RESPONSE: Re: On the Use of Familial Aggregation in Population-Based Case Probands for Calculating Penetrance

In a recent issue of the Journal (1), Begg describes potential biases in our (2) and other population-based casescreening studies that estimate mutation penetrance through kin–cohort methods (3). The main issue raised by Begg is that a case series represents individuals selected to have risk factors that place them at excess disease risk. To the extent that any risk factors are overrepresented among case relatives—because of either genetic or familial reasons— family members of both carrier and noncarrier probands will show excess disease incidence, and thus mutation penetrance will be overestimated. We agree with this theoretical argument but question the degree of bias actually present in the published studies of breast and ovarian cancer and mutations in BRCA1 or BRCA2 to which Begg refers (1). In particular, our penetrance estimates of breast cancer based on an ovarian cancer case series are unlikely to be biased (2). Disease risk to a given age among carrier and noncarrier relatives is found by treating the relatives as a cohort and performing a Cox regression analysis on it, with age at diagnosis, death, or end of follow-up as the time variable and with proband mutation status as the exposure. This method yields an estimate of the relative risk (RR) of disease among relatives associated with proband mutation status, as well as a product–limit estimate of the survivor function for the noncarrier relatives (S0). The estimated mutation penetrance is 1 + S0 – 2(S0) . This method is robust in the usual circumstances of one (or sometimes two) affected first-degree relatives per family, but generalized estimating equation methods can also be used. As Begg notes (1), the estimated RR is not subject to the bias; only S0 is subject to bias. In fact, 1 + S0 – 2(S0) RR estimates the penetrance for any similar base population to which the RR would apply. Therefore, for studies such as ours (2) that use this method, and in the very usual circumstances where mutation frequency is low in the population, the observed RR values can certainly be used