Effect of longer-term modest salt reduction on blood pressure (2013)

The study author then goes on to say that, since high blood pressure over an entire lifetime (not 4 weeks!) is associated with cardio vascular diseases, less salt should reduce the incidence of cardio vascular diseases. But that conclusion is a logical leap on the part of the authors. The studies in their meta-analysis did not show that reducing salt intake also reduces cardiovascular diseases. That may seem like a reasonable logical conclusion, but there are about a billion reasons why it might be wrong.

To go bigger picture, there are a lot of problems with nutrition science, but these are three that stand out to me (this study falls prey to numbers 1 and 2):

1. There are too many short term studies.

The problem with short-term studies is obvious: most health issues that we believe are influenced by diet are a long term and develop over a person's entire life.

2. Too much measuring measuring of risk factors, not enough measuring of health outcomes.

Blood pressure and lipids are mostly risk factors for health problems (read that wikipedia article again; blood pressure is mostly mentioned as a risk factor and the only mention of it being important for treatment is for people who are considered hypertensive, which is not the same thing as having high blood pressure). Most nutrition science research is measuring how diet affects these risk factors, not how diet affects actual health outcomes (morbidity). The reason is that it's cheaper and easier to measure risk factors than to measure actual outcomes. But the problem is that the risk factors are not what you actually care about.

If you never measure actual outcomes, then all of your dietary recommendations are based on connecting a whole bunch of knowledge about individual parts of the system that is the human body, but not based on knowledge about how the system works when taken as a whole. It's basically the same problem as unit testing vs. integration testing. Unit tests are great. But at the end of the day, you need to know if the whole thing you made works the way it's supposed to. Only integration tests get you that.

You may not think this is such a big issue, but American diet recommendations since the 60s have been based on such unit-test style knowledge and the results have been a staggering disaster. The incidence of all the diet related diseases have gone up by a huge amount.

3. Most nutritional research is epidemiological.

Epidemiology is a useful tool, but it can fundamentally only tell you whether two things are correlated. But correlation, at best, can only give you negative knowledge about causality. That is, if two things are not correlated, then you know that they don't have a causal relationship. But correlation can never give you positive knowledge, because both things can be caused by a third, unmeasured, factor. Hence, the famous adage:

Correlation != Causation

For a great dive into why you should basically not draw any positive conclusions from epidemiology, there's a great article by Gary Taubes:

http://www.nytimes.com/2007/09/16/magazine/16epidemiology-t....

I was tempted to answer that I eat very little cheese and not much bread, but OK, this was about average and I have low blood pressure anyway.

What I found extrange is that most people I know are much more concerned with salt intake than I am, often lecturing me that I should be using less salt, and then I read that I am using half of what's considered "current levels".