The Mythical Race Between Diet and Exercise

You can’t outrun a bad diet. It’s a clever turn of phrase that resonates. But like many things that resonate about diet, exercise, and obesity, it might be a little too clever. In a very gentle way, David Allison, Dennis Bier, and Julie Locher point this out in a brief commentary appearing this week in the International Journal of Obesity. The race between diet and exercise is more mythical than literal.

A Gentle Caution

Allison, Bier, and Locher raise their note of caution based on an article in Science about the work of Herman Pontzer. While agreeing that Pontzer’s work is “highly important, interesting, and exemplary,” they suggest that Pontzer may go a little too far when he says “exercise generally fails to aid weight loss.”

Their point is important for two reasons. First, it’s worth remembering that carefully controlled studies have shown that exercise indeed can cause weight loss. The effect might not be a potent as many people believe, but it’s not nothing.

Correlation and Causality

The second reason takes us back to a very basic truth. Correlation does not prove causality – nor does the lack of correlation disprove the potential for causality. Allison et al point out that Pontzer makes his comments about exercise and weight loss based on very large, high quality sets of observational data. But they write that we must remember this:

“No amount of measurement quality, new technology, or large sample size can substitute for randomization of experimental units to different levels of the independent variable (i.e., the postulated causal factor) in allowing causal inferences to be made. This has been demonstrated theoretically and illustrated empirically. This is not merely a pedantic point, but an increasing concern in science where trust in ‘communication of science’ is at stake. It is also of concern in some fields of human investigation where observational studies prevail and where causal inferences are often defended on the basis of large sample size.”

In other words, no matter how big and fancy your data and methods are, correlations that arise from them still do not establish a causal effect.

Maybe it’s time to retire the well-worn metaphor of a mythical race between diet and exercise.

Click here for the article from Science and here for the commentary from Allison et al. For more on the complicated relationship between exercise and body weight, click here.

Long Distance Runners, Ancient Greece, image from a Greek amphora / Wikimedia Commons

Subscribe by email to follow the accumulating evidence and observations that shape our view of health, obesity, and policy.

December 3, 2022