FLAVONOIDS, PROANTHOCYANIDINS AND CANCER RISK. STATISTICAL METHODS TO DISENTANGLE THE EFFECTS OF FLAVONOIDS, THEIR SOURCES AND OTHER DIETARY COMPOUNDS ON CANCER RISK.

A diet rich in vegetables and fruit has been associated with a reduced risk of various common cancers, particularly of the respiratory and digestive tracts. Further investigations have tried to understand whether such a favorable effect may be attributed to specific vitamins, micronutrients or bioactive compounds contained in plant foods. Flavonoids - a large group of phytochemicals with a similar structure, naturally occurring in vegetables, fruit, and beverages of plant origin - have shown antioxidant, antimutagenic, and antiproliferative properties in vitro, and have thus been suggested to have a potential protective effect on common cancers. Given the high correlation between dietary factors, it is difficult to disentangle the effect of various factors and detect which specific component or group of components is the responsible for a protective role against cancer. Applications of innovative statistical methods are needed to address this problem. Moreover, since several antioxidants may influence cancer risk and act synergistically against oxidative stress to prevent carcinogenesis, examining overall antioxidant exposure rather than individual antioxidants has also recently been proposed. I investigated the role of dietary flavonoids - including proanthocyanidins – and total antioxidant capacity (TAC) on cancer risk using data from a network of multicentric Italian case-control studies including about 10 000 incident, histologically confirmed cases of selected cancers, and over 16 000 hospital controls. For this purpose, I developed a standardized method based on flavonoid food composition tables published by the US Department of Agriculture (USDA) in order to estimate the intakes of six major classes of flavonoids, i.e., flavanols, flavanones, flavonols, anthocyanidins, flavones, and isoflavones, and of proanthocyanidins (by their degree of polymerization) from a reproducible and valid food frequency questionnaire (FFQ) of each subject. Similarly, I quantified the TAC of the diet by computing TEAC (Trolox Equivalent Antioxidant Capacity), TRAP (Total Radical-trapping Antioxidant Parameter), and FRAP (Ferric Reducing-Antioxidant Power) through Italian food tables in terms of these three assays, recently published from the National Institute for Food and Nutrition. Odds ratios (ORs) contrasting the highest to the lowest quintiles and tertiles of the variables of interest were estimated by multiple logistic regression models, including major confounding factors for each cancer. Adjustment for energy intake was performed entering the term in the model (with or without energy from alcohol intake) and using the residual method I then applied factor analysis and residual method in order to deal with the problems of collinearity between variables. In particular, I worked on the collinearity between: 1) classes of flavonoids, 2) flavonoids and wine, 3) flavonoids and other bioactive compounds contained in plant foods. At last, I used a new approach of the residual method to take into account the over-reporting of cases. I found evidence of a protective role of flavanones on upper aerodigestive tract cancers, flavonols, proanthocyanidins on stomach and pancreatic cancers, anthocyanidins and proanthocyanidins on colorectal cancer, flavonols and flavones on breast cancer, isoflavones on ovarian cancer, and flavonols on renal cancer. For most investigated neoplasms, adjustment for flavonoids reduced the strength of the inverse association between vegetables or fruit consumption and the risk of cancer, whereas allowance for fruit and vegetables consumption only moderately changed the observed associations with flavonoids. Misclassification may play a role, but it appears that a diet rich in fruit and vegetables does not alone account for the protections of flavonoids on the risk of several cancer sites, whereas the inverse relation of cancer with fruit and vegetables is not totally explained by flavonoid intake. With reference to TAC, the analysis on colorectal cancer found that TAC was inversely associated to the risk of this tumor. However, a diet rich antioxidants does not alone account for the protections of flavonoids on the risk of colorectal cancer. FA allowed to explain the variance of the entire structure of flavonoids trough five factors, and to compute the ORs of these factors for colorectal cancer. I found that three factors were inversely related to colorectal cancer risk taking into account also the possible confounding effect of all other flavonoids. These analyses highlight the importance to investigate individual flavonoids or groups of flavonoids besides their classes, and allowed to resolve – in part – the problem of the strong confounding effect of wine consumption. Moreover, these results suggest to include flavonoids and TAC in the estimation of dietary patterns in nutritional epidemiologic research, since no study have investigated specifically flavonoids, neither included flavonoids in their analyses up to now. When I computed the residuals on TAC, water, and fruit and vegetables, I found that the residual method on TAC gave the same results of the residuals on water for oral and pharyingeal cancer risk. In both analysis in fact, the relations for flavanones and flavonols persisted whereas the association for proanthocyanidins disappeared, providing evidence of a protective effect of flavanones and flavonols on oral and pharyngeal cancer. The proposal to use residuals as adjustment for over-reporting of cases in case-control studies leaves open an epidemiological question, given the criticism that the case-control studies still face in the epidemiological literature, but can be of help since new statistical methods are needed to resolve these limitations.