A 40-Year-Old Woman With a Strong Family History of Breast Cancer

DR DALEY: Mrs T is a 40-year-old married mother of 3 children with a strong family history of breast cancer who is considering genetic testing and tamoxifen therapy. She lives on Cape Cod, Mass, and works at home caring for her 3 children, ages 7, 5, and 2. Her health insurance is provided through a managed care plan.

Mrs T is in good health. She is of Dutch and English ancestry. Her maternal grandmother was diagnosed with breast cancer in her seventh decade and was treated with surgery. She died of causes unrelated to breast cancer. Mrs T's mother developed breast cancer and died of complications related to breast cancer metastatic to bone. Her sister developed breast cancer at age 26 and died at age 38 of complications related to metastatic breast cancer to lymphatics and lung. Mrs T has had yearly mammography since age 23 and semiannual breast examinations. She practices breast self-examination. Mrs T has decided not to have genetic testing because she is aware that negative test results do not assure her that she will not develop breast cancer. She says that having a prophylactic mastectomy would be a "big" decision. After several months of consideration, she recently began a course of tamoxifen and has not experienced any noticeable adverse effects. She is ambivalent about genetic testing, wants to know what the future of "gene therapy" is, and is curious about what the future holds for her 2-year-old daughter.

Her past medical history is not significant. She drinks socially and smokes about one-half pack of cigarettes daily. Her only current medication is tamoxifen.

Physical examination in the fall of 1998 revealed a thin woman in no acute distress. General physical examination findings were normal. Breast examination findings were normal without palpable masses and no axillary masses. Mammogram was negative for masses or calcifications.

MRS T: At 23, I started going for my yearly mammogram. I told my gynecologist that I wanted to be checked by him twice a year instead of once a year. When I heard about other options, specifically the tamoxifen program, I thought maybe I shouldn't be so passive, that I should try one more step. But there's always that fear in the back of my mind that maybe I'm going to get it someday. Or, I think "When am I going to get it?" I think many women feel like that who have histories—when am I going to get it and how prepared will I be?

There are several reasons why I decided not to have genetic testing. I don't know if I could handle making a decision like having a radical mastectomy on both sides. I thought the tamoxifen was a middle of the road decision, where at least my chances might be like the ordinary population that doesn't have a history. Why go through an operation which might be invasive to the cancer if you had it anyhow? Sometimes when people are operated on, it makes things go along a little faster. It would be a very tough decision. If my doctors thought it would give me a better chance, I definitely would do it. But I don't know if I want to face that decision now while I'm healthy. I don't know if I could do it, unless I knew something was wrong.

What percentage of healthy women who decide to have their breasts removed still get cancer? What would my daughter's success rate be, if I got cancer, what would be her chance of survival? If I had the gene testing and I found that I did have some kind of mutation, should I have her tested? Would she want to have children knowing that she's going to pass this down? But then again, you don't always have to get the cancer just because you have the gene. And should you be denied a life?

[Editor's note: Mrs T's oncologist was unavailable for interview.]

What data are available regarding the genetic determinants of different types of cancer? What are the operating characteristics of the tests? How do you counsel patients considering such tests? What are the roles of tamoxifen, raloxifene, or prophylactic mastectomy? What are the risks and benefits of these treatments? What data are available on the impact of a patient's decision to be tested? How does genetic predisposition to cancer affect a patient's health, life, or insurance status? What role does confidentiality play in genetic testing?

Mrs T's situation highlights both the remarkable power and significant limitations of currently available information for women at increased risk of breast cancer on the basis of a strong family history. Like many women dealing with breast cancer risk, she is forced to make the best decisions she can with incomplete information, while recognizing that the information and her decisions may change over time.

Breast cancer has been the subject of extensive epidemiologic investigation. Established risk factors fall into several categories. Reproductive factors, such as early age at menarche, nulliparity, late age at first pregnancy, and late age at menopause, all speak to the fundamentally hormonal nature of breast proliferation and neoplasia.¹ Particular histologic diagnoses of biopsied breast tissue are associated with increased risk of subsequent breast tissue neoplasia, including atypical hyperplasia, lobular carcinoma in-situ, and most recently, radial scar.^{2 - 3} Particular lifestyle factors may increase breast cancer risk, including alcohol intake and weight gain after menopause.⁴ Family history has long been recognized as a risk factor and is the most prominent issue for the patient under discussion today.

One of the major challenges breast cancer researchers have faced has been estimating risk for individual patients. Gail and colleagues⁵ at the National Cancer Institute developed a multivariate model of risk estimation from data from the Breast Cancer Detection and Demonstration Project, a study of women undergoing annual mammographic screening. The model includes reproductive factors, family history, and a novel factor, number of breast biopsies, with modifications for biopsy result and race. The Gail model is distributed by the National Cancer Institute in an enhanced format with information about risk (order at http://cancertrials.nci.nih.gov/) and has been validated in other large cohorts of women undergoing regular surveillance.^{6 - 7} The model is not a good way to assess risk for women whose risk appears to be primarily based on hereditary factors, because the breast cancer history it utilizes is restricted to first-degree relatives (mothers, sisters, and daughters); it does not incorporate information about ethnicity or age at diagnosis, and it considers breast cancer exclusively. Its utility is therefore limited for women with paternal inheritance of susceptibility and those with ovarian cancer in close relatives or other cancers that have been linked to breast cancers in syndromes of inherited cancer predisposition.

Since only an estimated 5% of breast cancers have a strong hereditary component, the majority of family clusters will not be attributable to 1 of the currently recognized susceptibility genes. The model developed by Claus et al⁸ is the best way to estimate breast cancer risk for the women whose risk appears, on review of the pedigree, to derive primarily from family history. In the Claus model, the ages at which close relatives (first- and second-degree in 1 lineage) were diagnosed with breast cancer are used to estimate the probability that a hereditary component underlies the pattern of cancers in the family and provide the basis for determining absolute breast cancer risk estimates for an individual patient. The model is generally not intended for families considered likely to carry predisposing mutations in highly penetrant susceptibility genes like BRCA1 or BRCA2 because it can underestimate risk if the family history is paternal, if the patient has no sisters, or if there is ovarian cancer instead of breast cancer among affected relatives.

Patients concerned about family history of breast cancer often overestimate their own risk of developing the disease,⁹ and the Claus model may be useful in providing more reassuring quantitative estimates. The Claus model is applicable to Mrs T because her family history features breast cancer only and the affected relatives are in the maternal lineage. The risk of breast cancer for Mrs T according to the Claus model would be 14.6% by age 49 years and 43.4% by age 79 years.⁸ These relatively high figures are influenced primarily by the remarkably young age (26 years) at diagnosis of her sister.

Mrs T's family breast cancer history should also be evaluated for the probability that a mutation in a currently recognized susceptibility gene might underlie the cancer pattern in the family. The first step is to construct an extended pedigree, including both maternal and paternal relatives, that should extend to include second-degree relatives (grandparents, aunts, and uncles) and, if possible, cousins. Data collection should include cancer types, sites of disease, and age at diagnoses for each affected relative. A medical record or death certificate documentation generally should be obtained, particularly for nonbreast cancer diagnoses.¹⁰ Studies have demonstrated that individuals are more accurate in recounting cancer history for first-degree than for more distant relatives and for breast than for other primary sites of disease.¹¹

Certain features of a pedigree suggest that a mutation in the BRCA1 or BRCA2 genes might be present in affected individuals. These features are derived from the model of inherited cancer susceptibility proposed by Knudson¹² and validated by the identification of tumor suppressor genes. They include early age at breast cancer diagnosis, multifocal or bilateral disease, clustering of breast or breast and ovarian cancers among family members in 1 lineage, and breast and ovarian cancers as multiple primary cancers in a single family member. Unaffected male relatives generally provide less information in the case of a sex-limited trait like breast cancer. However, the presence of a male relative with breast cancer can suggest the presence of a BRCA2 mutation, since male breast cancer is part of the complex phenotype associated with mutations in that gene,¹³ but not with other currently identified breast cancer susceptibility genes.

Another characteristic useful in the evaluation of cancer families is ethnic derivation. Founder mutations have been identified in both BRCA1 and BRCA2. The founder effect is the occurrence of an increased frequency of 1 or more specific mutations in a population because of the presence of such mutations in an ancestor. Factors contributing to a founder mutation include migration, population isolation (a founder mutation in BRCA2 accounts for the vast majority of hereditary breast cancer in Iceland¹⁴ ), or rapid population shrinkage followed by reexpansion in which the mutation is carried by a survivor. This last mechanism is thought to underlie the highly prevalent founder mutations (2 in BRCA1 and 1 in BRCA2) in the Jewish population, in which their combined prevalence is 2.5% compared with less than 0.1% frequency of any BRCA1 or BRCA2 mutation in the general population.^{15 - 16} Knowledge that a patient has a particular ethnic heritage may guide strategies for mutation analysis by targeting assessment to the common founder mutations before full gene evaluation. Among Jews of Eastern European descent, for example, the 3 founder mutations are thought to account for more than 90% of BRCA1/BRCA2 mutations.¹⁶

Models now exist to identify individuals and families more likely to carry a BRCA1 or BRCA2 mutation. Couch et al¹⁷ published the first useful tables, derived from individuals visiting risk assessment clinics, for estimating the probability that a BRCA1 mutation is present in a family. A few percentage points must be added for BRCA2, which was not measured in the study. The model considers average age at breast cancer diagnosis, Ashkenazi Jewish ancestry, and the presence of ovarian cancer in the kindred and provides a set of tables with probabilities that a mutation is present in a cancer patient. Because the tables are based on empiric data, but do not take advantage of the full information present in a pedigree, they are most useful as first approximations. A second empiric model has been derived by Frank et al¹⁸ using data from Myriad Genetics Laboratories, Salt Lake City, Utah.

A computer program, BRCAPRO, has been generated to calculate Bayesian probabilities from more complete pedigree data.¹⁹ The model uses epidemiologic estimates of mutation frequency rather than empiric data in its calculations and considers both affected and unaffected family members. These Bayesian calculations are conditional probabilities, as are used by geneticists to estimate carrier probabilities for most hereditary disorders. The program currently is undergoing validation (David Euhus, MD, written communication [http://www.swmed.edu/home_pages/cancergene/Accessed October 26,1999]).

There is no uniform agreement as to what constitutes a sufficient probability to make testing appropriate. The American Society of Clinical Oncology has recommended a 10% prior probability, which at least provides 1 landmark to consider.²⁰ Probabilities much below 10% should suggest nongenetic explanations for the pattern of cancer in the family under consideration.

Analysis of the BRCA1 and BRCA2 genes is technically demanding because they are very large genes with some unusual properties. In the United States, because of patent issues, most testing is performed in a single commercial laboratory (Myriad Genetics Laboratories), that performs direct analysis of the nucleotide sequences of the genes. The technology is highly accurate but can miss certain kinds of mutations, such as large deletions. Several other techniques for mutation analysis have their own advantages and limitations.²¹ A persistent technical challenge for the laboratory is the interpretation of the large number of unique mutations identified. The laboratory must distinguish the deleterious mutations (those conferring increased cancer susceptibility) from normal variants (polymorphisms) and possible "low penetrance" mutations, which would be associated with smaller increases in cancer risk. This may require additional blood samples from close family members to assess whether the "abnormality" came from the affected lineage or review of the mutation database to ascertain whether the mutation has been noted in other affected families (in which case, it is likely deleterious) or is seen with a deleterious mutation in the same individual (in which case, it is likely not deleterious). This dilemma arises most often for missense mutations, in which a base substitution changes an amino acid but does not clearly result in premature truncation of the BRCA1 or BRCA2 protein.

Precise estimates of the sensitivity and specificity of commercially available BRCA1/BRCA2 assays are not available but are likely to be high. However, it is important that usual rules guiding clinical practice apply to the evaluation of genetic tests. That is, the test should be performed only when the prior probability that an individual whose genes are to be analyzed carries a germline BRCA1 or BRCA2 mutation is sufficient to minimize the rate of false-positives. Genetic test results are rarely confirmed independently, and because germline mutations do not change, each test needs to be performed only once, carefully, in the lifetime of an individual.

The interpretation of genetic tests may be limited because more than 1 gene can account for familial clustering. Before BRCA1 was identified, segregation analyses of breast cancer kindreds revealed an autosomal dominant pattern of inheritance of breast cancer susceptibility with an allele frequency of 0.0033 (carrier frequency, 1 in 152) in the population.²² To date, autosomal dominantly inherited syndromes of breast cancer susceptibility have been associated with mutations in several genes, including BRCA1, BRCA2, p53, and PTEN.^{23 - 24} A sizeable portion of hereditary disease is not accounted for by these genes, so the search continues for BRCA3 and other strongly predisposing genes.²⁵ Lower penetrance variants of hormone or carcinogen-metabolism genes are expected to exert much less effect on individual breast cancer risk, but, because of their relative frequency in the population, they are likely to account for a greater proportion of breast cancer overall.²⁶

The exact proportion of hereditary breast cancer accounted for by the known genes has not yet been determined. Although initial studies of highly selected families estimated that BRCA1 and BRCA2 would account for the great majority of breast cancer families, subsequent analyses have found much lower mutation frequencies. Among younger women, Claus et al²² estimated that 36% of breast cancers diagnosed in women aged 20 through 29 years and 15% in women diagnosed at age 30 through 39 years would be associated with autosomal dominantly inherited predisposition genes, although in women diagnosed with breast cancer before age 35 years, only 6% to 8% have had identifiable mutations in BRCA1.^{27 - 28}

If BRCA1 and BRCA2 are excluded, cancer family syndromes that include breast cancer have each accounted for less than 1% of hereditary breast cancer. The syndromes and their associated genes are listed in Table 1. Commercial testing is available for TP53 (Li-Fraumeni syndrome), and Clinical Laboratory Improvement Act (CLIA)–approved testing will become available for PTEN (Cowden syndrome). Because of their low prevalence among individuals identified by breast cancer alone, a clinician would only consider doing genetic testing for mutations in these genes when other characteristic features of their syndrome might be present. For Li-Fraumeni syndrome, testing might be considered more often when an early onset breast cancer occurred in a woman with a personal or family history of soft tissue or osteosarcoma, or a childhood cancer.²³ Nonmalignant features suggesting Cowden syndrome, including skin tricholemmomas and thyroid abnormalities, would be required for most clinicians to consider PTEN analysis.²⁴

The genetic heterogeneity of breast cancer mandates that the affected family member be tested first when possible, and only if a pathogenic mutation in BRCA1 or BRCA2 is identified, then other blood relatives should be tested. If a mutation has not been identified in an affected family member, a negative BRCA1/BRCA2 test result is not necessarily reassuring.

Since neither Mrs T's mother nor her affected sister is living to provide a blood sample for genetic analysis, their full gene analysis cannot be performed, and the germline mutation underlying their breast cancers cannot be identified, if one exists. This situation limits the value of genetic testing for Mrs T. If she were to have her BRCA1 and BRCA2 genes analyzed and a deleterious mutation were identified, the result would be informative, and, given the rarity of germline mutations, would be highly likely to be the same mutation presumed to have been carried by Mrs T's sister and mother. However, mutation-specific cancer risk information might not be available because most mutations identified remain unique to specific families.²⁹ A negative genetic test result, however, would provide only limited information to Mrs T, since it would not distinguish among the absence of the BRCA1 or BRCA2 mutation carried by her sister (a true-negative test), the absence of a mutation in BRCA1, BRCA2, or a different gene actually present but not detected (a false-negative test), and the absence of an underlying hereditary risk in this kindred (all breast cancers in the relatives may have been sporadic). Therefore, genetic testing cannot bring Mrs T a definitive negative result. She can receive a positive result, which she would consider bad news, or a qualified negative result with no real relief from uncertainty. Clearly her oncologist had counseled Mrs T about these issues, because they formed the expressed basis for her decision not to undergo testing at this time. In addition, the possibility of another kind of indeterminate result, a missense mutation for which pathogenicity cannot be clearly determined, was likely discussed with Mrs T. This result is frequently encountered in analysis of the large BRCA1 and BRCA2 genes, which are expected to contain numerous variants that may confer either no or little increase in breast cancer risk.³⁰

In addition to medical considerations, in the course of genetic counseling women learn about the potential risks and benefits of testing, as well as potential limitations of genetic information and the status of genetic privacy protection in their state of residence. Although most states do not have specific genetic privacy legislation, only a few cases of health insurance discrimination have been documented to date. Women are often asked to consider the effects of genetic information on their psychological state and self-image, on other family members and family dynamics, and on their own insurability and employability.³¹ Life and disability insurance coverage is expected to be more vulnerable than health insurance for mutation carriers.³² In classic genetic counseling, women are asked to rehearse the impact of a positive or negative test result on their concerns for their children (testing is rarely performed on minor children, for example) or their interactions with other at-risk family members. Discordant results (1 negative, 1 positive) for sisters can be difficult for all concerned.³³ Mrs T has had the opportunity to learn quite a bit about genetic information and to understand its potential advantages and disadvantages. It is not unusual for women to express ambivalence about their decisions regarding genetic testing or to remember or incorporate only part of the information they receive in the course of risk counseling.³⁴ For example, Mrs T expresses the conviction that breast cancer is inevitable for her, although she surely was counseled about the incomplete penetrance of predisposing mutations. Most (85%) women participating in studies of genetic counseling demonstrate good adjustment to genetic information over the ensuing 3 to 9 months after testing, assessed by validated psychological measures.^{33 - 34} Unfortunately, there are no established means of identifying women in advance who will have difficulty coping with their test result, whether positive or negative. However, women from high-risk families who elect to forego testing have been found to exhibit more distress on follow-up, perhaps because of the persistence of uncertainty or heightened concern about a presumed positive result.³⁵

Several options are available for managing the treatment of women with strong family histories of breast and/or ovarian cancer or women with predisposing mutations.^{36 - 37} Options for medical management of increased breast and ovarian cancer risk fall into 3 categories: surveillance, prophylactic surgery, and other prevention strategies. Ironically, the data are most limited for surveillance. The efficacy of screening mammography is virtually unknown for women younger than 40 years and limited for women aged 40 through 49 years.³⁸ Whether more recent technologies—digital mammography or breast magnetic resonance imaging—will be more sensitive than standard mammography for high-risk women in particular remains to be proven. While mammography reduces breast cancer mortality for women aged 50 years and older, no data show that ovarian ultrasound scans with any level of sophisticated technology reduces mortality from ovarian cancer—or even downstages cancers—in women in any group.³⁹ Measuring cancer antigen (CA) 125 blood levels in addition to ultrasound imaging may improve the probability of detecting early ovarian cancer, but the data in high-risk populations remain sparse.⁴⁰ Nonetheless, many women at risk and their physicians will choose more intensive surveillance (clinical breast examinations, mammograms, vaginal ultrasound, and serum CA 125 assays) as strategies for some portion of the women's lives.

Prophylactic mastectomy has been studied in high-risk patients, although not in women with BRCA1 or BRCA2 mutations. Among 214 high-risk women and their 403 sisters,⁴¹ prophylactic mastectomy was found to reduce the incidence of breast cancer by approximately 90% (95% confidence interval [CI], 76.6%-98.3%). Furthermore, prophylactic mastectomy reduced the risk of death from breast cancer by 80.9% (95% CI, 31.4%-97.7%). The procedures performed were predominantly subcutaneous mastectomies, in which the nipple-areola complex is retained. More recently, this procedure has been replaced by the skin-sparing mastectomy to reduce remaining breast tissue still further. These data provide the most reliable estimate of the efficacy of prophylactic mastectomy and should be made available to women who wish to consider the range of options available to them.⁴¹

Unfortunately, the protective effect of prophylactic oophorectomy has not yet been quantified. Even if oophorectomy is performed, tumors that are indistinguishable from serous carcinoma of the ovary can arise, presumably from cells in the peritoneum.⁴² Studies published to date are inconsistent; Piver et al⁴³ identified only 6 cases of primary peritoneal carcinomatosis among 324 women followed up for 1 to 27 years after prophylactic oophorectomy of histologically normal ovaries, while Struewing et al⁴⁴ observed 8 cases of carcinomatosis among 44 oophorectomized women, a 50% risk reduction in the 460 person-years of follow-up. Despite the lower risk of ovarian cancer for BRCA1/BRCA2 mutation carriers at every age compared with breast cancer risk, the perception of ovarian cancer as an incurable and singularly unpleasant disease, coupled with the relative ease of laparoscopic oophorectomy, has made oophorectomy a popular option among mutation carriers in most series.³⁴ Among our patients, 40% of mutation carriers had undergone prophylactic oophorectomy by 6 months after receipt of their genetic test result.

Mrs T has chosen an option only recently shown to reduce breast cancer risk: chemoprevention. Tamoxifen, a synthetic estrogen receptor modulator (agonist/antagonist), reduced the risk of estrogen-positive breast cancer by 49% in the recently published Breast Cancer Prevention Trial (BCPT).⁴⁵ In this multicenter, randomized, placebo-controlled trial, 13,338 eligible women were assigned to take either tamoxifen or placebo for 5 years. The trial was stopped early by the data safety monitoring committee because of a significant effect of tamoxifen on breast cancer incidence and because of important data on tamoxifen-associated adverse effects. The findings were remarkably close to the 47% reduction in contralateral breast cancer incidence demonstrated for tamoxifen in the overview analysis of data from the adjuvant breast cancer treatment trials.⁴⁶

For Mrs T, several additional pieces of data would be considered in the decision to administer tamoxifen to reduce breast cancer risk. First, has the efficacy of tamoxifen been demonstrated for women with a strong family history of breast cancer?

In the double-blind, randomized, placebo-controlled Royal Marsden Trial of tamoxifen for breast cancer prevention in women with a strong family history, no effect of tamoxifen was demonstrated among 2494 participants.⁴⁷ However, in the larger National Surgical Adjuvant Breast and Bowel Project P-1 study (NSABP), approximately 19% of the cohort had 2 or more first-degree relatives with breast cancer, and the effects of tamoxifen in those groups were similar to the overall risk reduction for the full cohort: risk ratios 0.55 (95% CI, 0.30-0.97) for women with 2 affected first-degree relatives, and 0.51 (95% CI, 0.15-1.55) for the group with 3 or more affected family members.⁴⁵ Mrs T would have been eligible for participation in the BCPT trial based on her family history of breast cancer in 2 close relatives and her delayed age at first pregnancy (Gail model risk estimate, 1.7% over 5 years⁵ ).

Breast cancers among women with germline BRCA1 mutations are more often hormone receptor negative.⁴⁸ As a result, tamoxifen might reduce a smaller proportion of the breast cancers among women with BRCA1 mutations. A molecular substudy of the BCPT will evaluate the effect of tamoxifen for BRCA1/BRCA2 mutation carriers, but data are not yet available. Recent data from a cohort of BRCA1 and BRCA2 mutation carriers suggest that prophylactic oophorectomy reduces the risk of breast cancer as well, suggesting a role for ovarian hormones in breast cancer development in mutation carriers.⁴⁹ For BRCA1/BRCA2 carriers, oral contraceptive agents have been shown to reduce ovarian cancer risk.⁵⁰ However, concerns exist as to whether oral contraceptives might increase breast cancer risk in this high-risk group.^{50 - 51}

Since tamoxifen use as a chemopreventive is limited to 5 years, in which 5 years should it be taken? The potentially serious adverse effects of tamoxifen—uterine cancer and blood clots—occurred more frequently among women older than 50 years in the BCPT, so tamoxifen may be preferred for younger women when feasible. Given these factors, the decision by Mrs T to take tamoxifen to reduce her risk of breast cancer is certainly reasonable. However, by taking tamoxifen, premenopausal women also lose the potential benefits of estrogen, which may have an effect on their quality of life. The decision to take tamoxifen is, therefore, an individual decision that requires careful balancing of potential risks and benefits in the face of incomplete data.

Recent data have demonstrated that raloxifene, another of the agents now called synthetic estrogen receptor modifiers, can similarly reduce breast cancer incidence.⁵² These data are exciting, but preliminary, particularly since the risk of uterine cancer with raloxifene appears to be lower than has been observed with tamoxifen. However, raloxifene would not be an option for Mrs T at this juncture because she is premenopausal, and there are no data on the use of raloxifene in populations other than postmenopausal women. Furthermore, raloxifene has not been evaluated in a population with increased risk of breast cancer, although studies are ongoing.

Mrs T raised questions about the ability of gene therapy to correct the genetic alteration conferring such marked cancer risk. This remains a goal of the future, but one that is the subject of intense study at this time.

Genetic testing for cancer predisposition is increasingly becoming a part of medical care. Physicians are expected to recognize families in which hereditary risk may exist and to inform their patients of the risk. They will also find themselves being asked to help guide their patients through rather complex medical decisions. Physicians should consider referral to genetic counselors or other genetic professionals for detailed discussion of the risks and benefits of testing. The set of options facing Mrs T were complex in the absence of genetic information, but would not necessarily have been simplified by test results. Nonetheless, it seems reasonable to be optimistic that the power of genetic information will ultimately result in effective and acceptable strategies for cancer prevention and perhaps for improved cancer treatment. The societal obligation to make it safe for individuals to use this information remains a priority.

A PHYSICIAN: What is the approximate cost of the testing?

DR GARBER: For Mrs T, who is not of Jewish descent, the cost would be approximately $2400. She could request reimbursement from her insurance company. Many companies cover at least a portion of test fees. However, many people choose not to request reimbursement because of concerns about privacy. If she were of Jewish descent, a more limited analysis would cost about $475.

AN INTERNIST: It's hard to process material like this, even for a sophisticated audience. Is anything known about how different populations of people process this information and how stable they are in that?

DR GARBER: Because of the high cost of testing, most people studied for testing tend to come from a single socioeconomic background. They are the more educated people in any of the studies. For less educated and less socioeconomically advantaged individuals, there often are other issues. In minority populations, genetic testing carries heavy baggage from the Tuskegee and sickle cell days. Most evaluations of the understanding of genetic information occur in highly structured programs that incorporate the time to give this information appropriately. At the moment, if the question is how well people do with genetic information in general, I think you'd be better off studying genetic conditions other than cancer to look for how people learn.

A PHYSICIAN: It's often tempting to shift the burden of decision making from the doctor to the patient: then the doctor can stop worrying. But a patient's decision about such difficult things is often even more difficult because the patient is emotionally involved, and the decision might not be correct. Mrs T seems to have resolutely avoided the decision to have genetic testing. I wonder whether she received advice about that from an intelligent, impartial doctor or not. I wonder what your advice would be.

DR GARBER: The decisions about cancer risk management may only begin with decisions about genetic testing. For breast cancer, mutation carriers must consider surveillance as well as prophylactic mastectomies. How directive can a physician be about such personal decisions? The whole field is struggling with this issue. In our group, we have tried not to force people to have testing, but to help them focus on what the information will provide them, and what they'll do differently if they know one way or the other. We encourage people to consider testing as part of shared decision making. We have become more directive about what we tell people to do if they test positive. The data show that the 85% of people tested do well psychologically thereafter, but they decided themselves that they want to be tested. When a patient tells me she is ready, then it's fine. It's easier for us to discourage testing. For people who come and say, "I want to be tested—I need to know," you can often tell them that this test is not that informative and discourage them from being tested if it does not seem appropriate.

A PHYSICIAN: May I directly ask you, if this were you, what would you do about genetic testing?

DR GARBER: What's right for me is not necessarily right for somebody else. I hope this patient will have the opportunity to reconsider her decisions over time. She may change her mind some day.