The so-called mediterranean diet is perhaps the best publicized “healthy diet”, and has been for decades. It was “discovered” by Ancel Keys back in the 1970’s, and is based on the diets that were typical in the mediterranean region during the 60’s and 70’s.
Ancel Keys is largely responsible for the current slate of official diet recommendations promulgated by health authorities around the world, most of which are dead wrong. So I’m not a huge fan of the man. I’m also not sure where the idea that the mediterranean diet was healthier came from in the first place. Average life expectancy in Greece back in 1970 was 71 years. In Italy it was also 71 years. In Spain it was 72 years.
In contrast, in the United States, life expectancy at the time was 71 years. Hmm. And in the United Kingdom, it was 72 years. So, the Americans who were stuffing their faces with hamburgers, and the Brits who were stuffing their faces with marmite on toast, were not actually living any shorter lives than the South Europeans who were eating lots of vegetables and chugging gallons of olive oil. Strange.
Here in Sweden, the life expectancy in 1970 was 74 years. So Swedes were actually living three years longer than both Anglo-Saxons and Southern Europeans. Someone should have told Ancel Keys. Then maybe instead of the Mediterranean diet, we’d be hearing all about the Swedish diet, and the health benefits of meatballs and rapeseed oil.
One major problem with the mediterranean diet as a health intervention is that it contains a very large number of different parts. Even if some parts are good for your health, and the overall diet is beneficial, there could easily also be other components that are bad for you.
Here are the ten major components of the mediterranean diet:
- High consumption of olive oil.
- High consumption of nuts.
- High consumption of fruits.
- High consumption of vegetables.
- High consumption of legumes.
- High consumption of wholegrain cereals (pasta, bread, etc.)
- High consumption of fish
- Low consumption of other types of meat
- Moderate consumption of dairy products
- Moderate consumption of alcohol
So, that’s ten different pieces of advice you’re supposed to follow, if you want to follow the mediterranean diet. It’s not hard at all to imagine that some parts might be good for you, while other parts might be bad for you, and still others might be neutral. As I wrote about recently, red meat appears to be neutral from a health perspective. And there is today plenty of evidence that a high consumption of carbohydrates is bad for you, and that a low carb diet will make you healthier.
Anyway, let’s look at what the evidence actually says. We’ll start off with the cohort studies, and then move on to the randomized trials. Each has its advantages. Cohort studies are observational, so they can only show correlation, not causation, but they are often big, and they’re cheaper and easier to do, which means there’s lots of them. Randomized trials are more expensive and more complicated, so they’re fewer, and smaller, but the data they provide is higher quality and much less likely to be affected by confounders.
A systematic review and meta-analysis was published in Advances in Nutrition in November 2019, that looked at the cohort studies of the mediterranean diet that had been carried out up to that point. In order to be included in the review, studies had to report on mortality data, since the primary purpose of the review was to see if there was evidence that a mediterranean diet would allow people to live longer. The authors didn’t receive any specific funding for the work, and reported no conflicts of interest.
In total, 29 cohort studies were included in the review, with a total of 1,7 million people. The follow-up period varied from four years to 32 years. The different studies used various different methods of assessing the extent to which participants were aligned with the mediterranean diet. The authors of the systematic review translated these different scoring systems to a single scoring system called the MedDiet Score, which gives you from zero to nine points based on how well you are adhering to a mediterranean diet. In this scoring system, a score of zero means your diet is not in any way aligned with the mediterranean diet, while a score of nine means your diet is in perfect alignment.
Just to give a basic understanding of how this scoring system works, here’s a graph you can use to calculate it for yourself.
Personally, I score a three…
Anyway, back to the study. For each two point increase in alignment with a mediterranean diet on the MedDiet Score, there was a 10% reduction in mortality over the course of follow-up. The result was highly statistically significant. If you extrapolate fully, that would mean going from a zero on the MedDiet score to a nine would result in a 45% relative reduction in risk of dying over the course of follow-up.
That is a big reduction. But how then can we explain that people in the mediterranean countries weren’t living any longer than people in the US and UK back in 1970?
With a reduction that big, there should have been a noticeable difference in life expectancy, and yet the difference was exactly zero percent.
My guess would be because these results are based on cohort studies, so the scope for confounding is enormous. The earliest of the included studies was published in 1995. All but two of the 29 included studies were published in 2005 or later. The idea that a mediterranean diet is healthy has been widespread since at least the early 1990’s, so people who choose to eat a mediterranean diet are for the most part people who are health conscious. Apart from taking care with what they eat, they probably exercise more, smoke less, and do a thousand other things that affect their health positively.
As I mentioned in my earlier article about red meat consumption, telling people that something is unhealthy will cause those who care about their health to avoid it, and telling people that something is healthy will cause these same people to seek it out, which will create a self-fulfilling prophecy. Things that have been deemed unhealthy will correlate with a shorter life expectancy, while things that have been deemed healthy will correlate with a longer life expectancy in observational studies.
For this reason, we should be highly skeptical of anything but very large effect sizes. Really, anything less than a halving or doubling of risk should be viewed with the utmost skepticism until it has been confirmed in randomized controlled trials.
So let’s move on, and look at what the randomized trial data shows. The Cochrane Collaboration published a systematic review of the randomized trials of a mediterranean diet in March 2019. The review included 30 trials, with a total of 12,461 participants. There was a lot of variation between the trials, in terms of the groups studied and the control groups used.
The biggest study by far was PREDIMED, with 7,747 participants. It compared a mediterranean diet with a low fat diet. Since it towers over the other studies in the review in terms of size (contributing more than 60% of the participants), we’re going to spend some extra time discussing it.
PREDIMED was conducted in Spain. It ran for eight years, from 2003 to 2011, with an average follow-up of five years per participant, and was published in the New England Journal of Medicine in 2013. It quickly became one of the most cited studies in history, because it showed an impressive 30% reduction in the relative risk of a combined end point that consisted of cardiovascular death, stroke, or myocardial infarction among people eating a mediterranean diet a compared with a low fat diet.
In case you haven’t heard of combined end points before, a combined end point is just when you add a bunch of different end points together. Personally, I hate combined end points, because they are highly misleading. In fact, I think serious scientific journals should refuse to publish studies with combined end points.
If you get in to the nitty-gritty of the study, you find that there was actually no statistically significant reduction in deaths, nor in any any of the other end points studied, except for stroke. So, the reduction in the combined end point seen in the mediterranean diet group was almost entirely driven by a reduction in strokes. When you look at enough different end points, one is going to pop out as being statistically significant just by chance. So the results were actually nowhere near as impressive as they seemed at first.
Additionally, the trial was stopped early, another practice that really should be banned. Why? Because letting researchers stop a trial early gives them the opportunity to cherry pick the time point at which the study is stopped. This pretty much guarantees inflated results.
But neither of these is the main problem with the PREDIMED study. The main problem with PREDIMED is that it became known in 2018, five years after publication, that not all the participants had been properly randomized. This resulted in the paper being retracted, and then re-published with corrected results.
The re-published results were funnily enough no different than the initial results, but the fact that they were forced to retract and re-publish makes me wonder what else they might have screwed up, that we simply don’t know about. The only reason the mess with the randomization became known was because an outside team of statisticians noticed that some of the numbers being presented were highly implausible.
Anyway, that’s enough background about PREDIMED. Let’s get back to the Cochrane review. Only four out of the thirty studies actually reported clinically relevant endpoints such as heart attacks, strokes and deaths. The two largest of these were PREDIMED and the Lyon Diet Heart Study.
As mentioned, PREDIMED found a 30% relative reduction in the combined endpoint, but the reduction was almost entirely driven by a reduction in strokes. There was a 40% reduction in the relative risk of stroke that was statistically significant. There was also a 21% reduction in the relative risk of having a heart attack, but it wasn’t statistically significant. There was no difference in overall mortality between the mediterranean diet and the low fat diet. In fact, there wasn’t even a hint of a difference.
The Lyon Diet Heart Study was much smaller than PREDIMED, with just 605 participants who were followed for an average of four years. It recruited people who had recently had a heart attack and randomized them to either a mediterranean diet or just the standard dietary advice given to heart attack patients at the time (the study was carried out in the late 80’s and early 90’s and published in 1994). Just like PREDIMED, the trial was stopped early, which means we can be pretty certain that the results are inflated. And just like PREDIMED, it used a combined endpoint. In this case the combined endpoint was cardiac death and non-fatal cardiac infarction, for which the study found a whopping 72% relative risk reduction. When it came to overall mortality, the study found a 56% reduction in relative risk, which was also statistically significant.
Of the two remaining studies that reported clinical outcomes, the first was a study of 180 patients with high blood pressure. Half were randomized to a mediterranean diet, and half to a low fat diet. The study showed a 67% reduction in relative risk of stroke over two years of follow-up, but the results were nowhere near statistically significant. Basically, the study was just too small to show a difference even if one does exist.
The second was even smaller, with just 101 heart attack patients, who were randomized to either a mediterranean diet or a low fat diet, and followed for an average of four years. The groups did not differ in any clinically meaningful outcomes, which again isn’t surprising considering how small the study was.
So, of the four studies that did follow clinical outcomes, the two biggest ones both do suggest significant benefit, although both have major flaws that could easily be making the benefits of a mediterranean diet seem much bigger than they really are. Additionally, we need to think about what they were comparing to. PREDIMED was comparing with a low fat diet. The Lyon Diet Heart study was comparing with pretty much just letting people go on doing whatever had caused them to have their heart attacks in the first place. I have no trouble at all believing that a mediterranean diet is healthier than these two alternatives.
The remaining 26 studies were only looking at indirect markers of health, such as cholesterol levels and blood pressure. The mediterranean diet didn’t lower cholesterol levels noticeably, and didn’t lower blood pressure to any meaningful extent either. This suggests that any positive effects on health that accrue from a mediterranean diet are not thanks to any beneficial effect on cholesterol levels or blood pressure.
Which is a little bit funny, when you consider that the father of the cholesterol hypothesis (i.e. the belief that saturated fats in the diet increase cholesterol in the blood, which causes cardiovascular disease) was Ancel Keys, who was also, as already mentioned, the father of the mediterranean diet. Why do I find this funny?
Because the whole basis for both the cholesterol hypothesis and the healthiness of the mediterranean diet was Ancel Keys’ famous Seven Countries study, in which it was found that Italians and Greeks had lower cholesterol levels than Americans, and also lower rates of heart disease. This led Ancel Keys to surmise that a mediterranean diet induced low cholesterol levels, which caused low rates of heart disease.
But thanks to the Cochrane review, we now have clear evidence that the lower cholesterol levels seen in the mediterranean countries in the Seven Countries study cannot possibly have been due to the mediterranean diet. So the whole underlying basis for the idea that the mediterranean diet is healthy is clearly wrong. Oops. The world is a funny old place.
Anyway, what can we conclude from all these studies?
The evidence quality isn’t great, and there is a lot to suggest that the results seen both in the observational studies and the randomized trials are inflated. But overall I think it’s fair to say the evidence suggests that a mediterranean diet is healthier than both a low fat diet and a standard western diet. No big surprise there. I’d have been more surprised if that wasn’t the case. Both the low fat diet and the standard western diet are probably two of the least healthy diets you can possibly choose to eat.
The main problem is, as mentioned, that the mediterranean diet has so many different components that it’s hard to know which parts are responsible for the benefit. And, as mentioned earlier, it’s perfectly possible that some parts of the diet are actually harmful, and that the diet could be made even healthier if some parts of it were removed.
This is actually the basis for the Pioppi diet developed by cardiologist Asseem Malhotra. The Pioppi diet is basically just the mediterranean diet with the high carbohydrate parts removed. Which seems pretty sensible to me. If anyone ever did a study comparing the traditional mediterranean diet vs the Pioppi diet, I’m reasonably certain the Pioppi diet would win. Hopefully that will happen at some point.