Flavonoids belong to a vast group of polyphenols widely distributed in all foods of plant origin. Because of their antioxidant, antimutagenic and antiproliferative properties, they have been hypothesized to contribute to the favorable effects of fruit and vegetables against cancer. The aim of this study is to investigate the relation of 6 classes of flavonoids (flavan-3-ols, flavanones, flavonols, flavones, anthocyanidins and isoflavones) with ovarian cancer risk, using data from a multicentric case-control study conducted in Italy between 1992 and 1999. The study included 1,031 cases with incident, histologically confirmed epithelial ovarian cancer and 2,411 controls admitted for acute, nonneoplastic conditions to major hospitals in the same catchment areas. In logistic regression models including study center, education, year of interview, parity, oral contraceptive use and family history of ovarian or breast cancer or both, an inverse relation with significant trend in risk was found between ovarian cancer and flavonols [odds ratio (OR), 0.63; 95% confidence intervals (CI) 0.47-0.84] as well as isoflavones (OR, 0.51; 95% CI, 0.37-0.69), comparing the highest versus the lowest quintile. Further adjustment for fruit and vegetable intake did not modify these associations, suggesting that isoflavones and flavonols may have a distinct role in explaining the effect of fruit and vegetable against ovarian cancer. On the basis of our findings and the relevant literature, we infer that isoflavones, and perhaps flavonols, may have favorable effects with respect to ovarian cancer risk.

A role of hormone-related factors in renal cell cancer (RCC) etiology has been hypothesized, but the epidemiological evidence is inconsistent. The present study aimed at evaluating the effect of reproductive, menstrual and other gender-specific variables on RCC risk among women. This study is part of a larger hospital-based, case-control study on RCC risk, conducted in northern, central and southern Italy. Cases were 273 women, below age 80, with histologically confirmed, incident RCC. Controls were 546 women hospitalized for acute, nonneoplastic conditions, frequency-matched to cases by age and center. Odds ratios (OR) and 95% confidence intervals (CI) were computed using multiple logistic regression models. RCC risk was inversely related to age at first birth (OR = 0.7, 95% CI 0.5-1.0, for >/= 25 years vs. <25 years). Hysterectomy was found to double RCC risk (OR = 2.3, 95% CI 1.3-4.2). A negative association of borderline-statistical significance emerged for age at menarche, whereas, no associations were found between RCC risk and parity, menopausal status, age at menopause and use of hormone replacement therapy or oral contraceptives. Our findings give support to a role of hysterectomy in increasing RCC risk without corroborating, however, a major role of female hormone-related factors.

Dear Sir, ’We derived our pooled risk estimate of mesothelioma from nonoccupational exposure to asbestos from a previous publication authored by some of us, in which the methods for the meta-analysis were described in detail (1). In addition to the summary relative risk, our estimate of the number of non-occupational mesothelioma cases was based on frequencies in the populations experiencing exposures comparable with those in which the relevant epidemiological studies have been carried out. This parameter is subject to great uncertainty, and as a result, we selected it in our commentary as an example of the possible lack of temporal correspondence between risk and exposure data. Mirabelli and Merler suggest a different approach to estimate the number of mesothelioma cases attributable to non-occupational exposure to asbestos, based on the reasonable assumption that all cases with a recorded exposure are attributable to it. Their results of the Italian mesothelioma registry are of interest, and a recent analysis of the same data reported that 4.2% of cases in the registry had environmental asbestos exposure (2). It should be kept in mind that, although all reasonable efforts should be made to reduce and eliminate environmental asbestos exposure, this would have a limited effect on the overall cancer burden. Mesothelioma accounts each year for 1000 cases out of 300 000 total cancer cases (0.3%) in a country such as Italy, and hence 10% of mesothelioma cases represent 0.03% of all cancer cases.

Family history of urogenital cancers has been associated with an increased risk of cancer at the same sites. To evaluate the risks of bladder, prostate and renal cell neoplasms for family history of urogenital cancers in first-degree relatives, we analysed data from a case-control study conducted in northern Italy between 1985 and 1992 and including a total of 1,356 incident, histologically confirmed cancer cases: 727 bladder, 348 renal cell and 281 prostate cancer cases. Controls were 1,067 patients admitted to the same network of hospitals for acute, nonneoplastic conditions. Unconditional logistic regression was used to estimate odds ratios (ORs) of urogenital cancers, adjusting for age, sex, study centre, education, body mass index, smoking habit, alcohol consumption and number of brothers and, where appropriate, sisters. ORs of urogenital cancers for subjects with at least 1 first-degree relative with cancer at the same site were 6.1 [95% confidence interval (CI), 2.3-16.6] for bladder, 2.0 (95% CI: 0.6-6.2) for renal cell and 2.0 (95% CI: 0.9-4.1) for prostate cancer.

Epidemiological studies provide definite evidence that alcohol drinking is an independent risk factor for laryngeal cancer. This risk increases with the amount of alcohol consumed: in a meta-analysis of 20 studies conducted in North America, Europe, Japan and Korea the multivariate relative risks were about 2 for 50 g ( approximately 4 drinks)/day and about 4 for 100 g/day compared to nondrinkers, in the absence of evidence of a threshold. Genetic polymorphisms in the alcohol-metabolizing enzyme aldehyde dehydrogenases have been found to be associated with upper aerodigestive tract cancer, including the larynx. Further, the risk increases by concomitant tobacco smoking, each agent approximately multiplying the effect of the other. In the absence of smoking, the relative and absolute risks are small for moderate alcohol consumption, but there is an increased risk for elevated alcohol consumption. After stopping drinking, some fall in risk becomes apparent only in the long term. The supraglottis is more closely related to alcohol consumption, as compared to the glottis/subglottis. In various populations, the most commonly used alcoholic beverage appears to be the one most strongly associated with laryngeal cancer risk, suggesting that no meaningful difference exists for different types of alcoholic beverages.

In this issue of HEPATOLOGY, the article by Hu et al.1 provides relevant information to our understanding of the relation between coffee drinking, markers of liver disease, and liver cancer risk. CI, confidence interval; GGT, gamma-glutamyltransferase; HBV, hepatitis B virus; HCV, hepatitis C virus; RR, relative risk. First, it adds original data to our knowledge on the coffee and liver disease relationship from a cohort study of 60,323 subjects who underwent seven subsequent surveys between 1972 and 2002, and were followed-up to June 2006 (mean time of follow-up was 18 years). The study was conducted in Finland, a country with high coffee consumption.1 Previously available data on coffee and liver cancer were based on at least 10 studies, six of which were case-control (from Greece, Italy, and Japan) and four of which were cohort investigations (all of these from Japan). Overall, the pooled relative risk (RR) was 0.54 (95% confidence interval [CI] 0.39-0.76) for case-control studies, based on 1199 coffee drinker and 352 non–coffee drinker cases, and 0.64 (95% CI 0.56-0.74) for cohort studies, based on 404 coffee drinker and 305 non–coffee drinker cases.2 All the cohort studies, however, came from Japan, a country where coffee consumption is lower than in western countries2 and much lower than in Finland.1 Thus, available data from cohort studies did not allow investigation of the the dose-risk relation between coffee and liver cancer for persons who consumed >1 cup per day. The pooled RRs from Japanese cohort studies were 0.70 (95% CI 0.62-0.82) for <1 cup per day, and 0.50 (95% CI 0.38-0.66) for ≥1 cup per day. Only data from case-control studies, mainly derived from southern Europe, allowed estimation of a pooled RR for persons who consumed ≥3 cups per day (RR = 0.42, 95% CI 0.32-0.55,2). This estimate is largely consistent with the RR of liver cancer for persons consuming >3 cups per day in the study by Hu et al.,1 which was around 0.4. This consistency of the results across study designs and populations strongly supports the existence of a real inverse relation between coffee drinking and liver cancer risk. A second important and original finding of the study by Hu et al.1 is the observation that the inverse relation between coffee and liver cancer is independent of serum gamma-glutamyltransferase (GGT). Serum GGT levels were available only for about 63% of cases who underwent more recent surveys (from 1982-2002), and the RRs for subsequent levels of coffee consumption, further adjusted for serum GGT, were close to those unadjusted for that variable. The study was also able to consider the combined effect of coffee and GGT on liver cancer risk. Compared with heavy coffee drinkers in the three lowest quartiles of GGT, the RR was 2.7 for those who were nondrinkers or moderate coffee drinkers (<1 cup per day) on the same lowest quartiles of GGT, but rose to 11.0 for those in the highest quartile of GGT for nondrinkers or moderate coffee drinkers. Given the small number of subjects in each stratum, no CI for these estimates was given, but the pattern of risk across strata is suggestive of an independent effect of both serum GGT and coffee drinking on subsequent liver cancer risk. Case-control studies from Italy3-5 considered the separate effect of coffee and the two major risk factors for liver cancer, that is, clinical history of hepatitis or serum markers for hepatitis B (HBV) and C (HCV) viruses, and alcohol drinking, and reported consistent inverse relations between coffee drinking and liver cancer risk across strata of these covariates. It is now clear, therefore, that the inverse relation of coffee is independent from major established risk factors for liver cancer. Another recognized risk factor for liver cancer is tobacco smoking,3-5 which however tends to be directly correlated with coffee consumption, Furthermore, the association between tobacco and liver cancer is much less strong than those of HBV, HCV, and alcohol. Thus, tobacco is unlikely to exert a relevant confounding effect on the inverse relationship between coffee and liver cancer. Further, the RR provided by most studies,2-5 including that of Hu et al.,1 were adjusted for tobacco, and if anything, allowance for tobacco and several other variables such as education, history of diabetes, and chronic liver disease led to a stronger inverse relation with coffee. More important, the study by Hu et al.,1 was able to show that the inverse relation between coffee and liver cancer is similar across strata of smoking habits (current smoker versus never smoked or exsmoker), as well as of body mass index, besides alcohol drinking, as already reported by case-control studies from Italy and Greece.2-5 With reference to potential information bias, recall of coffee drinking has been shown to be satisfactorily reproducible and valid.6, 7 Thus, although all observational epidemiological studies (case-control and cohort) are prone to various sources of bias and confounding, these are unlikely to explain such a consistent inverse association as that reported between coffee and liver cancer. Likewise, it is unlikely that publication bias, that is, selective publication of studies with positive findings, may account for such a strong inverse relation. We can now, therefore, conclude that the inverse relation between coffee and liver cancer risk is real. Various components of coffee, including caffeine, the coffee oils kahweol and cafestol, antioxidant substances from coffee beans, and chlorogenic acid, have been related to such an inverse relation, because they may favorably influence glucose metabolism and prevent hyperglycemia,1, 2, 8 but no definite evidence on biologic mechanisms of the inverse relation between coffee and liver cancer is available. Coffee drinking has been related to decreased levels of liver enzymes, including GGT and alanine aminotransferase activity,8-11 and to reduced risk of type II diabetes,12 which is a recognized risk factor for liver cancer.13, 14 Coffee drinking has also been inversely related to the risk of cirrhosis, with RRs around 0.2-0.3 for heavy coffee drinkers versus nondrinkers in studies conducted in Italy.15, 16 Cirrhosis is a major correlate of hepatocellular carcinoma,17, 18 but allowance for clinical history of cirrhosis did not account for the inverse association between coffee and liver cancer.8 There is therefore a continuum of clinical and epidemiological evidence, which ranges from liver enzymes to cirrhosis to hepatocellular carcinoma, to support a favorable effect of coffee on liver function and liver diseases, including cirrhosis and also primary liver cancer. Despite the apparent consistency of these findings, it is difficult to establish whether such an inverse relation between coffee and liver cancer is causal on the basis of observational (case-control and cohort) epidemiological studies alone. The inverse relation observed may be spurious, and attributable simply to the fact that subjects with a broad spectrum of digestive tract disorders, liver diseases, or cirrhosis may reduce their coffee consumption. Indeed, caffeine metabolism is impaired in patients with cirrhosis, who could therefore reduce coffee intake in order to avoid side effects of caffeine. Thus, bias due to reduction of coffee drinking in unhealthy subjects cannot be excluded. However, in a study on hepatocellular carcinoma that used patients with chronic liver disease as controls,19 a reduced risk for coffee drinkers was found, similar to that observed in studies using subjects with no liver disease as controls. Avoidance of coffee, moreover, is not commonly recommended to patients with chronic liver disease, and an inverse relation between coffee and liver cancer was also observed among subjects with self-reported2 or serological evidence of hepatitis.3, 5 The study by Hu et al.8 adds relevant evidence that such an inverse relation is also observed, and is indeed apparently stronger, in subjects in the highest quartile of GGT, although the interaction test between coffee consumption and serum GGT on the risk of liver cancer was far from significant. Furthermore, the study by Hu et al.1 provides original and important quantitative evidence that GGT levels are related to subsequent incidence of liver cancer, with an overall RR of 2.3 (95% CI 1.0-5.2). This is not surprising, because GGT is known to be directly correlated to chronic liver disease and cirrhosis, which is a major risk factor and probably a pathogenic step in liver carcinogenesis.17, 18 It remains difficult, however, to translate the inverse relation between coffee drinking and liver cancer risk observed in epidemiological studies into potential implications for prevention of liver cancer by increasing coffee consumption. Liver cancer is the third most common cause of cancer death worldwide, with an estimated number of about 600,000 deaths per year in the early 2000s.20 Together with avoidance of lung cancer through tobacco control, primary liver cancer is the other common neoplasm which is most largely avoidable, through HBV vaccination, control of HCV transmission, and reduction of alcohol drinking. These three measures can, in principle, avoid more than 90% of primary liver cancers worldwide. Whether coffee drinking has an additional role in liver cancer prevention remains open to discussion, but in any case any such role would be limited—if not negligible—as compared to that achievable through control of HBV, HCV, and alcohol consumption, which are the major recognized risk factors for liver cancer. This work was conducted with the contribution of the Italian Association for Cancer Research and the Italian League Against Cancer. The work of this article was undertaken while the author was a Senior Fellow at the International Agency for Research on Cancer.

Ultraviolet radiation is the major cause of skin cancer, but promotes vitamin D synthesis, and vitamin D has been inversely related to the risk of several common cancers including prostate, breast and colorectum. We therefore computed the incidence of prostate, breast and colorectal cancer following skin cancer using the datasets of the Swiss cancer Registries of Vaud and Neuchâtel. Between 1974 and 2005, 6,985 histologically confirmed squamous cell skin cancers, 21,046 basal cell carcinomas and 3,346 cutaneous malignant melanomas were registered, and followed up to the end of 2005 for the occurrence of second primary cancer of the prostate, breast and colorectum. Overall, 680 prostate cancers were observed versus 568.3 expected (standardized incidence ratio (SIR) = 1.20; 95% confidence interval (CI): 1.11-1.29), 440 breast cancers were observed versus 371.5 expected (SIR = 1.18; 95% CI: 1.08-1.30) and 535 colorectal cancers were observed versus 464.6 expected (SIR = 1.15; 95% CI: 1.06-1.25). When basal cell, squamous cell and skin melanoma were considered separately, all the SIRs for prostate, breast and colorectal cancers were around or slightly above unity. Likewise, the results were consistent across strata of age at skin cancer diagnosis and location (head and neck versus others), and for male and female colorectal cancers. These findings, based on a population with a long tradition of systematic histologic examination of all surgically treated skin lesions, do not support the hypothesis that prostate, breast and colorectal cancer risk is decreased following skin cancer.