Fax: (617) 525-2008
Moving forward with breast cancer prevention
Article first published online: 26 APR 2007
Copyright © 2007 American Cancer Society
Volume 109, Issue 12, pages 2387–2391, 15 June 2007
How to Cite
Chen, W. Y., Rosner, B. and Colditz, G. A. (2007), Moving forward with breast cancer prevention. Cancer, 109: 2387–2391. doi: 10.1002/cncr.22711
- Issue published online: 4 JUN 2007
- Article first published online: 26 APR 2007
- Manuscript Accepted: 13 FEB 2007
- Manuscript Revised: 7 FEB 2007
- Manuscript Received: 15 DEC 2006
- National Cancer Institute. Grant Number: P01 CA 087969
The potential for breast cancer chemoprevention is supported by results from several large randomized trials of selective estrogen receptor modulators (SERMs), such as tamoxifen and raloxifene.1–4 Both tamoxifen and raloxifene have been shown to reduce the incidence of invasive breast cancer by approximately 50%, with the benefit confined to estrogen receptor‒positive (ER+) tumors. The first results from the National Surgical Adjuvant Breast and Bowel Project trial (NSABP P-1) of tamoxifen versus placebo for primary breast cancer prevention were published in 1998 and greeted with much fanfare.5 However, the use of tamoxifen for primary breast cancer prevention has been limited due to concerns regarding adverse effects (including uterine cancer and thromboembolic events) and the unfamiliarity of the majority of primary care providers with tamoxifen, which is primarily considered to be an oncology drug. Conversely, primary care providers are already familiar with raloxifene, which has been approved by the U.S. Food and Drug Administration for the prevention and treatment of osteoporosis in postmenopausal women. In addition, rates of adverse effects from placebo‒controlled raloxifene trials appeared to be lower than those observed for tamoxifen (Table 1).6, 7 These differences were confirmed in the large, randomized STAR trial (Study of Tamoxifen and Raloxifene), which reported a lower risk of thromboembolic events (relative risk [RR] of 0.70; 95% confidence interval [95%CI], 0.54–0.91) and cataracts (RR of 0.79; 95% CI, 0.68–0.92) in the raloxifene arm compared with patients treated with tamoxifen. Although there was no statistically significant differences noted with regard to rates of uterine cancer (RR of 0.62; 95% CI, 0.35–1.08), raloxifene clearly had fewer effects on the uterus, with a lower incidence of uterine hyperplasia (RR of 0.16; 95% CI, 0.09–0.29) and fewer hysterectomies performed during follow‒up (RR of 0.44; 95% CI, 0.35–0.56).4 In contrast to hormone replacement therapy, raloxifene did not appear to significantly increase the risk of cardiovascular disease.7, 8
|Tamoxifen versus placebo (NSABP P-1)||Raloxifene versus placebo||Raloxifene versus tamoxifen (STAR)|
|relative risk (95% CI)1||Event rates (P value for comparison)6*||Hazards ratio (95% CI)7||Relative risk (95% CI)4|
|Thromboembolic disease (total)||NA||1.1 vs 0.5%† (P = .0003)||1.44 (1.06–1.95)||0.70 (0.54–0.91)|
|Deep venous thrombosis||1.44 (0.91–2.30)||0.8 vs 0.3%† (P = .006)||1.37 (0.94–1.99)||0.64 (0.41–1.00)|
|Pulmonary embolus||2.15 (1.08–4.51)||0.4 vs 0.2%† (P = .051)||1.49 (0.89–2.49)||0.74 (0.53–1.03)|
|Stroke||1.42 (0.97–2.08)||NA||1.10 (0.92–1.32)||0.96 (0.64–1.43)|
|Uterine cancer||3.28 (1.87–6.03)||0.3% vs 0.3%† (P = .863)||0.5 vs 0.4% (P = .53)†||0.62 (0.35–1.08)|
|Cataract||1.21 (1.10–1.34)||NA||7.4 vs. 7.7% (P = .56)†||0.79 (0.68–0.92)|
|Death||1.10 (0.85–1.43)||NA||0.92 (0.82–1.03)||0.94 (0.71–1.26)|
Rather than using a “one size fits all” approach to chemoprevention, we propose following the cardiovascular disease paradigm in which interventions are targeted according to the estimated risk of disease development. For cardiovascular disease, both biomarker levels (eg, C-reactive protein) and epidemiologic models (eg, Framingham risk score) have been used for risk assessment. Conversely, to our knowledge, no single method of risk assessment for breast cancer has been widely adopted in clinical practice to date. In postmenopausal women not receiving hormone replacement therapy, estradiol levels have been shown to be reasonable predictors of breast cancer risk, with relative risks of approximately 2 reported when the highest and the lowest quintiles of plasma estradiol levels are compared.9 Currently, the clinical application of estradiol levels has been limited by the fact that most clinical laboratories do not have assays that are sensitive and reliable enough to detect these small differences.
However, with the ubiquity of computers in medical practices, epidemiologic risk models provide a fast and simple way to estimate breast cancer risk. The Gail model is to our knowledge the most widely used to determine eligibility for various prevention trials and it estimates the 5‒year absolute risks of invasive breast cancer using age, race, age at menarche, age at first live birth, number of first‒degree relatives with breast cancer, number of benign breast biopsies, and history of atypical hyperplasia.10 The Gail risk score was utilized for the entry criteria for the NSABP P-1 and STAR trials and subsequently has been modified to include mammographic density and weight.11 Other epidemiologic models utilizing other cohorts also have been developed using a larger number of variables, including the use of hormone replacement therapy, menopausal status, and body mass index.12, 13
Epidemiologic risk models can help to optimize the use of chemoprevention agents such as raloxifene by identifying a population at a high enough risk of breast cancer such that the potential benefits of prevention outweigh the risks of adverse events. First, we need to identify the target population. Because raloxifene is not approved for use in premenopausal women, the target population will be postmenopausal women and therefore we will begin our analysis with women aged 50 years or older. Because the risk of thromboembolic events and competing causes of mortality other than breast cancer increase significantly after age 70 years and mammography screening recommendations are less clear for women aged older than 70 years, we will limit the analysis to women aged younger than 70 years.14 In addition, it is less likely that a chemoprevention agent would significantly add to life expectancy in women aged older than 70 years.
Next, we need to estimate the risk of adverse events. Because raloxifene does not appear to be associated with an increased risk of uterine cancer, we will only account for thromboembolic risk. From the 2 large, randomized, placebo‒controlled trials of raloxifene with breast cancer as a primary or secondary outcome (the MORE [Multiple Outcomes of Raloxifene Evaluation]6 and RUTH [Raloxifene Use for The Heart])7 studies), the excess risk of thromboembolic events was 19 cases and 12 cases, respectively, per 10,000 women per year of use of raloxifene. Women in both trials were postmenopausal and on average aged 67 years at the time of study entry, with a mean Gail score of 1.73 in both the placebo and raloxifene arms on the RUTH trial (these data were not available for the MORE study). No data were available from these 2 trials to indicate whether thromboembolic risk varied by age for raloxifene. Although the risk of thromboembolic events with tamoxifen use is well known to vary with age, the largest differences are observed in women younger than compared with those older than 50 years of age; the increased risk of stroke (RR of 1.47; 95% CI, 0.97–2.22) and pulmonary embolus (RR of 2.16; 95% CI, 1.02–4.89) was only observed in women aged 50 years or older, but not in women aged younger than 50 years.1 Because our target population is women aged older than 50 years, we will not account for the potentially lower adverse event rate noted in women aged younger than 50 years. Instead, we have chosen the higher rate of thromboembolic events reported in the MORE trial, which more closely reflects the higher risk noted among women between 60 and 70 years of age (because the average age at the time of study entry was 67 years on the MORE trial) and may slightly overestimate the event rates for women in their 50s.
Thirdly, we need to estimate the benefits of raloxifene. The MORE and RUTH trials reported RR reductions of invasive breast cancer of 72% (RR of 0.28; 95% CI, 0.17–0.46) and 44% (RR of 0.56; 95% CI, 0.38–0.83), respectively. The absolute number of cases of breast cancer prevented depends on the underlying breast cancer risk of the target population. For the cases of breast cancer prevented to outweigh the number of thromboembolic events, the underlying rate of breast cancer among women receiving raloxifene must be greater than 380 cases per 100,000 women per year. This number was calculated assuming conservatively that raloxifene will prevent 50% of invasive breast cancers but will cause 190 excess cases of thromboembolic disease per 100,000 women (using the higher event rate from MORE and assuming uniform thromboembolic event rates in the target population).6 When the breast cancer rate is 380 cases per 100,000 women per year, the number of breast cancers prevented (380 × 0.5 = 190 cases) would equal the number of excess thromboembolic events (19 cases per 10,000 women-years = 190 cases per 100,000). This may be an overestimate of risks because we chose the higher rate from the MORE trial (19 cases per 10,000 women-years) rather than the RUTH trial (12 cases per 10,000 women-years) and assumed a uniform rate of thromboembolism across age groups even though the average age of the study participants was 67 years and our target population was aged 50 to 70 years.
Rather than using the Gail model, we have previously shown that the Rosner and Colditz model improves on the Gail model and classifies women significantly more accurately15 and, accordingly, we use this to estimate deciles of risk to stratify the population. The RR comparing the top‒to‒bottom decile of breast cancer risk was >6.5. In Table 2, we show the rates of breast cancer according to the percentile of estimated breast cancer risk and 5‒year age groups. For comparability to the 5‒year risk score generated by the Gail model, we have included a 5‒year estimated risk of invasive breast cancer for women in the highest decile in each age group according to the Rosner and Colditz model. Among women ages 50 to 54 years, only those in the highest decile of estimated risk are at a sufficiently high enough risk of breast cancer for cancer prevention benefits to outweigh thromboembolic risks. As age increases, the proportion of women for whom benefits outweigh risks in each age group increases because age is a strong predictor of breast cancer risk.
|Age, years||D1||D2||D3||D4||D5||D6||D7||D8||D9||D10||5-Year risk of breast cancer for the top decile*|
There are an estimated 26.7 million women ages 50 to 69 years living in the U.S. (U.S. Census Bureau, Census 2000, summary file 1, available at: http://factfinder.census.gov/[Accessed February 5, 2007]). Excluding the 18.9 million women whose estimated breast cancer risk is <380 per 100,000 women leaves 7.8 million women for whom the benefits in terms of breast cancers prevented will exceed the number of thromboembolic events caused by treatment with raloxifene. Among these 7.8 million women, we estimate that 42,971 cases of breast cancer will be diagnosed each year. Assuming the reduction in risk with raloxifene is comparable to that observed in the randomized trials, then 50%, or 21,486 cases, could be prevented every year. It should be noted that our risk‒benefit estimates have not taken into account the numerous osteoporotic fractures that would also be prevented with the use of raloxifene.16 We also chose the higher rate of thromboembolic events from the MORE trial, which may be an overestimate. Finally, for purposes of clarity, we did not take into account the increasing incidence of ER+/progesterone receptor (PR)+ tumors with increasing age, which would increase the chances of a benefit for raloxifene therapy in older women. The Rosner and Colditz model also can estimate the incidence for ER+ cancers separate from ER‒negative cancers, which would further improve the targeting of such tumors for raloxifene, which mainly prevents the development of ER+ cancers.
The number of women needed to be treated with raloxifene for 5 years to prevent 1 case of breast cancer varies by underlying cancer risk, which varies with age and other factors. Among women in the top decile of breast cancer risk, the number needed to be treated ranges from 79 women in the 50‒to‒54 years age group to just 43 women in the 65‒to‒69 years age group (Table 3). These estimates are similar to those for statin therapy as a primary prevention for cardiovascular disease.17 Given that 5 years would be sufficient for a reduction in breast cancer risk, raloxifene also promises to be more cost‒effective and time‒ efficient than many interventions in cardiovascular disease that need to occur over a lifetime.
|Age, years||Incidence /100,000 women/year||NNT*|
Although a number of studies support the use of raloxifene for osteoporosis,18–20 to our knowledge the risks and benefits of raloxifene therapy for the prevention of breast cancer have not been thoroughly explored. One previous evaluation used MORE trial data and assessed trade‒offs for major chronic diseases with a global health index similar to that used in the Women's Health Initiative (counting events of cardiovascular disease, stroke, pulmonary embolism, invasive breast cancer, endometrial cancer, colorectal cancer, hip fracture, or death from other causes). The annualized rate of events was 1.83% in the placebo group and 1.39% in the raloxifene groups, corresponding to a hazards ratio of 0.75 (95% CI, 0.62‒0.92), an overall risk‒benefit ratio favoring raloxifene.21
Other investigators have also performed detailed analyses of the risk‒benefit ratio for tamoxifen and estimates of the number of women who could benefit from tamoxifen.14, 22 For example, Gail et al. estimated that for women with a uterus, the risk‒benefit ratio for tamoxifen was most favorable for those women aged younger than 50 years, a finding that was mainly driven by a large increase in tamoxifen‒associated uterine cancers after the age 50 years. However, for women without a uterus (which would more closely parallel the effects of raloxifene), a favorable risk‒benefit ratio for tamoxifen was predicted in the majority of white women aged younger than 70 years, with the exception of those in the lowest risk groups for invasive breast cancer, a finding that is similar to our results.14 We have not estimated the risk‒benefit ratios for non-white women separately because the Nurses' Health Study (NHS) cohort, on which the Rosner and Colditz model is based, is predominantly white. Our estimates have also assumed that the distribution of breast cancer risk factors among white women ages 50 to 69 years in the U.S. roughly parallels that of women in the NHS cohort. This appears to be a reasonable assumption because the observed number of cases in the NHS cohort was found to be equal to expected rates in the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) population (RR of 1.0; 95% CI, 0.99–1.03) using the period between 1988 and 1992 as a representative midpoint for our analysis period (the NHS study with the nutrition questionnaire first began in 1980 and is ongoing).
Clearly, the use of breast cancer risk prediction models to identify higher risk subgroups within the population would optimize the risk‒benefit ratio for raloxifene or any chemoprevention agent. Given the burden of breast cancer in Western societies, the established efficacy of SERMs to reduce the incidence of breast cancer, and the magnitude of risk reduction with raloxifene, the use of breast cancer risk prediction models in the clinical setting to target interventions at the appropriate populations and maximize the risk‒benefit ratio has been underutilized. Implementation of breast cancer risk modeling could potentially prevent 21,486 cases of breast cancer each year among women ages 50 to 69 years, in addition to decreasing the risk of fracture. Until the “perfect” prevention drug is discovered, global recommendations regarding the utilization of SERMs will not efficiently decrease the burden of disease immediately. However, the identification of targeted populations who could derive maximum benefit from current interventions will.
Supported by Grant P01 CA 087969 from the National Cancer Institute.