We analysed the relation between coffee consumption and hepatocellular carcinoma in two case-control studies conducted between 1984 and 1998 in Italy and Greece, including 834 cases and 1912 controls. Compared to non coffee drinkers, the multivariate odds ratio was 0.7 for drinkers of three or more cups per day. The possible relation between coffee drinking and the risk of several cancers, particularly of cancers of the urinary bladder, pancreas and colorectum, has been widely investigated. These have shown a direct relation with bladder, an inverse one with colorectum, but no consistent association with other major sites, including liver cancer (IARC, 1991;Tavani and La Vecchia, 2000).Coffee drinking has been inversely related to the risk of liver cirrhosis in several studies (Klatsky and Armstrong, 1992;Klatsky et al, 1993;Corrao et al, 1994Corrao et al, , 2001Gallus et al, 2002). Although cirrhosis is a major correlate of hepatocellular carcinoma (Adami et al, 1992;La Vecchia et al, 1998, Kuper et al, 2000, the relation between coffee drinking and risk of hepatocellular carcinoma has been examined in only two studies which provided, however, no definite results. An Italian case-control study (La Vecchia et al, 1989), based on 151 cases with hepatocellular carcinoma, reported a multivariate odds ratio (OR) of 0.78 for drinkers of 53 cups of coffee per day, compared to non coffee drinkers. In a Greek casecontrol study (Kuper et al, 2000), based on 333 cases, the age-and sex-adjusted OR was 0.7 for drinkers of 520 cups per week compared to non drinkers.To clarify the role of coffee drinking in hepatocellular cancer, we updated and re-analysed the Italian and Greek studies.
MATERIALS AND METHODS
Selection of cases and controlsThe present data derived from two case-control studies of hepatocellular carcinoma. The first was conducted between 1984 and 1997, in a network of teaching and general hospitals in the Greater Milan area (La Vecchia et al, 1989. Cases were 501 subjects with incident, histologically confirmed hepatocellular carcinoma. Of these, 378 were males and 123 were females; the age range was 20 -75 years, median age 60. Controls were 1552 patients with acute non neoplastic conditions, unrelated to long-term changes in diet or coffee drinking habit (i.e., gastritis or other chronic digestive tract disorders), from the same catchment areas as the cases, and admitted to the same hospitals. Of these, 1141 were males and 411 were females; the age range was 18 -75 years, median age 56.The second study was conducted between 1995 and 1998 in three teaching hospitals from Athens (Kuper et al, 2000). Cases were 333 subjects with incident hepatocellular carcinoma. Of these, 283 were males and 50 were females; the age range was 31 -79 years, median age 65. Controls were 360 patients admitted to the same hospitals as the cases, for injuries, or eye, ear, nose or throat conditions. Of these, 298 were males and 62 were females; the age range was 24 -79 years, median age 65.Thus, a total of 834 cases of hepatocellular...