Is 'age at first drink' a useful concept in alcohol research and prevention? We doubt that.