Breast cancer risk assessment: use of complete pedigree information and the effect of misspecified ages at diagnosis of affected relatives.

Reliable risk estimates for hereditary breast cancer are important for the genetic counseling of women who have one or more first- and/or second-degree relatives affected by the disease. If no mutation analysis of known high-penetrance breast cancer genes is performed, risk estimation is often based on published reference tables. These tables express a woman's age-specific risk of breast cancer as a function of the ages at diagnosis of one or two affected relatives with different degrees of relationship to the counselee. However, unaffected relatives are not taken into account when these estimates are derived. We report here the extent to which risk estimation is influenced by the number and ages of any unaffected relatives and by the exact genealogical relationship between the proband and affected relative rather than merely the degree. Additionally, we describe the sensitivity of risk estimates when ages at diagnosis of affected relatives are misspecified because of inaccurate information supplied by the counselee. We determined a proband's probability of being a carrier of a highly penetrant breast cancer susceptibility gene, such as BRCA1 or BRCA2, by likelihood calculations that take into account information from the entire pedigree. This genetic risk was used to estimate a phenotypic lifetime breast cancer risk, which was compared with the risks derived from the published reference tables. We demonstrate numerically that the tabulated values tend to over-estimate the probands risk and that the extent of over-estimation depends greatly on the number and ages of unaffected relatives. The validity of the relatives ages at diagnosis can affect risk predictions considerably in small families with two or three affected relatives. Furthermore, the magnitude of the estimated breast cancer risks depends upon the assumed genetic model and can therefore vary appreciably when different penetrance estimates are used.