Mathematical Modeling for Breast Cancer Risk Assessment

August 1, 2002

Rubinstein and colleagues provide an excellent review of mathematical models for estimating breast cancer risk, including the risk of carrying inherited mutations of BRCA1 and BRCA2. Since we and others reviewed early models to predict the likelihood of inherited susceptibility to breast cancer,[1] newer quantitative tools, most notably by Parmigiani and colleagues,[2] have been developed. These models have been made available on CD-ROM, over the Internet, and in other electronic versions that are accessible to most clinicians and researchers. These quantitative resources constitute useful and important aids in genetic counseling.

Rubinstein and colleagues provide an excellent review of mathematical modelsfor estimating breast cancer risk, including the risk of carrying inheritedmutations of BRCA1 and BRCA2. Since we and others reviewed early models topredict the likelihood of inherited susceptibility to breast cancer,[1] newerquantitative tools, most notably by Parmigiani and colleagues,[2] have beendeveloped. These models have been made available on CD-ROM, over the Internet,and in other electronic versions that are accessible to most clinicians andresearchers. These quantitative resources constitute useful and important aidsin genetic counseling.

With this commentary, I will provide additional perspective to the excellentoverview presented by Rubinstein et al, addressing several areas that theauthors did not fully touch upon. These topics include (1) the importance ofbeing aware of genetic testing guidelines propagated by insurers, (2) theprobability of detecting missense variants of unknown significance as a resultof genetic testing, (3) the psychological implications of testing unaffectedprobands, and (4) special aspects of testing individuals of Ashkenazi ancestry.

Finally, I will review a general caution that affects all quantitativemodeling for hereditary breast cancer. This relates to the highly selected (ie,biased) nature of the ascertainments that have been used to generate risk(penetrance) information.

Quantitative Estimates and Insurance Reimbursement

Perhaps the most clinically relevant application of quantitative riskestimates relates to the use of quantitative models by third-party carriers. Incontrast to the early dire forecasts regarding insurance abuse of geneticinformation, several large carriers include BRCA testing in their coverage plans(without penalty) if specific family history criteria are met. For example, BlueCross/Blue Shield has issued centralized guidelines on BRCA testing.[3] However,Blue Cross guidelines vary according to the policies of local plans in eachstate.

As part of an American Medical Association conference, the Kaiser systemcirculated proposed criteria for BRCA testing,[4] and guidelines have also beenissued by Aetna/US Healthcare.[5] These policies may be of as much interest tohealth-care providers as the theoretical models presented in this excellentreview. Citation of the theoretical models may be useful for clinicians seekingto obtain insurance coverage for testing services provided to those insured bycompanies without established policies.

Detecting Missense Mutations of Unknown Significance

A surprisingly overlooked aspect of BRCA testing relates to the frequentoccurrence of "ambiguous" results. Missense mutations of unknownsignificance are found in up to 10%-15% of patients tested. The probability ofdetecting these variants depends on the ethnic origin of the proband, the genetested, and the test method utilized. None of the current quantitative modelsaddress these aspects of the prediction of these variants.

Counseling of individuals with missense mutations of unknown significance canbe psychologically challenging. Such counseling can also be complex because theclinician must be prepared to embark upon cosegregation analysis within thesefamilies.

Psychological Implications of Testing Unaffected Probands

Another specific aspect raised in this review relates to the desirability(from both an economic and counseling perspective) of initiating testing inrelatives who are affected by early-onset disease. Although the strong rationalefor this approach is well summarized, there are psychological aspects thatshould also be recognized. Affected sisters, mothers, or aunts may feel coercedinto undergoing testing by their unaffected relatives. If the affected relativeis receiving treatment for advanced disease, testing may prove an added burden.

In such circumstances, identification of a mutation in an unaffectedindividual may obviate the need for approaching the affected relative. Thiscourse may be reasonable if accompanied by appropriate counseling, and if theposterior probability of finding a mutation is sufficiently increased (forexample, in a patient of Ashkenazi background with a strong family history).

Testing Individuals of Ashkenazi Ancestry

In those of Ashkenazi origin, certain caveats are appropriate regardingtesting in the setting of a negative screen for the three founder mutations(case C in the article by Rubinstein et al). First, the recalculation of"residual" risk for nonfounder mutations, assuming the family is notAshkenazi (as suggested in the review), has not been formally evaluated in largestudies. Such an approach may, in fact, represent an overestimation of mutationfrequency.[6] The testing of paraffin blocks for specific mutations, althoughtechnically feasible, is also not recommended outside the context of a researchstudy. The sensitivity and specificity of clinical testing of paraffin blockshas not been established, and there is the possibility of false-positiveresults.[7]

Overestimation of Penetrance

On a more general note, the major models cited in this review—particularlythose developed at the University of Pennsylvania by the Myriad organization inUtah, at the National Cancer Institute (NCI), and at Duke University—werebased on, and potentially biased by, ascertainments that were highly selectedfor the presence of a family history of breast cancer. As Rubenstein et alcomment, with the exception of the Gail model developed at the National CancerInstitute, none of the models has been validated in a population-based cohort.However, several validation studies of the BRCAPRO model are ongoing.

Population-based models that were not cited,[8] notably those developed inAustralia, New York, and Toronto/New Haven, have resulted in markedly lowerestimates of breast and ovarian cancer risk associated with mutations of theBRCA1 and BRCA2 cancer susceptibility genes. In some of these studies, familyhistory did not predict likelihood of carrying a mutation.

These observations suggest that other genetic factors may not only modify thelikelihood of detecting a BRCA mutation, but may themselves constituteindependent risk factors for the development of breast or ovarian cancer. AsRubinstein et al point out, model-based risk assessments can be an importantadjunct to comprehensive cancer genetic counseling. However, it is important tonote that no currently available model is an adequate substitute for a detailedfamily evaluation conducted by an appropriately trained cancer care provider.


1. Offit K, Brown K: Quantitating familial cancer risk: A resource forclinical oncologists. J Clin Oncol 12(8):1724-1736, 1994.

2. Parmigiani G, Berry D, Aguilar O: Determining carrier probabilities forbreast cancer-susceptibility genes BRCA1 and BRCA2. Am J Hum Genet62(1):145-158, 1998.

3. Seidenfeld J et al: Technology Evaluation Center, Blue Cross and BlueShield Association. Chicago.

4. AMA (medical science) genetics and managed care—genetic services inKaiser Permanente in the 90s. Chicago, American Medical Association. Availableat

5. BRCA testing, prophylactic mastectomy, tamoxifen, and prophylacticoophorectomy for women at risk for breast and ovarian cancer. Coverage policybulletin 0227. Hartford, Conn, Aetna. Available at June 24, 2002.

6. Kauff ND, Perez-Segura P, Robson ME, et al: Incidence of non-founder BRCAmutations in high risk Ashkenazi breast and ovarian cancer families. J MedGenetics (in press).

7. Wong C, DiCioccio RA, Allen HJ, et al: Mutations in BRCA1 from fixed,paraffin-embedded tissue can be artifacts of preservation. Cancer GenetCytogenet 107(1):21-27, 1998.

8. Hopper JM, Southey M, Dite G, et al: Population-based estimate of theaverage age-specific cumulative risk of breast cancer for a defined set ofprotein-truncating mutations in BRCA1 and BRCA2. Australian Breast Cancer FamilyStudy. Cancer Epidemiol Biomarkers Prev 8(9):741-747, 1999.