Sequence-based testing of disease-susceptibility genes has identified many variants of unknown significance (VUSs) whose pathogenicity is unknown at the time of their measurement. Female breast cancer cases aged 20-49 years at diagnosis and who have VUSs in BRCA1 and no mutations in BRCA2 have previously been identified through the population-based Los Angeles County Cancer Surveillance Program. These nominal BRCA1 VUSs have been classified as "low," "medium," and "high" risk by four classification methods: Align-GVGD, Polyphen, Grantham matrix scores, and sequence conservation in mammalian species. Average hazard ratios (HRs) for classes of variants, i.e., the age-specific incidences of cancer for carriers of such variants divided by the population incidences, were estimated from the cancer family histories of first- and second-degree relatives of the index cases using modified segregation analysis. The study sample comprised 270 index cases and 4,543 of their relatives. There was weak evidence that the risk of breast cancer increases with the degree of sequence conservation (P = 0.03) and that missense variants at highly conserved sites are associated with a 5.6-fold (95 % confidence interval 1.4-22.2; P = 0.05) increased incidence of breast cancer. An upper bound of 2.3 is given for the average breast cancer HRs corresponding to variants classified as "low risk" by any of the four VUS classification methods. In summary, we have given a method to estimate cancer risks for groups of VUSs by combining existing classification methods with traditional penetrance analyses. This analysis suggests that classification methods for BRCA1 variants based on sequence conservation might be useful in a clinical setting. We have shown in principle that our method can be used to classify VUSs into clinically useful risk categories, but our specific findings should not be put into clinical practice unless confirmed by larger studies.