IJC International Journal of Cancer

Transcription

1 IJC International Journal of Cancer Dietary flavonoid, lignan and antioxidant capacity and risk of hepatocellular carcinoma in the European prospective investigation into cancer and nutrition study Raul Zamora-Ros 1,2, Veronika Fedirko 2,3, Antonia Trichopoulou 4,5, Carlos A. Gonzalez 1, Christina Bamia 4, Elisabeth Trepo 6, Ute N othlings 7, Talita Duarte-Salles 2, Mauro Serafini 8, Lea Bredsdorff 9, Kim Overvad 10, Anne Tjønneland 11, Jytte Halkjær 11, Guy Fagherazzi 12, Florence Perquier 12, Marie-Christine Boutron-Ruault 12, Verena Katzke 13, Annekatrin Lukanova 13, Anna Floegel 14, Heiner Boeing 14, Pagona Lagiou 4,15,16, Dimitrios Trichopoulos 5,15,16, Calogero Saieva 17, Claudia Agnoli 18, Amalia Mattiello 19, Rosario Tumino 20, Carlotta Sacerdote 21, H. Bas Bueno-de-Mesquita 22,23, Petra H.M. Peeters 24,25, Elisabete Weiderpass 26,27,28,29, Dagrun Engeset 26, Guri Skeie 26, Marcial Vicente Arg uelles 30, Esther Molina-Montes 31,32, Miren Dorronsoro 32,33, Marıa Jose Tormo 32,34, Eva Ardanaz 32,35, Ulrika Ericson 36, Emily Sonestedt 36, Malin Sund 37,38,39, Rikard Landberg 40, Kay-Tee Khaw 41, Nicholas J. Wareham 42, Francesca L. Crowe 43, Elio Riboli 25 and Mazda Jenab 2 1 Unit of Nutrition, Environment and Cancer, Catalan Institute of Oncology (ICO), Bellvitge Biomedical Research Institute (IDIBELL), Barcelona, Spain 2 Section of Nutrition and Metabolism, International Agency for Research on Cancer (IARC), Lyon, France 3 Department of, Rollins School of Public Health, Emory University, Atlanta, GA 4 WHO Collaborating Center for Food and Nutrition Policies, Department of Hygiene, and Medical Statistics, University of Athens Medical School, Athens, Greece 5 Hellenic Health Foundation, Athens, Greece 6 Centre de Biologie Republique, Lyon, France 7 Department of Nutrition and Food Sciences, Nutritional, University of Bonn, Germany 8 Antioxidant Research Laboratory, National Institute for Food and Nutrition Research, Rome, Italy 9 National Food Institute, Technical University of Denmark, Moerkhoej, Denmark 10 Section for, Department of Public Health, Aarhus University, Aarhus, Denmark 11 Institute of Cancer, Danish Cancer Society, Copenhagen, Denmark 12 INSERM U1018, Center for Research in and Population Health (CESP), Villejuif, France 13 Department of Cancer, German Cancer Research Center, Heidelberg, Germany 14 Department of, German Institute of Human Nutrition Potsdam-Rehbr ucke, Nuthetal, Germany 15 Department of, Harvard School of Public Health, Boston, MA 16 Bureau of Epidemiologic Research, Academy of Athens, Greece 17 Molecular and Nutritional Unit, ISPO Cancer Prevention and Research Institute, Florence, Italy 18 Nutritional Unit, IRCCS Foundation, National Cancer Institute, Milan, Italy Key words: flavonoids, lignans, dietary intake, antioxidant capacity, hepatocellular carcinoma, EPIC Abbreviations: EPIC: European Prospective Investigation into Cancer and Nutrition; FRAP: ferric reducing antioxidant capacity; HBV: hepatitis B virus; HCC: hepatocellular cancer; HCV: hepatitis C virus; HR: hazards ratio; NEAC: nonenzymatic antioxidant capacity; TRAP: total radical-trapping antioxidant parameter Additional Supporting Information may be found in the online version of this article. Grant sponsor: the French National Cancer Institute (L Institut National du Cancer, INCA); Grant number: ; Grant sponsors: the European Commission: Public Health and Consumer Protection Directorate 1993 to 2004; Research Directorate-General 2005; Ligue contre le Cancer, Institut Gustave Roussy, Mutuelle Generale de l Education Nationale, Institut National de la Sante et de la Recherche Medicale (INSERM) (France); German Cancer Aid; German Cancer Research Center (DKFZ); German Federal Ministry of Education and Research; Danish Cancer Society; Health Research Fund (FIS) of the Spanish Ministry of Health (RTICC); Grant number: DR06/0020; Grant sponsors: the participating regional governments from Asturias, Andalucıa, Murcia, Navarra and Vasco Country and the Catalan Institute of Oncology of Spain; Cancer Research UK; Medical Research Council, UK; the Stroke Association, UK; British Heart Foundation; Department of Health, UK; Food Standards Agency, UK; the Wellcome Trust, UK; the Hellenic Health Foundation; Italian Association for Research on Cancer; Compagnia San Paolo, Italy; Dutch Ministry of Public Health, Welfare and Sports; Dutch Ministry of Health; Dutch Prevention Funds; LK Research Funds; Dutch ZON (Zorg Onderzoek Nederland); Statistics Netherlands (The Netherlands); Swedish Cancer Society; Swedish Scientific Council; Regional Government of Skane, Sweden; Nordforsk Centre of Excellence programme DOI: /ijc History: Received 19 Feb 2013; Accepted 16 Apr 2013; Online 6 May 2013 Correspondence to: Raul Zamora-Ros, Unit of Nutrition, Environment and Cancer, Catalan Institute of Oncology, Bellvitge Biomedical Research Institute (IDIBELL), Avda Gran via L Hospitalet de Llobregat, Spain. Tel.: , Fax: , rzamora@iconcologia.net

2 2430 Flavonoids, antioxidant capacity and hepatocellular cancer 19 Department of Clinical and Experimental Medicine, Federico II University, Naples, Italy 20 Cancer Registry and Histopathology Unit, "Civile M.P. Arezzo" Hospital, Ragusa, Italy 21 Center for Cancer Prevention (CPO-Piemonte), and Human Genetic Foundation (HuGeF), Torino, Italy 22 National Institute for Public Health and the Environment (RIVM), Bilthoven, The Netherlands 23 Department of Gastroenterology and Hepatology, University Medical Center Utrecht (UMCU), Utrecht, The Netherlands 24 Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, The Netherlands 25 School of Public Health, Faculty of Medicine, Imperial College, London, United Kingdom 26 Department of Community Medicine, University of Tromsø, Norway 27 Cancer Registry of Norway, Oslo, Norway 28 Department of Medical and Biostatistics, Karolinska Institutet, Stockholm, Sweden 29 Samfundet Folkh alsan, Helsinki, Finalnd 30 Public Health Directorate, Asturias, Spain 31 Andalusian School of Public Health, Granada, Spain 32 CIBER Epidemiologıa y Salud Publica (CIBERESP), Spain 33 Public Health Division of Gipuzkoa, Basque Regional Health Department, San Sebastian, Spain 34 Department, Murcia Health Council, Murcia, Spain 35 Public Health Institute of Navarra, Pamplona, Spain 36 Diabetes and Cardiovascular disease, Genetic, Department of Clinical Sciences, Lund University, Malm o, Sweden 37 Department of Surgical and Perioperative Sciences, Nutrition Research, Umeå University, Umeå, Sweden 38 Department of Surgery, Nutrition Research, Umeå University, Umeå, Sweden 39 Department of Public Health and Clinical Medicine, Nutrition Research, Umeå University, Umeå, Sweden 40 Department of Food Science, BioCenter, Swedish University of Agriculture Science, Uppsala, Sweden 41 Department of Public Health and Primary Care, University of Cambridge, United Kingdom 42 MRC Unit, Cambridge, United Kingdom 43 Cancer Unit, University of Oxford, United Kingdom Limited epidemiological evidence suggests a protective role for plant foods rich in flavonoids and antioxidants in hepatocellular cancer (HCC) etiology. Our aim was to prospectively investigate the association between dietary intake of flavonoids, lignans and nonenzymatic antioxidant capacity (NEAC) and HCC risk. Data from the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort including 477,206 subjects (29.8% male) recruited from ten Western European countries, was analyzed. Flavonoid, lignan and NEAC intakes were calculated using a compilation of existing food composition databases linked to dietary information from validated dietary questionnaires. Dietary NEAC was based on ferric reducing antioxidant capacity (FRAP) and total radical-trapping antioxidant parameter (TRAP). Hepatitis B/C status was measured in a nested case control subset. During a mean follow-up of 11-years, 191 incident HCC cases (66.5% men) were identified. Using Cox regression, multivariable adjusted models showed a borderline nonsignificant association of HCC with total flavonoid intake (highest versus lowest tertile, HR , 95% CI: ; p trend ), but not with lignans. Among flavonoid subclasses, flavanols were inversely associated with HCC risk (HR , 95% CI: ; p trend ). Dietary NEAC was inversely associated with HCC (FRAP: HR 0.50, 95% CI: ; p trend ; TRAP: HR 0.49, 95% CI: ; p trend ), but statistical significance was lost after exclusion of the first 2 years of follow-up. This study suggests that higher intake of dietary flavanols and antioxidants may be associated with a reduced HCC risk. What s new? Coffee, tea, fruits and vegetables, and certain other foods may protect against hepatocellular carcinoma (HCC), thanks to their antioxidant ingredients. This study lends fresh support to that idea, revealing specifically that dietary flavanols, which possess antioxidant activity, could play a favourable role in HCC prevention. Dietary antioxidant capacity from coffee intake in particular was found to be inversely associated with HCC risk, though statistical significance was lost after exclusion of the first two years of follow-up. Assessment of the bioavailability of flavonoids and other antioxidants is needed to confirm links between antioxidant intake and HCC risk. Liver cancers are the sixth most common cancer and the third most common cause of cancer mortality worldwide. 1 Hepatocellular carcinoma (HCC) is the most common type of liver cancer, with over half a million cases being diagnosed worldwide, each year. It is highly malignant, usually diagnosed at late stages and often has very poor prognosis with

3 Zamora-Ros et al limited treatment options. As a result, its 5-year survival rate was only 9% in Europe during the period of Established risk factors for HCC include chronic hepatitis B and C virus (HBV/HCV) infection, heavy alcohol drinking, tobacco smoking, type 2 diabetes, obesity and aflatoxin exposure. 3,4 However, dietary factors are strongly suspected to also contribute to HCC risk. 3,5 Several epidemiological studies suggest the potential protective effect of certain food groups, such as coffee, fruits and vegetables, tea and soya products, against the development of HCC. 3,6,7 All of these foods are abundant sources of dietary antioxidant compounds, particularly polyphenols, 8 which may play a role in HCC pathophysiology. Polyphenols are secondary metabolites widely distributed in plant foods. They are divided into four main classes: flavonoids (anthocyanidins, flavonols, flavanones, flavones, flavanols and isoflavones), phenolic acids, stilbenes and lignans. They have been extensively studied due to their antioxidant, anti-inflammatory and anticarcinogenic properties Indeed, polyphenols are the main in vitro contributors of food s nonenzymatic total antioxidant capacity (NEAC), a measurement taking into account the antioxidant activity of single ingredients as well as their synergistic interactions. 12 Modulation of estrogen signaling is another potential anticarcinogenic mechanism of action for lignan and isoflavone subclasses of polyphenols, which has been described mainly in breast tissue. 13 Several in vitro studies have suggested an antitumor effect of flavonoids in some hepatocarcinoma cell lines Animal model data also show that flavonoids modulate important cellular and molecular mechanisms related to proliferation, invasion, angiogenesis, survival and metastasis of tumor cells. 17 Although an inverse association between a high intake of flavonoid-rich foods and some gastrointestinal cancers has been observed, 18 very little is currently known about their association with HCC risk. In a Greek case control study, flavone intake was inversely associated with HCC risk in both chronic hepatitis virus positive (highest versus lowest quintile, OR , 95% CI: ; p trend ) and negative cases (highest versus lowest quintile, OR , 95% CI: ; p trend ). 19 Furthermore, there is some evidence suggesting that fruits, the main food source of flavonoids in Western diets, 20 protect against HCC. 5 To our knowledge, there are no studies evaluating the association between NEAC and HCC risk. Thus, further research assessing the potential effect of polyphenols and NEAC in HCC is warranted. Therefore, the aim of this study was to evaluate the association of dietary intakes of flavonoids, lignans and NEAC on the risk of HCC within the European Prospective Investigation into Cancer and Nutrition (EPIC) study, 21 a large prospective cohort with considerable variability in flavonoid, lignan and NEAC intakes among participants SUBJECTS AND METHODS Subjects EPIC is an ongoing multicenter prospective cohort study designed to investigate the role of diet, lifestyle and environmental factors in the etiology of cancer and other chronic diseases. The cohort consists of over 521,000 men and women, mostly aged years, who were recruited between 1991 and 2000 from 23 centers in ten European countries: Denmark, France, Germany, Greece, Italy, Norway, Spain, Sweden, the Netherlands and United Kingdom. 27 Approval for this study was obtained from the ethical review boards of the International Agency for Research on Cancer (IARC) and from all local institutions. All cohort members provided written informed consent to participate in the study. Follow-up and identification of cancer cases Follow-up for endpoint status was based mostly on population cancer registries, except for France, Germany and Greece, where a combination of methods including health insurance records, cancer and pathology hospital registries and active follow-up were used. The date of last complete followup (recorded in a central database at IARC) ranged from 2006 to 2010, depending on the center. HCC was defined as tumor in the liver (C22.0 as per the 10th Revision of the International Statistical Classification of Diseases, Injury and Causes of Death). For each case identified, the histology and the methods used for the diagnosis of HCC were carefully reviewed to exclude metastatic cases or other types of liver cancers. Cohort study A total of 477,206 participants were included in the present analysis after an exclusion of 23,818 with prevalent cancer other than nonmelanoma skin cancer, 4,380 with incomplete follow-up data or missing information on date of diagnosis, 6,192 with missing dietary information, 60 with missing lifestyle information, 9,596 who were in the top or bottom 1% of the distribution of the ratio of reported total energy intake to estimated energy requirement and 78 with metastasis in the liver or with ineligible histology code. A total of 191 HCC cancer cases were included in the present analyses. Nested case control study The design of the nested case control subset has been previously described in detail elsewhere. 4 The study population is based on EPIC subjects with available blood samples. Briefly, 125 HCC cases were identified during the period between participants recruitment and For each case, two controls were selected by incidence density sampling from all cohort members alive and free of cancer (except nonmelanoma skin cancer), and matched by age at blood collection (61 year), sex, study center, date (62 months) and time of the day at blood collection (63 hr) and fasting status at blood collection (<3/3 6/>6 hr). Women were additionally matched by menopausal status (pre-/peri-/postmenopausal), and hormone replacement therapy use at the time of blood collection (yes/no). HBV and HCV seropositivity was detected using the ARCHITECT HBsAg and anti-hcv chemiluminescent

4 2432 Flavonoids, antioxidant capacity and hepatocellular cancer microparticle immunoassays (CMIAs; Abbott Diagnostics, France). The laboratory personnel were blinded as to the disease status of the subjects whose blood samples they analyzed. Participants with missing blood sample or failed laboratory assay for HBV/HCV were excluded (N 5 11). Therefore, the final sample size included 122 cases and 242 controls. Data collection Dietary intake during the year before recruitment was measured with country-specific validated food-frequency questionnaires (FFQs) or diet histories that were designed to capture local dietary habits Face-to-face interviews were done in Greece, Ragusa-Italy, Naples-Italy and Spain, whereas, in all other centers, questionnaires were self-administered. Daily food intakes were calculated in grams per day. Ethanol (g/ day), fiber (g/day) and energy (kcal/day) intakes were computed using the EPIC Nutrient Database. 30 Anthropometric measures and data on sociodemographic characteristics, lifestyles, smoking history, physical activity and medical history were collected. 27 At recruitment, blood samples were also taken from most participants and stored at 2196 C in liquid nitrogen. Flavonoid, lignan, FRAP and TRAP intakes Dietary flavonoid and lignan intakes were estimated by matching food items on the country-specific questionnaires with a comprehensive food composition database on flavonoids based on US Department Agriculture, Phenol- Explorer, 34 and the UK Food Standards Agency databases. 35 Our food composition database contains 1877 food items and composition data on lignans and the six flavonoid subclasses: anthocyanidins, flavanols (including flavan-3-ols monomers, proanthocyanidins and theaflavins), flavonols, flavones, flavanones and isoflavones Dietary NEAC intake was estimated using a food composition database on ferric reducing antioxidant capacity (FRAP) and total radical-trapping antioxidant parameter (TRAP) data on foods. FRAP assay is an in vitro measure of the reducing antioxidant power, whereas TRAP assay evaluates the chain-breaking antioxidant capacity. 36 This database is based on analytical data from Italy and contains 210 and 207 food items, mostly raw foods, for FRAP and TRAP, respectively. 37,38 Because of the high contribution of Maillard products, the bioavailability of which is still unclear, 26 in determining more than 90% of in vitro NEAC of coffee, we calculated FRAP and TRAP intake with and without coffee data. For food items for which flavonoids, lignans, FRAP or TRAP data were unavailable, the value of a similar food was assigned as previously explained. 20,26 Statistical analysis Descriptive analysis Distribution of the population s main characteristics according to tertiles of either total flavonoids, TRAP or FRAP intakes were examined using two-sided chi-square and Kruskal Wallis tests, as appropriate. Spearman correlations were calculated between coffee consumption and intake of total flavonoids, TRAP and FRAP (with and without incorporation of coffee data). Cohort study Multivariable Cox proportional hazard models were used to evaluate the association of intake of total and subclasses of flavonoids (including flavanols [flavan-3-ol monomers, proanthocyanidins and theaflavins], anthocyanidins, flavonols, flavanones, flavones and isoflavones), lignans, FRAP and TRAP with risk of HCC. Tests and graphs based on Schoenfeld residuals were used to assess the proportional hazards assumption. Age was used as the primary time variable, and the Breslow method was adopted for handling ties. 39 Time at entry into the study was age at recruitment; exit time was age at cancer diagnosis, death, loss to follow-up or end of follow-up, whichever occurred first. Crude models were run stratified by EPIC-participating center, to account for center effects related to different recruitment and followup procedures, by sex and by age at recruitment in 1-year categories to reduce sensitivity to any violations of the proportional hazards assumption and adjusted for total energy intake (kcal/day). Multivariable models were additionally adjusted for body mass index (BMI; kg/m 2 ), self-reported diabetes (yes, no and not specified), educational level (no formal education, primary, technical/professional school, secondary, university and not specified), smoking status/ duration/intensity (never, former from <11 years, former from 11 to 20 years, former from >20 years, current 1 15 cigarettes/day, current cigarettes/day, current >25 cigarettes/day, current occasional and unknown), sex-specific physical activity level (inactive, moderately inactive, moderately active, active and unknown), alcohol intake at recruitment (g/day), pattern of lifetime alcohol intake (never drinkers, former light drinker, former heavy drinkers, light drinkers, never heavy drinkers, periodically heavy drinkers, always heavy drinkers and unknown), dietary fiber (g/day) and coffee (ml/day). Multivariate Cox models evaluating FRAP and TRAP including coffee data were not adjusted for coffee intake. Flavonoid, lignan, FRAP and TRAP intakes were analyzed as categorical variables based on tertiles of the distribution among the entire EPIC cohort. Tests for linear trend were performed by assigning the medians of each tertile as scores. Moreover, flavonoid, lignan, TRAP and FRAP intakes were analyzed as continuous, after a log 2 transformation that indicates a doubling in dietary intakes. Interactions between flavonoid, TRAP and FRAP (with and without coffee data) intakes and sex, BMI, tobacco smoking and alcohol consumption were also evaluated in separate analyses. The statistical significance of interactions was assessed using likelihood ratio tests based on the models with and without the interaction terms. Sensitivity analyses

5 Zamora-Ros et al Table 1. Distribution of subjects and hepatocellular carcinoma (HCC) cases during the 11 years of follow-up in the 10 countries participating in the EPIC study Country Cohort size Person-years No of HCC cases France 67, ,125 3 Italy 44, , Spain 39, ,614 9 United Kingdom 75, , The Netherlands 36, ,852 4 Greece 26, , Germany 48, , Sweden 48, , Denmark 54, , Norway 35, ,086 2 Total 477,206 5,415, were performed by repeating the models after the exclusion of HCC cases diagnosed during the first 2 years of follow-up (n 5 26) or excluding the self-reported cases of diabetes at recruitment (n 5 12,518) and subjects with missing data on diabetes at baseline (n 5 39,188). Nested case control study Conditional logistic regression was used to estimate the associations between HCC risk and dietary flavonoids, lignan, FRAP and TRAP intakes. Crude conditional logistic models, with matching factors, were only adjusted for total energy intake (kcal/day). Multivariable models were additionally adjusted for the same confounders as described above for the cohort analyses. These analyses were conducted on HBV/ HCV negative individuals. All statistical tests were two-sided, and p-values < 0.05 were considered statistically significant. p-values between 0.05 and 0.08 were considered borderline significant. All statistical analyses were conducted using SAS version 9.2 software (SAS Institute, NC). RESULTS The final cohort of 477,206 subjects (29.8% men) was followed for an average of 11.4 (standard deviation, SD 2.8) years with a total accumulation of 5,415,385 person-years. During this period, 191 incident cases of HCC (66.5% men) (Table 1) were diagnosed. Table 2 shows the mean (SD), median, percentiles (5th and 95th) and main food sources of both total and subclasses of flavonoids, lignans. As indicated by the large differences between means and medians, the distributions of flavonoid, lignan, FRAP and TRAP intakes were skewed to higher values. Flavanols were the most important contributor (74%) to total flavonoid intake (proanthocyanidins 54%, flavan-3-ols monomers 19%, theaflavins 1%), followed by flavanones (8.8%), anthocyanidins (8.7%) and flavonols (7.3%). The main food sources of the total flavonoids were fruits (39.5%), tea (19.1%), wine (11.9%), fruit juices (5.7%) and vegetables (3.3%), whereas, as expected, the main source of FRAP and TRAP was coffee (71 and 77%, respectively) (Table 2). However, without considering coffee, tea ( %), fruits ( %) and vegetables ( %) were the main dietary source for FRAP and TRAP. Participants in the highest tertile of total flavonoid intakes were more likely to be older, nonsmoker, more physically active, and to have a lower BMI, less diabetes and a higher schooling level, as well as a higher intake of energy, alcohol and fiber compared with participants in the lowest tertile (Table 3). In contrast, participants in the top tertile of TRAP intake were more likely to be current smokers, and to have a higher coffee consumption than subjects in the first tertile. The characteristics of participants according tertiles of FRAP intake were similar to the TRAP intake (data not shown). Total flavonoid intake was moderately correlated with both FRAP (Spearman s q correlation ) and TRAP intake without coffee data (q ), but weakly with FRAP and TRAP with coffee (q < 0.10). Coffee consumption were strongly correlated to TRAP and FRAP intakes including coffee data (q > 0.98), but not to TRAP and FRAP intakes without coffee data (q < 0.10). The multivariable adjusted HRs for risk of HCC according to total and subclasses of flavonoid, lignan, FRAP and TRAP intakes were explored in the cohort study (Table 4). In models stratified by center, sex and age and only adjusted for energy intake, a statistically significant inverse association between total flavonoid intake and the HCC risk was found for the highest when compared to the lowest tertile of intake (HR 0.63, 95% CI: ; p for trend (p trend ) 0.036). In the multivariable model, this association was borderline nonsignificant (HR 0.65, 95% CI: ; p trend 0.065), although the association using the continuous variable, after a log 2 transformation, was not statistically significant. Flavanols was the only flavonoid subclass significantly and inversely associated with HCC risk comparing extreme tertiles, but not using the continuous variable. Inverse trends were also observed between risk of HCC and flavan-3-ol monomers and theaflavins. For total FRAP and TRAP intake, significant inverse associations were found with the multivariable adjusted HCC risk comparing extreme tertiles (HR 0.50, 95% CI: ; p trend 0.001; and HR 0.49, 95% CI: , respectively). In models stratified by center, sex and age and only adjusted for energy intake, a significant inverse association between TRAP intake excluding coffee data and the risk of HCC was observed, but not in the multivariable model. We did not observe any statistically significant interactions of the dietary exposures and HCC risk by sex, BMI, smoking status and baseline alcohol intake (Table 5). In a sensitivity analysis, in which 26 HCC cases diagnosed within the first 2 years of follow-up were removed, the inverse associations between total flavonoids and their subclasses and lignan intakes were almost identical to the results

6 2434 Flavonoids, antioxidant capacity and hepatocellular cancer Table 2. Dietary intake 1 of flavonoids, lignans, TRAP and FRAP and their main food sources in the EPIC study Mean Standard Deviation Median Percentile 5 Percentile 95 Main food sources 2 Total flavonoids ,138.2 Fruits (39.5%), tea (19.1%), wines (11.9%), fruit juices (5.7%) Flavanols ,013.4 Tea (44.1%), fruit (29.1%), wines (8.8%), chocolate products (3.8%) Flavan-3-ols monomers Tea (83.9%), fruits (5.9%), wines (3.4%), chocolate products (1.8%) Proanthocyanidins Fruits (52.6%), wines (14.1%), chocolate products (5.9%), tea (4.0%) Teaflavins Tea (100%) Anthocyanidins Fruits (52%), wines (20.8%), vegetables (8.1%), fruit juices (6.6%) Flavonols Tea (25.9%), vegetables (23.0%), fruit (13.4%), soups and bouillons (12.2%) Flavanones Fruits (50.2%), fruit juices (40.9%), wines (4.8%), vegetables (1.1%) Flavones Herbal tea (29.5%), wine (17.8%), fruits (16.4%), vegetables (15.1%) Isoflavones Soya products (40.2%), cakes and sweets (18.4%), bread and cereals (11.0%), coffee (7.8%) Lignans Vegetables (23.6%), fruits (16.5%), bread and cereals (16.4%), tea (10.1%) FRAP 59,837 45,362 48,021 11, ,485 Coffee (71.0%), tea (6.7%), fruits (5.4%), vegetables (4.1%) TRAP 28,233 22,839 22,345 4,012 69,989 Coffee (77.0%), tea (6.5%), fruits (4.0%), vegetables (3.3%) FRAP without ,699 17,869 Tea (23.0%), fruits (18.6%), vegetables (14.0%), wines (9.4%) coffee data TRAP without coffee data ,364 6,378 Tea (28.3%), fruits (17.5%), vegetables (14.2%), wines (12.6%) 1 Intake of flavonoids and lignans in mg/day, TRAP in mmol Trolox equivalents/day and FRAP in mmol Fe 21 equivalents/day. 2 The first 4 main food sources in each food group are given; a lower number of foods indicates the absence of other food sources that contained polyphenols in the food group considered. Abbreviations: EPIC: European Prospective Investigation into Cancer and Nutrition; FRAP: ferric reducing antioxidant capacity; TRAP: total radical-trapping antioxidant parameter. based on the whole cohort (Supporting Information Table 1). However, the results for TRAP and FRAP, especially with coffee data, were attenuated and became statistically nonsignificant (for FRAP HR for log , 95% CI: ; and for TRAP HR for log , 95% CI: ). In another sensitivity analysis, 12,518 self-reported diabetics and 39,188 subjects with missing data on diabetes at baseline were excluded resulting in an attenuation of the associations for all dietary exposures of interest and they became not significant (Supporting Information Table 1). In the nested case control study, 122 HCC cases were diagnosed, on average, 5 years (standard deviation 5 2.9) after recruitment. Thirty-one percent of HCC cases and only 4% of matched controls had either an HBV or HCV infection, or both (Supporting Information Table 2). In the current analyses, we included 84 case control sets (84 cases and 162 controls), of these, 78 case control sets relied on two matched control subjects, whereas six set relied on one matched control subject because the second control subject was excluded due to a HBV and/or HCV infection. Table 6 shows the associations between HCC risk and flavonoid, lignan, TRAP and FRAP intakes limited to HBV- and HCVnegative subjects. Overall, the results for HBV/HCV-free participants were similar in direction and magnitude to those reported for the entire cohort, except dietary isoflavones and lignans that were borderline nonsignificantly associated with HCC risk in the nested case control study (p trend and for isoflavones and lignans, respectively), but not in the entire cohort (p trend 0.83 and 0.17 for isoflavones and lignans, respectively). DISCUSSION The results from this large cohort of participants from 10 Western European countries suggest that higher dietary intakes of flavanols and antioxidants may be associated with a decreased HCC risk. Moreover, similar results were observed in the nested case control study taking into account hepatitis infection status in the analysis.

7 Table 3. Baseline characteristics according to tertiles of intake of total flavonoids and TRAP Total flavonoids 4 TRAP 5 Baseline characteristics Tertile 1 Tertile 2 Tertile 3 Tertile 1 Tertile 2 Tertile 3 Number of participants 159, , , , , ,666 Men (%) Age at enrollment (years) (9.3) 51.0 (9.6) 51.7 (10.8) 50.6 (10.6) 51.1 (10.1) 51.9 (8.9) No. with diabetes (%) BMI (kg/m 2 ) (4.4) 25.6 (4.3) 25.0 (4.1) 25.7 (4.6) 25.1 (4.1) 25.4 (4.0) Educational level (%) None Primary school Technical/professional Secondary school University or higher Not specified Physical activity (%) 3 Inactive Moderately inactive Moderately active Active Unknown Smoking status and intensity (%) Never Current, 1 15 cig/day Current, cig/day Current, >25 cig/day Former, quit <11 years Former, quit years Former, quit >20 years Current, occasional Unknown Alcohol lifetime pattern (%) Never drinkers Former light drinkers Former heavy drinkers Light drinkers Never heavy drinkers Periodically heavy drinkers Always heavy drinkers Unknown Daily dietary intake 1 Total energy (kcal/day) 1870 (562) 2143 (601) 2209 (639) 2015 (602) 2066 (621) 2142 (628) Alcohol (g/day) 8.4 (13.4) 12.2 (16.8) 14.2 (19.3) 8.9 (14.3) 12.0 (17.5) 13.8 (18.1) Fiber (g/day) 19.2 (6.1) 23.3 (6.7) 25.9 (8.6) 22.6 (7.8) 22.6 (7.5) 23.2 (7.8) Coffee (ml/day) 468 (403) 362 (350) 323 (332) 59 (56) 300 (118) 795 (329) Total flavonoids (mg/day) 155 (55) 341 (64) 818 (310) 447 (323) 469 (348) 398 (331) TRAP (mmol Trolox equivalents/day) (25759) (21984) (19905) 8155 (3418) (5507) (21014) FRAP (mmol Fe 21 equivalents/day) (51076) (43788) (39780) (7221) (11421) (41168) 1 Mean (SD). 2 Self-reported data. Number of participants with missing data: For total flavonoids T1:6.6%; T2: 3.5%; T3: 14.6%; for TRAP T1: 5.7%; T2: 9.6%; T3: 9.4%. 3 Physical activity categories were sex-specific. 4 Tertiles of total flavonoid intake (mg/day): T1,<242.6; T2, to 472.6; T3, > Tertiles of TRAP intake (mmol Trolox equivalents/day): T1,<14,108; T2, 14,108 to 32,840; T3, > Abbreviations: FRAP: ferric reducing antioxidant capacity; TRAP: total radical-trapping antioxidant parameter.

8 2436 Flavonoids, antioxidant capacity and hepatocellular cancer Table 4. Hazard ratios and 95% confidence intervals for hepatocellular carcinoma, by tertiles of flavonoids, lignans, TRAP and FRAP in the EPIC study Hepatocellular carcinoma Crude 2 Multivariable 3 Intake 1 No of cases HR (95% CI) HR (95% CI) Total flavonoids Tertile 1 < (ref) 1.00 (ref) Tertile ( ) 0.91 ( ) Tertile 3 > ( ) 0.65 ( ) p trend Continuous (log 2 ) 0.92 ( ) 0.97 ( ) Flavanols Tertile 1 < (ref) 1.00 (ref) Tertile ( ) 0.76 ( ) Tertile 3 > ( ) 0.62 ( ) p trend Continuous (log 2 ) 0.92 ( ) 0.97 ( ) Flavan-3-ol monomers Tertile 1 < (ref) 1.00 (ref) Tertile ( ) 0.97 ( ) Tertile 3 > ( ) 0.65 ( ) p trend Continuous (log 2 ) 0.93 ( ) 0.93 ( ) Proanthocyanidins Tertile 1 < (ref) 1.00 (ref) Tertile ( ) 0.98 ( ) Tertile 3 > ( ) 1.10 ( ) p trend Continuous (log 2 ) 0.99 ( ) 1.13 ( ) Theaflavins Tertile (ref) 1.00 (ref) Tertile 2 > ( ) 1.06 ( ) Tertile 3 > ( ) 0.60 ( ) p trend Continuous (log 2 ) 0.96 ( ) 0.96 ( ) Anthocyanidins Tertile 1 < (ref) 1.00 (ref) Tertile ( ) 0.95 ( ) Tertile 3 > ( ) 1.20 ( ) p trend Continuous (log 2 ) 0.94 ( ) 1.04 ( ) Flavonols Tertile 1 < (ref) 1.00 (ref) Tertile ( ) 1.05 ( ) Tertile 3 > ( ) 0.79 ( ) p trend Continuous (log 2 ) 0.90 ( ) 0.91 ( )

9 Zamora-Ros et al Table 4. Hazard ratios and 95% confidence intervals for hepatocellular carcinoma, by tertiles of flavonoids, lignans, TRAP and FRAP in the EPIC study (Continued) Hepatocellular carcinoma Crude 2 Multivariable 3 Intake 1 No of cases HR (95% CI) HR (95% CI) Flavanones Tertile 1 < (ref) 1.00 (ref) Tertile ( ) 0.86 ( ) Tertile 3 > ( ) 1.10 ( ) p trend Continuous (log 2 ) 0.97 ( ) 1.03 ( ) Flavones Tertile 1 < (ref) 1.00 (ref) Tertile ( ) 0.94 ( ) Tertile 3 > ( ) 1.06 ( ) p trend Continuous (log 2 ) 0.96 ( ) 1.00 ( ) Isoflavones Tertile 1 < (ref) 1.00 (ref) Tertile ( ) 0.60 ( ) Tertile 3 > ( ) 0.74 ( ) p trend Continuous (log 2 ) 0.79 ( ) 0.86 ( ) Lignans Tertile 1 < (ref) 1.00 (ref) Tertile ( ) 0.94 ( ) Tertile 3 > ( ) 1.37 ( ) p trend Continuous (log 2 ) 0.89 ( ) 1.14 ( ) FRAP Tertile 1 < (ref) 1.00 (ref) Tertile ( ) 1.04 ( ) Tertile 3 > ( ) 0.50 ( ) p trend Continuous (log 2 ) 0.77 ( ) 0.74 ( ) TRAP Tertile 1 < (ref) 1.00 (ref) Tertile ( ) 0.92 ( ) Tertile 3 > ( ) 0.49 ( ) p trend Continuous (log 2 ) 0.79 ( ) 0.76 ( ) FRAP without coffee data Tertile 1 < (ref) 1.00 (ref) Tertile ( ) 1.15 ( ) Tertile 3 > ( ) 1.06 ( ) p trend Continuous (log 2 ) 0.99 ( ) 1.12 ( )

10 2438 Flavonoids, antioxidant capacity and hepatocellular cancer Table 4. Hazard ratios and 95% confidence intervals for hepatocellular carcinoma, by tertiles of flavonoids, lignans, TRAP and FRAP in the EPIC study (Continued) Hepatocellular carcinoma Crude 2 Multivariable 3 Intake 1 No of cases HR (95% CI) HR (95% CI) TRAP without coffee data Tertile 1 < (ref) 1.00 (ref) Tertile ( ) 0.82 ( ) Tertile 3 > ( ) 0.77 ( ) p trend Continuous (log 2 ) 0.74 ( ) 0.84 ( ) 1 Intake of flavonoids and lignans in mg/day, TRAP in mmol Trolox equivalents/day, and FRAP in mmol Fe 21 equivalents/day. 2 Stratified by center, sex, age and adjusted for total energy intake. 3 Stratified by center, sex, age and adjusted for total energy, educational level, smoke intensity, alcohol lifetime and alcohol baseline, BMI, selfreported diabetes at baseline, physical activity, fiber and coffee intake. Models evaluating FRAP and TRAP including coffee data were not adjusted for coffee intake. Table 5. Adjusted 1 hazard ratios and 95% confidence intervals for hepatocellular carcinoma for a doubling of total flavonoid and TRAP intakes 2 in the EPIC study Total flavonoids TRAP No of cases HR (95% CI) p for interaction HR (95% CI) p for interaction Sex Men ( ) 0.72 ( ) Women ( ) 0.91 ( ) BMI (kg/m 2 ) < ( ) 1.00 ( ) ( ) 0.64 ( ) > ( ) 0.79 ( ) Smoking status Nonsmokers ( ) 0.73 ( ) Former smokers ( ) 0.73 ( ) Current smokers ( ) 0.82 ( ) Alcohol consumption None to low ( ) 0.84 ( ) Moderate ( ) 0.82 ( ) High ( ) 0.60 ( ) Former drinkers ( ) 1.20 ( ) 1 Stratified by center, sex, age and adjusted for total energy, educational level, smoke intensity, alcohol lifetime and alcohol baseline, BMI, selfreported diabetes at baseline, physical activity, fiber and coffee intake. Models evaluating TRAP including coffee data were not adjusted for coffee intake. 2 Intake of flavonoids in mg/day, TRAP in mmol Trolox equivalents/day. 3 Alcohol consumption: None to low: men (0 to <10) g/day and women (0 to <5) g/day; Moderate: men (10 to <40) g/day and women (5 to <20) g/day; High: men (40) g/day and women (20) g/day. To our knowledge, there is only one case control study published from Greece evaluating the association between total flavonoid intake and HCC cancer risk, by hepatitis status. 19 The study showed that flavones and their main food sources (spinach and peppers) were significantly and inversely associated with HCC risk. However, in our cohort study, total flavonoids and flavanols (flavan-3-ol monomers and theaflavins), but not flavones, were inversely related to HCC. Moreover, flavones contributed to 0.4 and 1.3% of the total flavonoid intake in the Greek study 19 and EPIC cohort, 20 respectively. In a recent review on dietary factors and liver cancer, Chuang et al. suggested that there is a potential

11 Zamora-Ros et al Table 6. Odds ratios and 95% confidence intervals for hepatocellular carcinoma, by tertiles of dietary intake of flavonoids, lignans, FRAP and TRAP equivalents among HBV and HCV free subjects, within the EPIC nested case control study Crude 1 Multivariable 2 No of cases No of controls OR (95% CI) OR (95% CI) Total flavonoids Tertile (ref) 1 (ref) Tertile ( ) 0.75 ( ) Tertile ( ) 0.29 ( ) p trend Continuous (log 2 ) 0.86 ( ) 0.81 ( ) Flavanols Tertile (ref) 1 (ref) Tertile ( ) 0.92 ( ) Tertile ( ) 0.27 ( ) p trend Continuous (log 2 ) 0.90 ( ) 0.86 ( ) Flavan-3-ol monomers Tertile (ref) 1 (ref) Tertile ( ) 1.40 ( ) Tertile ( ) 0.58 ( ) p trend Continuous (log 2 ) 0.91 ( ) 0.81 ( ) Proanthocyanidins Tertile (ref) 1 (ref) Tertile ( ) 0.59 ( ) Tertile ( ) 1.04 ( ) p trend Continuous (log 2 ) 1.00 ( ) 1.17 ( ) Theaflavins Tertile (ref) 1 (ref) Tertile ( ) 1.37 ( ) Tertile ( ) 0.67 ( ) p trend Continuous (log 2 ) 0.93 ( ) 0.94 ( ) Anthocyanidins Tertile (ref) 1 (ref) Tertile ( ) 2.08 ( ) Tertile ( ) 0.76 ( ) p trend Continuous (log 2 ) 0.88 ( ) 1.03 ( ) Flavonols Tertile (ref) 1 (ref) Tertile ( ) 0.82 ( ) Tertile ( ) 0.70 ( ) p trend Continuous (log 2 ) 0.95 ( ) 0.73 ( )

12 2440 Flavonoids, antioxidant capacity and hepatocellular cancer Table 6. Odds ratios and 95% confidence intervals for hepatocellular carcinoma, by tertiles of dietary intake of flavonoids, lignans, FRAP and TRAP equivalents among HBV and HCV free subjects, within the EPIC nested case control study (Continued) Crude 1 Multivariable 2 No of cases No of controls OR (95% CI) OR (95% CI) Flavanones Tertile (ref) 1 (ref) Tertile ( ) 0.68 ( ) Tertile ( ) 0.84 ( ) p trend Continuous (log 2 ) 0.89 ( ) 0.89 ( ) Flavones Tertile (ref) 1 (ref) Tertile ( ) 0.83 ( ) Tertile ( ) 0.59 ( ) p trend Continuous (log 2 ) 0.91 ( ) 0.82 ( ) Isoflavones Tertile (ref) 1 (ref) Tertile ( ) 0.23 ( ) Tertile ( ) 0.16 ( ) p trend Continuous (log 2 ) 0.80 ( ) 0.62 ( ) Lignans Tertile (ref) 1 (ref) Tertile ( ) 0.44 ( ) Tertile ( ) 0.19 ( ) p trend Continuous (log 2 ) 0.61 ( ) 0.29 ( ) FRAP Tertile (ref) 1 (ref) Tertile ( ) 0.59 ( ) Tertile ( ) 0.72 ( ) p trend Continuous (log 2 ) 0.80 ( ) 0.78 ( ) TRAP Tertile (ref) 1 (ref) Tertile ( ) 0.47 ( ) Tertile ( ) 0.62 ( ) p trend Continuous (log 2 ) 0.81 ( ) 0.81 ( ) FRAP without coffee data Tertile (ref) 1 (ref) Tertile ( ) 0.66 ( ) Tertile ( ) 0.51 ( ) p trend Continuous (log 2 ) 1.03 ( ) 0.64 ( )

13 Zamora-Ros et al Table 6. Odds ratios and 95% confidence intervals for hepatocellular carcinoma, by tertiles of dietary intake of flavonoids, lignans, FRAP and TRAP equivalents among HBV and HCV free subjects, within the EPIC nested case control study (Continued) Crude 1 Multivariable 2 No of cases No of controls OR (95% CI) OR (95% CI) TRAP without coffee data Tertile (ref) 1 (ref) Tertile ( ) 0.64 ( ) Tertile ( ) 0.49 ( ) p trend Continuous (log 2 ) 0.61 ( ) 0.63 ( ) 1 Adjusted for total energy. 2 Adjusted for educational level, smoke status (never, former, current, unknown), alcohol lifetime and alcohol baseline, BMI, self-reported diabetes at baseline, physical activity, total energy, fiber and coffee intake. Models evaluating FRAP and TRAP including coffee data were not adjusted for coffee intake. Abbreviations: FRAP: ferric reducing antioxidant capacity; TRAP: total radical-trapping antioxidant parameter. protective effect from fresh fruit, whereas the evidence for vegetables is not conclusive. 3 It is important to highlight that the main food sources of total flavonoids in EPIC were fruits (44%), followed by tea (19.1%) and wines (11.9%), while vegetables only accounted for 3.3% of total flavonoids. There was no evidence of heterogeneity either by sex, BMI, tobacco smoking or alcohol consumption, although these interaction analyses had low statistical power due to the small number of HCC cases in each category. In the nested case control study, participants infected with HCV/HBV, the main risk factor for HCC, were excluded. 4 The effect estimates were similar to those reported in all-cohort analyses, but the confidence intervals were wider, and the results were attenuated or lost the significance. In our cohort study, no statistically significant association between isoflavone intake and HCC risk was observed. A Japanese cohort-based, case control study has shown an inverse relationship between soy consumption, the main food source of isoflavones and HCC. 6 In contrast, a Japanese cohort study has found a significant positive association between genistein or daidzein and HCC in women, but not among men. 40 However, the consumption of isoflavones in the Japanese population on average is 25 times higher than in the EPIC cohort. 25 NEAC intake, measured by TRAP and FRAP, was inversely associated with HCC in the EPIC cohort study. However, when the HCC cases diagnosed in the first 2 years of the follow-up were excluded, the association with HCC risk became attenuated. Therefore, the possibility of reverse causality cannot be excluded. Participants with undiagnosed HCC might decrease their consumption of coffee due to nonspecific HCC symptoms, such as loss of appetite and pain or discomfort in the abdomen. Dietary FRAP and TRAP intakes were highly correlated with coffee consumption because coffee contributed to >70% of both dietary FRAP and TRAP. A pooledanalysis has shown a potentially favorable effect of coffee consumption on liver cancer risk. 5 In our study, we were not able to differentiate between the effect of coffee and the effect of FRAP or TRAP. Caffeine, phenolic acids and/or other contributors of TAC could be the potential compounds that are driving this protective effect of coffee on liver cancer risk, although further research on this topic is needed. Maillard products (melanoidins) are the main contributors to the NEAC of coffee in vitro, 41 but their bioavailability is still unclear but, due to the large molecular weight, very limited. 42 For this reason, we also examined the association between NEAC excluding coffee data and HCC risk. TRAP intake without coffee was borderline significantly and inversely related to HCC, but not FRAP. TRAP and FRAP were only moderately correlated with total flavonoids. TRAP is a measure of the ability to scavenge peroxyl radicals, whereas FRAP is an indicator of the ability to reduce iron. Therefore, these results suggest that nonenzymatic antiradical compounds might be more related to a reduction on HCC risk than nonenzymatic iron reducing activity, although coffee seems to be the main important factor in the inverse association between both FRAP and TRAP and HCC risk. Interestingly, an experimental study has demonstrated that coffee consumption protects the liver from damage caused by a high-fat diet through a reduction in hepatic fat accumulation, in systemic and liver oxidative stress and in liver inflammation. 43 Furthermore, after the exclusion of the self-reported cases of diabetes at baseline, the associations between FRAP and TRAP and HCC risk also became nonsignificant. Therefore, a possible mediating effect of diabetes on this association may not be excluded, because diabetes is associated with oxidative stress 44 and low levels of plasma antioxidants are more pronounced in diabetic subjects than nondiabetic subjects. 45 In a recent review, Hollman et al. 46 suggested that the evidence for any direct antioxidant properties of polyphenols in vivo is questionable, particularly against cardiovascular diseases, although phenolic compounds might also indirectly up-regulate antioxidant pathways, by inducing antioxidant