Mendelian randomization studies also support the notion that less alcohol intake is probably better for everyone (from a strictly cardiovascular health standpoint):
Isn't Mendelian Randomization prone to pleiotropy?
I've never understood how people can control for that without a complete understanding of the genome.
Do you know any good overviews of Mendelian randomization?
I'm afraid not, my knowledge is all 2nd hand (wife is scientist) :)
Yes, MR is prone to biased estimates due to pleiotropy. Attempting to unbias the estimators is a relatively hot topic right now.