A May 2016 paper in in the Journal of the American Medical Association found that the BMI correlated with lowest levels of all-cause mortality is approximately 27, a level traditionally defined as overweight. The simple inference that being overweight protects people from death is a prime example of confounding variables. Just two months later, a much more careful analysis came out in The Lancet. This study found that after controlling for conditions which cause you both to lose weight and to die, the optimal BMI is much lower than 27.
This later study starts with a sample of 10.6 million people from 239 prospective trials. The researchers start by considering every observed person and then sequentially remove people who had known diseases at the beginning of their trials, then people who died within five years of the beginning (which they take as a rough proxy for the presence of unknown diseases at the trial beginning), and finally smokers. When they consider all participants, they reproduce the result from the May 2016 paper that the healthiest BMI is in the range 25-30. When they remove people who had known diseases at the beginning of their trials, they find that nothing changes, contrary to their expectation. When they further exclude people who died in the first five years of their trials and introduce a linear adjustment for smoking status they find that the overweight group is just barely at higher risk of death than the lighter groups. Finally, when they restrict the sample to never-smokers, they find that the overweight group is it at a highly significant 11% increased risk of all-cause mortality. So, their hypotheses about pre-existing disease explaining premature death of lightweight people don’t really pan out, but their hypotheses about smoking do.
Beyond the headline that being overweight is not, in fact, good for you, I was curious to see if the study found a lower limit to healthy weight, which would be relevant to the health benefits of caloric restriction. It did. At the sub-overweight end of the spectrum, they break BMI into intervals of 15-18.5, 18.5-20, 20-22.5 and 22.5-25. The entire range of 20-25 has an indistinguishable risk of all-cause mortality. My main take away, then, is that anywhere in the BMI range 20-25 you’re the perfect weight. For me at 5’8″, that’s 132 – 164 pounds. To make it more precise, they find that the exactly optimal BMI moves up as we age: it’s 22 when we’re 35-49, 23 when we’re 50-69, and 24 when we’re 70-89. I wouldn’t read too much into that last claim though. It strikes me as too fine a point to put on a study based on such simple measurements.
Below a BMI of 20, health deteriorates. People in the BMI range 18.5-20 had a 13% higher risk of death and people in the range 15-18.5 were at 51% higher risk of death, which was actually worse than people in the range 30-35. Of course, they’re not observing much about these people besides their diagnosis with a few diseases, smoking status, and some basic demographics. Maybe you could salvage the benefits of caloric restriction by finding that these very light people are nutritionally deficient or have some other unobserved unhealthy habit.