Attributable risk for familial breast cancer

In a comprehensive study of the Swedish Family Cancer Database (including 10.2 million individuals, with 190,000 mothers' and 26,000 daughters' breast cancers), Hemminki et al.1 estimated a population-attributable fraction (PAF) of 7.05% of all breast cancers (3.6% for mother history, 3.0% for sister, 0.4% for both). These figures represent a reference standard for familiar breast cancer, given the unbiased design of the study and its uniquely large dataset. Using data from a hospital-based case-control study of 2,569 incident, histologically confirmed breast cancer cases and 2,588 controls from 6 Italian geographic areas,2, 3 including information on family history of breast cancer in first-degree relatives and age at occurrence of each neoplasm, we estimated an overall odds ratio (OR) for familial history of 2.4, corresponding to an overall PAF of approximately 7%. Table I gives more detailed analyses of the same dataset in strata of age and history of mother, sister and other/more than one relative with history of breast cancer. Multivariate ORs of breast cancer were derived from unconditional multiple logistic regression models, including terms for age, study center, education, age at first birth and age at menopause. A term for number of sisters was also included to obtain ORs and PAFs for breast cancer history in sisters, which is a function of number of sisters. Using the distribution of risk factors in the cases and the ORs from the models, PAFs were computed for various combinations of family history and age, using the method described by Bruzzi et al.4 This implies knowledge of the risk estimates and of the distribution of the risk factors of interest in the population of cases only and can therefore be applied to data of hospital-based case-control studies. PAFs were expressed as percent terms, together with the corresponding 95% CI.5 The PAF at all ages was 2.86% for mothers' history, 3.15% for sisters' and 1.11% for other/combined. As in the Swedish dataset, the PAF tended to decline with increasing probands' age at diagnosis for maternal history from 4.01 at age <45 to 1.25 at age ≥60, but to increase with sister history from 1.10 at age <45 to 5.74 at age ≥60, reflecting the low proportion of sisters with breast cancer in younger women. The overall multivariate PAF for any familial breast cancer was 7.12% (8.62% below age 45). In epidemiology, replication of findings using different methodologic approaches is a key factor for any quantitative inference. The strict consistency of the PAF estimates from a population-based record-linkage study and a hospital-based case-control study gives reassuring evidence of the reliability and validity of information on family history of breast cancer collected through interviews.6 Yours sincerely, The authors thank Ivana Garimoldi for editorial assistance. Claudio Pelucchi*, Eva Negri*, Alessandra Tavani*, Silvia Franceschi , Carlo La Vecchia* , * Istituto di Ricerche Farmacologiche “Mario Negri”, Milano, Italy, Field and Intervention Studies Unit, International Agency for Research on Cancer, Lyon Cedex, France, Istituto di Statistica Medica e Biometria, Università degli Studi di Milano, Milano, Italy.