Fruit and vegetables, and human cancer

A large body of evidence indicates that high intakes of fruit and vegetables are associated with a reduced risk of cancer at several sites. The association is generally most marked for epithelial cancers, apparently stronger for those of the digestive and respiratory tracts, and somewhat weaker for hormone-related cancers. The relationship between frequency of consumption of vegetables and fruit and cancer risk was analysed using data from a series of case-control studies conducted in northern Italy since 1983. The relative risks (RRs) for most common neoplasms ranged from 0.2 to 0.5 for the highest compared with the lowest tertile of vegetable intake. Protective effects were highest for epithelial neoplasms, but were also observed for hormone-related neoplasms. Fruit was related to reduced RRs for cancers of the oral cavity and pharynx, oesophagus, stomach, larynx, as well as of the urinary tract. There was a specific and consistent pattern of protection by tomatoes, a typical Mediterranean food, with RRs between 0.4 and 0.7, most notably for gastrointestinal neoplasms. No significant association was observed between fruit and vegetable consumption and non-epithelial lymphoid neoplasms. For digestive tract cancer, population attributable risks for low intake of fresh vegetables and fruit ranged from 15 to 40% of all cases in this Mediterranean population. Combined with tobacco and alcohol, the population attributable risks exceeded 85% for men and 55% for women for upper digestive and respiratory tract neoplasms. Thus, from a public health viewpoint, epidemiological evidence indicates that a substantial reduction in epithelial cancer risk can be obtained by increasing fruit and vegetable consumption.

Olive oil, other seasoning fats, and the risk of colorectal carcinoma

BACKGROUND

An association between fats and colorectal carcinoma has been suggested, but the epidemiologic evidence by type of dietary fat is far less clear. Colorectal carcinoma rates have been relatively low in Mediterranean countries compared with most other Western countries, but the components of the Mediterranean diet responsible for this favorable pattern are unclear.

METHODS

The relationship between various added (seasoning) fats and colorectal carcinoma risk was investigated using data from a case-control study conducted between January 1992 and June 1996 in six Italian areas. Cases were 1953 patients with incident, histologically confirmed colorectal carcinoma (1225 of the colon and 728 of the rectum) admitted to the major teaching and general hospitals in the study areas. Controls were 4154 subjects with no history of cancer who were admitted to hospitals in the same catchment areas for acute, nonneoplastic diseases unrelated to the the digestive tract and requiring no long term modifications of diet. Dietary habits were investigated using a validated food frequency questionnaire including 78 items. Lipid intake was estimated by taking into account the content of seasoning lipids in different dishes, the frequency of consumption and portion size of each dish, and individual fat intake patterns.

RESULTS

The odds ratios (OR) for successive tertiles of olive oil intake, compared with the lowest one, were 0.87 (95% confidence interval [CI], 0.75-1.01) and 0.83 (95% CI, 0.70-0.99) (chi2trend = 4.49, P = 0.03) when colorectal carcinoma was analyzed as a whole, 0.82 (95% CI, 0.68-0.98) and 0.81 (95% CI, 0.66-0.99) (chi2trend = 4.05, P = 0.04) for colon carcinoma, and 0.96 (95% CI, 0.77-1.19) and 0.88 (95% CI, 0.66-1.12) for rectal carcinoma. For specific seed oils (including sunflower, maize, peanut, and soya), the OR for colorectal carcinoma was 0.91 in the highest tertile of intake, and the corresponding values were 1.01 for mixed seed oils and 0.93 for butter. None of these estimates was significantly different from the unity. Allowance for vegetable intake attenuated the apparent protection from olive oil consumption (OR, 0.94 for colon and 0.97 for rectum for the highest tertile), which still was apparent in younger subjects (OR, 0.82 for colon and 0.69 for rectum).

CONCLUSIONS

In this study, seasoning fats did not appear to increase the risk of colorectal carcinoma, and there was little evidence for a differential effect by fat type. If such a differential effect exists, it is minor and could favor olive oil.

Trends in cancer mortality in Argentina, 1966-91

Trends in death certification rates for 12 major cancer sites and total cancer mortality in Argentina were analysed for the period 1966-91 on the basis of the World Health Organization database. In the late 1960s, total cancer mortality rates in Argentina (184/100,000 men, 117/100,000 women, world standard) were among the highest in the world. Over the 25-year period considered, however, cancer mortality in Argentina declined by 15% in both sexes, to reach 157/100,000 in men and 99/100,000 in women, for 1990-91. These rates were somewhat lower than those of North America and, particularly for women, relatively low on a worldwide scale. The favourable trends, observed mostly between the 1960s and the 1980s, reflect the steady decline in gastric cancer rates in both sexes, together with some decline in oesophageal, lung and other tobacco-related neoplasms, mostly in men, following some decline in tobacco consumption over the last two decades. The fall in oesophageal cancer may be related to decreased consumption of hot maté, too. Colorectal cancer rates were high in the 1960s, but declined by 17% in men and 35% in women. An approximately 50% decline was observed for skin cancer mortality, which was among the lowest in the world in the early 1990s, and some decline was observed also for leukaemias and uterine cancer, while breast and prostate neoplasms showed a general stability. The two major unfavourable features of cancer mortality in Argentina were the persistently high rates for oesophageal in men, and for uterine cancer mortality in women. These are likely a result of hot maté drinking for oesophageal cancer and inadequate screening for cervical cancer.

Coffee consumption and the risk of breast cancer

On the basis of clinical observations that some women with fibrocystic breast disease experienced resolution of the disease on eliminating methylxanthines from their diet, it has been suggested that coffee intake might be related to breast carcinogenesis. The relationship between coffee (mostly expresso and mocha), decaffeinated coffee and tea intake and breast cancer risk was therefore considered, combining data from two case-control studies, conducted in Italy between 1983 and 1994. Cases were 5,984 women, below age 75, with histologically confirmed breast cancer, and controls were 5,504 women admitted to hospital for a wide spectrum of acute, non-neoplastic, non-hormone-related diseases. The odds ratios (ORs) were estimated from multiple logistic regression equations including terms for study/centre, age, education, body mass index, smoking status, total alcohol intake, age at menarche and menopause, parity and age at first birth, use of oral contraceptives, use of hormone replacement therapy, history of benign breast disease and family history of breast cancer. No relationship was observed between coffee intake and the risk of breast cancer. The multivariate ORs were 1.17 (1.03-1.33), 1.17 (1.04-1.33), 1.21 (1.06-1.37) and 0.96 (0.83-1.11) for women drinking < 2, 2, > 2 to < 4 and > or = 4 cups/day compared to non-drinkers. Decaffeinated coffee was consumed only by 6-7% of cases and controls and the corresponding OR was 0.84 (0.72-0.98). Tea consumption was also low and not associated with the risk of breast cancer (OR 0.94, 95% CI 0.85-1.03). No significant heterogeneity was found for coffee intake across strata of age at diagnosis, education, body mass index, smoking status, total alcohol intake, age at menarche and menopause, parity, age at first birth, ever use of oral contraceptives, hormone replacement therapy, history of benign breast disease and family history of breast cancer. Thus, this study, based on a large data set, allows us to exclude the hypothesis that coffee intake is related to breast cancer risk in this Italian population.

Oral contraceptive use and benign gynecologic conditions. A review

The following review considers the main epidemiologic data on oral contraceptives (OC) and the risk of major benign gynecologic diseases. Earlier studies investigating the relationship between OC use and functional ovarian cysts indicated that the use of high-dose combination OC have a protective effect, whereas low-dose monophasic or multiphasic pills have little or no effect. With regard to seromucinous cysts, the scanty data available are consistent with the notion of some protection of OC in current but not in former users, although possible effects of diagnostic and selection bias should be considered. Published data on the relationship between fibroids and OC appear too scattered to allow a precise quantification of risk. In any case, any relationship is moderate and, hence, of limited clinical and public health relevance. The data on OC and benign gynecologic conditions are reassuring in their absence of any consistent excess risk.

Trends in mortality from major diseases in Europe, 1980-1993

Trends in age-standardized death certification rates from all causes, coronary heart disease (CHD), cerebrovascular diseases, all neoplasms and lung cancer were analysed over the period 1980-1993 in 20 major European countries. There were steady and substantial declines of overall mortality in all western European countries for both sexes, although appreciable geographic differences persisted. These favourable trends reflect a decline in CHD mortality in most western countries, besides a persisting fall in cerebrovascular disease, and a substantial stability (with some decline in a few northern and central European countries) in cancer mortality. In contrast, in eastern European countries appreciable rises were registered in mortality from major causes of death considered for males. For females, only moderate declines were observed in Eastern Europe. In the early 1990s, overall mortality was 30 to 100% higher for males and 20 to 100% higher for females as compared to Western Europe. As indicated by the trends in lung cancer death rates, this reflects a major impact of the tobacco-related disease epidemic in subsequent cohorts, as well as more unfavourable lifestyle factors (i.e. aspects of diet, other environmental factors), and a delayed control of hypertension in Eastern Europe, together with a substantial excess of suicides, (road) accidents, homicides and alcohol-related diseases, and the delayed introduction of rational treatment for some conditions. An indication of reversal of mortality trends was evident in the early 1990s only in Poland. In conclusion, there is ample scope for intervention on avoidable mortality in eastern European countries.

Cancer mortality in Europe: effects of age, cohort of birth and period of death

Death certification data for 19 cancers or groups of cancers, plus total cancer mortality, in 16 major European countries were analysed using a log-linear Poisson model with arbitrary constraints on the parameters to disentangle the effects of age, birth cohort and period of death. Three major patterns emerged including: first, the prominent role of cohort of birth in defining trends in mortality from most cancer sites (except testis or Hodgkin's disease, where newer treatments had a major period of death effect); and second, the major role of lung and other tobacco-related neoplasm epidemics in determining the diverging pattern of cancer mortality, for each sex and in various European countries and geographic areas. In most countries, the peak male cohort values were reached for generations born between 1900 and 1930. This was observed in women only for Denmark and the U.K., i.e. the two countries where lung and other tobacco-related neoplasm epidemics had already reached appreciable levels. This confirms the importance of cigarette smoking in subsequent generations as a major cause of cancer deaths in Europe. Further, there is a persistent rise in several cancer rates, again chiefly on a cohort basis, in Eastern Europe, which calls for urgent intervention to control the cancer burden in these countries.

Family history of cancer and risk of colorectal cancer in Italy

Subjects with a family history of colorectal cancer (CRC) are at increased risk of CRC, but quantification of the risk in different populations, the possible differences in risk according to localization of the cancer and the association of family history of other cancers with CRC risk are still open issues. We have therefore analysed data from a multicentric case-control study conducted in six Italian areas between 1992 and 1996 of 1225 incident cases of colon cancer, 728 cases of rectal cancer and 4154 controls admitted for acute conditions to the same network of hospitals as the cases. Unconditional logistic regression models including terms for gender, age, study centre, years of education and number of siblings were used to estimate the odds ratios (ORs) of CRC according to various aspects of history of CRC and other cancers in first-degree relatives. The OR for family history of CRC was 3.2 (95% confidence interval, CI, 2.5-4.1) for colon cancer and 2.2 (95% CI 1.6-3.1) for rectal cancer. Colon cancer was significantly associated with a family history of stomach (OR 1.4), bone (OR 2.1) and kidney (OR 2.3) cancers, while rectal cancer was significantly associated with a family history of lymphomas (OR 2.8). There was a 30% higher risk of colon and rectal cancer in subjects with a family history of any cancer, excluding intestine. The ORs for family history of CRC were 5.2 for colon and 6.3 for rectum when the proband's age was below 45 years. The ORs were similar when the affected relative was a parent or a sibling and in different strata of age of relative(s). For subjects with two or more first-degree relatives with CRC, the risk was 6.9 for the right colon, 5.8 for the transverse and descending colon, 3.8 for the sigma, 3.2 for the rectosigmoid junction and 1.9 for the rectum. This study confirms that a family history of CRC in first-degree relatives increases the risk of both colon and rectal cancer, the association being stronger at younger ages and for right colon.

Incidence, mortality and survival from prostate cancer in Vaud and Neuchâtel, Switzerland, 1974-1994

BACKGROUND

Prostate cancer incidence has been increasing in most developed countries in the absence of similar trends in mortality, and with variable patterns in different areas of the world.

MATERIALS AND METHODS

Trends in incidence and mortality from prostate cancer for the period 1974-1994 were analysed using data from the Cancer Registries of the Swiss Cantons of Vaud and Neuchâtel. Of 5,010 cases registered, 80% were histologically or cytologically confirmed.

RESULTS

Age-standardized incidence rates increased from 33.1 to 48.6 per 100,000 (+47%). The upward trends were greater in the most recent calendar periods, and in the younger age groups (+77% at age 45 to 54; +57% at age 55 to 64). In contrast, mortality was stable, with an overall increase of only 3% in age-standardized rates (from 20.4 to 21.0 per 100,000), due to some increase in men aged 65 or above. Consequently, the incidence/mortality rate ratios increased from 1.6 in 1974-1979 to 2.3 in 1990-1994. Five-year observed and relative survivals increased from 26% to 41% and from 46% to 58%, respectively. Ten-year observed and relative survival for cases diagnosed in 1985-1989 were 19% and 42%, respectively. Survival improvements were greater below age 75.

CONCLUSION

The pattern of trends in incidence, mortality and survival confirms the influence of improved diagnosis of prostate cancer over the last few years in this European population. Still, while Swiss prostatic cancer mortality rates are the highest in the world (20.3 per 100,000, world standard), i.e., about 30% higher than in the United States, all races combined, incidence rates are still half as much. On account of the steady increase of prostate-specific antigen testing in Switzerland, further incidence increases are likely.

Diabetes mellitus and colorectal cancer risk

The relationship between diabetes mellitus and the risk of colorectal cancer was investigated in a multicenter case-control study, conducted in Italy between 1992 and 1996 on 1225 cases of incident, histologically confirmed colon cancer, 728 cases of rectal cancer, and 4154 controls, who were in the hospital for acute, nonneoplastic diseases. Overall, 66 (5.4%) cases of colon cancer, 50 (6.9%) cases of rectal cancer, and 185 (4.4%) controls reported a history of diabetes. The corresponding multivariate odds ratios (ORs) were 1.2 [95% confidence interval (CI), 0.8-1.6] for colon, 1.5 (95% CI, 1.1-2.2) for rectal, and 1.3 (95% CI, 1.0-1.6) for all colorectal cancers. No association was observed for subjects who were diagnosed with diabetes at ages of < 40 years (7 cases and 27 controls, OR = 0.9). The OR was 1.4 (95% CI, 1.1-1.7) for subjects who were diagnosed with diabetes at ages of > or = 40 years and were likely to have non-insulin-dependent diabetes. The association was also stronger (OR = 1.6; 95% CI, 1.1-2.3) among subjects whose diabetes was diagnosed 10 or more years in advance and who were > or = 60 years old at the time of colorectal cancer diagnosis. None of the other covariates, including sex, education, body mass index, physical activity, energy intake, alcohol drinking, and fiber intake, showed any appreciable modifying effect. Thus, this uniquely large case-control study of colorectal cancer confirms that subjects with non-insulin-dependent diabetes mellitus have a slightly increased risk of colorectal cancer. More importantly, allowance for a large number of identified potential confounding factors, including body mass index, diet, and physical activity, could not explain the excess colorectal cancer risk among subjects with diabetes mellitus. These findings have plausible biological correlations because insulin-like-growth factor-I is a promoter of colon tumor cell growth in vitro.

Intake of selected micronutrients and risk of colorectal cancer

The relationship between various micronutrients and colorectal cancer risk was investigated using data from a case-control study conducted between January 1992 and June 1996 in Italy. Cases were 1,953 incident, histologically confirmed colorectal cancers (1,225 of the colon and 728 of the rectum), admitted to the major teaching and general hospitals in the study areas, and 4,154 controls with no history of cancer, admitted to hospitals in the same catchment areas for acute, non-neoplastic diseases unrelated to the digestive tract and requiring no long-term modifications of the diet. Dietary habits were investigated using a validated food-frequency questionnaire. Odds ratio (ORs) were computed after allowance for age, sex and other potential confounding factors, including physical activity, total energy and fibre intake. For most micronutrients, ORs were below unity with increasing quintile of intake. The most consistent protective effects were for carotene, riboflavin and vitamin C (Multivariate ORs from the continuous model, with unit set as the difference between the upper cut-point of the 4th quintile and that of the 1st one, were 0.65, 0.73 and 0.80, respectively). Inverse relationships were observed also for calcium and vitamin D (ORs of 0.85 and 0.93, respectively). When the combined effect of calcium and vitamin D and selected anti-oxidants was considered, the OR reached 0.46 in subjects reporting high calcium/vitamin D and high anti-oxidant intake compared to those reporting low intake of both groups of micronutrients. Most results were apparently stronger for colon cancer and among females. Our results provide further support for a protective effect of several micronutrients on colorectal cancer risk and some indications for a specific and stronger effect of selected anti-oxidants.

Incidence of invasive cancers following squamous cell skin cancer

The authors describe the incidence of new primary cancers among 4,639 cases of squamous cell skin cancer (SCC) diagnosed between 1974 and 1994 in the cancer registries of the Swiss cantons of Vaud and Neuchâtel (total person-years at risk = 23,152). Overall, 729 metachronous cancers were observed versus 527.6 expected, corresponding to a standardized incidence ratio (SIR) of 1.4 (95% confidence interval (CI) 1.3-1.5). After exclusion of skin cancers, however, 384 second primary neoplasms were observed versus 397.2 expected (SIR = 1.0). Excesses were observed for cancers of the lip (SIR = 3.1) and lung (SIR = 1.3), for basal cell (SIR = 4.3) and melanomatous skin cancers (SIR = 3.3), and non-Hodgkin's lymphomas (SIR = 1.7). Rates were elevated for cancers of the salivary glands (SIR = 4.3) and for Hodgkin's disease (SIR = 2.7), and, below age 65 years, for cancers of the lung (SIR = 1.6), breast (SIR = 1.5), and prostate (SIR = 1.8), for Hodgkin's disease (SIR = 15.8), as well as for all neoplasms except skin (SIR = 1.2; 95% CI 1.0-1.5). The cumulative risk of basal cell skin cancer reached 17% after 15 years. The authors believe that the excesses for basal cell carcinomas and melanomas of the skin following SCC, and possibly of lymphomas, were likely attributable to common phenotypic characteristics and exposure to UV radiation. The elevated rates of lung cancer are suggestive for a role of tobacco as a cause of squamous cell skin cancer.

Low-risk diet for breast cancer in Italy

To define a low-risk diet for breast cancer in Italy, a multicentric case-control study of 2569 incident cases of breast cancer and 2588 controls from Italy was analyzed. A logistic regression model was applied to the estimated intake of five macronutrients and used to compute a diet-related risk score (RS). The pattern of macronutrient and food group intake across RS deciles was defined. The mean of diet-related RSs across subsequent risk deciles ranged from 0.83 to 1.44. Total energy intake first decreased slightly, from the first to the second decile, and then increased, mostly in the last three risk deciles. Intake of starch increased in absolute and relative terms, whereas saturated fat intake rose in absolute terms but remained stable as a proportion. A relative decline was observed for unsaturated fat and sugars, with a hint, however, of U-shape effect. From a food group viewpoint, there was a marked increase in the intake of bread and cereal dishes, cakes and desserts, and refined sugar across subsequent deciles, whereas the consumption of vegetables, olive and seed oils, and fruit decreased.

Intake of selected foods and nutrients and breast cancer risk: an age- and menopause-specific analysis

The relationship between selected foods and nutrients and breast cancer risk was investigated in strata of age and menopausal status using data from a case-control study on breast cancer conducted between June 1991 and April 1994 in six Italian areas. Cases were 2,569 women with histologically confirmed incident breast cancer admitted to the major teaching and general hospitals of the study areas; controls were 2,588 women with no history of cancer admitted to hospitals in the same catchment area as cases for acute, nonneoplastic, nongynecological conditions unrelated to hormonal or digestive tract diseases or to long-term modifications of diet. Dietary habits were investigated using a validated food frequency questionnaire, including 78 foods or food groups. Among food groups, bread was directly and significantly related to breast cancer risk in older women and, consequently, in postmenopause, whereas the protection conferred by fish consumption was stronger in postmenopause and that exerted by raw vegetables was stronger in premenopause. Among nutrients, unsaturated fatty acids were inversely related to breast cancer risk, the association being stronger in postmenopausal and elderly women. The pattern was similar for total fats. For starch, available carbohydrates, and total proteins, no heterogeneity emerged across strata of age and menopausal status. Among micronutrients, protection diminished with increasing age for beta-carotene and calcium, whereas no heterogeneity emerged for vitamin E. Thus this age-specific analysis of the largest investigation to date on diet and breast cancer did not show any consistent pattern of breast cancer risk in relation to selected dietary factors across strata of age and menopausal status.

Fibers and breast cancer risk

Data from a multicenter case-control study on breast cancer conducted in Italy were used to analyze the relationship between various types of fibers and breast cancer risk. Cases were 2,569 women with histologically confirmed, incident breast cancer; controls were 2,588 women admitted to the same network of hospitals for acute, nonneoplastic, non-hormone-related diseases. Cases and controls were interviewed between 1991 and 1994 using a validated food frequency questionnaire. The data were modeled through multiple logistic regression, controlling for demographic and reproductive breast cancer risk factors. The continuous odds ratios for the difference between the upper cut point of the fourth and the first quintile of intake were 0.90 [95% confidence interval = 0.82-0.98, p (for trend) < 0.05] for cellulose and 0.94 (95% confidence interval = 0.86-1.02) for soluble fibers. The protection tended to be stronger in premenopausal women. No material association was found for noncellulose polysaccharides and lignin. This study, based on a large data set from various Italian regions, suggests that fiber intake may confer some protection against breast cancer, particularly for cellulose and also for soluble fibers, i.e., those of vegetable origin. This possible protection has been related to an influence of fibers on levels and availability of estrogens and other steroid hormones in breast carcinogenesis.

Diabetes mellitus and the risk of primary liver cancer

The relationship between diabetes mellitus and primary liver cancer was investigated in a case-control study conducted in Italy between 1984 and 1996 on 428 cases with incident, histologically confirmed hepatocellular carcinoma, 59 with gallbladder and bile duct cancer, and 1,502 control subjects in the hospital for acute non-neoplastic diseases. Sixty-four cases of hepatocellular carcinoma vs. 87 controls reported a history of diabetes, corresponding to an odds ratio (OR) of 2.3 after allowance for age, sex and area of residence, and of 2.1 [95% confidence interval (CI) = 1.4-3.2] after further allowance for alcohol and tobacco consumption, history of hepatitis and liver cirrhosis, body mass index and history of liver cancer in first-degree relatives. The ORs were similar both for subjects diagnosed with diabetes below age 45, who most likely had insulin-dependent diabetes, and for those diagnosed later, who were likelier to have non-insulin-dependent diabetes. The OR was 2.3 for subjects whose diabetes was diagnosed <5 years before diagnosis of liver cancer, 1.9 for those diagnosed 5-9 years in advance and 2.2 for those diagnosed since 10 years or more. Five cases of gallbladder and bile duct cancer reported a history of diabetes: the corresponding OR was 1.2 (95% CI 0.5-2.9). The OR of hepatocellular carcinoma was 2.4 for males and 2.0 for females, 3.0 for subjects diagnosed with liver cancer under age 60 and 1.8 for those diagnosed at age 60 or over. None of the other covariates considered, including education, history of hepatitis, liver cirrhosis and alcohol drinking showed any meaningful modifying effect or interaction. The potential pathogenic mechanisms include liver alteration-and consequent cell proliferation-in subjects with diabetes. Thus a history of diabetes mellitus could explain about 8% (95% CI 5-11) of cases of liver cancer in this population.

Diabetes mellitus and the risk of primary liver cancer

The relationship between diabetes mellitus and primary liver cancer was investigated in a case-control study conducted in Italy between 1984 and 1996 on 428 cases with incident, histologically confirmed hepatocellular carcinoma, 59 with gallbladder and bile duct cancer, and 1,502 control subjects in the hospital for acute non-neoplastic diseases. Sixty-four cases of hepatocellular carcinoma vs. 87 controls reported a history of diabetes, corresponding to an odds ratio (OR) of 2.3 after allowance for age, sex and area of residence, and of 2.1 [95% confidence interval (CI) = 1.4-3.2] after further allowance for alcohol and tobacco consumption, history of hepatitis and liver cirrhosis, body mass index and history of liver cancer in first-degree relatives. The ORs were similar both for subjects diagnosed with diabetes below age 45, who most likely had insulin-dependent diabetes, and for those diagnosed later, who were likelier to have non-insulin-dependent diabetes. The OR was 2.3 for subjects whose diabetes was diagnosed <5 years before diagnosis of liver cancer, 1.9 for those diagnosed 5-9 years in advance and 2.2 for those diagnosed since 10 years or more. Five cases of gallbladder and bile duct cancer reported a history of diabetes: the corresponding OR was 1.2 (95% CI 0.5-2.9). The OR of hepatocellular carcinoma was 2.4 for males and 2.0 for females, 3.0 for subjects diagnosed with liver cancer under age 60 and 1.8 for those diagnosed at age 60 or over. None of the other covariates considered, including education, history of hepatitis, liver cirrhosis and alcohol drinking showed any meaningful modifying effect or interaction. The potential pathogenic mechanisms include liver alteration-and consequent cell proliferation-in subjects with diabetes. Thus a history of diabetes mellitus could explain about 8% (95% CI 5-11) of cases of liver cancer in this population.