Alternative explanations for normal variability in antithrombin-III levels.