We live in a society totally dependent on science and technology, in which almost nobody knows anything about science and technology ” – Carl Sagan

The science of nutrition is complex. It seems evident that food influences life expectancy, but there is little evidence on the best strategy.

A recent study in the journal The Lancet caused great media impact when finding that low-carbohydrate diets were associated with higher mortality (as well as highs in this macronutrient).

Some headlines about the Lancet study

These headlines generate alarm in a society with little training on science in general and nutrition in particular.

My goal today is to deepen the reality of this specific study, but  also providing a global view on how these investigations are carried out and their important limitations . A better formed society is a less manipulated society.


In computer science, the term ” Garbage In, Garbace out ” is used to highlight the importance of the input data. No matter how good your statistical model is, the results will be wrong if the starting data is bad .

Garbage In, Garbage Out: If you feed any model with bad data, the result will be incorrect, even if the processing is perfect

In the study, for example, this questionnaire was used , where respondents had to remember the average quantities they had consumed from dozens of foods during the last year. There are two important problems with these forms:

People forget . Most do not remember what they ate last Tuesday, and much less during the last year. The great gap between memories and reality is an important limitation of nutritional science ( study , review ).

The people lie . We always try to show our best side to others, as seen in any social network . We publish the good, we omit the bad. And the same thing we do when filling out these forms. People tend to report fewer calories than they actually eat, and not to mention socially questioned foods.

How bad are the data in this case? It’s impossible to know, but a quick look at the daily calorie chart makes us suspect that something is terribly wrong.

Average calories in each quintile of carbohydrate consumption

In the United States, one of the leaders in global obesity, they report an average of between 1,550 and 1,660 calories a day , below the poorest countries in the world ( detail ). Unlikely.

If the input data is not credible, surely neither will the results be

Based on these limitations, many experts recommend not financing studies based on this type of data collection:

A review of the Mayo Clinic Proceedings considers this process ” pseudoscientific and inadmissible in scientific research .”

Something similar published in the Journal of Clinical Epidemiology , dismissing as fictitious the results obtained with these forms, and considering them invalid in studies that intend to establish policies based on scientific evidence.

These problems apply to all studies based on these questionnaires, but in this case there is an important additional limitation. Of the 25 years that the study lasted,  only information was recorded twice : at the beginning and a few years later.

If you are like most, your diet of 15-20 years ago is not a good reflection of what you eat today , but the study assumes that the participants’ diet did not change during all this time. That is, not only the initial data are questionable, but during most of the study not even any information was recorded.


Another important limitation of observational studies is the difficulty in identifying causality . As I explained when talking about the supposed risks of not having breakfast , that two variables move at the same time does not imply that one causes the other.

However, the media prefer simple and sensational ideas, even if they are wrong . That is why they opt for the interpretation that ” low carbohydrate diets raise mortality ” (alternative A), without stopping to think about more probable explanations, but which do not generate such good headlines.

The media prefer simple interpretations (such as A) with respect to less direct but probably more correct alternatives (such as B)

Another possible explanation would be B, where people with overweight or diabetes are more likely to experiment with a low carbohydrate diet, in addition to that being overweight is usually the result of a set of bad habits, which would also explain the higher mortality .

Do we see those bad habits in this study? Clearly. The group with the lowest carbohydrate consumption has the following characteristics :

  • Less physical activity
  • More tobacco
  • More overweight
  • More diabetes

Less consumption of fruits and vegetables .

More men Being a man is not a bad habit as such, but men have a higher mortality risk, probably affecting the final result.

They did not record information on alcohol consumption, but with high probability this group also drank more. As I always say, bad habits often go hand in hand .

Is the difference in mortality then due to reducing carbohydrates or all of these factors? Surely to the second, but to be sure we should evaluate clinical trials, which we will see in point 5.


In the previous breakfast article I also talked about the well-known bias of the healthy user , which turns official recommendations into self-fulfilling prophecies . If, for example, the idea that meat is bad is spread , those who care about their health will reduce their consumption, while the message will be ignored by the most neglected.

When carrying out any subsequent study, a correlation will be observed between higher consumption of meat and worse health, but the direction of causality will be confused again.

To mitigate this bias, we should study populations less influenced by modern dogmas , and luckily we have good examples. A recent study of more than 135,000 people, also published in The Lancet under the name PURE ( Prospective Urban Rural Epidemiology ), included different non-Western rural populations in its analysis.

Although it is also an observational study, with the same limitations as the others , it obtained different conclusions. It showed lower mortality by increasing fat consumption , and both animal protein and fat were associated with a protective effect.

The experts propose different theories to explain this phenomenon, but one of them is that precisely these populations are less influenced by the current recommendations, avoiding the bias of the healthy user ( detail ).

In fact, from the PURE study and four additional ( detail ), the Healthy Diet Score was recently published , which attributed positive or negative values ​​to different foods according to their likely impact on health. A few weeks ago this review was presented at  the annual congress of the European Society of Cardiology ( detail , detail ):

“People who consume a diet with emphasis on fruit, vegetables, nuts, legumes, fish, dairy and meat have the lowest rates of coronary heart disease and mortality. Regarding meat, we find that the unprocessed is associated with benefits. “- Dr Andrew Mente

In summary, the new evidence proposes something very similar to my pyramid of real food .

In fact, a recent study concludes that diets low in carbohydrate and high in fat / protein are not associated with more cardiovascular risk , evaluated with CAC (coronary arterial calcium), better predictor than cholesterol.


We do not eat macronutrients, but foods. We know hunter-gatherer populations with very different intakes of carbohydrate and fat, but none presents obesity nor the typical diseases of modernity.

If humans needed to consume macronutrients in very specific percentages, we would have been extinct for a long time.

It is valid to study the impact of varying the macronutrients, but without losing sight of the food. Any study that targets broccoli in the same category as white bread (both are carbohydrates) is bound to fail. The percentage of macronutrients in a diet says nothing about its quality .

The wars of macronutrients distract us from the real enemy: the food industry . In the United States 60% of food is ultraprocessed ( study ), and we are on the same path in Spain, already exceeding 30% ( study).

If you focus on the right foods, the macronutrients are secondary. A  recent clinical trial concluded that by reducing processed foods people lost weight, regardless of the macronutrients .

Our ancestral food has passed the rigorous quality tests of natural selection , and before the enormous uncertainty of nutritional science, it is the best option to increase longevity.


As we have seen, observational studies have important limitations. They can be used to pose initial hypotheses, but not to establish final recommendations .

Pyramid of evidence quality. The study analyzed is cohort (Cohort Studies), representing a low quality of evidence

Luckily, we have a lot of meta – analysis of clinical trials (experimental studies) demonstrating the effectiveness of diets low in carbohydrate when losing fat and improving metabolic health ( meta , meta , meta – analysis ,  study ), treat diabetes ( review , meta-analysis ) or improve coronary health  ( review , meta-analysis ).

Although these meta-analyzes of clinical trials represent the best quality evidence, they have their own weakness: short duration . While observational studies may extend for several decades, most clinical trials last less than a year.

But if the main indicators of health improve when adopting a diet of this style, it is difficult to believe that the long-term result will be negative.


The swings of the science of nutrition cause great confusion among the population. More and more researchers raise their voices demanding a new direction.

Ironically, in the same issue of The Lancet where the previous study was published, another article was included   that alerted of the poor foundations on which nutrition science is based , summarizing similar errors of the past (like the initial investigations that demonized the fats). About this study in particular warns of the following:

” The results should be interpreted with caution, since group thinking can bias information published from observational studies, using statistical methods to produce results that fit with current beliefs. “

How many great media echoed this article? Exactly none.

John Ioannidis , professor of medicine at Stanford, recently wrote in the prestigious journal JAMA about the need to reform nutrition research :

” Some nutrition scientists and a large part of the public often consider that epidemiological associations represent causal effects that can give indications about health recommendations. However, current epidemiology is difficult to reconcile with good scientific principles. This field needs radical reform . “


The onus probandi  refers in the legal field to the person responsible for providing the burden of proof: the normal is understood as proven, the strange must be proven . And the stranger a proposal is, the stronger its evidence must be.

In relation to our food, the normal thing is that we are well adapted to the usual foods, regardless of their macronutrients . The strange thing would be that these foods produced damage or that we needed very specific percentages of some to keep us healthy.

While you prioritize real food , your health will not suffer due to varying percentages of macronutrients , although we can make the following recommendations:

Cycle calories . Our ancestors alternated periods of abundance with times of scarcity, and we still require a certain balance between anabolic and catabolic processes . One way to achieve this is by including intermittent fasts , or simply alternating cycles of volume and definition .

Cicla macronutrients . Our foods varied according to the season, and with them our macronutrients. We ate more fat and protein in winter and more carbohydrate in summer . This variation also maintains our metabolic flexibility , without becoming too dependent on a single fuel .

Eat enough protein , but no more than the bill.

Consider your genetics and your preferences. Your evolutionary line conditions your individual requirements to a certain extent, and it is important to also consider your natural inclinations. Do you feel better with something more fat? Perfect, do not be afraid of it while it comes from natural foods ( eggs , meats, nuts , whole milk products …). Do you prefer to include more carbohydrate? Go ahead, it will not hurt you either, as long as you prioritize the good ones . At the sports performance level, you can use some periodization strategy .

And finally, remember that food is only part of the equation . To live more and better you should not forget to  do physical activity , respect your circadian rhythms and reconnect with your social group .


Please enter your comment!
Please enter your name here