|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
© 2002 American Society for Clinical Oncology BRCAPRO Validation, Sensitivity of Genetic Testing of BRCA1/BRCA2, and Prevalence of Other Breast Cancer Susceptibility GenesByFrom the Department of Biostatistics, University of Texas M.D. Anderson Cancer Center, and Breast Center, Baylor College of Medicine, Houston, TX; Institute of Statistics and Decision Sciences, Duke University, Durham, NC; Dana-Farber Cancer Institute, Harvard University, and Division of Surgical Oncology, Massachusetts General Hospital, Boston, and Lahey Clinic, Burlington, MA; Lombardi Cancer Center, Georgetown University Medical Center, Washington, DC; University of Pennsylvania Cancer Center, Philadelphia, PA; Creighton University, Omaha, NE; Center for Medical Genetics, Evanston Northwestern Healthcare, Evanston, IL; and Departments of Oncology and Biostatistics, Johns Hopkins University, Baltimore, MD. Address reprint requests to Donald A. Berry, PhD, Department of Biostatistics, The University of Texas M.D. Anderson Cancer Center, 1515 Holcombe Blvd, Box 447, Houston, TX 77030-4009; email: dberry{at}mdanderson.org
PURPOSE: To compare genetic test results for deleterious mutations of BRCA1 and BRCA2 with estimated probabilities of carrying such mutations; to assess sensitivity of genetic testing; and to assess the relevance of other susceptibility genes in familial breast and ovarian cancer. PATIENTS AND METHODS: Data analyzed were from six high-risk genetic counseling clinics and concern individuals from families for which at least one member was tested for mutations at BRCA1 and BRCA2. Predictions of genetic predisposition to breast and ovarian cancer for 301 individuals were made using BRCAPRO, a statistical model and software using Mendelian genetics and Bayesian updating. Model predictions were compared with the results of genetic testing. RESULTS: Among the test individuals, 126 were Ashkenazi Jewish, three were male subjects, 243 had breast cancer, 49 had ovarian cancer, 34 were unaffected, and 139 tested positive for BRCA1 mutations and 29 for BRCA2 mutations. BRCAPRO performed well: for the 150 probands with the smallest BRCAPRO carrier probabilities (average, 29.0%), the proportion testing positive was 32.7%; for the 151 probands with the largest carrier probabilities (average, 95.2%), 78.8% tested positive. Genetic testing sensitivity was estimated to be at least 85%, with false-negatives including mutations of susceptibility genes heretofore unknown. CONCLUSION: BRCAPRO is an accurate counseling tool for determining the probability of carrying mutations of BRCA1 and BRCA2. Genetic testing for BRCA1 and BRCA2 is highly sensitive, missing an estimated 15% of mutations. In the populations studied, breast cancer susceptibility genes other than BRCA1 and BRCA2 either do not exist, are rare, or are associated with low disease penetrance.
INDIVIDUALS WITH a family history of breast and/or ovarian cancer are at increased risk of carrying deleterious mutations of susceptibility genes BRCA1 and BRCA2.1-4 Genetic testing for the presence of these mutations is becoming commonplace. The basis for testing is usually an assessment of risk of carrying a mutation based on a family history of disease. Because genetic testing has important health and cost consequences, accurate risk assessment is important. Such assessment is complex, and most people overestimate their risk, of both breast cancer and of carrying a deleterious mutation.5-8 Two important risk models were developed before the discovery of BRCA1 and BRCA2: Gail et al9 incorporated the number of first-degree relatives with breast cancer into a comprehensive model of breast cancer risk; Claus et al10,11 estimated the risk of breast cancer caused by susceptibility genes depending on the number of familial cases of breast cancer and the ages at diagnosis of affected members. After testing became available, new models were developed using regression or classification approaches in which various family characteristics were used to predict results of genetic testing.12-15 BRCAPRO1,3 is a computer program that implements a statistical model for calculating an individual’s probability of carrying a deleterious mutation of BRCA1, BRCA2, neither, or both on the basis of the individual’s cancer status and the history of breast and ovarian cancer among her first- and second-degree relatives. (We use "carrier probability" to mean the probability of carrying a deleterious mutation of either BRCA1 or BRCA2.) The model uses the autosomal dominant Mendelian characteristics of the genes, incorporates prevalence and penetrance on the basis of the published results as described in detail in Iversen et al,16 and uses Bayesian updating.17 BRCAPRO has been licensed to approximately 1,000 users worldwide. It is distributed free of charge as part of the genetic counseling package CancerGene (in a Web-based format available through David M. Euhus, MD, Division of Surgical Oncology, The University of Texas Southwestern Medical Center at Dallas, at http://www.swmed.edu/home_pages/cancergene/). In a clinical setting, BRCAPRO can be used to aid in advising individuals whether to undergo genetic testing. Testing usually has no value for someone whose carrier probability is small. However, the threshold depends on the individual’s attitudes toward testing, her family’s attitudes toward testing, and her attitudes about prophylactic procedures.18 Less intuitively, testing may not be worthwhile for a woman whose carrier probability is large (see Discussion). Testing will generally be informative and potentially valuable for women whose carrier probability is between these extremes. BRCAPRO also has research uses. Family histories are usually easier and less expensive than genetic testing, and sometimes genetic testing is not possible. BRCAPRO can be helpful in two categories of application: (1) as a guide to determine who to test and to judge the utility of such testing; and (2) as a substitute for testing. Regarding the first application, a researcher who contemplates genetic testing for subjects involved in a clinical trial or subjects in a research database can use BRCAPRO to determine the expected number of mutations in each subset of study participants. This allows for judging whether the scientific question can be addressed with sufficient statistical power. It also allows for estimating the cost of testing per mutation expected. In addition, it allows for selecting the most informative subset of participants for testing when funding for genetic testing is limited. Regarding the second application, using a procedure called multiple imputation, each subject can be assigned a mutation with probability calculated using BRCAPRO and the conclusions drawn on the basis of these assignments, assuming that the assignment is actual. This procedure is repeated many times, with potentially different conclusions each time. The result is a frequency distribution of conclusions. This frequency distribution may have great variability, indicating uncertainty regarding the scientific hypothesis being addressed. However, it may have little variability, especially if the number of participants is large, thereby obviating the need to perform genetic testing. Applications of BRCAPRO for research include the following: (1) quantifying the association between breast cancer risk and family history that is not attributable to major autosomal dominant genes19; (2) using multiple imputation to analyze the association between familial susceptibility and severity of prognosis in a large population-based case series obtained by linking Surveillance, Epidemiology, and End-Results (SEER) follow-up to Cancer and Steroid Hormone study family history information20; (3) developing a classification algorithm for determining which missense mutations have a deleterious effect21; (4) estimating the prevalence of nonfounder BRCA1 and BRCA2 mutations in Ashkenazi families at high-risk for breast and ovarian cancer (Kauff et al, manuscript submitted for publication); and (5) validating simple family history assessment tools.22 For a carrier probability model to be useful it must be accurate and widely applicable; the substance of this article is an assessment of the limitations and practical application of BRCAPRO and other carrier probability models. Potential limitations of BRCAPRO include the following: (1) it uses published penetrance/incidence and prevalence estimates, which may be inaccurate; (2) it considers only first- and second-degree relatives; (3) it accounts for breast and ovarian cancer only; and (4) it considers only the two known susceptibility genes, BRCA1 and BRCA2. If there are other susceptibility genes with disease penetrances comparable to those of the two known genes, then some individuals with large BRCAPRO carrier probability will harbor mutations of the other genes but will test negative for BRCA1 and BRCA2. In view of the wide applicability and potential limitations of BRCAPRO, we undertook the present study to validate its performance, extending three previous studies.23-25 These three studies considered subsets of data from two of the six centers that are considered here. Inferences from those studies were consistent with ours, but firm conclusions about validation, sensitivity, and so on are not possible because of the small sizes of the previous studies.
Study Description We collected data retrospectively from families for which at least one member was tested for mutations at both BRCA1 and BRCA2. Six cancer genetic counseling centers participated: Dana-Farber Cancer Institute, Georgetown University Medical Center, Creighton University, The University of Texas Health Science Center at San Antonio, University of Pittsburgh Medical Center, and Lahey Clinic. There was no uniform set of eligibility criteria across centers, reflecting current practice. At some of the centers, models used to help guide whether to carry out genetic testing included Frank et al,12 Couch et al,13 Shattuck-Eidens et al,15 and FitzGerald et al.26 Characteristics of eligibility criteria from each center are listed in Table 1.12,13,15,26
We included every family for which at least one member had been tested, regardless of family history. No genetic testing results were used in determining eligibility for the study. Therefore, the numbers of cancers reported in families represent the standard in these six centers. We selected one member (the proband) from each family; this was the member who had the first genetic test for both BRCA1 and BRCA2 in the family. A family was excluded if the proband had not completed testing of both genes. Information was reported to genetic counselors by family members. Errors in reporting are possible, particularly for second-degree relatives,27 but these errors are also likely to occur in the practical use of BRCAPRO outside of this study. Documentation was obtained when possible. We compared each proband’s carrier probability and testing result. A better comparison of the accuracy of the program would be to compare the probability with the actual genotype, but genetic testing is imperfect and so the genotype may be unknown. "Carrier probability" refers to the probability of genotype and not to test results. Genetic alterations reported as polymorphisms and uncertain variants were scored as negative.
Description and Application of BRCAPRO
The population-based inputs of BRCAPRO are mutation prevalence and disease penetrance. The prevalence of mutations for individuals of AJ descent is estimated to be 1.22% for BRCA1 and 1.36% for BRCA2.28,29 These figures are estimated to be 0.12% and 0.044% for non-AJ individuals.30 Penetrance is the probability of diagnosis of cancer up to and including each age. Published penetrance estimates are similar in the two ethnic groups, and the associated published standard deviations are large, especially at the two extremes of age. Therefore, we estimated penetrance for breast and ovarian cancer by combining estimates for AJ31 and non-AJ30 individuals. For each age, the result is a weighted average of the two (specifics of the estimation method are given in Iversen et al16). Penetrances at age 70 for both AJ and non-AJ carriers of mutations of BRCA1 and BRCA2 are .691 and .671 for breast cancer and .296 and .191 for ovarian cancer. Differences are greater at some other ages. Penetrances for noncarriers are taken from the SEER database32 and adjusted to exclude cases estimated to be attributable to BRCA1 and BRCA2. (BRCAPRO penetrance files are available at http://biosun01.biostat.jhsph.edu/
Genetic Testing
Description of Families
Accuracy of BRCAPRO We repeat the above comparison for AJ and non-AJ families separately. We also compare mean carrier probabilities between the probands who test positive and those who test negative, using t tests. The ability of any model to discriminate between BRCA1 and BRCA2 on the basis of family history is limited because both types of mutations are associated with similar increases in the incidence of breast and ovarian cancer. Therefore, we combine the two genes in most of our analyses. However, small differences in penetrances and distribution of age of onset for these two genes30,31,33,34 give BRCAPRO a modest ability to discriminate between them. We assess this ability by calculating the average probabilities of carrying mutations for BRCA1 and BRCA2 separately for those probands who tested positive for BRCA1 and BRCA2 and neither gene. We assess the dependence of the accuracy of BRCAPRO in predicting testing results on the assumed penetrance of BRCA1 and BRCA2. In particular, we consider fractions from 20% to 100% of the published estimates of penetrance and calculate the linear correlation coefficient between BRCAPRO and the observed testing results for each fraction. The higher the correlation, the more the evidence favors the respective penetrance, but we do not quantify the strength of this evidence.
Sensitivity and Specificity of Genetic Testing
Description of Study Sample Our sample consisted of 301 probands and their families. Two examples that are representative of our study families are shown in Fig 1. The BRCAPRO carrier probability is 99.9% for the proband depicted in Fig 1A. This proband was found to carry a mutation on BRCA1. For the proband depicted in Fig 1B, the BRCAPRO carrier probability is 1%. No mutations were found on either gene for this individual.
Of the 301 families in our sample, 126 (42%) were AJ. Table 2 lists the distribution of breast and ovarian cancers in the probands and their families. The average number of cases (± SD) of breast cancer in the 301 families was 3.0 ± 1.8, the average number of ovarian cancers was 0.8 ± 1.0, and the average number of bilateral breast cancers was 0.5 ± 0.7.
Table 3 lists the numbers of cancers in the families categorized by carrier probability. The numbers of cancers varied greatly among the families. The majority (216 [71%]) of the probands have three or more cases of familial breast or ovarian cancer, and might be deemed at "high risk" for carrying mutations for this reason.4 Conversely, a fair number of probands had rather low carrier probabilities. There was an average of 3.8 cancers (breast or ovarian) per family, including probands: 3.9 for the 267 (89%) families with affected probands and 2.8 for those with unaffected probands. As expected, there was a tendency for probands with extensive family histories of breast and ovarian cancer to have higher carrier probabilities. However, many probands with extensive family histories of disease had low carrier probabilities and some with only a few cancers had high carrier probabilities. This indicates that the number of familial cancers alone cannot be used to give an accurate assessment of genetic risk. It is important to consider ages of onset, types of cancers, number and ages of unaffected family members, and the exact relationships among all family members.
The distribution of carrier probabilities is listed separately for affected and unaffected probands in Table 3. Most unaffected probands had carrier probabilities of less than .5 (the probability that the proband would inherit a mutation harbored by a particular parent). Carrier probability could be greater than .5 for unaffected probands if there is cancer on both sides of the family or if the proband has both descendants and ancestors who are affected. There were different levels of genetic risk among tested individuals at the six centers, with numbers of cancers per family averaging 2.4, 3.0, 3.5, 3.5, 3.6, and 5.2. Similarly, they differ by carrier probability, with respective averages of 32%, 24%, 65%, 58%, 71%, and 72%. The overall proportion of probands testing positive was 168 (56%) of 301: 139 (46%) for BRCA1 and 29 (10%) for BRCA2. Corresponding frequencies among AJ probands were 78 (62%) of 126: 62 (52%) for BRCA1 and 13 (10%) for BRCA2. Among non-AJ probands, the frequencies were 90 (51%) of 175: 74 (42%) for BRCA1 and 16 (9%) for BRCA2.
Validation of BRCAPRO
AJ Subgroup Figure 2B displays the same scheme as shown in Fig 2A but separated into AJ (n = 126) and non-AJ (n = 175) subgroups, and corresponding to the data in Table 5. Each open circle in Fig 2B represents a sextile of the proband data set that is AJ, and each solid square in Fig 2B represents a sextile of the proband data set that is not AJ. The area of each circle and each square is approximately proportional to the sample size. The model fit is similar in the two subgroups. For the AJ, the correlation is .445 (P < 10-10) and for the non-AJ it is .577 (P < 10-10). The estimated sensitivities are not statistically different in the two subgroups: 83.3% (95% CI, 73.3% to 90.6%) among the AJ and 87.5% (95% CI, 78.1% to 93.9%) among the non-AJ.
High-Risk Families
Low-Risk Probands
Unaffected Probands
Male Probands
Discriminating BRCA1 and BRCA2 Mutations From Family History
Sensitivity of Genetic Testing/Prevalence of Other Genes The estimated sensitivity of genetic testing was 85.4% (95% CI, 78.7% to 90.5%). The undiscovered mutation proportion of about 15% includes any mutations on susceptibility genes other than BRCA1 and BRCA2 (see Discussion).
BRCAPRO is designed to provide the probability of carrying a mutation on the basis of family history. Our study shows that the program is effective in predicting the risk of testing positive. It is especially accurate in predicting testing results when the carrier probability is less than 70%. Larger carrier probabilities overestimate the frequency of a positive test result in our sample by about 15%. Assuming that the penetrance and prevalence functions used by BRCAPRO are accurate, we conclude that genetic testing for BRCA1 and BRCA2 is highly sensitive, missing at most an estimated 15% of mutations. Included are mutations of other breast and ovarian cancer susceptibility genes that have approximately the same penetrances as BRCA1 and BRCA2. The fit of the actual testing results to carrier probability in Fig 2A is good but not perfect. Individuals with small carrier probabilities are unlikely to carry mutations at BRCA1 or BRCA2: among the 37 probands having carrier probability less than 5%, only two (5.4%) were carriers. Conversely, individuals with high carrier probabilities are likely to carry mutations: among the 90 probands having carrier probability more than 95%, 75 (83%) were carriers. For predicting whether an individual tests positive for BRCA1 or BRCA2, these calculations and Fig 2 suggest that carrier probability is a good estimate unless it is large—greater than 70%. For such large carrier probabilities, taking 85% of the carrier probability to account for imperfect test sensitivity improves the estimated probability of a positive test result. Estimated sensitivity of genetic testing in our study was at least 85%. This is an average in the sense that it applies over the spectrum of testing methods and laboratories used; individual methods may have higher sensitivity and others may have lower sensitivity.30 Our estimate of sensitivity of testing is substantially greater than the 63% of Ford et al.30 A likely reason is that they used estimated odds of linkage and their testing methodology was less sensitive than full sequencing. Our results lead to the conclusion that any undiscovered breast cancer genes in the population studied are likely to have low prevalence or low penetrance. Genes having penetrance similar to BRCA1 and BRCA2 are probably rare, with total prevalence no greater than approximately one quarter that of BRCA1 and BRCA2 combined. Figure 3 is a schematic showing the trade-off between false-negatives in testing for BRCA1 and BRCA2 and the existence of other moderately penetrant susceptibility genes, with the sum being 15%. If the sensitivity of testing for BRCA1 and BRCA2 is 85%, then there is no room for other such genes in our study populations. If the sensitivity of testing is 100%, then other susceptibility genes could have total prevalence one quarter that of the total of these two genes.
BRCAPRO uses published estimates of penetrance. Population-based evidence regarding a particular mutation of BRCA135 suggests that these estimates may be too high. Our study has little ability to address penetrance. However, to investigate the effect of overestimating penetrance, we recalculated all carrier probabilities after lowering penetrance input to the model to various proportions of published estimates. Mean carrier probability dropped and estimated sensitivity increased, as shown in Table 7. Cases with carrier probability near 100% have the most influence in estimating sensitivity. Large values of carrier probability tend to decrease more in noncarriers, resulting in increased sensitivity if a smaller penetrance is assumed. Of the 90 probands having carrier probability more than 95% when using published penetrance, 75 (83%) tested positive. Decreasing penetrance to 50% of the published estimates means that only 47 had carrier probability more than 95%, and 43 (91%) of these tested positive. Decreasing the estimate of penetrance to 20% of published figures means that only 14 probands had carrier probability more than 95%, and all tested positive. The correlation coefficient is reasonably high and nearly constant over the range of penetrances considered, which means that Table 7 provides little evidence concerning the accuracy of published estimates.
Our results suggest that genetic testing is highly specific. Although the slight excess in the proportions of positives we found for probands having low carrier probability could be because of wrongly calling a case positive, other explanations are more plausible. One is ascertainment bias. For example, families with affected third-degree relatives, relatives with colon cancer, and so on, may be more likely to present at clinics, and this information is not used in BRCAPRO. Conversely, the excess may be due to sampling variability, incorrect omission of deleterious missense mutations, inaccurate penetrance and prevalence estimates, or inaccurate family histories. The family history in Fig 1A illustrates incomplete information, which is a type of inaccuracy that is common in our study. The current ages of many unaffected relatives were not reported. Ignoring unaffected relatives in BRCAPRO overestimates carrier probability. The effect can be substantial but is not great in this particular family: If all unknown ages of siblings of the proband were assumed to be 42 and all aunts were assumed to be age 69, then the proband’s carrier probability would drop to 99.8%, only slightly reduced from the value 99.9% indicated in Fig 1. We compared BRCAPRO with an easy-to-use model for predicting testing results that was developed by Myriad Genetics, Inc.12,15 We used the updated version of the model,12 available at http://www.myriad.com. The Myriad model is not based on Mendelian inheritance but instead uses logistic regression fitted to various characteristics of family history: whether any relatives have been diagnosed with ovarian cancer, whether the proband has a second primary cancer, and an interaction term between diagnosis at age less than 40 and whether any relative had early-onset breast cancer. The Myriad model applies only for women with early-onset breast cancer (age < 50) and at least one first- or second-degree relative with ovarian or early-onset breast cancer. In our study, the Myriad model applied for 144 (48%) of the 301 probands. Of these, 91 (63%) tested positive, which compares with a proportion of 46% positive tests predicted by the Myriad model (50% among positive tests and 39% among negative tests; r = .29; P = .0003). The Myriad probability tends to be smaller than that of BRCAPRO, in part because the former predicts results of testing rather than carrier status. However, the Myriad probability is larger than BRCAPRO when many relatives are unaffected (unaffected relatives are not considered by Myriad) and when a family’s cancers are not consistent with a heredity pattern. An example of the latter is member 1 of family 4 of Berry et al3: BRCAPRO carrier probability is 2.4% and Myriad’s is 59%. The Myriad model is not as accurate as BRCAPRO. For the 144 probands where Myriad applied, the mean BRCAPRO probability was 78% (90% among those testing positive and 58% among those testing negative; r = .49; P = 10-10). BRCAPRO more accurately predicted testing result than did Myriad for 110 (76%) of the 144 probands: of the 25 probands for which BRCAPRO was less than Myriad, 22 (88%; P = .0002) tested negative; of the 119 probands for which BRCAPRO was more than Myriad, 88 (74%; P = 10-7) of them tested positive. Therefore, BRCAPRO was more accurate whether it pointed more in the negative direction or more in the positive direction. Also, BRCAPRO applies generally, whereas Myriad’s application is restricted. Of the 157 probands for which the Myriad model did not apply, the BRCAPRO fit to the genetic testing results was just as good (r = .56) as for the 144 cases for which Myriad did apply. The utility of BRCAPRO as compared with simply counting the number of familial cancers is evinced by the cross-tabulation in Table 3. It is easy to assess genetic risk in families that are cancer-free and those that have more than six cancers. For such families, computer programs are of little use. However, when a family has from one to six cancers, finding carrier probability can be difficult, and it is not necessarily intuitive. For a given number of familial cancers, carrier probabilities range from small to large. For example, for probands with three familial cancers, about one in six had probabilities smaller than 10%, whereas one in three had probabilities greater than 90%. The spread is not as large for unaffected probands, who constituted a small proportion (11%) of our sample. As indicated above, 50% is usually the upper limit of carrier probability in unaffected women. Most unaffected probands in our study were at high genetic risk, with about two thirds having carrier probability greater than 40% (Table 3). However, there is still substantial variability, with about 15% of unaffected probands having probability less than 10%. We indicated in the Introduction that most people overestimate their risk of carrying mutations. This overestimation can be substantial. For example, in the Bluman et al study,6 and restricting to the 78 women whose carrier probability as assessed by BRCAPRO was less than 10%, the women estimated their carrier probability to be 65% on average. Moreover, none of the 78 women thought that their probability was less than 10% and 16 thought that it was 90% or more.6 Using BRCAPRO can provide a service to women by helping to correct unreasonable overestimates. In the opposite direction, BRCAPRO can also correct underestimates. Of the 31 women in this study whose BRCAPRO carrier probability was greater than 90%, their own assessment of their probability averaged only 79%; 19 of the 31 thought that their probability was less than 90% and one estimate was as low as 20%. The importance and clinical application of BRCAPRO depends on the individual’s carrier probability, but it also depends greatly on the individual’s attitudes. If she will not make changes in her life as a result of learning her carrier probability, then the information has no value. However, some women will make decisions on the basis of the level of associated risk. A low probability might not prompt any changes in lifestyle or health care decisions, whereas a probability in the range of 5% to 10% might prompt the individual to seek genetic counseling and a high probability might lead to prophylactic surgery. An individual with a 10% probability might base subsequent health care treatment options on the outcome of genetic testing, opting for a surgical solution if the test finds a deleterious mutation, and avoiding surgery otherwise. The benefit of testing depends on the individual’s attitudes and carrier probability, and for some it can result in additional life expectancy. Genetic counseling should include the ramifications of positive genetic test results and of negative test results. BRCAPRO facilitates counseling by addressing the probabilities of both types of results. It also elucidates the individual’s circumstances in the case of a negative genetic test. If the test has 100% specificity, then an individual whose test is positive has a 100% probability of carrying a mutation. However, if her test is negative, then her carrier probability is not 0; it depends on the test’s sensitivity and on her carrier probability before the test (eg, from BRCAPRO). Assume sensitivity equals 85%. Her carrier probability after a negative test ("negative predictive value") is shown in Table 8.
A woman with a 20% pretest carrier probability will go to 100% with a positive test and 4% with a negative test. However, a woman with a large carrier probability will still have a large carrier probability even after a negative test. Such women with a negative test will be difficult to counsel, and our study shows that they are not rare (15 probands had carrier probabilities > 95% and yet tested negative). Is it a case of a missed mutation? A mutation at another susceptibility gene? A woman with a pretest probability of 99% would reasonably consider prophylactic interventions even if her test is negative. If her decisions will be the same for both positive and negative results, then the test has little value. Probands of special interest in counseling are those unaffected by disease. Unaffected women usually have carrier probabilities less than 50%. Moreover, for any given family history, a woman’s carrier probability gets smaller with increasing age as long as she is disease-free. As Table 8 shows, testing is most informative for women who have moderate carrier probabilities, and because unaffected women with moderate carrier probabilities tend to be young, they also have the most to gain from genetic testing in terms of expected incremental quality-adjusted life-years,18 depending on their attitudes toward prophylactic procedures. The average carrier probability over the entire population is the sum of the prevalences of BRCA1 and BRCA2. Therefore, in the general population, most women’s carrier probabilities are less than 1%. These women would probably be advised to not have genetic testing, at least in part because of their low carrier probability. However, the number of such women is large and so many who have been advised against testing are indeed carriers. Some will discover their carrier status after being diagnosed with disease. This is an inescapable consequence of genetic counseling, with or without providing carrier probabilities. We conclude, first, that for individuals who present at genetic counseling clinics, BRCAPRO provides an accurate assessment of probability of carrying a deleterious mutation of BRCA1 and BRCA2. Multiplying this probability by the sensitivity of the testing method used gives an accurate estimate of the probability of testing positive. Moreover, although an individual’s carrier probability is affected by the estimates of penetrance and prevalence used, the accuracy of BRCAPRO is not dependent on having accurate estimates of these quantities. Second, methods for genetic testing for BRCA1 and BRCA2 used in our study have high sensitivity, which we estimate to be at least 85%. Finally, if there exist breast cancer susceptibility genes with penetrance comparable to those of BRCA1 and BRCA2 in the population we studied, then mutations at these genes are rare.
Supported in part by grant no. P50 CA68438 from the National Cancer Institute through the Specialized Program of Research Excellence in breast cancer at Duke University.
1. Parmigiani G, Berry DA, Aguilar O: Determining carrier probabilities for breast cancer susceptibility genes BRCA1 and BRCA2. Am J Hum Genet 62: 145-158, 1998[CrossRef][Medline] 2. Parmigiani G, Berry DA, Iversen E, et al: Modeling risk of breast cancer and decisions about genetic testing, in Gatsonis C, Carlin B, Carriquiry A, et al (eds): Case Studies in Bayesian Statistics IV. New York, NY, Springer, 1999, pp 173-268
3.
Berry DA, Parmigiani G, Sanchez J, et al: Probability of carrying a mutation of breast-ovarian cancer gene BRCA1 based on family history. J Natl Cancer Inst 89: 227-238, 1997 4. Szabo CI, King M-C: Population genetics of BRCA1 and BRCA2. Am J Hum Genet 60: 1013-1020, 1997[Medline] 5. Iglehart JD, Miron A, Rimer BK, et al: Overestimation of hereditary breast cancer risk. Ann Surg 228: 375-384, 1998[CrossRef][Medline]
6.
Bluman LG, Rimer BK, Berry DA, et al: Attitudes, knowledge and risk perceptions of women with breast cancer considering testing for BRCA1 and BRCA2. J Clin Oncol 17: 1040-1046, 1999
7.
Ponder B: Genetic testing for cancer risk. Science 278: 1050-1054, 1997
8.
Lerman C, Lustbader E, Rimer B, et al: Effects of individualized breast cancer risk counseling: A randomized trial. J Natl Cancer Inst 87: 286-292, 1995
9.
Gail MH, Brinton LA, Byar DP, et al: Projecting individualized probabilities of developing breast cancer for white females who are being examined annually. J Natl Cancer Inst 81: 1879-1886, 1989 10. Claus EB, Risch N, Thompson WD: Genetic analysis of breast cancer in the Cancer and Steroid Hormone Study. Am J Hum Genet 48: 232-242, 1991[Medline] 11. Claus EB, Risch N, Thompson WD: Autosomal dominant inheritance of early-onset breast cancer: Implications for risk prediction. Cancer 73: 643-651, 1994[CrossRef][Medline] 12. Frank TS, Manley SA, Olopade OI, et al: Sequence analysis of BRCA1 and BRCA2: Correlation of mutations with family history and ovarian cancer risk. J Clin Oncol 16: 2417-2425, 1998[Abstract]
13.
Couch FJ, DeShano ML, Blackwood MA, et al: BRCA1 mutations in women attending clinics that evaluate the risk of breast cancer. N Engl J Med 336: 1409-1415, 1997 14. Hartge P, Struewing JP, Wacholder S, et al: The prevalence of common BRCA1 and BRCA2 mutations among Ashkenazi Jews. Am J Hum Genet 64: 963-970, 1999[CrossRef][Medline]
15.
Shattuck-Eidens D, Oliphant A, McClure M, et al: BRCA1 sequence analysis in women at high risk for susceptibility mutations. JAMA 278: 1242-1250, 1997 16. Iversen ES, Parmigiani G, Berry DA, et al: Genetic susceptibility and survival: Application to breast cancer. J Am Stat Assoc 95: 28-42, 2000[CrossRef] 17. Murphy EA, Mutalik GS: The application of Bayesian methods in genetic counseling. Hum Hered 19: 126-151, 1969 18. Tengs TO, Winer EP, Paddock S, et al: Testing for BRCA1 breast-ovarian cancer susceptibility gene: A decision analysis. Med Decis Making 18: 365-375, 1999
19.
Claus EB, Schildkraut J, Iversen ES Jr, et al: Effect of BRCA1 and BRCA2 on the association between breast cancer risk and family history. J Natl Cancer Inst 90: 1824-1829, 1998 20. Iversen ES Jr, Parmigiani G, Berry DA, et al: Genetic susceptibility and survival: Application to breast cancer. J Am Stat Assoc 95: 28-42, 2000 21. Zhou X, Parmigiania G, Iversen E: Disease causality of missense mutations: A hierarchical approach. Genet Epidemiol 19: 279, 2000 (abstr) 22. Gilpin CA, Carson N, Hunter AG: A preliminary validation of a family history assessment form to select women at risk for breast or ovarian cancer for referral to a genetics center. Clin Genet 58: 299-308, 2000[CrossRef][Medline] 23. Hiller EH, Parmigiani G, Berry DA, et al: Validity testing of a computer model for calculating carrier probabilities of BRCA1 and BRCA2 mutations. Presented at the Annual Meeting of the American Society of Human Genetics, Baltimore, MD, October 28-November 1, 1997 24. Hilsenbeck SG, Wenger CR, Reynolds LP, et al: Early results of genetic testing for breast cancer susceptibility. Breast Cancer Res Treat 46: 38, 1997 (abstr) 25. Iversen ES, Parmigiani G, Berry DA: Validating Bayesian prediction models: A case study in genetic susceptibility to breast cancer, in Gatsonis C, Carlin B, Carriquiry A, et al (eds): Case Studies in Bayesian Statistics IV. New York, NY, Springer, 1999, pp 321-338
26.
FitzGerald MG, MacDonald DJ, Krainer M, et al: Germ-line BRCA1 mutations in Jewish and non-Jewish women with early-onset breast cancer. N Engl J Med 334: 143-149, 1996 27. Love RR, Evans AM, Josten DM: The accuracy of patient reports of a family history of cancer. J Chronic Dis 38: 289-293, 1985[CrossRef][Medline] 28. Oddoux C, Struewing JP, Clayton CM, et al: The carrier frequency of the BRCA2 6174delT mutation among Ashkenazi Jewish individuals is approximately 1%. Nat Genet 14: 188-190, 1996[CrossRef][Medline] 29. Roa BB, Boyd AA, Volcik K, et al: Ashkenazi Jewish population frequencies for common mutations in BRCA1 and BRCA2. Nat Genet 14: 185-187, 1996[CrossRef][Medline] 30. Ford D, Easton DF, Stratton M, et al: Genetic heterogeneity and penetrance analysis of the BRCA1 and BRCA2 genes in breast cancer families. Am J Hum Genet 62: 676-689, 1998[CrossRef][Medline]
31.
Struewing JP, Hartge P, Wacholder S, et al: The risk of cancer associated with specific mutations of BRCA1 and BRCA2 among Ashkenazi Jews. N Engl J Med 336: 1401-1408, 1997 32. National Cancer Institute: Surveillance, Epidemiology, and End Results (SEER). Bethesda, MD, National Cancer Institute, http://www.seer.cancer.gov, 1996 33. Ford D, Easton DF: The genetics of breast and ovarian cancer. Br J Cancer 72: 805-812, 1995[Medline] 34. Ford D, Easton DF, Peto J: Estimates of the gene frequency of BRCA1 and its contribution to breast and ovarian cancer incidence. Am J Hum Genet 57: 1457-1462, 1995[Medline] 35. Thoriacius S, Struewing JP, Hartge P, et al: Population-based study of risk of breast cancer in carriers of BRCA2 mutation. Lancet 352: 1337-1339, 1998[CrossRef][Medline] Submitted May 29, 2001; accepted March 15, 2002.
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2002 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
gparmigi/brcapro.html.) Prevalence and penetrance estimates used in BRCAPRO in the present validation study were arrived at independent of the families considered in this study. 
, male; 
, female; /, deceased; 
, cancer; BC# (or OC#), age at breast (or ovarian) cancer diagnosis; age#, current age or age at death (when known).
