of theology. It’s a religion. How many low-fat diets interventions can dance in research paper?
    What was your approach to idiots that don’t respond to data? Shame them? Just imagine, doctors handing out murderous nutritional advise for decades! Will shame be enough? Or should the smart doctor invest in pitch-fork factories now? Lucky you, that the nutritional BS of decades is so ingrained that people believe whole-heartly what is and isn’t a heart-healthy diet. It’s the eeeeeviiiiiil saturated fat. No amount of data, and no amount of shame will change that for most of the people who have swallowed that.
    Personally, I find ridiculing idiots better than shaming them – it both is ineffective in changing their opinions, but at least I get something out of it and have my sarcastic fun.
    Please, keep us updated from the highly fascinating world of nutritional science. As I said, can’t wait to hear more news how to tackle the saturated fat problem!

  5. I´m glad to read this news! Today olive oil is one of the most popular edible oils worldwide both for its nutritional value as for its high gastronomic qualities. Its high content of unsaturated fatty acids, vitamin E, natural antioxidants and other nutrients make medical science considers as one of the healthiest.
    There are many studies showing the benefits of the Mediterranean diet on weight control, cancer prevention breast, cerebro-vascular accidents, from heart attacks and anaemia, among other diseases. The Mediterranean diet besides being varied and balanced nutritional intake, is rich in fiber, unsaturated fatty acids and antioxidants. More than a diet, is said to be a way of life, which to generate a positive effect obviously be combined with moderate exercise daily. goo.gl/I5FKH
    I invite you to learn more about the health benefits of the Mediterranean diet by visiting: goo.gl/39RVR
    Regards!

  6. PhotoRecipe, if you are even a real person, you shouldn’t be glad.
    1. The participants in the study did not adopt the mediterranean diet. They only increased their consumption of olive oil, fish, and nuts, they didn’t adopt the avoidance of red meat, butter, chicken etc.
    2. Mortality was not affected [ed. in the med + nuts arm – there was a statistically insignificant decrease in the med + olive oil arm – per David’s point]
    3. The “med diet” plus nuts arm was identical in mortality to the ‘low fat’ control
    4. The low fat control wasn’t a low fat diet! It was a “do nothing” control.
    This study showed nothing about the efficacy of the Mediterranean diet, because no one was actually on it. It showed nothing about superiority of this diet to low fat diet, because no one was on a low fat diet. This study, really didn’t show much of anything at all, except bumping up your olive oil intake may decrease cardiovascular events, but not your mortality. Color me unimpressed.

  7. So people can’t stick to a low-fat diet, even in a highly controlled setting?
    Maybe that in itself is an important result.

    Tony, this was not a “highly-controlled” setting. They basically sent a leaflet every year to the low-fat diet group telling them what a low fat diet was. After that proved ridiculously ineffective, in 2006 they upped the protocol to in person sessions with a nutritionist, still, a relatively low level of intervention. Nothing compared to what they did with the other groups, which included sending them the relevant foodstuffs in the mail weekly.

    Maybe the christian image of the human being as gluttonous and lazy (you know, calories in, calories out, thermodynamics, yadayada), and the only animal that needs to be forced to eat less, and forced to exercise more, maybe that image is wrong? And have you considered that saturated fat – the most basic of the fatty acids – might actually be *not* unhealthy? Unlike say a consumption of supra-natural amounts of n-6 fatty acids, that gets metabolized among others in supra-natural amounts of series 2 prostaglandins? And I’m sure if you really wanted to know, you would find other evolutionary novel foods besides seed oils that have the potential to be involved in those pesky diseases of civilization.

    It’s the Christian version of human physiology that suggests that we are a biological machine that is subject to the laws of physics? I don’t get this bizarre association you have with this, or who you’re arguing with. I’m not sure what idea you’re railing against. That eating less makes you lose weight? Yes, but it’s more complicated than that, we’ve discussed obesity science here extensively. I don’t know when I’ve specifically attacked saturated fats as you think I have. So I’m mostly just confused what you’re so angry about, as always.

    Whatever it is, I am highly optimistic that another couple of decades of more of the same kind of research (that tries so hard to frame fat and especially saturated fat) will find the answer. Can’t wait for medical science to come up with an answer that is based in reality! I’m holding my breath! Until then, try to half-starve yourself and ignore evolution!

    What ever your dislike of specific nutritional recommendations, obesity is bad for you and represents a public health crisis. The effects of these other diet issues beyond excess weight aside, the best thing you can do is avoid becoming obese. As far as “starving” yourself, if you read some of the links above, you can see why this is problematic and usually ineffective. We have to focus on prevention, because it’s extremely different to modify obesity once it’s established. I don’t know why, also, you think evolution comes to bear on this discussion, we’re talking about diseases of people far beyond their reproductive years, evolution has no meaningful input for the diseases of the elderly. They’ve already done what natural selection expects of them.

    Mark, I really need to apologize that I am so lousy in hiding my contempt for idiots. It’s just that nutritional science reminds me so much of theology. It’s a religion. How many low-fat diets interventions can dance in research paper?

    I tend to agree, most nutritional science is total crap. This paper is another example if you ask me, and it made the front page of the New York Times. Maybe this can be a starting point for rational discussion?

  8. Mark, I think you’re reading too much into a secondary endpoint (mortality), when the primary endpoint of the study reached an unambiguously positive result. Coupled with the low mortality rate, early termination meant that the study is clearly not adequately powered to support valid inference about differences in mortality between arms.
    Also, you’re mixing “absence of evidence” (eg, no finding of a statistically significant difference) with “evidence of absence.” This study was not designed to rule out a meaningful difference in mortality, and is not adequate to do so. The only thing about mortality that should be inferred is that whatever difference actually exists is too small to be detected with the current study (given the sample size, and the shortened duration). It’s not right to conclude that “mortality was not affected” – you should limit the statement to “the effect on mortality was smaller than some statistically determined lower limit” and that is not the same thing at all.

  9. David, that’s why they shouldn’t have ended the study early. Maybe I’m being too dismissive of studies such as this that don’t use mortality as a primary endpoint. And I agree, I can’t say there was no effect on mortality, but critically, the effect was only seen in one of the two Mediterranean diet arms with Med+nuts tracking more or less the same as the no intervention “low fat” diet. Not to mention, no one was really on the Mediterranean diet! What does that mean?
    It means excess olive oil intake > 4 tbsp a day appears to have a meaningful effect on cardiovascular effects, and might decrease mortality, with a larger n, this could have been determined. The Mediterranean diet, or at least the aspects the patients were compliant with, on it’s own did not have an effect, at all, on mortality, as the med+nut arm didn’t decrease mortality. The mixture of effects between the two arms on the secondary endpoint, as well as the compliance data in the supplemental figures S5-S7 makes me think this study is basically meaningless about the Mediterranean diet.
    I can’t help but conclude then that this study is being overblown. It doesn’t show superiority to low fat, since that was not even studied. The patients didn’t adopt fully half of the diet parameters. And when divided between those that consumed more nuts vs olive oil, one group had identical mortality to the shoddy control, and one had a statistically-insignificant decrease.
    I just can’t get excited about that.

  10. IMHO, it was entirely proper to end the study early, and not doing so would have been unethical. The link between the combined CV endpoint (MI, stroke, or death) is firmly established and it is clear that a treatment that reduces that endpoint is highly likely to also reduce mortality, in the absence of a non-cardiovascular treatment risk. Running the study longer, or recruiting more subjects (either approach to getting more patient-years) would have exposed more subjects to avoidable MI or stroke. The only counter-balancing benefit would be to confirm that MI & stroke are associated with death, which we already know. Hence continuation would have exposed subjects to risk without meaningful benefit, and would have been unethical. Basic Helsinki stuff.
    Essentially every modern CV safety trial I am aware of is event-driven and uses an endpoint structure similar to this trial with a safety monitoring board and a similar stopping rule (though the details of stopping rules can be mind-numbingly complex, depending on the trial structure). No modern IRB would approve such a trial without an interim safety stopping rule. You’ve probably guessed by now that I design clinical trials for a living (for a pharma company) – I should add I have no conflict of interest with nuts or oils.
    I do agree with you that interpreting the study is complex. In part, as you point out, because the base case wasn’t really a mediterranean diet, and it’s not clear if the added nuts/oils led to reduction in any source of calories. Also, it seems to me that there’s no way to estimate the optimal dose of nuts or olive oil. Nonetheless, the study strongly supports at least one clear conclusion: our diets should contain some amount of nuts and or olive oil.
    As for your statement about “superiority over low fat” – reality is that treatments are evaluated based on intent-to-treat. You can tell your patients to eat a low fat diet until you’re blue in the face, and they won’t do it. But if you tell them to put a handful on nuts on their breakfast cereal each day, you are providing an intervention that will likely have good compliance. That’s another valid inference from this study. So, what advice are you going to give – advice that is proven (in the past) to be unlikely to be followed to an effective degree, or advice with good compliance and proven outcome efficacy?
    Finally, I could point out that there is at best marginal evidence that a low-fat diet is associated with reduced mortality.

  11. I agree that intention to treat failure is valid, but the level of intervention was quite different between groups. There’s a big difference between sending people food for free, and yearly counseling sessions, and basically sending another group a pamphlet in the mail. It seems to have been almost designed to fail.

  12. You comment on “whether or not physicians should make new recommendations” based on this diet study. If you look at the flimsy evidence available in all the literature, the real issue though is not about “new” recommendations but “any” recommendations whatsoever. For example what evidence is there that low-fat is any good? Are they of good quality? No they aren’t if it has been suggested that this Mediterranean diet study is now the gold standard. Yet the assumption is made that the low-fat diet is the correct diet. Why? 30 years of low-fat diet prominence has only led to increasing obesity rates. Look at the most prominent studies of the past in support of the low-fat mantra and you see confounding left and right mainly because insufficient accounting of carbohydrate intake. Most of it is garbage.

  13. The relative reduction in cardiovascular events with the “active comparator” diet as compared to the standard Western diet is actually greater than the expected relative reduction with statins. Statins in primary prevention generally do not significantly reduce total mortality even for middle-aged high-risk men, much less for women and the very old. Yet a 20% or so reduction in heart attacks is considered adequate reason for lifelong medication with potential side effects. It is not possible to state that the same type and magnitude of benefit is worthless if a dietary intervention is responsible.
    Likewise, the JUPITER trial of statins was famously ended after less than two years “for ethical reasons”, allowing unusually good (and indeed suspicious) early results to be reported as the final results. Would you describe that action as “absurd”? If not, ending a dietary trial early after five years is not absurd either. Enough events had been accumulated to allow the authors to demonstrate that the dietary intervention significantly reduced risk. Therefore, there was no longer equipoise. Trial participants had a right to be informed of that fact.

  14. Jane, I guess it’s just my prejudice against mortality as secondary endpoint rather than a primary. I realize it’s hard to make it a primary endpoint. But my unwillingness to automatically infer benefits from the primary endpoint to the one that matters most to me might just have to go down as a personal failing.
    Then there’s the issue that I keep bringing up that no one can explain to me. Why did one arm have identical mortality to the controls, while the other trended (insignificantly)away? Shouldn’t we mention that this suggests that the whole trial has nothing to do with the Mediterranean diet (since there weren’t really any difference in critical criteria between controls and experimental arms), and maybe has more to do with olive oil as a dietary fat? I don’t know what to make of that.
    I love olive oil, I think it’s great that this study appears to show it’s less likely to kill me. Was this a study of the Mediterranean diet? Not really. Will a bunch of olive oil lengthen your life? Maybe. Will a bunch of nuts? Eh, maybe not. Is that NYT-worthy front page material? Not to me.

  15. Hello, first of all, of course I am a person! I use this nickname because it´s easier for me to be associated with our web application of Mediterranean food that we are about to bring to the market.
    Thanks for your comment, I understand your point of view, of course there variations in the study on the groups, and should be more studies on this, but let me leave you this article that demonstrates increased life expectancy that is achieved by following the diet Mediterranean. http://www.sciencedaily.com/releases/2005/04/050425111008.htm
    Regards,
    M.V.D María de los Angeles Tort

  16. We see no hint that non-cardiovascular causes of death increased; there were just very few deaths, period, and so no difference in cardiovascular mortality, much less total mortality, could be seen. But it seems likely that if you avoid cardiovascular events, you will somewhat reduce premature cardiovascular mortality. Again, people who, reasonably or not, have particular fear of heart attacks think they will be well served if they get a reduced risk of heart attack from statins, despite the lack of a mortality benefit. Whether they are being well served or not depends upon individuals’ values and the side effects they suffer. The “side effect” of sauteing stuff in olive oil is that it tastes better. If olive oil is as potent as a statin or better, people who take statins might also want to take olive oil. Informed consent!

  17. I like that read Jane, and considering side effects of statins like muscle soreness, and expense, I’d take the olive oil any day.

  18. A lot of American rhetoric misses the point that eating real, traditional food is not a burden – it’s a pleasure. Having been forced to go on a home-cookin’ diet to reverse a family member’s serious health problem, I’d never choose to go back to the fat-food-in-bags diet. Processed food now tastes too salty, too sugary, and otherwise flavorless and blah.

  19. Great article. There’s a lot of bad nutritional science and it’s important to read every health story that makes headlines with a heavy dose of skepticism. It seems the findings in this study were negligible, so it shouldn’t have been hyped in the New York Times.
    This kind of reporting has an insidious effect because not only is it bad science, but it makes good public health advice appear less credible. Maybe journalists should hold back from reporting on studies until they are endorsed by a number of public health organisations? With this in mind, one story that seems trustworthy for this reason affirms what we’ve known for a while: eating large amounts of processed meat is bad for our health and may lead to increased mortality. But yes, the reporting in the NYT was completely overblown.

Comments are closed.