The tricky thing with correlation is that behavior can still act as a "middle-man" for genetic correlation. For instance, a genetic susceptibility to addictive

The tricky thing with correlation is that behavior can still act as a "middle-man" for genetic correlation. For instance, a genetic susceptibility to addictive behavior could lead to higher rates of smoking, and therefore lung cancer. (While some of the more obvious factors might be corrected for, there's no way that all of them would be.)