Correlation and causality, are there really any good and bad foods?

A recent article published in Medpage subjects observational studies to harsh criticism.

The author, Larry Husten, laments that some observational studies have so such media exposure that they can be accused of being “poisonous, enthusiastic and superstitious” of the science of nutrition. In particular, he refers to a recent article published in JAMA which estimates the number of deaths from cardiovascular disease, stroke and type 2 diabetes to be the result of 10 different nutritional factors.

Neolife medical management

“Correlation is not causation” is a common argument used by prestigious doctors such as Neil Rouzier who remains critical of certain observational studies.

A recent article published in the medical information portal Medpage (1) has attracted widespread attention due to the manner in which the topic was presented and the approach to the eternal debate on correlation and causality in Science. Every day scientific articles are published on observational studies that correlate different variables, such as taking more or less salt and the risk of becoming ill as a result, but this correlation does not necessarily imply a causal link: such as salt produces disease.

“Correlation is not causation.” This is one of the “tag lines” that has been repeated by one of the world leaders in Bioidentical Hormone Replacement Therapy, Dr. Rouzier.

Longitudinal observational studies obtain their information, their variables (biomarkers, surveys etc.) from databases that have accumulated over time, retrospectively or prospectively. If this information is limited to a certain moment in time, the observational study is referred to as transversal. Both longitudinal and transversal studies involve statistical analysis in order to determine whether there is a correlation between two or more variables whilst eliminating the influence of other variables in the chosen correlation as far as possible. For example, an excess of salt in the diet has been associated with a higher incidence of disease regardless of weight. In order to establish a causal relationship it is necessary for observational studies to establish correlations between the different variables and confirm them through randomised studies, double blind studies, controlled studies and, if possible, cross-over studies, referred to as RCTs (Randomized Controlled Trials).

Correlation and causality, are there really any good and bad foods?

The double-blind, randomised, controlled and cross-over studies are designed prior to implementation. The sample is randomly selected (women, men, a certain age, a certain race etc.) and assigned to two or more participation groups. Each group is then assigned a variable using a controlled manner. For example, we could make two random groups of women (we have already selected the sample), in which one group ingests more salt than the other. None of the women know if they are in the group that is ingesting more or less salt, which is why the study is a blind study. After a certain period of time we then collect the data we want to study, such as the presence or absence of a disease, and these results are analysed by researchers who do not know if the origin of the data has been collected from the group that has ingested more or less salt. Now here the study is a double blind: neither the women nor the researchers who are analysing the data know which group is which. The study is then “opened”, that is to say, the results of one or other of the groups is sorted and analysed to determine whether taking more or less salt has a causal relationship to the appearance of the chosen disease. In certain cases the study conducted is a cross-over which means that during a second phase the study is repeated so that the women in the group that have ingested more salt, will then ingest less sale, and vice versa.

Needless to say, doing RCTs is a very laborious and complicated process. This is principally because it is difficult to obtain large enough samples that are representative, to continue them for a sufficient amount of time and this is notwithstanding how difficult human trials are to complete. That is why many results from observational studies have not been able to confirm the causality of a correlation.

The authors of the article that is the subject of criticism confirmed that 45% of cardiometabolic deaths were related to an excess or deficiency in the intake of one of the 10 foods under analysis.

Returning to the news published in Medpage, the author Larry Husten criticises a recent article published in JAMA (2) in which the number of deaths due to cardiovascular disease, stroke and type 2 diabetes is estimated to be the result of 10 different nutritional factors: the intake of fruit, vegetables, nuts and seeds, whole grains, unprocessed red meats, processed meats (e.g. sausages), sweetened drinks and/or sugary substances, polyunsaturated fats, omega 3 fats from the sea and salt. Using a sophisticated calculation model, the authors of the study affirmed that 45% of deaths from the cardiometabolic causes previously reported are shown to be the result of an excess or deficiency in the intake of any of the 10 previously mentioned foods. In order of their negative impact, the result was as follows:

  1. Excessive salt consumption.
  2. Insufficient consumption of nuts and seeds.
  3. Excessive consumption of processed meats.
  4. Insufficient consumption of omega 3 from sea animals.
  5. Insufficient consumption of vegetables.
  6. Insufficient consumption of fruit.
  7. Excessive consumption of sugary and/or sweetened drinks.

The authors of the study have highlighted that these nutritional circumstances correlate with a higher incidence of cardiometabolic death, but that they can not state that the causal relationship is an absolute one. However, they do believe that these correlations should be taken into account when public health campaigns relating to nutrition are prepared by the state. The impact that this article has had on the general media has been significant, with headlines in the news from NBC, confirming “The 10 foods that most affect your heart health”.

Larry Husten laments that some observational studies such as this have such a widespread impact upon the media and has accused such articles as being “poisonous, enthusiastic and superstitious” of the science of nutrition. To argue his position Husten cites the recent “salt debate”, already referred to in this blog (3), in which a study published the apparent health risks of taking too little salt or the well-known PURE study (The Prospective Urban Rural Epidemiology Study) (4), in which the benefits of a diet rich in vegetables and fruits were questioned.

The opinion of Neolife, as an Antiaging Preventive Medicine clinic, is that moderation and common sense should be applied to the field of nutrition. And of course, the control of biomarkers.

As in many other areas of life, the answer is to apply moderation and common sense. It is sensible to understand and accept the results of the JAMA study, even if it is not an RCT; but equally it is also important to understand that whilst there are not many RCTs, the results from observational studies are valid as a means to alter attitudes to our health. Notwithstanding this it is also common sense not to assume that the results are beyond question and one should always refer to other evidence and not just a single study, whether the study is an observational study or an RCT. On the one hand, neither salt, red meat, nor sugary drinks are quintessentially bad; on the other hand, neither are fruits, vegetables, nor nuts quintessentially good. Properly interpreted science and logic tells us that we should not over indulge the former but that also the latter is not a guarantee to a healthy life. We have always considered that a good or bad food will be demonstrated in the result from the biomarkers measured. For example, if a patient with a supposedly poor nutritional habit presents with a body mass index, a fat percentage, a glycosylated haemoglobin, a vitamin D, an LDL-cholesterol and many other biomarkers that are deemed to be excellent then we can not attribute risks to your health to your poor nutritional habit, but common sense and science requires us to recommend that you modify your habit.