top of page
 FOLLOW Nutrition Wonk: 
  • Facebook B&W
  • Twitter B&W
 A Nutrition wonk: 


A person who loves nutrition science as much as they love food.


No tags yet.

The 5 Continent (PURE) Study: A Prospective Cohort Study


A new study came out on August 29th and all of Twitter is abuzz! Why?

Study authors, and indeed the media hype surrounding the study claims that the study UPENDS everything we believe about nutrition and everything we thought was wrong! (How many times can we upend all of nutrition before we're right back at the beginning, I wonder?)

So what is it this time?

Another paper came out from the PURE study, which is an 18-country, epidemiological study examining associations between diet and total mortality, CVD mortality, CVD events, and non-CVD mortality. Overall, the study found that diets very high in carbohydrates (>60%), and therefore low in fat, were associated with a statistically significantly higher risk of DEATH. Yes, really.

Of the study, which included 135,335 individuals:

  • Those who were eating diets highest in carbohydrates (74.4-80.7% of daily calories from carbs) had a hazard ratio 1.28 (1.12-1.46) times greater of dying over the median follow-up period of 7.4 years.

  • If you ate 65.7-69.7% of daily calories from carbs? Hazard ratio of 1.17 (1.03-1.32) compared to the lowest carb intake.

  • Intake of 59.3-62.3% of daily intake (or below) from carbs? The association became nonsignificant.

That was when you compared lowest to highest quartile of total mortality. The significant association also existed when you compared lowest to the 4th and 5th quintiles looking at non-cardiovascular death.

Cardiovascular disease? Myocardial infarction? Stroke? Cardiovascular disease death? Non-significant.

Perhaps unsurprisingly, since there was a significant association of carbohydrate with death, there was also a significantly observed protective effect of total fat, both in total and non-cardiovascular deaths. There is also an interesting protective effect of total and saturated fat on risk of stroke (more on that later).

Ok, so carbs won't give me a heart attack but they will - gulp - kill me? Well, also no. First, correlation does not equal causation, and all that 140 character epidemiology-couching language.

But secondly, we have to consider the population that's being examined and how this study was done. Because despite the headlines, and like our "fake nutrition controversies" article, it turns out the findings of this study are actually... fairly consistent with what we already know.

Then why not report that? I don't know! Maybe because this back-and-forth is the key force on the pushcart that keeps nutrition science in the news. Maybe it's because it's Tuesday. Don't ask me, it's late at night!


How the sausage gets made, or, how the white rice gets polished:

This study is unique and important because it recruited from 18 different countries across the globe, and across diets, and income/poverty spectrum.

Individuals from communities (either very small rural towns, or a series of adjacent postal codes in larger urban areas) were chosen and given a baseline Food Frequency Questionnaire (FFQ) and asked to estimate how often they ate certain foods. (If you've never seen one of these things before, it asks questions like: How often do you drink milk each day? A. Never B. Less than once a day C. Once a day D. Twice a day, up to let's say... H. Seven times per day or more, and you pick one. It repeats this for a LOT of food items.)

Woah, you might be thinking, that sounds really vague! It is and it isn't. In people who eat traditional diets, researchers can narrow down the food items most people are most likely to eat day-to-day. And while people's TOTAL amounts eaten can be way off (most people underestimate how much they eat), the macronutrient ratios can be fairly accurate in validated surveys (PURE researchers did validate the surveys by doing repeated 24-hr diet recalls in 60-250 people in each country and comparing those recalls to the FFQ data). Plus, you have to consider than in >130,000 people, if one guy slightly underestimates his milk intake, someone else will likely overestimate it and the macronutrient percentages will balance out.

Ok, so how was the other data collected? From reading, it seems researchers thoroughly collected health updates, death certificates, and autopsy information on the recruited population over time. Individuals were recruited for the study from the year 2003-2013, so some people were in the study longer than others, but the mean time for follow-up was 7.4 years. Cardiovascular disease and deaths were measured by standardized means (in FACT, the study used the Minnesota Code for categorizing possible and probable heart attacks - developed by Blackburn and Rose and first used for... THE SEVEN COUNTRIES STUDY). Other deaths were likewise collected via thorough follow-up.

Then researchers compared macronutrient intakes with risk of measured outcomes: Death from all causes, Death from CVD, CVD incidence, and Non-CVD death.

That's where you get the results at the top showing that the high carb diets are bad bad bad!


Why do I sense a but coming?

Does the study really show that high carb diets are bad? Well firstly, we have to remember that correlations are (generally) hypothesis-generating and (generally) don't provide strong proof for cause and effect.

But to a certain extent, sure, this study provides a hint that getting 80% of your calories from carbohydrates isn't great. Additionally, the authors note that the vast majority of carbohydrates captured in the study were refined carbohydrates, and we already have pretty strong reasons for limiting those.

Ok, but who is getting 80% of calories from refined carbohydrates?


1. A supermodel in the 80's


2. Someone at risk for malnutrition (not getting enough to eat)

What makes me say that?

We like to think of all "Western Diet" junk food as being chock full of carbs: snack foods, cheese curls, potato chips, ice cream, cakes, etc. And it is! But these foods also have fats as well. Unless you're mainlining Coke and Jolly Ranchers (which someone, somewhere certainly is), the junk food you enjoy is likely NOT 80% carbohydrate. And even if you DO like tons of soda, you'd need to drink a TON to offset the % of calories you're getting from foods you eat daily like meats and oils in processed food.

In fact, people eating 80% carbs are either eating extremely low fat, or quite low protein... and perhaps not eating that much at all.

And this is where the countries included in the study become an important factor.

PURE study is groundbreaking because it focuses on food patterns in NON-Western countries - and honestly - it's high time for that.

And the countries included have a wide range of socioeconomic, food availability, and cultural food intakes that differ widely from those in the US (which was not included country in the study).

The authors included three countries they termed "High Income":

Canada, Sweden, UAE

Eleven "Middle Income"

Argentina, Brazil, Chile, China, Colombia, Iran, Malaysia, Palestine, Poland, South Africa, Turkey

And four "Low Income"

Bangladesh, India, Pakistan, Zimbabwe

Authors also included a supplement that showed the "top five" contributors of each major macronutrient in each country. For some, it shows what you might expect: Carbs are carbs, meats are proteins, meats and butter provide fat, etc.

But in some countries that isn't what you see.

For example, in Bangladesh, white rice is listed as the top contributor of carbohydrates, protein, AND total fat. Not exactly a varied diet.

And you can imagine that people aren't eating mostly white rice totally by choice.

Now there is poverty and hunger in EVERY country, including countries that can have high percent fat intakes (like Greece after WWII), so this shouldn't necessarily mean the findings here can't be extrapolated.

But in these cases, it's difficult to assess whether increased risk of total mortality has anything to do with macronutrient intake, or if it has more to do with differential access to things like adequate total energy intake, medical care, proper screening for chronic disease, or differential propensity for infectious disease.

And that is sometimes why focusing on death from diseases we know are linked to lifestyle (like CVD) can make more sense than focusing on total mortality.

Though they corrected for "education" in their statistical model to attempt to correct for socioeconomic status in the study, researchers acknowledged that "high-carbohydrate and low-fat diets might be a proxy for poverty or access to healthcare" but, they said, "all of our models adjusted for education and study centre (which tracks with country income and urban or rural location) and would be expected to account for differences in socioeconomic factors across intake categories."

This would be true if family socioeconomic status between countries was completely comparable, but it likely isn't. Same with access to preventive or acute medical care. Though certainly the adjustments corrected for some of the problem, it is likely that the higher end of the spectrum re: carb consumption were from the lower income countries and that these countries experience disparities in more areas that just food.

Authors then suggest that people eating >60% of calories from carbohydrates daily should consider upping their fat intake and lowering their carbs. This recommendation feels a little "let them eat cake" in light of the collected data.


Ok but you just said CVD might be important! What about that?

Yes, ok - so if refined carbohydrates (and most of the carbohydrates measured were refined, though researchers did not delineate between refined and non-refined sources) are inherently bad for your heart, we will SEE it because people eating tons of rice not by choice will develop heart disease.

So do they get heart disease?

Err, no.

When authors compared the highest quintile of carb-eaters (74.4-80.7% daily calories from carbs) with the lowest (42.6-49%), there was no significant association. (In fact, HR slightly dropped the higher up you went, but did not get even close to reaching significance, so it's better to just ignore it).

But wait! This study said that carbs are bad and FATS ARE GOOD! What happened with the fats?

Very generally, the fats that most consider "good fats" (aka unsaturated, or monos and PUFAs) there were no significant differences in CVD events, but significantly decreased risks for total death.

And! Tra la! This protective effect was seen among saturated fats as well! VINDICATION!?

PURE study saturated fat, picture of salami

Yes! And no! It's true total mortality and non-CVD mortality had a significant inverse correlation with saturated fat intake. And, very interestingly, saturated fat was inversely correlated with stroke risk!

All that being said, the bad news is this. The absolute highest quintile of saturated fat intake consisted of people eating 13.2% (11.9-15.1) of their calories from saturated fat. The current guidelines typically recommend <10%, so this intake is over those recommendations, but not by much.

(In a 2,000 calorie diet going from 10-15% kcals from saturated fat is equivalent to adding the saturated fat from ~2 deli slices of cheese to your diet. 10-13% is *roughly* 1 oz of cheese difference... Yum cheese!)

Also it's worth noting that almost the exact same reduction in total mortality was seen in people eating 9.5% (8.9-10.2%) of their daily calories from saturated fat as those eating 13.2%. This is within the range of the current dietary guidelines.

And what about the stroke risk? Stroke risk showed a statistically significant reduction when comparing the highest to lowest quintile of saturated fat intake (HR 0.79 (0.64-0.99). This inverse correlation has been seen in at least one other study in Japan where saturated fat intake is also traditionally very low, but seems to be specific for type of stroke - something that was not reported by the authors in this study. Also, this association was just barely significant, but the study is still progressing and hopefully they'll have more years to follow up. Nevertheless, an interesting result to keep an eye on.

So what's all this about overturning dietary advice?

In the study, the authors state that "Global dietary guidelines should be reconsidered in light of these findings."

Should they?

I am not about to get into how dietary guidelines do and don't get made (and do and don't work), but I can see how an argument could be made to establish a maximum suggested percent intake of carbohydrate or a minimum suggested percent intake of fat.

The authors made these observations and suggestions:

"Our findings indicate that limiting total fat consumption is unlikely to improve health in populations, and a total intake of about 35% of energy with concomittent lowering of carbohydrate intake might lower risk of mortality."

My nitpick would be to suggest that most people in this study weren't limiting fat consumption by choice, but this is a fair point and was suggested by their evidence. Further, at least in the USDA diet guidelines the suggested upper limit on total fat no longer exists, so check one off the list for the US!

They also state:

"Importantly, a certain amount of carbohydrate is necessary to meet short-term energy demands during physical activity and so moderate intakes (eg. 50-55% of energy) are likely to be more appropriate than either very high or very low carbohydrate intakes."



The study has only been following people for a mean of 7.4 years, so I will be really excited to see what else they come out with. For other posts discussing aspects of the study, check out: Kevin Klatt's take, Pauli Ohukainen's hot take on twitter, and this response from the Science Media Centre.

Overall, the study suggests that having an extremely limited diet very high in refined carbohydrate may lead to increased risk of death, but it might also be a sign of differential access to healthcare.

Join our mailing list

& Get Weekly Updates


See Nutrition Wonk disclosures

bottom of page