Risk/Benefit Assessment of Tamoxifen to
Prevent Breast Cancer --
Still a Work in Progress?

By Anne L. Taylor, Lucile L. Adams-Campbell, Jackson T. Wright, Jr.
Journal of the National Cancer Institute
November 3, 1999

Affiliations of authors: A. L. Taylor, J. T. Wright, Jr., Department of Medicine, Case Western Reserve University, Cleveland, OH; L. L. Adams-Campbell, Howard University Cancer Center, Division of Epidemiology and Biostatistics, Howard University College of Medicine, Washington, DC.

Correspondence to: Jackson T. Wright, Jr., M.D., Ph.D., Department of Medicine, Case Western Reserve University, 10900 Euclid Ave., Wood Bl W-165, Cleveland, OH 44106-4982 (e-mail: jxw20@po.cwru.edu).

The selective estrogen receptor modulators (SERMs), e.g., tamoxifen and raloxifene, represent an exciting new class of therapeutic agents, which may prove to be effective against a wide variety of conditions, including breast cancer, osteoporosis, and potentially other conditions. There are two principal limitations to the use of the currently available SERMs. First, the specificity of the mix of agonist and antagonist actions is limited. Newer agents with greater specificity hold the promise for highly targeted therapeutic actions with fewer undesirable side effects. Second, clinical trial data showing which populations benefit or suffer excessive risk from treatment with SERMs are very limited. It is with respect to this second limitation that Gail et al. (1), reporting in this issue of the Journal, have proposed estimates of benefits and risk for the use of tamoxifen for the prevention of breast cancer.

The authors use data, based primarily on the Breast Cancer Prevention Trial (BCPT) (2), to construct guidelines to assess the benefits and risks of tamoxifen therapy for use in counseling women at risk for breast cancer. In this trial, 13 388 women at increased risk for breast cancer (projected absolute risk of at least 1.66% over 5 years) were randomly assigned to treatment with either tamoxifen or placebo and followed for an average period of 4.1 years. Tamoxifen was associated with an overall 49% reduction in risk of invasive breast cancer. However, increased risks of endometrial cancer, stroke, pulmonary embolism, and deep venous thrombosis were noted with tamoxifen, with relative risks of 2.53, 1.59, 3.01, and 1.60, respectively, compared with placebo. The authors estimated the reduction in risk of invasive breast cancer as well as the increases in risk of endometrial cancers, stroke, venous thromboembolic events, and fractures associated with tamoxifen treatment. The relative risk versus benefit of tamoxifen treatment for a woman of a given age and ethnicity was derived from the BCPT data in combination with background incidences of each condition derived from population-based surveys, hospital discharge databases, and death certificate databases.

Risk/benefit assessment of any therapeutic tool is an extremely important process, particularly when a preventive care strategy is to be recommended, because most therapies have beneficial effects coupled with potentially harmful side effects. Risks and benefits to a population from any therapy may differ substantially, depending on age, concomitant illnesses and treatments, ethnicity, socioeconomic status, educational level, living environment, and health behaviors of the population. Therefore, the efforts of Gail et al. are commendable. Furthermore, the BCPT is a formidable database upon which to draw in formulating guidelines.

However, there are important shortcomings in utilizing these guidelines. To assess risks and benefits of tamoxifen, the authors used data from outside the BCPT to estimate background incidence rates of breast cancer, endometrial cancer, stroke, and venous thromboembolic events, whereas estimates of the effects of tamoxifen on these health outcome rates were obtained from the BCPT. Although the BCPT trial had more than 13 000 participants, only 220 were African-American and 249 were of other ethnic minorities. While the incidences of breast cancer, endometrial cancer, venous thromboembolic events, and stroke are well documented in white women from many population databases, including the BCPT, the incidences of these conditions in African-American women and other subpopulations could not be derived because of the paucity of data. Similarly, the small numbers of minority women in the BCPT prevented validation of the estimates of tamoxifen effects in minority subgroups.

To calculate the potential for the adverse effects of tamoxifen in minority women, the authors utilized estimates of risk derived from databases other than the BCPT. Thus, they created multipliers for risks of endometrial cancer, stroke, and thromboembolic events for African-American women from mortality and hospital discharge databases (3) and epidemiologic studies, such as the Atherosclerosis Risk in Communities (ARIC) (4), the Cardiovascular Health Study (CHS) (5,6), and the Surveillance, Epidemiology, and End Results (SEER) Program¹ of the National Cancer Institute (NCI) (7). Use of mortality databases may be misleading because there may be divergence in incidence of and mortality from a disease. For example, African-American women have a lower incidence of breast cancer but a significantly higher mortality from breast cancer when compared with white women (7).

Furthermore, the patients from population-based surveys would be expected to have very different demographic and risk profiles than the "healthy volunteer" population in the BCPT. The National Hospital Discharge Survey data are collected from a sample of inpatient records acquired from a national probability sample of hospitals (3). Individuals with multiple hospital discharges during the year could be sampled more than once; thus, the estimates produced represent estimates for discharges not persons. The SEER data reflect cancer incidence data from selected sites in the United States, in which there is underrepresentation of African-Americans and other ethnic groups (7). The ARIC study consisted of four sites, with the vast majority of the African-American cohort derived from a single site in Jackson, MS (4). There are no follow-up data in any of the databases on the impact of tamoxifen or SERMs or the risk/benefits estimated by the Gail Model. They are certainly not randomized, controlled trials such as the ones used to derive the risk/benefit assessment in white women. In addition, it is unknown whether any excess drug-induced risk from tamoxifen is estimated by a multiplier, as suggested by the authors, or is simply additive to the excess population-based risk in African-Americans. It is interesting that, despite similar data limitations in Hispanic and Asian women, they were assumed to have risks comparable to those of white women, an assumption that also may or may not be warranted.

Thus, estimates of tamoxifen effects, both positive and negative, in minority women have a much lower level of precision than those described for white women similar to this study population. While the authors qualify their conclusions regarding use of these guidelines in African-Americans (and other minorities), they still conclude, on the basis of the presentations in Tables 10 and 11 in their article (1), that African-American women should be counseled against (and potentially denied) preventive therapy with tamoxifen at ages approximately 10 years younger than non-African-American women.

Because such guidelines can have a dramatic impact on medical care (e.g., used to construct care-paths [i.e., algorithms used in medical practice to guide physicians in the treatment of specific diseases], educate providers, or even determine payor patterns), it is extremely important to exercise caution in promoting guidelines for preventive or therapeutic care in the absence of adequate data. This is especially true in those populations in whom both the benefits and risks may be exaggerated. Although available data suggest a lower breast cancer incidence in African-American women, mortality is increased. This might suggest the need for more intensive preventive strategies, including more extensive use of tamoxifen in this group. By contrast, if there is truly an increase in adverse effects from tamoxifen in minority women, extremely well defined population-specific limits for the use of this agent as a preventive strategy are warranted. However, the adverse effects may be defined more correctly by specific risk profiles than simply by race.

Therefore, the guidelines proposed by Gail et al. can clearly be used with more confidence for white women than for African-American women and other understudied ethnic groups of women. In fact, with the available data, it is not possible to assess the magnitude of additional risk, if any, of tamoxifen use in African-American women. While it is the responsibility of those promoting an intervention to document a favorable risk/benefit ratio in those for whom it is prescribed, with tamoxifen as with many other agents, we are currently obligated to accept data on the benefit of these agents in minorities from data obtained on other populations. At present, we must accept the benefit of tamoxifen's use that is suggested from the BCPT population. We must also accept that there is a possible greater risk associated with its use in African-American women. However, to suggest that we can quantify this risk/benefit ratio as precisely as defined in the risk tables of the article by Gail et al. is misleading.

There is obviously a desperate need for more data to define estimates of the risks and benefits associated with tamoxifen in African-American women. Empirical epidemiologic and clinical trial data in minority populations, rather than statistical modeling with sparse data, are needed to determine the risks/benefits of tamoxifen. In the BCPT, as in other NCI studies, the lack of inclusion of minorities has been the rule, as was recently confirmed by the Institute of Medicine (8). The dilemma now facing the medical community posed by the article by Gail et al. in the use of tamoxifen reaffirms the need to even more aggressively recruit minority populations in NCI-sponsored clinical trials in proportion to the disease load in the population. It is imperative that recruitment of minorities in NCI trials, such as STAR (i.e., Study of Tamoxifen and Raloxifene) and other National Institutes of Health trials using SERMs, is monitored to ensure that their minority recruitment goals are met. It would be unfortunate if questions of their efficacy in minority populations remain at the conclusion of otherwise definitive trials.

NOTES

¹ Editor's note: SEER is a set of geographically defined, population-based, central cancer registries in the United States, operated by local nonprofit organizations under contract to the National Cancer Institute (NCI). Registry data are submitted electronically without personal identifiers to the NCI on a biannual basis, and the NCI makes the data available to the public for scientific research.

REFERENCES

1. Gail MH, Costantino JP, Bryant J, Croyle R, Freedman L, Helzlsouer K, et al. Weighing the risks and benefits of tamoxifen treatment for preventing breast cancer. J Natl Cancer Inst 1999;91:1829-46.
2. Fisher B, Costantino JP, Wickerham DL, Redmond CK, Kavanah M, Cronin WM, et al. Tamoxifen for prevention of breast cancer: report of the National Surgical Adjuvant Breast and Bowel Project P-1 Study. J Natl Cancer Inst 1998;90:1371-88.[Summary]
3. National Center for Health Statistics. Vital statistics of the United States, 1992, volume II, mortality, part A. Washington (DC): Public Health Service; 1996.
4. Rosamond WD, Folsom AR, Chambless LE, Wang CH, McGovern PG, Howard G, et al. Stroke incidence and survival among middle-aged adults: 9-year follow-up of the Atherosclerosis Risk in Communities (ARIC) cohort. Stroke 1999;30:736-43.[Summary/Full Text]
5. Fried LP, Borhani NO, Enright P, Furberg CD, Gardin JM, Kronmal RA, et al. The Cardiovascular Health Study: design and rationale. Ann Epidemiol 1991;1:263-76.[Medline]
6. Kuller L, Fisher L, McClelland R, Fried L, Cushman M, Jackson S, et al. Differences in prevalence of and risk factors for subclinical vascular disease among black and white participants in the Cardiovascular Health Study. Arterioscler Thromb Vasc Biol 1998;18:283-93.[Summary/Full Text]
7. Miller BA, Kolonel IN, Bernstein L, Young JL Jr, Swanson GM, West D, et al. Racial/ethnic patterns of cancer in the United States, 1988-1992. Bethesda (MD): National Institutes of Health, National Cancer Institute; 1996 Report No. 96-4104.
8. National Academy of Sciences. The unequal burden of cancer: an assessment of NIH reserves and programs for ethnic minorities and the medically underserved. Washington (DC): National Academy Press; 1999.