There have been a few adoption based studies that work to correct for confounders. Nothing conclusive obviously, but it's not like no one has thought of that stuff.