. "These differences persisted after adjustment for such potentially confounding variables as educational level, alcohol use, and adiposity./.../" . . .