Genetically Tailored Preventive Strategies: an Effective Plan for the Twenty-first Century?¹

Genetically tailored therapy has become essential for treating cancer and other chronic diseases. For example, the antileukemic agent TG³ exerts its cytotoxicity only after methylation by the enzyme TPMT to form thiopurine nucleotides. Individuals with relatively inactive variants of the TPMT enzyme metabolize TG poorly. As a result, their hematopoietic cells accumulate high levels of TG nucleotides, which can lead to severe toxicity (1). Thus, the dose and type of chemotherapy must be tailored to accommodate individual genetic variation in TPMT activity.

Less well established is genetically tailored intervention to prevent the occurrence of disease. As more genes with predisposing alleles are identified and as people become increasingly aware of genetic developments and are interested in knowing about their own genes, there is need to offer them options for preventing the diseases to which they are particularly susceptible by inheritance. This article reviews our present arsenal of strategies for the primary prevention of cancers of the colorectum, female breast, and ovary, with particular reference to strategies available to those at high risk because of inherited susceptibility.

Imagine a patient/physician conference on options for cancer prevention that might occur at a routine annual physical examination in the year 2025. Fig. 1 shows a possible scheme for such decision-making. The patient would be classified according to his or her level of risk for a particular cancer. Those at low risk would consider lifestyle changes to reduce their risk even further. They would not, however, undertake specific chemopreventive regimens unless they could be reasonably sure that the regimes impose no unacceptable side effects. Those at high risk, in contrast, would consider one or more of three types of preventive options: lifestyle changes, chemoprevention, and prophylactic surgery. Because a person’s risks and responses to chemopreventive agents may depend on his or her unique genetic constitution, the risk-benefit tradeoff will be unique to each patient. For example, the magnitudes of breast cancer risk reduction afforded by a preventive agent such as tamoxifen may be substantially larger for BRCA1 mutation carriers than for noncarriers. Thus, the potential side effects of the agent may be acceptable to carriers but unacceptable to others. To make informed decisions, patients will need detailed information on the magnitudes of the cancer risks both with and without a proposed intervention and detailed information on the risks associated with any adverse effects of the intervention.

This paper describes our present progress in obtaining this information and where we need to go in the future. To facilitate the discussion, it is helpful to distinguish two types of variation in inherited susceptibility to cancer. The first type arises from common polymorphisms of genes governing the metabolism of exogenous and endogenous substances and genes governing specific DNA repair mechanisms. For example, the family of NAT enzymes affects metabolism of the mutagenic heterocyclic amines found in well-cooked red meat. There are data suggesting that meat consumption may be a colorectal cancer risk factor for individuals with rapid-metabolizing variants of the NAT1 and NAT2 enzymes (2, 3). Unlike carriers of defective tumor suppressor genes, individuals with this type of susceptibility are at increased risk only if exposed to the particular substances (e.g., heterocyclic amines from red meat) for which their metabolism is unsuited. For this type of susceptibility, it is particularly helpful to understand the nature of these specific gene-environment interactions. Having done so, it may be possible and feasible to prevent cancer in these individuals by reducing or eliminating the environmental exposures.

The second type of inherited susceptibility to cancer stems from germ-line mutations of genes that defend cells against malignancy, including genes controlling the stability and repair of cellular processes, such as BRCA1, BRCA2, APC, and p53. Individuals who have inherited such mutations are at risk of additional genetic events that destroy proteins needed for the control of cellular proliferation and differentiation. To prevent malignancies in such individuals, we need to understand the functions of these proteins so that we can devise compensatory mechanisms for them.

This paper focuses on the second type of susceptibility. I will not discuss present recommendations for surveillance and screening to facilitate early detection, although such recommendations are critical for the care and management of individuals with inherited predispositions to cancer. Instead, I will discuss the data presently available on options for lifestyle changes, chemopreventive agents, and prophylactic surgery to prevent neoplastic development before the premalignant or malignant lesion becomes detectable by screening. The paper is organized as follows. For each of the three cancer sites (colorectum, female breast, and ovary), I shall describe the inherited mutations that increase risk and outline our present knowledge of why these mutations increase risk. I will then describe existing data concerning lifetime cancer risks in mutation carriers. This is followed by a summary of the options now available for reducing risk and, for each of the options, our present knowledge about the magnitudes of risk reduction and the possible adverse effects of the intervention. The paper concludes with a discussion of research needs to develop effective and safe preventive strategies tailored to individual genetic variation, accompanied by data to allow patients and their physicians to make informed decisions.

Cancers of the colon and rectum are estimated to account in 1999 for 10% of all incident male cancers and 10% of male cancer deaths and 11% of incident female cancers and 11% of female cancer deaths (4). Inherited syndromes account for <10% of all incident colorectal cancers. These syndromes can be classified into three categories: familial adenomatous polyposis, germ-line mutations of mismatch repair genes, and inflammatory bowel disease, such as ulcerative colitis and Crohn’s disease.

Approximately 1% percent of all colorectal cancers are associated with FAP, an autosomal dominant disorder characterized by the occurrence of large numbers of intestinal polyps at an early age (5). Adenomatous polyposis was first documented in the 18^th century, and its inherited nature was recognized by the early 20^th century. However, the specific genetic lesion and its molecular pathogenesis have been elucidated only in the last decade. FAP individuals inherit a mutation of the APC gene. Although such individuals develop numerous intestinal polyps, the polyps represent only a small fraction of the intestinal epithelial stem cells at risk of generating a polyp. The data suggest that a polyp forms only when somatic mutation of the APC allele inherited from the wild-type parent causes complete loss of APC-related tumor suppressor activity (6). Thus, the data on polyp formation support Knudson’s two-hit model (7). The APC protein appears necessary for orderly cell replication, adhesion, and migration (8, 9, 10). These activities involve β-catenin (which both binds E-cahedrin and activates transcription) and Tcf, a downstream transcriptional activator. A dysfunctional APC or β-catenin gene causes loss of proper cellular adhesion-migration and the transcription of a proliferative signal (11, 12, 13). FAP individuals are at high risk of developing colorectal cancer in young adulthood. It has been hypothesized that polyp cells are at risk of additional genetic events leading to malignancy because the polyps extend into the fecal stream, exposing their cells to fecal mutagens.

A second type of inherited colorectal cancer susceptibility, accounting for some 6% of United States colorectal cancer incidence (5, 14), is due to germ-line mutations of genes involved in the MMR system. This system repairs DNA poly-merase errors that occur during replication. A common error is to increase the number of dinucleotide or trinucleotide repeats in a run of such repeats, i.e., a microsatellite. The MMR system recognizes the mismatched sequence, binds to it, excises it, and helps synthesize the correct sequence (15). To date, investigators have identified five MMR genes (hMLH1, hMSH2, hMSH6, hPMS1, and hPMS2). Individuals with a germ-line mutation in one of these genes have increased risks of cancers of the colorectum, endometrium, ovary, and biliary tract. The colorectal cancers, called hereditary nonpolyposis colorectal cancers, demonstrate characteristic features of microsatellite instability. The pathway by which the cellular instability leads to cancer appears not to involve damage to the APC gene but rather to other genes, e.g., the gene encoding the type II tumor growth factor β receptor (16), and the BAX gene (17), the products of which regulate cellular proliferation and apoptosis. Because the APC gene is undamaged, individuals with an MMR defect are no more likely to develop adenomas than the general population. However, once an adenoma develops, its progression to carcinoma is more rapid as the colonic environment induces irreparable damage. Individuals without germ-line MMR mutations can, nevertheless, develop sporadic colorectal cancers showing microsatellite instability (6). The molecular mechanisms underlying the instability in such sporadic cancers are not well understood.

In addition to FAP and germ-line MMR mutations, the inherited IBD syndrome increases susceptibility to colorectal cancer (18). Approximately 1% of colorectal cancers occur in individuals with IBD (5). The specific genetic factors responsible for this syndrome are unknown. The carcinogenic pathway does not appear to include disabling of the APC gene, because somatic APC mutations are found in <10% of colorectal cancers in IBD patients. Instead, the systemic destruction of the intestinal epithelium characteristic of IBD apparently insures that stem cells are exposed to fecal mutagens, even without polyp formation. An early genetic event seems to be a p53 mutation (19). Thus, a loss of p53-mediated cell-cycle checkpoint seems necessary for transition from IBD-related dysplasia to malignancy.

In the absence of prophylactic colectomy, the colorectal cancer risks to individuals with inherited mutations of FAP and MMR genes are high. It is thought that virtually all FAP patients would develop colorectal cancer during their lifetimes. Cumulative colorectal cancer risk by age 70 years has been estimated at ≈80% among 414 Finnish putative carriers of germ-line mutations of the MMR genes (20). Individuals with IBD have ≈20-fold increased risk of colorectal cancer, compared with the general population (18). All of these risk estimates are imprecise, because of limited sample sizes, and difficult to apply to carriers of particular mutations, because of aggregation of different mutation sites within a given gene. There is need for detailed data on risks associated with specific mutation sites of the APC gene and with specific mutations at each of the MMR genes. Different APC mutation sites have been associated with different polyp phenotypes (21, 22, 23). Moreover, the same mutation in the same family can be associated with different polyp phenotypes (24), suggesting roles for other genes in expression of the phenotype.

Little is known about risk-modifying effects of specific lifestyle characteristics and chemopreventive agents in individuals with inherited susceptibility to colorectal cancer. The low population prevalence of these individuals makes it difficult to evaluate such risk modification, even with large, multicentered collaborative studies. However, the available data suggest that the same pathways, involving inactivating mutations of the same genes, may cause both inherited and sporadic colorectal cancers. This similarity may allow us to extrapolate findings from epidemiological studies and randomized trials in the general population to individuals with inherited susceptibility. Accordingly, I will review briefly our present knowledge of these relationships as they pertain to the general population.

The modifiable lifestyle characteristics of greatest potential in preventing colorectal cancer are dietary factors and physical activity levels. A recent review of the epidemiological data concerning the association between colorectal cancer risk and intakes of specific dietary nutrients concluded that, overall, none of the associations is completely convincing (25). The observed relationships of fiber to risk, although often inverse, are nevertheless inconsistent. Yet there does seem to be consistent evidence that high vegetable intakes are protective (26). There also is evidence, albeit not entirely consistent, that higher dietary intakes of calcium are associated with reduced risk of colorectal cancer (25). Most, but not all, of 19 studies evaluating risk in relation to intakes of total/saturated/animal fat showed some evidence of elevated risk associated with higher intakes. Overall however, the data suggest a stronger association with red meat than with any of the associated nutrients. Moreover, although both processed meat and saturated/animal fat may be associated with an increased risk of colorectal cancer, neither total fat nor total protein seems to play a major role.

Of relevance to these issues is the Women’s Health Initiative, a large, controlled trial of a low-fat dietary pattern in 67,000 postmenopausal women at risk for cancers of the colorectum and breast (27). This trial, described in the subsequent section on breast cancer, has been designed to evaluate whether change to a low-fat dietary pattern in later adulthood alters female risk of these cancers.

Epidemiological studies have consistently reported decreased colorectal cancer risk among individuals who are physically active, either on the job or at leisure. The consistency is greater for the colon than the rectum (25). However it is difficult to extrapolate these results from the general population to individuals with inherited susceptibility, because we do not understand the mechanisms for protection. It has been hypothesized that physical activity stimulates colon peristalsis, thereby decreasing the time that colonic contents are in contact with the epithelium. However, fecal transit time is not a well-established risk factor for colorectal cancer. Despite the consistency of the epidemiological data, there is need for a randomized trial of the effects of physical activity on colorectal cancer risk, particularly in genetically susceptible individuals.

The two most promising chemopreventive agents for colorectal cancer are the NSAIDs and calcium supplements. Both are thought to exert their anticarcinogenic effects by inhibiting the cyclooxygenases. These enzymes catalyze the metabolism of lipid arachidonic acid to prostaglandins, which are mitogenic to intestinal epithelial cells. It has been hypothesized that cyclooxygenase-2 inhibitors, such as the NSAIDS and calcium, might prevent cancer in individuals at elevated risk because of inherited mutations of APC, the MMR genes, or the (as yet unknown) genes responsible for IBD (25).

Epidemiological studies have consistently found NSAIDS, including aspirin, to be associated with reduced risk of colorectal cancer. Seven case-control studies (28, 29, 30, 31, 32, 33, 34) and four cohort studies have reported lower incidence (35, 36, 37) or lower mortality (38). In contrast, one cohort study (39) and one intervention trial (40) showed null results. Inverse associations between adenomatous polyps and regular aspirin use have been noted in several studies (30, 36, 41, 42). Moreover, an NSAID called sulindac induces partial regression of adenomas in FAP patients (43).

A recent randomized, double-blind, placebo-controlled trial of calcium supplements to prevent the occurrence of adenomatous polyps in the general United States population showed a statistically significant 15–20% reduction in the incidence of metachronous colorectal adenomas in patients with preexisting polyps (44). Although these were not FAP patients, the similarities between inherited and sporadic disabling of APC in adenoma formation suggests that calcium supplements may prevent polyp development and enhance polyp regression in FAP patients. Presently, there are few data on the efficacy of calcium supplements for individuals with germ-line mutations of the MMR genes or for those with IBD.

At present, colectomy is the only certain way to reduce colorectal cancer risk in those with inherited susceptibility to the disease. However, because some surgical procedures (e.g., ileorectal anastomosis) leave part of the rectum intact, some colectomized individuals remain at risk and must be monitored closely throughout their lives. Thus, data on the preventive efficacy of agents such as the NSAIDs and calcium supplements are needed for the management of these colectomized individuals.

Breast cancer will account in 1999 for an estimated 29% of all new cancers and 16% of all cancer deaths in United States women (4). Less than 10% of these cancers are thought to be related to hereditary syndromes. These include germ-line mutations of the genes BRCA1, BRCA2, and p53. In addition, there is evidence that heterozygote carriers of mutations of the ataxia telangiectasia (ATM) gene have elevated breast cancer risk (45). The available data suggest that mutations of p53 and ATM account for a relatively small proportion of all hereditary breast cancer (46, 47). Accordingly, I shall focus here on risks and preventive options for carriers of mutations of BRCA1 and BRCA2.

Fig. 2 shows estimated cumulative breast cancer risks among carriers of germ-line mutations of BRCA1 or BRCA2. Two of these curves are based on positive identification of BRCA mutations in DNA from peripheral lymphocytes, either of the individuals themselves or their relatives, and two are based on segregation analysis of reported breast cancer occurrence in relatives of participants in case-control studies. The disparity of these curves, a measure of the uncertainty in our estimates, could reflect different types of mutations or different genetic backgrounds among the different groups of women. There is need for population-based data on breast cancer risks associated with specific mutations of the BRCA genes.

There is compelling evidence that endogenous estrogen levels are critical determinants of breast cancer risk (48, 49). Asian women at low breast cancer risk have lower serum and urinary estrogen levels than do Caucasian women at high risk for the disease (49, 50). Cohort studies (51, 52) have found that higher postmenopausal estrogen levels are associated with increased risk of subsequent breast cancer. The metabolically active form of estrogen, E2, exerts its mitogenic activity by diffusing through the plasma membrane of cells and binding to the ER. The E2/ER complex homodimerizes and interacts with EREs in the promoter regions of estrogen-activated cells. These events trigger gene expression that leads to proliferation (Fig. 3).

Efforts to reduce breast cancer risk have focused largely on lifestyle changes and chemopreventive agents that either would reduce circulating estrogen levels or would compete with estrogen in binding to the ERs of breast epithelial cells. The modifiable lifestyle characteristics most studied for their relationship to breast cancer risk are dietary fat, physical activity levels, and body size. The hypothesis that dietary fat may increase the risk of breast cancer by increasing the levels and availability of estrogen has been explored in basic, epidemiological, and dietary-intervention studies for many years and still remains controversial. In a recent meta-analysis of randomized trials of dietary fat reduction and serum estrogen levels, Wu et al. (53) found statistically significant reductions in serum estradiol (E2) among both pre- and postmenopausal women who had been randomized to a low-fat diet. The Women’s Health Initiative (27) will evaluate the possibility of a direct relationship between dietary fat intakes and breast cancer risk in postmenopausal women. This initiative includes a randomized trial involving 67,000 postmenopausal United States women ages 50–79 years at baseline. Its goal is to evaluate a low-fat dietary pattern to prevent breast cancer, colorectal cancer, and coronary heart disease. The trial also will evaluate potential side effects of dietary fat reduction and any concomitant estrogen reduction on risk of osteoporosis. The trial began in 1991 and will be completed in 2006. Participants have been randomized to intervention and control arms for dietary modification (<20% of calories in total fat and <7% of calories in saturated fat, five servings of fruits and vegetables, and six servings of grain products per day), calcium/vitamin D supplementation (1000 mg/day calcium and 400 IU/day vitamin D), and hormone replacement therapy (0.625 mg/day premarin and 2.5 mg/day progesterone), in a 2 × 2 × 2 factorial design.

Vitamin A and its metabolites play a crucial role in regulating the differentiation and proliferation of epithelial cells and are potent inducers of apoptosis. The effects of retinoids are thought to be mediated through the metabolite retinoic acid, which binds to nuclear retinoic acid receptors that then interact with specific RAREs. There is a large body of evidence suggesting that vitamin A is associated with reduced breast cancer risk (reviewed in Refs. 54, 55, 56).

Available data suggest that increased weight-for-height, as measured by BMI, is inversely associated with breast cancer risk in premenopausal women but positively associated in postmenopausal women. In particular, 12 of 13 case-control studies of postmenopausal women found a relative risk greater than unity when the highest category of BMI was compared with the lowest category (reviewed in Ref. 57). However, the data from prospective studies are much less supportive of these associations (57).

Physical activity has been hypothesized to reduce breast cancer risk, but an inverse association has not been reported consistently. Results from seven of nine studies suggest that higher levels of occupational physical activity may be associated with reduced risk, at least among a subgroup of women. Eleven of 16 investigations on recreational exercise reported a 12–60% decrease in risk among premenopausal and postmenopausal women, although a dose-response trend was not evident in most of the studies. The risk reduction associated with exercise was more likely to be observed in case-control studies than in cohort studies (reviewed in Ref. 58).

Considerable effort has focused on synthesizing compounds called AEs, which compete with estrogen in binding to the ER in breast epithelial cells. Type I AEs, such as tamoxifen and raloxifene, are antiestrogenic to breast epithelial cells because the AE/ER complex fails to dimerize and thus cannot bind to the EREs and initiate transcription (Ref. 59; Fig. 4). Type I AEs exert different effects in the cells of different tissues. This tissue specificity is thought to occur because different target EREs have different tolerances for the nondimerized conformation of the AE/ER complex (59). Alternatively, the tissue specificity may reflect tissue-specific activity of other proteins bound to the AE/ER complex that themselves affect transcription. Table 1 shows the agonist and antagonist activity of the Type I AEs tamoxifen and raloxifene in various tissues. Type II AEs, such as ICI64,384 and ICI 182,780, in contrast, apparently bind to the ER outside the nucleus and prevent it from entering the nucleus (59). As seen in Table 1, type II AEs act as estrogen antagonists in all four tissues.

Tamoxifen, the most widely studied AE, is antiestrogenic to the breast and partially estrogenic to the endometrium (Table 1). Tamoxifen inhibits estrogen-dependent secretion of TGF-α and epidermal growth factor and stimulates production of TGF-β (60, 61). Both TGF-α and epidermal growth factor bind to cell membrane receptors and promote breast cell proliferation. TGF-β, in contrast, inhibits the growth of many epithelial cell lines (62).

The National Surgical Adjuvant Breast Project Breast Cancer Prevention Trial was designed to evaluate the efficacy of tamoxifen in preventing breast cancer in asymptomatic women (63). It was motivated by clear evidence of reduced risk of second primary cancers in the contralateral breasts of postmenopausal breast cancer patients receiving tamoxifen therapy. This randomized, double-blind, placebo-controlled trial enrolled 13,288 women with a risk of breast cancer equivalent to that of a 60-year-old woman, based on the model of Gail et al. (64). Few of the women were likely to carry BRCA mutations. The treatment arm consisted of 20 mg/day of tamoxifen for 5 years. Breast cancers occurred in 22% of those on tamoxifen and 43% of those on placebo, for a risk reduction of 50%. However, the incidence of endometrial cancer was increased (13% of the tamoxifen group versus 5.4% of the placebo group), and the incidence of pulmonary embolism and deep-vein thrombosis, although infrequent in both groups, was higher in the tamoxifen group. The elevated endometrial cancer incidence among women treated with tamoxifen is consistent with its partial agonist activity in that tissue. There is need for alternative AEs, with the beneficial effects of tamoxifen on breast tissue but without its risk to the endometrium.

The possibility that raloxifene may be antiestrogenic to both breast and endometrial cells, while estrogenic in its effects on bone and lipid metabolism (65), has raised hope that this compound may reduce the risk of both breast and endometrial cancer, while protecting against osteoporotic fractures and heart disease. Preliminary evidence from the Multiple Outcomes of Raloxifene trial suggests that this may be the case. This randomized, double-blind, placebo-controlled trial is designed to evaluate the efficacy of raloxifene in preventing fractures and breast and endometrial cancer in postmenopausal osteoporotic women (66). A total of 7740 such women were randomized into one of three treatment arms: 120 mg/day raloxifene, 60 mg/day raloxifene, or placebo. Interim results from this trial suggest no difference in cancer risk between the two dose groups of raloxifene, but that the combined raloxifene group had reduced risk of both cancers (the estimated relative risks of breast and endometrial cancers are 0.26 and 0.16, respectively). A new trial, the Study of Tamoxifen and Raloxifene, will compare the two agents directly with respect to their effects on risks of breast and endometrial cancer.

Another class of potential breast cancer chemopreventive agents consists of the synthetic retinoids. Several epidemiological studies have suggested that the retinoids, including vitamin A and its synthetic derivatives, reduce cancer incidence or recurrence at various sites, including the breast (55, 56, 67). They may exert their anticarcinogenic effects by binding to retinoic acid receptors and acting as transcription factors, binding to multiple target genes, including TGF-β (68). The synthetic retinoid 4-HPR (or fenretinide) has shown promise as a chemopreventive agent in breast cancer (54). A total of 2849 women with localized breast cancer has been enrolled in a randomized, double-blind, placebo-controlled trial of 4-HPR for prevention of contralateral breast cancer. The treatment arm consists of 200 mg/day of 4-HPR for 5 years (69, 70).

Bilateral mastectomy to prevent the future occurrence of breast cancer has been controversial because of reports that the disease can occur in residual breast cells or in lymph node micrometastases remaining after surgery. The data needed to quantify the risk reductions have been sparse. In decision analyses, Schrag et al. (71) and Grann et al. (72) estimated the gains in life expectancy associated with bilateral prophylactic mastectomy among women who carry mutations of the BRCA1 or BRCA2 gene. The two sets of authors used largely the same available data on the incidence of breast cancer, the prognosis for women with breast cancer, and the efficacy of prophylactic mastectomy in preventing breast cancer to estimate the effects of this intervention on life expectancy. The authors found that, on average, 30-year-old carriers of BRCA1 or BRCA2 mutations gain 3–5 years of life expectancy from prophylactic mastectomy. More recently, Hartmann et al. (73) conducted a retrospective study of all women with a family history of breast cancer who underwent bilateral prophylactic mastectomy at the Mayo Clinic between 1960 and 1993. The data showed that mastectomized women experienced ≈90% reduction in breast cancer incidence, relative to that of their sisters or as predicted by the Gail model (64). Thus, the available data suggest that prophylactic mastectomy can significantly reduce the incidence of breast cancer in these high-risk women.

It is expected that ovarian cancer will account in 1999 for about 4% of all new cancers and 5% of cancer deaths in United States women (4). Most ovarian cancers originate in the epithelial cells of the ovary. The pathogenesis of these cancers is poorly understood. Their origins are thought to involve the chronically repeated formation of stromal-epithelial clefts and inclusion cysts after ovulation and/or hormonal stimulation to ovarian epithelial cells, either on the ovarian surface or within the inclusion cysts. The nature of the relevant hormones is unknown, but roles for gonadotropins and androgens have been conjectured (74, 75). It also has been hypothesized that ovarian cancer risk may be decreased by factors related to greater progesterone stimulation (76). These hypotheses are difficult to distinguish on the basis of available pathological, epidemiological, and clinical evidence.

It has been estimated that some 5–10% of all ovarian cancers among United States women are due to hereditary factors (77). Increased ovarian cancer risk has been associated with germ-line mutations of BRCA1, BRCA2, and the MMR genes. In the United States and the United Kingdom, some 5% of ovarian cancers have been attributed to BRCA1 (77, 78). In Israel, however, BRCA1 mutations have been identified in 40% of ovarian cancer cases with a family history and 13% of those without such a history (79).

Fig. 5 shows estimated cumulative epithelial ovarian cancer risks among carriers of germ-line mutations of BRCA1 and BRCA2. The curve showing highest risks is based on positive identification of BRCA mutations in DNA from peripheral lymphocytes, either of the individuals themselves or their relatives, in families selected for multiple cases of breast cancer (80). The middle curve is based on segregation analysis of reported ovarian cancer occurrence in first-degree relatives of participants in three case-control studies (77). The curve showing the lowest risks is based on reported occurrence of ovarian cancer in first-degree relatives of Ashkenazi Jewish individuals who tested positive for BRCA mutations (81). As with the curves for breast cancer risk, the disparity of these curves indicates the uncertainty in our estimates, due in part to sparse numbers. It also could reflect different types of BRCA mutations or, if other genes modify risk in carriers, different genetic backgrounds among the different groups of women. There is need for population-based data on ovarian cancer risks associated with specific mutations of the BRCA genes, and with specific MMR genes.

The epidemiological data do not support strong etiological roles in ovarian cancer for diet, physical activity, and body size (82). Some data have suggested a positive association between ovarian cancer risk and galactose-1-P derived from the lactose in dairy products (83), but data from a large case-control study do not support this association (84).

There is strong and consistent evidence that premenopausal oral contraceptive use reduces the risk of epithelial ovarian cancer in the general population (85). Recently, a collaborative retrospective analysis of oral contraceptive use in carriers of BRCA1 and BRCA2 mutations (86) showed dose-specific risk reductions strikingly similar to those observed in the general population (Ref. 85; Fig. 6). These results need confirmation in other populations of BRCA mutation carriers. If confirmed, they provide encouraging evidence that ovarian carcinogenesis in these high-risk women can be blocked by an agent observed to be effective in the general population. However, data from a small study of young breast cancer patients suggest that long-term oral contraceptive use before first full-term pregnancy increases ovarian cancer risk more in BRCA mutation carriers than in noncarriers (87). Clearly, these results need replication with larger numbers. Nevertheless, they indicate the need for data on the full spectrum of effects of oral contraceptives and other potential preventive agents on those with inherited susceptibility.

In addition to oral contraceptives, there is hope that synthetic retinoids may offer protection against ovarian cancer. This hope was sparked by preliminary results from a trial of the synthetic retinoid 4-HPR (fenretinide) discussed in the previous section on breast cancer. This trial was designed to prevent primary cancers in the contralateral breasts of patients with localized breast cancer (69). After 7 years of follow-up, only two 4-HPR patients developed epithelial ovarian cancer, compared with six patients in the placebo group (P = 0.15). Although this reduction in incidence was not statistically significant, it provides a basis for further investigation. Molecular support for the preventive efficacy of 4-HPR has been provided by data showing that the compound inhibits proliferation of ovarian cancer cells in vitro (88).

Bilateral oophorectomy to prevent the future occurrence of ovarian cancer is common among women with a family history of the disease. Uncertainty about the benefits of this surgery stems from reports of subsequent extraovarian abdominal cancers histologically similar to ovarian cancer. The first report noted that 3 of 28 oophorectomized women with a family history of ovarian cancer (11%) developed intraabdominal carcinomatosis in 1–11 years after the surgery (89). Two other reports indicated the occurrence of these cancers in 6 of 324 (1.9%; Ref. 90) and 2 of 44 (4.5%; Ref. 91) oophorectomized women with a family history of ovarian cancer. In decision analyses, Schrag et al. (71) and Grann (72) estimated the gains in life expectancy from bilateral prophylactic oophorectomy among women who carry mutations of the BRCA1 or BRCA2 gene. To do so, the authors used available data on the incidence of ovarian cancer, the prognosis for BRCA mutation carriers with ovarian cancer, and the efficacy of prophylactic oophorectomy in preventing ovarian and other extraovarian peritoneal cancers. The authors found that, on average, 30-year-old carriers of BRCA1 or BRCA2 gain 0.5–2.6 years of life expectancy from prophylactic oophorectomy. These estimates are crude. We need more data on outcomes among oophorectomized carriers of BRCA mutations.

In conclusion, our efforts to prevent epithelial ovarian cancer are still in their infancy. Additional basic research is needed to advance our understanding of the pathogenesis of this disease enough to design potential interventions.

What can we hope for patients as they meet their physicians in the year 2025 for their annual physical examinations, armed with a battery of information about their own unique genetic constitutions? We can hope for a selection of safe, effective choices for preventing the cancers that might otherwise strike them. We also can hope that prophylactic surgery is an item for the history of medicine, replaced by drugs that do not disfigure, yet do not themselves cause illness. To achieve these goals, we must understand the functions of proteins encoded by tumor suppressor genes so that we can devise ways to compensate for them when they are dysfunctional, to eliminate as much of the guesswork as possible in designing and testing new interventions.

In anticipating future genotype-specific efforts at cancer prevention, it is helpful to distinguish variation in the common, low penetrance genes from variation due to carriers of highly penetrant mutations that disable protective genes. The future looks promising for prevention with respect to the first type of variation, particularly if the carcinogenic pathways involve specific, avoidable, exogenous exposures. Individuals with high-risk genotypes can be counseled to avoid the exposures that are particularly risky for them.

The future of preventive research is less clear for carriers of mutations of genes that protect against cancer. Because prophylactic surgery of the target organs is presently the most certain way to prevent cancers in these individuals, many of them opt for this strategy, forming a catch-22 in which it will be difficult to amass enough at-risk individuals to evaluate less disfiguring options. There are two potential solutions to this problem. The first is the development of reliable intermediate end points to monitor efficacy in the intact organ before cancer develops. Indeed, the existence of adenomatous polyps as precursors to colorectal cancer has allowed this research to progress relatively rapidly.

The second potential solution is the possibility that agents found effective in the general population also will protect those with inherited susceptibility. For example, the APC and MMR genes are often disabled in both sporadic and inherited cancers, suggesting similar colorectal cancer pathways in at least some patients with and without inherited susceptibility. In contrast, the functions of BRCA1 and BRCA2 proteins and their roles in protecting against cancers of the breast and ovary are still poorly understood. We need better understanding of the genetic basis and molecular pathogenesis of inherited cancers and their relevance to sporadic cancer.

If we are successful in finding effective chemopreventive agents, patients are apt to face a bewildering array of choices, each with its own benefits and risks. It is difficult for people to internalize quantitative risk information, much less pit one risk against another in cost-benefit analyses. If, for example, the patient is at high risk for both breast cancer and osteoporosis, how can she decide between postmenopausal estrogen replacement therapy, tamoxifen, raloxifene, and no therapy at all? To help her with these decisions, we will need precise and accurate data on cancer risks pertinent to her unique combination of genes and lifestyle characteristics. We also will need data concerning the risk reductions she can expect from a given intervention, the possible side effects, and the chances that they might occur specifically to her, given her genetic makeup. These needs strike at the heart of the epidemiological method, which bases inferences on observations of average effects in large populations rather than unique genetic profiles in individual patients. It will not be possible to stratify individuals into enough cells to accommodate the many variations on genetic susceptibility. Instead, epidemiological studies will have to use the biology and genetics of the mutations and cancers to aggregate people so as to provide accurate estimates of genotype-specific risks and benefits. It is a challenging charge, but one that we must accept.