Did the non-drinking couples get regressed out of the confounders? I would guess a lot of people stop drinking after a health scare, so you'd have to remove the

Did the non-drinking couples get regressed out of the confounders? I would guess a lot of people stop drinking after a health scare, so you'd have to remove the effect of the underlying medical issue to compare them with the moderately drinking people.