Accuracy of the BRCAPRO Model Among Women With Bilateral Breast Cancer

Accuracy of the BRCAPRO Model Among Women With Bilateral Breast Cancer Kaylene J. Ready, MS 1, Kristen J. Vogel, MS 2, Deann P. Atchley, PhD 1, Kristine R. Broglio, MS 1, Kimberly K. Solomon, MBA 1, Christopher Amos, PhD 1, Karen H. Lu, MD 1, Gabriel N. Hortobagyi, MD 1, and Banu Arun, MD 1 BACKGROUND: The likelihood of identifying a BRCA mutation was often calculated using the BRCAPRO model. A previous study suggested that this model may overestimate the chance of detecting a BRCA mutation among women diagnosed with bilateral breast cancer. Studies also suggested that few patients with bilateral breast cancer whose age at first diagnosis is >40 years were mutation carriers. The objectives of this study were to determine the accuracy of the BRCAPRO model among women with bilateral breast cancer and to determine whether their mutation status was dependent on their age at first diagnosis. METHODS: A retrospective chart review was performed. Women who were diagnosed with bilateral or unilateral breast cancer and who had undergone comprehensive BRCA1 and BRCA2 genetic testing at M. D. Anderson Cancer Center between 1997 and 2006 were included in the study. RESULTS: For individuals with pre-test carrier probabilities >31%, the proportion of positive tests was significantly lower than predicted by the BRCAPRO model (P <.05). In addition, the carrier rate of BRCA mutations was significantly higher (P ¼.002, Fisher exact test) in women with bilateral breast cancer whose age at first diagnosis was 40 years compared with those diagnosed >40 years. CONCLUSIONS: The BRCAPRO model was overestimating the relative contribution bilateral breast cancer had on the likelihood of detecting a BRCA1 or BRCA2 mutation. Bilateral breast cancer did not appear to be a good indicator of mutation status, particularly for women whose age at first diagnosis is >40 years. Cancer 2009;115:725 30. VC 2009 American Cancer Society. KEY WORDS: BRCA1 gene, BRCA2 gene, breast neoplasms, genetic models. The presence of bilateral breast cancer is used as 1 of the indicators for hereditary breast and ovarian cancer syndrome, which is caused by mutations in the BRCA1 and BRCA2 genes. Women with a mutation in BRCA1 or BRCA2 have an approximately 43% to 87% lifetime risk of developing breast cancer and a 27% to 39% lifetime risk of developing ovarian cancer. 1-5 In addition, women with a BRCA1 or BRCA2 mutation have a 3% to 4% per year, or 40% to 60% lifetime risk of developing a second primary breast cancer. 3,6 Thus, using the presence of bilateral breast cancer as an indicator for hereditary breast and ovarian cancer and, subsequently, genetic testing of BRCA1 and BRCA2 is logical. Nonetheless, the question Corresponding author: Kaylene J. Ready, MS, The University of Texas M. D. Anderson Cancer Center, Department of Breast Medical Oncology, Unit 1354, P.O. Box 301439, Houston, TX 77230-1439; Fax: (713) 563-0909; kready@mdanderson.org 1 The University of Texas, M. D. Anderson Cancer Center, Houston, Texas; 2 Evanston Northwestern Healthcare Center for Medical Genetics, Evanston, Illlinois Received: April 30, 2008; Revised: September 4, 2008; Accepted: September 10, 2008 Published online: January 6, 2009, VC 2009 American Cancer Society DOI: 10.1002/cncr.24102, www.interscience.wiley.com Cancer February 15, 2009 725

remains as to how much importance to assign to the presence of bilateral breast cancer when calculating the likelihood of detecting a BRCA1 or BRCA2 mutation. Risk estimates for the likelihood of identifying a BRCA1 or BRCA2 mutation can be calculated using the BRCAPRO risk assessment program, among others. BRCAPRO is a Bayesian probability model that incorporates a patient s personal and family history features to determine the likelihood of identifying a BRCA1 or BRCA2 mutation. 7,8 A comparison of the BRCAPRO model to other models demonstrated that the BRCAPRO model performed better than the other risk assessment models. 9 This distinction, and the wide availability of BRCAPRO on the Internet, has made BRCAPRO 1 of the most commonly used models among clinical cancer genetics programs and the BRCAPRO model is often used in clinical practice to make decisions regarding the appropriateness of genetic testing. However, it has been previously suggested that the BRCAPRO model may overestimate the chance of detecting a BRCA1 or BRCA2 mutation among women diagnosed with bilateral breast cancer. 9 Population-based studies have also suggested that there is a low proportion of mutation carriers among patients with bilateral breast cancer whose age at first diagnosis was older than 40 years, and that bilaterality of breast cancer alone was not strongly associated with mutations in BRCA1 and BRCA2. 10,11 Therefore, we may be overestimating the relative contribution the presence of bilateral breast cancer has on the likelihood of detecting a BRCA1 or BRCA2 mutation. To provide patients with the most accurate risk assessment possible, we retrospectively report the accuracy of the BRCAPRO model among patients with bilateral breast cancer and evaluate the mutation carrier rate among bilateral breast cancer patients based on their age at first diagnosis. MATERIALS AND METHODS Study Population We retrospectively analyzed the records of patients with a personal history of bilateral breast cancer who underwent genetic counseling and BRCA1 and BRCA2 testing between 1997 and 2006 at the cancer genetics clinic at The University of Texas M. D. Anderson Cancer Center at Houston. All of the bilateral breast cancer patients underwent genetic counseling and testing only after they had been diagnosed with both of their breast cancers. A selection of unilateral breast cancer patients was identified as controls. Unilateral controls were matched to the bilateral cases at a ratio of 2 to 1. Each unilateral control was matched in such a way that the number of years between cancer diagnosis and genetic testing equaled the number of years between first and second cancer diagnosis in the bilateral cases. Matching the controls in this way ensured that the controls did not develop a second breast cancer in the given increment of time. Cases and controls were not matched based on age at first diagnosis. Although age may be perceived as a potential bias, the BRCAPRO model includes age of breast cancer diagnosis as 1 of the variables and, therefore, corrects the potential bias. The institutional review board at The University of Texas M. D. Anderson Cancer Center approved the study protocol. Personal and Family History Information Personal and family history information for input into the BRCAPRO model was obtained from patients medical records. Patients cancer diagnoses were confirmed with pathology reports, which were reviewed by M. D. Anderson Cancer Center breast pathologists. Because MRI screening may affect the detection rate of bilateral breast cancer, the bilateral breast cancer cases were reviewed for MRI screening use. The rate of breast MRI screening was not expected to be high, as all patients at M. D. Anderson Cancer Center who have a history of breast cancer are followed up with yearly mammograms, not breast MRI. In addition, the American Cancer Society guidelines for breast MRI, which recommend breast MRI for patients with significant family histories of breast cancer, were not released until April 2007, and our patient population only included patients through 2006. Most family history information was by patient report. Although this method may result in potential error in the BRCAPRO calculation, patient report of family history more accurately reflects current clinical practice. Personal and family information regarding ancestry was also collected. Information was limited to Ashkenazi Jewish ancestry or non-ashkenazi Jewish ancestry, as the BRCAPRO program accounts for Ashkenazi Jewish ancestry as a risk factor in the likelihood of detecting a BRCA1 or BRCA2 mutation. 726 Cancer February 15, 2009

Accuracy of BRCAPRO Model/Ready et al Interpretation of Genetic Test Results All patients included in the study, including those of Ashkenazi Jewish ancestry, had comprehensive BRCA1 and BRCA2 genetic testing through Myriad Genetic Laboratories. Possible results included positive, negative, or variants of uncertain significance. Only positive and negative results were used to evaluate the sensitivity and accuracy of the BRCAPRO program. BRCAPRO Calculation The University of Texas Southwestern CancerGene program, version 4.3.1, was used to calculate the BRCAPRO scores. This program is available online at http://www. utsouthwestern.edu/cancergene/. Age, age at death, breast and/or ovarian cancer diagnoses, and ages at diagnoses were entered for patients and their first- and seconddegree relatives. Statistical Analysis To determine the accuracy of the BRCAPRO model, we calculated the proportion of BRCA1- and BRCA2-positive results and plotted them and their corresponding 95% confidence intervals against the estimated average carrier probability at defined intervals. 7,12 The intervals used were 0% to 10%, 11% to 30%, 31% to 60%, 61% to 90%, and 91% to 100%. Intervals were chosen so that the number of bilateral breast cancer cases in each interval was similar. To make a fair comparison, the same intervals were used for the unilateral breast cancer controls that were used for the bilateral breast cancer cases. If the BRCAPRO model provided a perfect estimate of an individual s chance to carry a BRCA1 or BRCA2 mutation, then the observed proportion of positive results would occur linearly along a diagonal. For example, at an average BRCAPRO of 50%, we would expect the actual observed proportion of positive results to be 50%. Chi-square tests were performed to test for significance. All tests were 1-sided. P values less than or equal to.05 were considered significant. To determine whether the presence of a BRCA1 or BRCA2 mutation in women diagnosed with bilateral breast cancer is dependent on the age of diagnosis of the first breast cancer, the proportion of positive test results among women diagnosed with a first breast cancer 40 years of age and >40 years were compared. Fisher exact test was used to test for significance. All tests were 2-sided. All P values less than or equal to.05 were considered significant. RESULTS Study Population A total of 75 women with bilateral breast cancer, who underwent genetic counseling were identified. Five patients were excluded from the study due to genetic test results that revealed a variant of uncertain significance, 2 were excluded because they only underwent multisite testing for the 3 Ashkenazi Jewish founder mutations rather than comprehensive testing, and 2 were excluded because they declined or cancelled testing. Thus, 66 women with a personal history of bilateral breast cancer were included in the study. None of these women had any history of breast MRI screening. A total of 137 women with unilateral breast cancer were identified as controls. A summary of the genetic test results and personal and family history for the bilateral breast cancer cases and unilateral breast cancer controls can be seen in Table 1. Accuracy of the BRCAPRO Model For individuals with pretest carrier probabilities >31%, the proportion of positive tests was significantly lower than predicted by the BRCAPRO model (P <.05). The BRCAPRO model predicted a 1.2- to 7-fold excess of mutations in this range. This same trend was not observed among unilateral controls (Table 2 and Figs. 1 and 2). A summary of this subgroup s genetic test results and personal and family history can be seen in Table 3. Of interest, only 14% of the total study population was Ashkenazi Jewish, whereas 21% of this subgroup s population was Ashkenazi Jewish. This made up 100% of the total Ashkenazi Jewish population included in the study. When the individuals of Ashkenazi Jewish ancestry were removed from the analysis, the predicted excess of mutations remained. Mutation Carrier Rate by Age at First Diagnosis When the cases and controls were divided by age at first diagnosis, the carrier rate of BRCA mutations was Cancer February 15, 2009 727

significantly higher (P ¼.002, Fisher exact test) in women with bilateral breast cancer whose age at first diagnosis was 40 years (58%; 14 of 24) compared with women whose Table 1. Genetic Test Results and Personal/Family History of the Study Population Characteristic Bilateral Cases, n566 Unilateral Controls, n5137 No. (%) No. (%) Mutation BRCA1 mutation carriers 13 (20) 17 (12) BRCA2 mutation carriers 8 (12) 11 (8) BRCA1 and BRCA2 1 (2) 0 mutation carriers Personal history Bilateral breast 65 (98) NA Bilateral breast 1 (2) NA and ovarian Unilateral breast NA 134 (98) Unilateral breast NA 3 (2) and ovarian Family history Breast only 46 (70) 77 (56) Bilateral breast 4 (6) 0 Male breast 1 (2) 3 (2) Ovarian only 3 (5) 5 (4) Breast and ovarian 5 (8) 22 (16) None 12 (18) 33 (24) Ancestry Ashkenazi Jewish 9 (14) 4 (3) Non-Ashkenazi Jewish 57 (86) 133 (97) Age of diagnosis Median age of 1 st breast cancer 43 [20-67] 42 [21-68] Median age of 2 nd breast cancer 48 [30-78] NA NA indicates not applicable. age at first diagnosis was >40 years (18%; 8 of 42) (Fig. 3). This same trend was not observed among unilateral controls (20%; 12 of 59 positive test results among unilateral breast cancer controls diagnosed 40 years; and 21%; 16 of 78 positive test results among unilateral breast cancer controls diagnosed >40 years; P ¼ 1, Fisher exact test). DISCUSSION Our results indicate that the BRCAPRO model is overestimating the relative contribution bilateral breast cancer has on the likelihood of detecting a BRCA1 or BRCA2 mutation. A previous study suggested that the BRCAPRO model predicted a 2-fold excess of mutations among women with bilateral breast cancer, but did not specify over which carrier probabilities this overestimate occurred. 9 Our study suggests that the BRCAPRO model predicts a 1.2- to 7-fold excess of mutations at carrier probabilities greater than 31%. One explanation for the predicted excess of mutations in individuals with carrier probabilities greater than 31% is the excess of Ashkenazi Jewish individuals. However, even when the individuals of Ashkenazi Jewish ancestry were removed from the analysis, the predicted excess of mutations remained. The previous study, which also demonstrated a predicted excess of mutations, did not include individuals of Ashkenazi Jewish ancestry in their analysis. 9 The BRCAPRO model was originally developed by Parmigiani et al in 1998 8 and was later validated in Table 2. Observed Versus Estimated Positive Results Among Bilateral Breast Cancer Cases and Unilateral Breast Cancer Controls* Defined Interval No. (%) Average BRCAPRO, % Observed Positive Results Estimated Positive Results P Bilateral cases, n¼66 0-10 11 (17) 5.4 1 1 1 11-30 13 (20) 21.5 2 3.5 31-60 15 (23) 44.6 1 7 <.01 61-90 11 (17) 76.2 5 8.04 91-100 16 (24) 97.2 13 15.04 Unilateral controls, n¼137 0-10 83 (61) 4.5 4 4 1 11-30 23 (17) 21.4 6 5.6 31-60 12 (9) 42.7 4 5.5 61-90 14 (10) 73.4 9 10.6 91-100 5 (4) 93.4 5 5 1 * Data in boldface text indicate statistically significant results. 728 Cancer February 15, 2009

Accuracy of BRCAPRO Model/Ready et al Table 3. Genetic Test Results and Personal/Family History of Bilateral Breast Cancer Cases With Pretest BRCAPRO Scores Greater Than 31% Characteristic Bilateral Cases, n542 No. (%) FIGURE 1. Observed versus estimated positive results for bilateral breast cancer cases. The diagonal line represents the expected proportion of positive results, whereas the squares represent the observed proportion of positive results at defined intervals. The 95% confidence intervals are also shown. The arrowed brackets demonstrate the overestimation as calculated by BRCAPRO. Mutation BRCA1 mutation carriers 11 (26) BRCA2 mutation carriers 8 (19) Personal History Bilateral breast 42 (100) Family history Breast only 29 (69) Bilateral breast 4 (10) Male breast 1 (2) Ovarian only 3 (7) Breast and ovarian 5 (12) None 5 (12) Ancestry Ashkenazi Jewish 9 (21) Non-Ashkenazi Jewish 33 (79) Age of diagnosis Median age of 1 st breast cancer 39 [20-66] Median age of 2 nd breast cancer 45 [30-78] FIGURE 2. Observed versus estimated positive results for unilateral breast cancer controls. The diagonal line represents the expected proportion of positive results, whereas the squares represent the observed proportion of positive results at defined intervals. The 95% confidence intervals are also shown. FIGURE 3. The carrier rate of BRCA mutations based on age at first diagnosis among bilateral breast cancer cases compared with unilateral breast cancer controls. 2002. 7 The model is a computer-based Bayesian probability model based on the proband s and her first- and second-degree relatives personal and family history of unilateral or bilateral breast cancer and ovarian cancer, including ages of diagnosis and ages of unaffected family members. During the initial development of the model, the authors admit that the rate of bilateral breast cancer used in the model is an approximation because the rate of bilateral breast cancer is difficult to estimate due to lack of good data. 8 Nonetheless, during the validation of the BRCAPRO model, the authors demonstrated that the model appeared to be accurate, especially at carrier probabilities less than 70%. 7 This validation study included women with all types of personal histories, including unilateral breast cancer, bilateral breast cancer, ovarian cancer, breast and ovarian cancer, and no cancer. The accuracy of BRCAPRO was determined as a whole, rather than based on the proband s personal history. In contrast, our study determined the accuracy of BRCAPRO based on the proband s personal history of unilateral breast cancer or bilateral breast cancer. By using this method, we have demonstrated that the BRCAPRO model appears to Cancer February 15, 2009 729

be accurate at all carrier probabilities for women with unilateral breast cancer, but only accurate at carrier probabilities less than 30% for women with bilateral breast cancer. Our study also demonstrated that the carrier rate of BRCA mutations among women with bilateral breast cancer is significantly higher among those whose first breast cancer is diagnosed 40 years of age. This observation is consistent with previous studies 10,11 and suggests that the age of first breast cancer diagnosis appears to be more relevant than the presence of bilateral breast cancer in estimating the likelihood of a BRCA mutation. Our results must be interpreted in light of the limitations of the study. First, our study population was small, limiting our power to detect smaller inaccuracies in the BRCAPRO model. It is possible that the BRCAPRO model overestimates at other pre-test carrier probabilities besides those greater than 31%, but these inaccuracies would only be detected by a larger study sample. Second, the CancerGene program is frequently updated, resulting in multiple versions. We used the most recent version of the program available at the time of our study. However, the initial validation study of the BRCAPRO model would have used a much earlier version, as the version we used would not have been available in 2002. Comparisons between our study and the initial validation study must be interpreted with caution. Finally, this study was performed retrospectively, which limits the availability of the most current clinical information for the subjects included, making it possible for some of the unilateral controls to develop bilateral breast cancer without our knowledge. In conclusion, as referrals and recommendations for BRCA1 and BRCA2 genetic testing increase, we must carefully evaluate the tools used to estimate a patient s risk of carrying a mutation. Our study provides evidence that 1 of our primary risk assessment tools, the BRCAPRO model, may be overestimating the likelihood of detecting a mutation among women with bilateral breast cancer. Furthermore, bilateral breast cancer does not appear to be a good indicator of mutation status, particularly for women whose age at first diagnosis is older than 40 years. We propose that further validation studies regarding the accuracy of the BRCAPRO model be performed and that the model be refined to provide patients with a more accurate risk assessment. Conflict of Interest Disclosures The authors made no financial disclosures. References 1. Cancer risks in BRCA2 mutation carriers. The Breast Cancer Linkage Consortium. J Natl Cancer Inst. 1999;91:1310-1316. 2. Chen S, Iversen ES, Friebel T, et al. Characterization of BRCA1 and BRCA2 mutations in a large United States sample. J Clin Oncol. 2006;24:863-871. 3. Ford D, Easton DF, Stratton M, et al. Genetic heterogeneity and penetrance analysis of the BRCA1 and BRCA2 genes in breast cancer families. The Breast Cancer Linkage Consortium. Am J Hum Genet. 1998;62:676-689. 4. King MC, Marks JH, Mandell JB. Breast and ovarian cancer risks due to inherited mutations in BRCA1 and BRCA2. Science. 2003;302:643-646. 5. Narod SA, Ford D, Devilee P, et al. An evaluation of genetic heterogeneity in 145 breast-ovarian cancer families. Breast Cancer Linkage Consortium. Am J Hum Genet. 1995;56:254-264. 6. Metcalfe K, Lynch HT, Ghadirian P, et al. Contralateral breast cancer in BRCA1 and BRCA2 mutation carriers. J Clin Oncol. 2004;22:2328-2335. 7. Berry DA, Iversen ES Jr, Gudbjartsson DF, et al. BRCAPRO validation, sensitivity of genetic testing of BRCA1/BRCA2, and prevalence of other breast cancer susceptibility genes. J Clin Oncol. 2002;20:2701-2712. 8. Parmigiani G, Berry D, Aguilar O. Determining carrier probabilities for breast cancer-susceptibility genes BRCA1 and BRCA2. Am J Hum Genet. 1998;62:145-158. 9. Marroni F, Aretini P, D Andrea E, et al. Evaluation of widely used models for predicting BRCA1 and BRCA2 mutations. J Med Genet. 2004;41:278-285. 10. Eccles DM, Englefield P, Soulby MA, Campbell IG. BRCA1 mutations in southern England. Br J Cancer. 1998; 77:2199-2203. 11. Gershoni-Baruch R, Dagan E, Fried G, Kepten I, Robinson E. BRCA1 and BRCA2 founder mutations in patients with bilateral breast cancer. Eur J Hum Genet. 1999;7:833-836. 12. Barcenas CH, Hosain GM, Arun B, et al. Assessing BRCA carrier probabilities in extended families. J Clin Oncol. 2006;24:354-360. 730 Cancer February 15, 2009