Much public health data comes in the form of “X increases your risk for developing Y”. Sometimes this is colloquially stated as “X causes Y”, but this is less accurate. For example, everyone by now should know that “smoking causes lung cancer” by which we mean that smoking increases the statistical risk of developing lung cancer. Most risk data is correlational, because we cannot ethically expose people to a potential risk to see how much harm it does. So really what we mean is that smoking positively correlates with the risk of developing lung cancer.
Health care providers, regulators, and the public are awash in such correlational data, with recommendations that can often be overwhelming. I previously wrote about a study in which 50 common ingredients were chosen at random out of cookbooks, and then the literature reviewed to find correlations (positive or negative) with cancer. They found that 80% of the ingredients were shown to either correlate with increased or decreased cancer risk. Often the same ingredient will have both health benefits and health risks.
Most of this is likely just statistical noise, too-eagerly reported by the media looking for sensational headlines (these common foods can kill you!!). In fact, in 2015, scientist and journalist John Bohannon deliberated carried out a crap study apparently showing that eating dark chocolate helps people lose weight. The media gullibly reported the results without the slightest investigation into the quality of the study.
There is a risk to reporting background scientific noise as if they are reliable findings that should guide our eating and health decisions. It creates a type of fatigue that can lead people to ignore actually good advice, or people who want to make an effort to be more healthy to waste time and energy on false leads. The system can be improved at every level – with better research design, clearly marking preliminary studies as such, more carefully crafted press releases (or more selective press releases), and better science journalism. But we can also use expert reviews that clear out the noise and focus attention on those risk factor correlations that deserve our attention.
The Institute for Health Metrics and Evaluation (IHME), a global health research organization at the University of Washington School of Medicine, has recently done that with a series of systematic reviews they call “Burden of Proof” reviews. They published their results in Nature Medicine, but also present their data in a useful interactive way on their website. You can, for example, look through the relative risk curves, matching possible risk factors with possible outcomes.
There are three big factors to consider when looking at any individual potential risk factor – what is the strength of the evidence, what is the magnitude of the risk, and what is the magnitude of the behavior necessary to trigger a risk. We can likely comfortably ignore alleged risks when the quality of the evidence or strength of the association is low, the magnitude of the risk is low, or the behavior would have to be extreme in order to trigger a meaningful risk.
Of course, there have been plenty of reviews looking at the evidence for apparent benefits or risks, but this effort conveniently puts a lot of information in one place in a user-friendly format with a more intuitive context.
With regard to the strength of the evidence, they use a five-star system with five stars being solid evidence, going down to strong, moderate, weak, and then possibly no association at one star. Based on this rating system alone, I think we should take four and five star associations seriously, comfortably ignore ones and twos, with three star associations being ambiguous (requires further study).
In their first crop of data they look at four risk factors: smoking, high blood pressure, eating too much red meat, and eating too few vegetables. As you can see, most of the four and five star associations have to do with smoking. One deals with high blood pressure and ischemic heart disease. There is one diet-related three-star rating, for too few vegetable and ischemic stroke, the rest of the diet associations have 1 or 2 stars. This should not be a surprise to regular readers here, or anyone very familiar with the evidence. There is very good evidence that smoking is bad for you, and that chronic high blood pressure is also a serious risk factor for vascular disease.
Diet-related risk factors, however, generally have very poor evidence. I discussed this previously with a 2019 systematic review that essentially concluded the same thing – nutritional studies tend to be of poor overall quality, mainly because they are observational with lots of noise and confounding factors.
In addition to the quality of the evidence being low, you can look at the risk curves here. If you look at smoking and cancer, there is a strong dose-response with relative risk increasing to 3-10 times baseline risk. If you look at eating red meat, the relative risks are in the 1.1-1.3 range. The “dose” also needs to be relatively high.
Looking through this data (and I am looking forward to seeing more data they plan to add) reinforces the bottom-line advice I have given in the past. Don’t go down the rabbit hole of chasing every small reported risk. Most of it is noise. What we actually know can be boiled down to some simple rules: Don’t smoke, don’t drink to excess, get sufficient sleep, exercise regularly, don’t eat too much, and eat your vegetables. When it comes to lifestyle factors, everything else is mostly noise.
