Characteristics of subjects from cases and controls^a.

aP values were adjusted for the history of smoking, diabetes and hypertension. bAn outlier due to the clerical error was eliminated. cLog-transformation was used. dNonparametric rank test was applied.