Three case-control studies and one cohort study have suggested that soy products reduce the risk of prostate cancer (1-4), but overall evidence from these epidemiologic studies is still limited. To investigate this issue further, we conducted a cohort study on the association of tofu (soybean curd) intake with prostate cancer among Japanese Americans in Hawaii. The annual mortality rate of prostate cancer among Japanese American men is 11.7 per 100,000, which is lower than the mortality rate of 24.1 per 100,000 among U.S. Whites (5). However, unlike clinical prostate cancer, occult prostate cancer, as determined by step sections of the prostate at autopsy, is as common among the Japanese Americans in Hawaii as among the U.S. Whites (6). This suggests the operation of factors that inhibits the progression of this tumor after induction among Japanese American men.Japanese American men (n = 8,006) were examined on the Hawaiian island of Oahu from 1965 to 1968, as described previously (7), and 6,860 men returned for another examination from 1971 to 1975. Because of delays in designing the dietary questionnaire, 983 (14.3% of 6,860) subjects did not fill it out. The remaining 5,877 participants recorded their weekly frequency of intake of tofu and the average serving size. A small portion was 60 g, medium portion 120 g, and large portion 180 g. Twenty-two men with prevalent prostate cancer were excluded from the study.Incident cases of prostate cancer were identified through surveillance of discharge records of all general hospitals on Oahu. To reduce the number of missing cases, a computer linkage file was established with the Hawaii Tumor Registry, a member of the Surveillance, Epidemiology and End Results Program of the National Cancer Institute. The surveillance should be nearly complete, because only 2.5% of the 6,860 reexamined men could not be located on Oahu during a survey completed in 1993.There were 304 cases of prostate carcinoma diagnosed from 1972 to 1995 and confirmed by tissue examination. Of the 304 cases, 82 had occult tumors, which were found incidental to transurethral resections in no more than three of all totally embedded tissue chips. These small occult tumors are comparable with the occult tumors found when step sectioning the prostate at autopsy. Of the remaining 222 with clinically apparent disease, 61 had metastastic disease to regional nodes or distant sites. An additional 29 cases were diagnosed clinically but were not confirmed histologically and these were excluded from the study.The risk of prostate cancer associated with tofu intake was assessed by the relative risk (RR) and 95% confidence interval (95% CI) estimated by the Cox proportional hazards regression model (8). Tofu intake was categorized into five groups (0, 1-60, 61-120, 121-240, and >240 g/wk). These groups were used to create a set of binary indicator variables with 0 g/wk as the reference group. The indicator variables and other confounding covariates (age, cigarette smoking, alcohol intake, total calories, arm muscle area, and body mass index) were used as explanatory variables in the model for the estimation of RRs. The test for trend was done using the class midpoints of tofu intakes as explanatory variables. All Ps and 95% CIs are based on a two-sided test. Statistical analysis was done with the SAS software.The 5,826 men in this study provided 113,159 person-years of follow-up. Table 1 shows the age-adjusted RRs of prostate cancer according to tofu intake. The RR in the highest group (>240 g/wk) was 0.80 for all prostate cancer cases. When the analysis was limited to the 222 clinical cases, the RRs were similar and the trend test was still not statistically significant (P = 0.37). Additional adjustment for cigarette smoking history, alcohol intake, total calories, arm muscle area, and body mass index did not substantially change the results (RR of 0.87 in the highest group).Next, analysis was further restricted to the 61 cases with metastatic disease and the five categories of tofu intake were collapsed into three groups (0, 1-120, and >120 g/wk). After multiple adjustment for the same six variables, the RRs (95% CIs) were 1.11 (0.59-2.09) for the low consumers and 1.26 (0.67-2.39) for the high consumers of tofu (>120 g/wk).After following 5,855 Japanese American men for over 20 years, we found no discernible association between tofu intake and prostate cancer risk. The highest group of tofu consumers had a RR of 0.80 for prostate cancer compared with the lowest group, but the inverse relation was not statistically significant. When the analysis was limited to the clinical cases, the RR was 0.85 in the highest intake group. Further restriction of the analysis to the metastatic cases did not provide any evidence of a protective effect of tofu against prostate cancer progression.There are several limitations in our prospective study. The tofu data were based on just two questions (i.e., frequency of tofu intake during the past week and the average serving size). The original dietary questionnaire in 1971 was limited to 32 other foods and was not designed to explore isoflavone intake in depth. Thus, the consumption of other sources of soy foods, such as soybean milk, cooked soybeans, fermented soybeans, and fried soybean curd, were not included in the diet questionnaire. This led to an underestimation of soy consumption in our study, although it has been suggested that >80% of soy intake among Japanese Americans has been from tofu (9). In addition, Japanese American men ate less tofu than that of men living in Japan. It is estimated that they consumed an average of 47 g of tofu daily,44H. Shimizu, personal communication. whereas the median intake was 120 g/wk or ∼17 g/d in our study. If the protective effect of tofu intake is more apparent at high levels of consumption, then our data are limited in this regard. However, other studies in populations that did not have a...