In animal models of colon cancer, n-3 polyunsaturated fatty acids (PUFA) have antineoplastic properties, whereas n-6 PUFAs may promote carcinogenesis. Prior epidemiologic studies have been inconsistent regarding the association of PUFAs and colorectal cancer. We prospectively evaluated the association between PUFA intake and colorectal cancer in a cohort of 73,242 Chinese women who were interviewed in person at the baseline survey for the Shanghai Women's Health Study. Dietary fatty acid consumption was derived using data collected from two food frequency questionnaires administered at baseline and 2 to 3 years later. The dietary total n-6 to n-3 PUFA ratio was strongly associated with colorectal cancer risk. Compared with women in the lowest quintile group, elevated relative risks (RR) were observed for the second [RR, 1.52; 95% confidence intervals (CI), 1.00-2.32], third (RR, 2.20; 95% CI, 1.41-3.45), fourth (RR, 1.65; 95% CI, 0.99-2.75), and fifth (RR, 1.95; 95% CI, 1.07-3.54) quintile groups. Arachidonic acid was associated with colorectal cancer risk with elevated RRs of 1.20(Q2-Q1) (95% CI, 0.87-1.64), 1.44(Q3-Q1) (95% CI, 1.05-1.98), 1.61(Q4-Q1) (95% CI, 1.17-2.23), and 1.39(Q5-Q1) (95% CI, 0.97-1.99; P(trend) = 0.03) with increasing dietary quintile. In a subset of 150 cancer cases and 150 controls, we found a statistically significant trend between an increasing n-6 to n-3 PUFA ratio and increasing production of prostaglandin E(2) (PGE(2)) as measured by urinary PGE(2) metabolites (P = 0.03). These results suggest that dietary PUFA and the ratio of n-6 to n-3 PUFA intake may be positively associated with colorectal cancer risk, and this association may be mediated in part through PGE(2) production.