The Japan Collaborative Cohort Study (JACC Study) for Evaluation of Cancer Risk sponsored by the Ministry of Education, Science, Sports and Culture of Japan (Monbusho) is a large prospective cohort study, which was mounted from 1988 to 1990 in 45 areas in Japan. The number of subjects is 110
465 males and 64
327 females) who were 40–79 years of age at the time of the baseline survey. Individual informed consent to participate in the study was obtained in 36 out of 45 areas (Tamakoshi et al, 2005
). The subjects were asked to complete a self-administered questionnaire about past medical history, various life style factors and marital status as baseline information. The detailed design of the JACC study has been described previously (Ohno and Tamakoshi, 2001
; Watanabe et al, 2005
). During the approximately 10 years of follow-up through December 31, 1999, there were 550 deaths from liver cancer that were coded as C22 in the International Classification of Diseases and Related Health Problems, 10th Revision. Those survey participants who underwent health-screening checks sponsored by municipalities were asked to donate blood samples during the same period as the questionnaire survey. Eventually, 39
242 subjects provided blood samples (Tamakoshi et al, 2005
). Baseline serum samples had been collected 120 of the 550 subjects who died of liver cancer. As the control group, sera of 11
543 subjects from the same geographical areas as the 120 deaths also were screened for anti-HCV. Deaths (nine) coded as C22.9 (liver cancer not otherwise specified) were excluded from this analysis, while deaths (five) coded as C22.1 (intrahepatic cholangiocarcinoma and cholangiohepatoma; included among non-HCC deaths), leaving 106 deaths from HCC that were regarded as the end point of this analysis. The total subjects were 11
654 but as the sera of 34 cases and controls could not be screened because of insufficient serum volume, the sera of 11
620 subjects were screened for antibody to HCV (anti-HCV) and then divided into anti-HCV positiveand anti-HCV negative groups. In each group, the cases were deaths due to HCC. The controls (living) were individually matched with cases for age (±3 years), gender and area, the cases or the controls that could not be matched being eliminated from this analysis; in the end, there were 3431 subjects.
We used SAS version 8.2 software (SAS institute, Cary, NC, USA) for the statistical analysis. Baseline information on smoking habits was divided into three groups. Matched multivariate-adjusted odds ratios (OR) and 95% confidence interval (CI) for risk factors for death due to HCC were estimated after adjusting for potential confounding factors (alcohol-drinking habit, past history of liver disease and past history of diabetes mellitus) using a conditional logistic model. With respect to interaction between HCV and smoking habits on HCC risk, we evaluated this by examining whether the odds ratio of death from HCC for each factor differed between anti-HCV positive group and anti-HCV negative group.