This study indicates that cigarette smoking is an indicator of elevated risk for FNH of the liver, whereas whole grain and, possibly, vegetable intake seems to be a favorable indicator.

Ovarian cancer is among the five leading sites for cancer incidence and mortality in women from developed countries. Its incidence and mortality rates have, however, been declining over the last few decades following the introduction of oral contraceptives, which - together with parity - are the best recognized protective factor for the disease. Late menopause and irregular menstrual cycles may also reduce the risk, while the role of hormone replacement therapy in menopause and fertility treatments is still unclear. Cosmetic talc use and some aspect of diet (i.e. saturated fats, refined carbohydrates) have been associated with increased risk, in some--though not all--studies), while vegetable consumption appears to be inversely related to risk. These issues remain open to debate. Women with a family history of ovarian and breast cancer in first-degree relatives are also at increased risk, but family history accounts for only 4-5% of cases. Most ovarian cancers are therefore environmental in origin and consequently, at least in principle, avoidable.

Between 1992 and 1997 we conducted a case-control study of oesophageal cancer in 3 areas of northern Italy. Cases were 304 patients (29 women), ages 39-77 years (median age 60 years), with a first incident squamous-cell carcinoma (SCC) of the oesophagus. Controls were 743 patients (150 women), ages 35-77 years (median age 60 years), admitted for acute illnesses, unrelated to tobacco and alcohol, to major hospitals of the areas under surveillance. We derived estimates of daily dietary intake of 6 macronutrients, cholesterol, and 20 micronutrients or minerals from a validated food-frequency questionnaire, including 78 food groups and recipes and 15 questions on individual eating patterns. After allowance for age, gender, area of residence, education, body mass index, physical activity, smoking habit, alcohol consumption and energy intake, most micronutrients were inversely associated with oesophageal SCC risk. Highly significant associations emerged for monounsaturated fatty acids [odds ratio (OR) in highest vs. lowest intake quintile = 0.5]; carotene (OR = 0.3); lutein + zeaxanthin (OR = 0.4); vitamin C (OR = 0.4); and niacin (OR = 0.5). Only retinol appeared to be positively related to risk (OR = 1.9). The effect of the above nutrients, expressed as ORs, appeared to be similar in non-smokers and smokers, and non/light drinkers and heavy drinkers.

Epidemiological studies on risk factors for colorectal cancer have focused mainly on diet. Weight and height have also been studied, partly because they reflect the balance between energy intake and expenditure in different age periods. Energy intake, body size, physical activity and colorectal cancer risk will be reviewed in this paper focusing mostly on recent data coming from Italian, English and Scandinavian studies. Overweight has long been recognised as a risk factor for hormone related and other cancers, including colorectal cancer. In addition, the epidemiological evidence consistently shows that physical activity reduces the risk of colon cancer. On the contrary, evidence on rectal cancer is less impressive. In conclusion, body size control along all life and physical activity represent important factors to prevent colon cancer and a wide range of chronic conditions. Therefore, strategies to favour these goals through counselling from health-care providers, regulatory changes, and programs aimed at individuals and communities should be implemented.

1Cancer Epidemiology Unit and Cancer Registries of Vaud and Neuchâtel Institut Universitaire e Médecine Sociale et Preventive, CHUV-Falaises 1, 1011 Lausanne, Switzerland 2Laboratory of Epidemiology, Istituto di Ricerche Farmacologiche ‘Mario Negri’, Via Eritrea 62, 20157 Milano, Italy 3Istituto di Statistica Medica e Biometria Università degli Studi di Milano, Via Venezian 1, 20133 Milano, Italy Correspondence to: F Levi. Fax: (+41) 21 323 0303 E-mail: [email protected]

Introduction The incidence of oesophageal adenocarcinoma has been reported to be rising in the United States (Blot et al., 1991, 1993Zheng et al., 1993;Devesa et al., 1998) and in several areas of Europe, including Norway (Hansen et al., 1997), Denmark (Moller, 1992), Sweden (Hansson et al., 1993), the West Midlands and the Oxford area of England (Harrison et al., 1992;Powell and McConkey, 1992), the Swiss Canton of Vaud (Levi et al., 1990a, 1998bLevi and La Vecchia, 1991), as well as in New Zealand (Armstrong and Borman, 1996) and Australia (Thomas et al., 1996;Lord et al., 1998). The upward trends were generally greater in males. Among White males in the US, the incidence of adenocarcinoma of the oesophagus rose by over fourfold between the mid-1970s and the mid-1990s (3.2/100 000 males), surpassing squamous cell cancers in around 1990. The upward trend was greater among older males (Devesa et al., 1998). A similar rise in incidence was observed for adenocarcinoma of the gastric cardia, too, and there are major difficulties and uncertainties in the classification of cancers arising at the gastro-oesophageal junction (Devesa et al., 1998;Jankowski et al., 2000). However, no clear increase in the incidence of oesophageal adenocarcinoma was observed during the 1980s in males from the three French cancer registries, and some decline was observed for squamous cell cancer in Calvados (Launoy et al., 1994;Desoubeaux et al., 1999), one of the areas with the highest oesophageal rates in France. Other areas showing no clear rise of adenocarcinomas of the oesophagus and gastric cardia in Europe were Basel, Switzerland, Iceland, Bas Rhin, France, southern Ireland and Eindhoven, the Netherlands (Botterweck et al., 2000). The incidence of squamous cell cancer declined in US White males after the mid-1970s, and in Black males after the mid-1980s (Devesa et al., 1998). The widespread unfavourable trends for oesophageal adenocarcinoma have been related to a number of risk factors which are, at least in part, different from those of squamous cell carcinoma of the oesophagus. Alcohol drinking and tobacco smoking, in fact, account for over 80% of squamous cell oesophageal cancers in developed countries (Negri et al., 1992). While tobacco smoking has been related to the risk of adenocarcinoma of the oesophagus and gastric cancers, too, the association is less strong than for squamous cell carcinomas. Alcohol drinking is not consistently related to the risk of oesophageal adenocarcinoma (Gammon et al., 1997;Zhang et al., 1997). A frequent consumption of vegetables and fruit appears to be related to both squamous cell and adenocarcinoma of the oesophagus (Morris Brown et al., 1995;Levi et al., 2000), whereas overweight and obesity have been consistently related to adeno- but not squamous cell carcinoma of the oesophagus (Morris Brown et al., 1995;Chow et al., 1998). Indeed, measures of body mass index seem to be inversely related to the risk of squamous cell oesophageal cancer (D’Avanzo et al., 1996;Chow et al., 1998). The influence of obesity on adenocarcinoma of the oesophagus and gastric cardia may be related to increased gastro-oesophageal reflux, since the risk of the disease is strongly related to Barrett's oesophagus (Levi et al., 1990b;Gammon et al., 1997). Social class indicators tend to be inversely related to risk of squamous cell and adenocarcinoma of the oesophagus (Gammon et al., 1997;Levi et al., 2000). In the Swiss Canton of Vaud, we reported a rise in the incidence of oesophageal adenocarcinomas, but not squamous cell cancers, between 1976 and 1994, in the absence of material change for gastric cardia cancers (Levi et al., 1998b). To further monitor these trends, we updated the analysis of the Vaud Cancer Registry dataset, to include cancers registered up to 1998. Materials and methods The Vaud Cancer Registry dataset includes data concerning incident cases of malignant neoplasms in the Canton, whose population, according to the 1990 census, was around 602 000 inhabitants (Levi et al., 1992). Information collected by the Registry for oesophageal cancer includes specification of the subsite of origin (ICD-O:T) and morphological type (ICD-O:M;World Health Organization, 1976). The following morphological categories were considered: squamous cell carcinoma (ICD-O:M 8050–8082); adenocarcinoma (8140–8573); and other or unspecified cancers. In 1976–1979, 87% of oesophageal cancers were histologically confirmed; corresponding values for 1994–1998 were 93.3%. The following ICD-O:T codes were considered and grouped into separate anatomical sites: (1) oesophagus: upper and middle third (ICD-O:T: 150.0,1.,3.,4); (2) lower third (150.2,.5); and (3) other or unspecified subsites (150.8,.9). Age-standardized incidence rates per 100 000 population (world standard) were calculated. Average annual percentage changes in incidence rates were estimated by fitting a log-linear regression model. Results and comments Table 1Table 1: Trends in age-adjusted incidence rates of oesophageal cancers according to sex, morphology and subsite. Vaud, Switzerland, 1976–1998gives the distribution of 1123 cases of oesophageal cancer registered in Vaud between 1976 and 1998 according to calendar period, sex, histological type and site of origin. For squamous cell cancer, the trends were inconsistent in both sexes, with no material or significant linear trend in annual change, and some decline in incidence was observed over the last few years. In contrast, the incidence of oesophageal adenocarcinoma rose steadily in males, from 0.5/100 000 at all ages in 1976–1979 to 1.93 in 1995–1998, with an average percentage annual rise of 5.1%. Incidence of oesophageal adenocarcinoma appeared to rise in women, too, although rates remained low (0.12/100 000) (Fig. 1Fig. 1: Trends in age-adjusted (world population) incidence rates of oesophageal squamous cell carcinomas and adenocarcinomas in Vaud, Switzerland, 1976–1998.). Thus, the male/female sex ratio in the late 1990s was almost 5 for squamous cell, but over 10 for adenocarcinoma. For males, there was a decline between 1976 and 1990 of other and unspecified oesophageal cancers. With reference to site of origin, no clear pattern of trend was observed over time, and the sex ratio remained around 5 for both upper-middle and lower third of the oesophagus. Table 2Table 2: Trends in age-adjusted incidence rates of oesophageal cancers according to morphology and subsite, per 100 000 males aged <65 and ≥65, in Vaud, Switzerland, 1976–1998gives comparable figures for two separate age groups for males. The rise in adenocarcinoma – as well as the recent decline in squamous cell cancer – was larger or restricted to men aged ≥65 years. With reference to main recognized risk factors, Table 3Table 3: Multivariate a odds ratios (ORs) and corresponding 95% confidence intervals (CIs) for oesophageal adenocarcinoma and squamous cell carcinoma, in relation to cigarette smoking, alcohol consumption and body mass index (BMI) in the US bgives – as an example and for comparative purposes – the multivariate relative risk (RR) of oesophageal adenocarcinoma and of squamous cell cancer from a large multicentre US study (Gammon et al., 1997;Chow et al., 1998) in relation to tobacco, alcohol and body mass index (BMI). The RRs of squamous cell cancer were 7.4 for the highest level of alcohol, 3.9 for tobacco, and 0.6 for BMI. Corresponding values for adenocarcinoma were 0.9, 2.1 and 2.9, respectively (Table 3). In a case–control study conducted in Vaud, Switzerland on 92 cases and 327 controls (Levi et al., 2000), the RRs of oesophageal squamous cell cancer were 96.4 for the highest level of alcohol, 13.0 for tobacco, and 0.15 for BMI, and all the trends in risk were significant (Table 4Table 4: Multivariate a odds ratios (ORs) and corresponding 95% confidence intervals (CIs) for oesophageal squamous cell carcinomas in relation to cigarette smoking, alcohol consumption and body mass index (BMI) in Vaud, Switzerland, 1992–1999). This confirms that alcohol and tobacco are consistently more strongly related to squamous cell than to adenocarcinoma, while BMI is directly related to adenocarcinoma, but inversely to squamous cell cancer. This pattern of risk factors is broadly consistent with the descriptive epidemiology of various histological types of oesophageal cancer. The falls in squamous cell cancer, in fact, should be attributed to the recent declines in alcohol and tobacco consumption in males, while the upward trends of adenocarcinomas may reflect the rising prevalence of obesity (Wietlisbach et al., 1997). A role of newer anti-ulcer drugs, including histamin-2 receptor antagonists and proton pump inhibitors on the risk of oesophageal adenocarcinoma has been suggested on the basis of animal data, but epidemiological evidence in humans is at present reassuring (Fioretti et al., 1997;La Vecchia and Tavani, 2000). The overall age-adjusted incidence of oesophageal adenocarcinoma in males in Vaud (1.8/100 000 in 1995–1998) is higher than reported in northern Europe, but lower than in the US (3.2/100 000 US Whites, US standard;Devesa et al., 1998). The incidence of oesophageal squamous cell cancer, though declining, remains comparatively high, thus contributing to an overall high incidence of oesophageal cancer in Vaud as well as in the French-speaking areas of Switzerland, which is probably related to the persistently high – in comparative terms – alcohol and tobacco consumption in these populations (Levi et al., 1998a). Acknowledgements— The contributions of Mrs C. Pasche, of the Vaud Cancer Registry's staff, and of the Swiss and Vaud Leagues against cancer are gratefully acknowledged.

Istituto di Ricerche Farmacologiche 'Mario Negri', Via Eritrea 62, 20157 Milano, and Istituto di Statistica Medica e Biometria, Università di Milano, Via Venezian 1, 20133 Milano, Italy Fax: (+39) 02 3546 277. E-mail: [email protected] *This paper was an invited presentation of the 18th Annual Symposium of European Cancer Prevention Organisation (ECP): Precancerous lesions of the digestive tract held in Maastricht, The Netherlands, 12–14 October, 2000

Mortality from 13 principal cancer sites, plus total cancer mortality at age 65-84 in 22 European countries, the United States, and Japan was analyzed. After earlier increases, total cancer mortality at age 65-84 has been declining between the late 1980s and the late 1990s in the European Union (-5.5% in males, -4.5% in females), in United States males (-2.3%), but not females (+4.4%), and in Japanese females (-5.6%), but not males (+6.3%). Cancer mortality in the elderly rose for both sexes in Eastern Europe between the late 1980s and the late 1990s. Gastric cancer mortality steadily declined in all the areas considered. Lung cancer rates at age 65-84 declined over the last decade by 8.5% in males in the European Union, and by 0.9% in the United States. Rates increased in Eastern Europe, in Japanese males, and in females in all areas. In women, an approximately threefold difference was evident between lung cancer rates of 67-77 per 100,000 in Japan and Europe in the late 1990s, and the rates of 212 per 100,000 in the United States. Likewise, pancreatic cancer mortality rates increased in the elderly of both sexes in the European Union and Japan up to the late 1980s, and in Eastern Europe up to the 1990s, whereas these rates decreased for United States males over more recent calendar periods, thus again reflecting the different spread of the tobacco-related cancer epidemic. After earlier increases in most areas, female breast cancer mortality in elderly women declined over the last decade by 8% in the United States and by 3% in the European Union, whereas it increased in Eastern Europe and Japan. Prostate cancer mortality declined in the European Union and the United States, whereas it rose in Eastern Europe and Japan. Most rates for breast, prostate, and ovarian cancers in the elderly remained comparatively low in Japan. Leukemia mortality in the elderly was stable in most areas, with some evidence of leveling off in the European Union. Mortality from multiple myeloma steadily increased by 10-20% over the last decade in both sexes in all geographic areas considered, and mortality from lymphomas increased, but improved diagnosis and certification may have played a relevant role in these trends.