Still, while there’s no scientific consensus on what range of normal BMI equates to lowest overall mortality, Stokes hazards a guess for the lower end, perhaps 20 to 22. He also points out that many studies (on rodents and primates, anyway) show calorie restriction can increase longevity. One population he thinks we should study more are those humans who have always had a very low BMI. Ping-ponging your weight around to try to get on the lowest possible end of the healthy threshold is definitely not a good idea, but living your entire life with a BMI on the lower end—maybe even under 20—could decrease risk of diseases like cancer and diabetes. It's an important question, Stokes says, and one that not enough researchers are asking.