The relationship between cholesterol and 28-year CHD mortality in women was evaluated in the Charleston Heart Study. Linear, quadratic, and cubic models were investigated using Cox proportional hazards regression analysis. In white women, the linear, quadratic, and cubic terms for cholesterol were significant suggesting an asymmetric J-shaped relationship. In black women, only the linear term in all three models was statistically significant suggesting an increasing CHD mortality rate with increasing cholesterol level. The lack of consistency of results by different statistical analyses in black women make conclusions concerning the nature of the relationship between cholesterol and CHD mortality less strong in black women than in white women. Compared with women having a cholesterol value equal to the mean of the group (241 mg/dl), white women having a cholesterol value one standard deviation above the mean (s = 52.5 mg/dl) had a 60% higher CHD mortality rate (hazard ratio = 1.6, 95% CI: 1.2-2.1). In black women, the estimated hazard ratio for a one standard deviation (s = 47.8 mg/dl) increase in cholesterol is 1.4 (95% CI: 1.03-1.8). The results suggest that the relationship of cholesterol to CHD mortality is different in white and black women. The relationship in white women appears to be curvilinear and represented by an asymmetric curve while the relationship in black women is not curvilinear and the overall pattern of association, while possibly linear, is equivocal.