I can't view the studies on SciHub (DOI not found) but this is what variable adjustment models are for. It's how we know that smokers aren't just people who don

I can't view the studies on SciHub (DOI not found) but this is what variable adjustment models are for. It's how we know that smokers aren't just people who don't exercise. And even when there is colinearity between variables and you're not sure how to adjust for it, then your confidence levels blow up because not literally everyone who A's also B's.