With that kind of methodology, I imagine the "moderate" drinkers are going to contain the group of people who said they (frequently) drink, but not that much. I'd be willing to bet there will be a relationship between reporting that you drink frequently and the negative effects of alcohol that will outweigh the attempt to self report how much you drink (because the latter won't be reported accurately, but people who don't drink will tend to more accurately report and select themselves out into another category. Not only is it easier to self report whether you are a non drinker/drinker than it is to report you are a moderate/heavy drinker, I imagine there's also another confounding effect coming through given that there generally has to be something exceptional about you to be a non-drinker in our societies.
Self-disclosure: practically a non-drinker, nothing ideologically against it. Might have a beer every two months socially with food.
Thoughts? Especially from anyone reading the actual study?