Tobacco and alcohol use are the main risk factors for oral and oropharyngeal cancers, yet, dietary habits may also be of importance. Data from a series of case-control studies conducted in 9 countries worldwide (1,670 cases and 1,732 controls) were used to investigate the role of several food groups and body mass index (BMI). Low BMI significantly increased the odds ratio (OR) of cancer more than 2-fold among ever- and never-tobacco users and ever- and never-alcohol drinkers. After adjustment for potential confounders, high intake of fruits and vegetables significantly reduced the OR of cancer compared to low intake among ever-tobacco users (OR 0.4, 95% confidence interval [CI] 0.3-0.6), although not among never-tobacco users (OR 1.1, 95% CI 0.6-2.0). Similarly, the protective effect of high fruit and vegetable consumption was present among ever-drinkers (OR 0.4, 95% CI 0.3-0.6), but not among never-drinkers (OR 1.0, 95% CI 0.6-1.6). In conclusion, low BMI increases the risk of oral cancer, and vegetables and fruits may modulate the carcinogenic effects of tobacco and alcohol.

Various aspects of the Mediterranean diet are considered favourable with regard to cancer risk. These aspects were analysed using data from a series of case-control studies conducted in northern Italy between 1983 and 2001 on over 12,000 cases of 20 cancer sites and 10,000 controls. For most epithelial cancers, the risk decreased with increasing vegetable and fruit consumption, with odds ratios (OR) between 0.3 and 0.7 for the highest versus the lowest tertile. Subjects reporting frequent red meat intake showed ORs above unity for several common neoplasms. Conversely, fish (and consequently, n-3 fatty acids) tended to be another favourable dietary indicator. Wholegrain food intake was related to reduced risk of several types of cancer, particularly of the upper digestive tract, probably on account of its high fibre content. Fibres were in fact found to be protective with regard to colorectal and other selected cancers. In contrast to wholegrain, refined grain intake, and consequently glycaemic load, was associated with an increased risk of different types of cancer, including those of the upper digestive tract, colorectum, breast and endometrium. These results thus suggest that a low-risk diet for cancer entails increasing vegetables and fruit, reducing meat, but also refined carbohydrate consumption. Furthermore, olive oil and other unsaturated fats, which may be a unique common characteristic of the Mediterranean diet, should be preferred to animal and saturated fats. A score summarizing the major characteristics of the Mediterranean diet was inversely and consistently related to the risk of selected cancer sites. Regular consumption of pizza, one of the most typical Italian foods, showed a reduced risk of digestive tract cancers. Pizza could however simply be an indicator of a typical Italian diet.

We reviewed the descriptive epidemiology of thyroid cancer using data from the Cancer Registry of the Canton of Vaud, Switzerland, a consistently well-surveilled population with relatively high rates of the disease, between 1974 and 1998, on the basis of a total of 596 registered cases. Overall thyroid cancer incidence tended to increase moderately in both genders over the 25-year period considered, to reach rates of 5.4 and 2.0 per 100,000 (world standard) in females and males, respectively. There were also changes in histologic classification, with some increases in papillary neoplasms and corresponding decreases in other and unclassified ones. Ten-year relative survival rates for cases diagnosed in 1988-1993 were 94% for papillary cancer in females and 69% in males, and 59% for follicular cancer in females. Corresponding figures were 12% for undifferentiated, and 31% for other and unspecified neoplasms in both sexes combined. Multivariate analysis confirmed the unfavorable influence of male gender (hazard ratio [HR] = 1.9), elderly age (HR = 16.5 for age > or = 65 vs. < 45 years) and undifferentiated histotype (HR = 3.5 vs. papillary) on the long-term prognosis of thyroid cancer, and showed no consistent evidence of appreciably improved prognosis over more recent calendar periods.

1Istituto di Ricerche Farmacologiche “Mario Negri”, via Eritrea 62, 20157 Milan, Italy 2Istituto di Statistica Medica e Biometria, Università degli Studi di Milano, via Venezian 1, 20133 Milan, Italy Correspondence to: C La Vecchia, Istituto di Ricerche Farmacologiche “Mario Negri”, Via Eritrea 62, 20157 Milano, Italy E-mail: [email protected]

To evaluate the strength of the evidence provided by the epidemiological literature on the association between alcohol consumption and the risk of 18 neoplasms, we performed a search of the epidemiological literature from 1966 to 2000 using several bibliographic databases. Meta-regression models were fitted considering linear and non-linear effects of alcohol intake. The effects of characteristics of the studies, of selected covariates (tobacco) and of the gender of individuals included in the studies, were also investigated as putative sources of heterogeneity of the estimates. A total of 235 studies including over 117 000 cases were considered. Strong trends in risk were observed for cancers of the oral cavity and pharynx, oesophagus and larynx. Less strong direct relations were observed for cancers of the stomach, colon and rectum, liver, breast and ovary. For all these diseases, significant increased risks were found also for ethanol intake of 25 g per day. No significant nor consistent relation was observed for cancers of the pancreas, lung, prostate or bladder. Allowance for tobacco appreciably modified the relations with laryngeal, lung and bladder cancers, but not those with oral, oesophageal or colorectal cancers. This meta-analysis showed no evidence of a threshold effect for most alcohol-related neoplasms. The inference is limited by absence of distinction between lifelong abstainers and former drinkers in several studies, and the possible selective inclusion of relevant sites only in cohort studies.

The hypothesis that the Mediterranean diet has a beneficial role on the risk of cancers of the upper aerodigestive tract has been evaluated using data from three case-control studies conducted in Italy between 1992 and 2000. The first study included 598 cases with incident, histologically confirmed cancers of the oral cavity and pharynx and 1491 hospital controls admitted to the same network of hospitals as cases for acute, nonneoplastic diseases. The second one included 304 subjects with squamous cell carcinoma of the esophagus and 743 controls. The third one included 460 laryngeal cancer cases and 1088 controls. A score summarizing eight of the major characteristics of the Mediterranean diet was used. Odds ratios and corresponding 95% confidence intervals (CIs) for increasing levels of this score were estimated using unconditional regression models, adjusted for age, sex, study center, years of education, tobacco consumption, body mass index, and total energy intake. For all cancers considered, a reduced risk was found for increasing levels of the Mediterranean score: the odds ratios for subjects with six or more Mediterranean characteristics, compared with those with less than three characteristics, were 0.40 (95% CI, 0.26-0.62) for oral and pharyngeal, 0.26 (95% CI, 0.13-0.51) for esophageal, and 0.23 (95% CI, 0.13-0.40) for laryngeal cancer. All of the estimates were consistent in strata of the major identified risk factors for these neoplasms. This study provides evidence that an a priori defined nutritional pattern, which includes several aspects of the Mediterranean diet, favorably affects the risk of cancers of the upper aerodigestive tract.

Istituto di Ricerche Farmacologiche ‘Mario Negri’, 20157 Milan, Italy and Istituto di Statistica Medica e Biometria, Università degli Studi di Milano, 20133, Milan, Italy Fax: (+39) 02 3330 0231, E-mail: [email protected]

In their paper on gender effects in familial cancer, Hemminki and Li1 reported that, among 15 cancer sites considered, only thyroid cancer showed a significant difference in the relative risk (RR) for concordant cancer by gender. Based on the nation-wide Swedish Family-Cancer Database, which included 10.2 million individuals and over 1 million cancers, the RR of thyroid cancer was 16.4 (95% confidence interval [CI] = 11.3–23.8) in the male but 6.6 (95% CI = 4.9–9.0) in the female offspring of thyroid cancer cases. Thus, the ratio observed for male compared to female offspring was 2.48 (95% CI = 1.5–4.0). Notably, women had a 2–3-fold higher thyroid cancer incidence rate compared to men in the general population of Sweden and most other countries.2 Because thyroid cancer is rare among men, the majority of our knowledge on thyroid cancer comes from data on women. In a pooled analysis of 12 case-control studies on thyroid cancer, which includes all ten published studies that have males, there were 2,094 female cases, but only 425 male cases (3,248 female and 928 male controls, respectively3). The only risk factor that appeared to be stronger in males than in females was a personal history of benign thyroid diseases (RR for goitre = 38.3, 95% CI = 5.0–291.2 in men and 5.9; 95% CI = 4.2–8.1 in women; RR for history of benign nodules/adenomas = infinity, 95% CI = 9.2-infinity in men and 29.9, 95% CI = 14.5–62.0 in women). The gender difference in the RRs resulted mainly from the rarity of goiter (1/836) and benign nodules/adenomas (0/531) among male controls. A history of goiter or benign nodules/adenomas was reported, respectively, by 1.8% and 0.3% of female controls.3 The similarity in the gender effect for family history of thyroid cancer and personal history of benign thyroid disease suggests that both factors can be considered markers of increased individual susceptibility to thyroid cancer. Lower RR for markers of individual susceptibility among women than men supports the suggestion4 that the greater risk of cancers of the thyroid in the general female population is chiefly attributable to female-specific environmental risk factors (e.g., female hormones, greater functional demand of iodine and thyroid hormones during reproductive years, etc.2). Conversely, in men, a greater proportion of thyroid cancer may be the result of genetic factors. The genetic bases for the commonest types of thyroid cancer (i.e., non-medullary forms) are not well understood.5 Linkage studies have identified a few genes (MNG1,6 TCO,7 PTC8), but single genes seem to explain only a minority of cases. Non-medullary thyroid cancer is thus likely to be a polygenic disease.9 Although surveillance bias may explain some of the findings related to benign thyroid disease,3 because women tend to undergo general medical examinations more frequently than men, it is unlikely to account for the gender-related difference in the influence of family history of thyroid cancer.1 Also in another cancer site, the breast, where the female excess is even greater than for the thyroid, a marker of individual susceptibility (history of contralateral breast cancer) showed a clear difference between the gender-specific RRs.10 The RR of contralateral breast cancer among men was 29.6 (95% CI = 15.5–52.4) compared to 1.8 (95% CI = 1.7–1.8) among women. In conclusion, the findings from Hemminki and Li1 and our own3 illustrate well the strong influence of non-genetic risk factors on the apparent strength of the RRs for genetic markers. As a consequence of a weaker “dilution effect” from non-genetic risk factors, family history of thyroid cancer and previous occurrence of goitre and benign nodules/adenomas can, however, distinguish high-risk men more accurately than high-risk women. The contribution of the Italian Association for Cancer Research, Milan, Italy is gratefully acknowledged. Yours sincerely, Silvia Franceschi*, Eva Negri , Carlo La Vecchia , * International Agency for Research on Cancer, Lyon, France, Istituto di Ricerche Farmacologiche “Mario Negri”, Milan, Italy, Istituto di Statistica Medica e Biometria, Università degli Studi di Milano, Milan, Italy