One week, a headline in the news blares that a particular food is dangerous. The next week, articles about that same food describe it as beneficial. Even more puzzling, the scientists may have used the same dataset to come up with their wildly different results. How can this be?

    Published research about diet and health outcomes usually comes from analysis of large datasets such as the National Health and Nutrition Examination Survey (NHANES) or the UK Biobank. Enrolled participants answer questionnaires and undergo many tests. Subsequently, over the years, information about any medical problems is added. Finally, scientists analyze the dataset, hoping to find meaningful connections that can help determine what tests and lifestyle choices are beneficial.

    Many scientists criticize conclusions about diet, because they question the accuracy of the collected information. It is unrealistic to ask people to track everything they eat. But it is also unrealistic to think that the information people provide about their diets is comprehensive — people forget, especially the unhealthy snacks. Plus, people’s food choices may change over the course of the observation period, which can last for decades.

    However, a recent publication in the Journal of Clinical Epidemiology addresses another problem. Statisticians use different techniques and assumptions to analyze these datasets. For example, if researchers wanted to assess connections between food and heart disease, they would want to control for cigarette smoking. But how to do so? And which variables to adjust for? How to account for missing data? Do these assumptions affect the conclusions drawn?

    To study this, the authors deliberately used many, many different ways of looking at the same data. They used as an example the question of whether eating red meat increases mortality. An overview on Medscape is easier to read than the highly technical original article.

    Depending on how the data was analyzed, the conclusions ranged from red meat being beneficial to red meat being harmful. Averaging them all out, the final verdict was that red meat might be mildly harmful. Given the overall results, I would not take that as proven. In the model Dr. Nicholas Gonzalez and I used, red meat is good for some and not for others.

    What to do with all this? Take the headlines about diet and health with a grain of salt, or better yet, ignore them. They are clickbait. Instead, look at the conclusions Dr. Weston Price reached as he traveled the world decades ago. Humans can thrive on a variety of different diets, as long as they avoid processed foods.