SYSTEMATIC POPULATION-BASED ASSESSMENT OF CANCER RISK IN FIRST-DEGREERELATIVES OF CANCER PROBANDS

Background: Cancer has long been recognized to have a familial compone nt. Elevated risks for cancers at the same site for relatives of cance r probands have been reported for both common cancers and a number of the rarer cancer sites. For a particular cancer site, however, the est imated risks to relatives have varied considerably depending on criter ia for selection of probands, how cancers were determined in relatives , and overall study design. Not surprisingly, the estimated risks of o ther cancers in relatives of probands with cancer at a given site have been subject to even more variation. Purpose: The aim of this study w as to use the Utah Population Database resource to systematically stud y familial clustering of 28 distinct cancer site definitions among fir st-degree relatives (parents, siblings, and offspring) of cancer proba nds. Methods: We estimated familial relative risks from the Utah Popul ation Database by identifying all cases of cancer in these first-degre e relatives. These observed values were compared with those expected b ased on cohort-specific internal rates calculated from 399 786 relativ es of all individuals in the Utah Population Database known to have di ed in Utah. Results: All sites showed an excess of cancers of the same site among relatives, with thyroid and colon cancers and lymphocytic leukemia showing the highest familial risks. When the analyses were re stricted to cases with early ages at diagnosis, increased familial com ponents for most cancer sites became evident. A significant difference in familial relative risk (FRR) between male (FRR = 4.04; 95% confide nce interval [CI] = 3.13-5.07) and female (FRR = 2.24; 95% CI = 1.54-3 .08) probands was found for colon cancer. Highly significant familial associations (one-sided; P<.001) were found among breast, colon, and p rostate cancers and between breast and thyroid cancers. Statistically significant (one-sided, P<.01) associations were also found between to bacco-associated sites (lung, larynx, lip, and cervix). Conclusions: T his study represents a unique comprehensive population-based study of familial cancer. The familial associations reported here will be usefu l in generating hypotheses about specific genetic and environmental fa ctors that can be tested in genetic linkage and case-control studies.