FAMILY HISTORY OF CANCER AND FAMILY HISTORY SCORES FOR ASSESSING THE LEVEL OF DISEASE RISK IN FAMILIES

In the last decades genetic factors are playing an increasingly important role in medical research, given the evidence for the existence of a heritable susceptibility for various diseases, including common cancers, based on reports of families with multiple affected relatives. Epidemiologists have utilized family history, usually of first-degree relatives, as a surrogate for genetic risk, aware that family history reflects the consequences of genetic susceptibilities, shared environment, and common behaviors. During my PhD I have dealt with two different aspects of family history, i.e., the role of family history of cancer in epidemiological cancer research (Chapter 1) and the use of complex family history score for assessing the level of disease risk in families (Chapter 2). In particular, I have systematically examined the extent to which a family history of cancer might be a risk factor for cancer within the same cancer site and across multiple cancer sites, analyzing a large and comprehensive dataset based on a network of integrated case-control studies, conducted in Italy and Switzerland since the early 90's. The database included 1468 cases of cancer of the oral cavity and pharynx, 198 of the rhinopharynx, 505 of the esophagus, 230 of the stomach, 2390 of the colorectum, 185 of the liver, 326 of the pancreas, 852 of the larynx, 3034 of the breast, 367 of the endometrium, 1031 of the ovary, 1294 of the prostate, 767 of the renal cell, and a total of 16022 corresponding controls. Unconditional multiple logistic regression models, adjusted for the major possible confounding factors, and a procedure for controlling for multiplicity using a false discovery rate were used. The risk of developing cancer at a particular site was increased, although not always significantly, in subjects with a first-degree relative affected by cancer at the same site, with odds ratios ranging from 1.4 for pancreatic cancer, to 7.4 for ovarian cancer. Several across sites associations emerged, some of which possibly due to shared environmental exposures or lifestyle practices among family members (e.g., alcohol, smoking, unhealthy diet, infections) or to the inheritance of one or more predisposing gene mutations (high penetrance gene mutations, such as BRCA1/2 in breast and ovarian cancer, and/or low penetrance polymorphisms, as those involved in carcinogens metabolism, such as GST genes in oral cancer) or to a combination of both. The analysis I performed confirmed that several associations were stronger for a younger age at diagnosis in relatives. A detailed discussion of the findings is reported in paragraph 4 of Chapter 1. In addition to the investigation of the role of family history of cancer in cancer etiology, I have performed a statistical evaluation of the performance of different family history scores to recommend the measure that performs best. Family history scores summarize familial information and are used for estimating the familiar risk, i.e. the level of risk for a particular disease among members of that family. The simplest and most common family history scores are the dichotomous measure indicator, positive in families that have at least one relative with the disease, the number of affected family members, and the proportion of affected relatives, which takes into account the size of the family. The other family history scores proposed in the literature are statistics that describe the deviation of the observed situation from the expected risk for each family. More detailed information on family members (affected and unaffected) as well as incidence rates of the diseases of interest in strata of selected covariates are needed to compute these more complex family history scores. To evaluate family history scores’ performance I used two different complementary approaches: a data-derived approach, using data from the Italian HI-WATE study, with the aim of examining the power of various family history scores in predicting a particular diseases (i.e., colorectal cancer), and a simulation approach to evaluate their accuracy of predicting the true familial risk. From 200 simulations for 48 different settings, Reed’s score and FHS2 seem to perform slightly better than the other scores. However, the simple proportion of affected relatives is not so far in terms of predictivity of the true familial risk. The use of this simple score seems therefore justified, at least until stronger evidence is brought for the advantages of using a more complex score.