Breast carcinoma chemoprevention in the community setting

Estimating risks and benefits

Authors

  • Abenaa M. Brewster M.D.,

    Corresponding author
    1. Department of Clinical Cancer Prevention, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
    • Department of Clinical Cancer Prevention, The University of Texas M. D. Anderson Cancer Center, Unit 1360 P.O. Box 301439, Houston, TX 77230-1439
    Search for more papers by this author
    • Fax: (713) 563-5746

  • Dana K. Christo M.P.H.,

    1. Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland
    Search for more papers by this author
  • Hong Lai Ph.D.,

    1. Department of Ophthalmology, Johns Hopkins University School of Medicine, Baltimore, Maryland
    Search for more papers by this author
  • Kathy Helzlsouer M.D.

    1. Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland
    2. The Prevention and Research Center, Mercy Medical Center, Baltimore, Maryland
    Search for more papers by this author

Abstract

BACKGROUND

The United States Preventive Services Task Force recommends that women who are at both high risk for breast carcinoma and low risk for adverse events should receive counseling regarding tamoxifen for chemoprevention. Estimates of the risks and benefits of tamoxifen based on results from clinical trials may not reflect the real-world experience. The authors determined the prevalence of women in a community-based cohort who would meet the definition of high risk for breast carcinoma and calculated the number of women needed to screen to determine one for whom the benefits of tamoxifen would outweigh the risks. Baseline incidence also was examined for adverse health events in this community-based cohort compared with participants in the Breast Cancer Prevention Trial.

METHODS

The study participants were women ages 40–70 years (n = 6048 women) who were members of the CLUE II cohort, which started in 1989, and who responded to questionnaire surveys in 1996 and 2000.

RESULTS

Eighteen percent of all women had a 5-year risk of invasive breast carcinoma ≥ 1.66%. The number of women needed to screen to find 1 woman for whom the benefits outweighed the risks of tamoxifen ranged from 26 women ages 40–49 years to 142 women ages 60–70 years. For women who had undergone a hysterectomy, the numbers needed to screen were lower. Baseline incidence rates of fracture and thromboembolic disease were higher in the community-based cohort compared with the rates observed among prevention trial participants; thus, fewer women had to be treated with tamoxifen to prevent one fracture. However, fewer women in the community also had to be treated to observe harm with a thromboembolic event.

CONCLUSIONS

Clinicians who counsel women about tamoxifen should take into consideration community-level risks and benefits. Cancer 2005. © 2005 American Cancer Society.

The United States Preventive Services Task Force (USPSTF) recommends that women who are at both high risk of developing breast cancer and low risk for adverse events should be counseled about chemoprevention for breast carcinoma.1 Currently, tamoxifen is the only drug approved by the United States Food and Drug Administration (FDA) for this indication, but raloxifene also was included in their assessment. Chemoprevention was not recommended for women at low or average risk for whom the adverse health events associated with tamoxifen may outweigh the benefit.1 Although no set definition of “high risk” was specified, the USPSTF referred to the entry eligibility for the Breast Cancer Prevention Trial (BCPT) and for the follow-up trial Study of Tamoxifen and Raloxifene (STAR), which defined high risk as a 5-year risk of developing invasive breast carcinoma ≥ 1.66%.2, 3

In the BCPT, which randomized high-risk women to receive either tamoxifen or placebo for an average follow-up of 4 years, a reduction of nearly 50% was observed in the risk of developing invasive breast carcinoma, and a decreased risk also was observed for hip, spine and Colles fractures among women who were assigned to the tamoxifen arm. However, tamoxifen also was associated with a 2.5-fold increased risk of endometrial carcinoma and an increased risk of stroke, deep venous thrombosis (DVT), pulmonary embolus, and cataracts. To assess the balance between the benefits of tamoxifen for breast carcinoma chemoprevention and the adverse health effects, risk-benefit indices for tamoxifen were calculated.4 Baseline incidence data for specific health events among women who were not taking tamoxifen were determined using Surveillance, Epidemiology, and End Results incidence rates for endometrial carcinoma and using the incidence rate of cataracts from the placebo arm of the BCPT. Otherwise, age-specific and race-specific incidence rates of stroke, pulmonary embolus, DVT, and fractures were estimated from other sources, such as United States mortality statistics and studies of the Rochester population. These risk-benefit indices incorporate baseline breast carcinoma risk and other factors, such as age, race, and hysterectomy status, to help identify women for whom the benefits from tamoxifen outweigh the risks. Using the Gail et al. risk-benefit index, it has been estimated that 4.9% of white women and 0.6% of black women in the general United States population would have a positive net benefit from tamoxifen for chemoprevention.5

However, the prevalence of women in a given community who may have a favorable net benefit from tamoxifen will vary according to the baseline incidence of tamoxifen-associated health events. Estimates of the risks and benefits of tamoxifen based on results from clinical trials may not reflect the real-world experience accurately due to a healthy volunteer bias for study participants.6, 7 Baseline incidences of adverse health events in a community are likely to be higher than the incidences observed among participants in the placebo arm of the BCPT due to restrictive eligibility criteria in a trial population. Thus, extrapolation of the clinical trial results to the general population may not predict accurately an average woman's risk of harm or benefit with tamoxifen treatment. This difference in outcome between the clinical trial setting and actual practice was demonstrated recently in the Randomized Aldactone Evaluation Study. When results from that clinical trial were applied to the community, there was an associated increase in the rate of spironolactone prescriptions and increased hyperkalemia-associated morbidity and mortality, but there were no decreases in the rates of death or readmission for heart failure.8

In the current study, a community-based cohort was used to determine the distribution of breast carcinoma risk among women ages 40–70 years and the prevalence of women who potentially are eligible for counseling regarding chemoprevention for breast carcinoma. Based on reported risk-benefit indices for the use of tamoxifen in the prevention setting, we examined the number of women needed to screen to determine one women for whom the benefits of tamoxifen outweighed the risks. The study provides meaningful data on the baseline comorbidities that may influence decisions regarding chemoprevention in a given community of women in the United States. A greater incidence of health events in the community compared with women who participate in prevention trials may influence the number of women needed to treat with tamoxifen to prevent or cause an adverse event. Knowledge of the incidence of tamoxifen-associated health events in varying populations of women is needed to help tailor individual treatment decisions for breast carcinoma chemoprevention.

MATERIALS AND METHODS

The study population consisted of female participants of the CLUE II study, which is an ongoing cohort based in Washington County, Maryland.9 The cohort was established in 1989 when residents of Washington County and surrounding communities were invited to participate in the formation of a research specimen bank. In 1996 and every 2 years afterward, participants were sent a follow-up questionnaire about health events, medication use, and cancer risk factors. The 1996 and 2000 questionnaires included extensive questions on health events. Participants were asked if they had been diagnosed by a physician or other health care profession with over 48 health conditions and when they were diagnosed. The overall response rates to the 1996 and 2000 questionnaires were 79% (10,465 women) and 70.9% (10,388 women), respectively.

The CLUE II cohort is predominately white, reflecting the racial demographic of the region. To determine risk status, women ages 40–70 years in 1996 were included in the analysis, because the breast cancer risk factor information needed to calculate their risk of breast carcinoma was obtained first in 1996. Of the 6333 study participants who met the age criteria for eligibility for the study, we excluded 284 women who had a history of invasive breast carcinoma and/or bilateral mastectomy prior to 1996, and 1 woman was excluded who reported an age at menarche that was greater than age at first birth. Thus, 6048 participants were included in the analyses to estimate their risk of invasive breast carcinoma.

The incidence of health events (stroke; transient ischemic attack [TIA]; DVT; hip, spine, wrist, or Colles fracture; endometrial carcinoma; and cataracts) was determined for the period from 1990 to 2000 among women in the CLUE II cohort. To be included in the incidence analysis, women in the CLUE II study had to be age ≥ 40 years in 1990 (n = 6066 women). Only white women were included, because they make up 98% of the cohort participants and because baseline incidence of health events could not be computed for the few black women in the cohort.

Statistical Analysis

The modified, validated model of Gail et al. was used to calculate the 5-year risk of developing invasive breast carcinoma (BCPT.FOR FORTRAN CODE; National Cancer Institute).10 The breast carcinoma risk factors used in the model include age, race, age at menarche, age at first birth, number of first-degree relatives with breast carcinoma, number of previous breast biopsies, and presence of atypical hyperplasia in a breast specimen. High risk was calculated according to the model of Gail et al. as 5-year risk of developing invasive breast carcinoma ≥ 1.66% based on eligibility criteria for the BCPT trial and the ongoing STAR trial.2, 3 The 1996 questionnaire included questions on breast biopsy but did not include questions on whether atypical hyperplasia had been diagnosed; thus, women who had a history of a breast biopsy were assigned to the atypical hyperplasia “unknown” category. The model of Gail et al. imputes a higher risk to this “unknown” category than the atypical hyperplasia “no” category. For women who had missing data, the model of Gail et al. imputes the lowest category of risk for the missing variable. Risks that were calculated by excluding women with missing data from the analytic cohort were similar to those with the imputed data. The number of women who had a positive benefit-risk index for tamoxifen was determined using net risk-benefit indices from Tables 10 and 11 in the study by Gail et al.4

The baseline annual incidence of tamoxifen-associated health events over the period 1990–2000 was calculated. The denominator or population at risk for each health event excluded women who had missing data for a given health event or who reported having a diagnosis of that specific health event prior to 1990. The numerator consisted of women who reported a new diagnosis of the health event between 1990 and 2000. The incidence of endometrial cancer was calculated by excluding women who had undergone a hysterectomy prior to 1990 from the denominator. The number of breast carcinoma cases expected among those women who had a 5-year risk ≥ 1.66% was projected using the weighted average 5-year risk of developing an invasive breast carcinoma as determined with the modified model of Gail et al.10 Statistical analyses were performed using SAS statistical software (version 8.2; SAS Institute Inc., Cary, NC).

We evaluated the number of participants in the CLUE II community cohort needed to treat (NNT) with tamoxifen to prevent or cause one additional health event. The NNT was estimated using the analyzed cohort baseline risk of each health event. The relative risk (RR) of the health event that could be expected with tamoxifen treatment was obtained from the BCPT results.2 The NNT among CLUE II cohort participants was estimated using the formula: NNT = 1/PEER (in which PEER is the participant's estimated event rate or the baseline risk) * 1 − RR (RR from the BCPT study).11 The NNT for each health event also was estimated for participants of the BCPT using the following formula: 1/absolute risk reduction (ARR) (in which ARR= placebo event rate − tamoxifen event rate).12

RESULTS

The age distribution of the cohort for whom the risk of developing breast carcinoma was determined according to the Gail et al. model was as follows: 40–49 years (32%), 50–59 years, (32%), and 60–70 years (36%). Risk factors for breast carcinoma were distributed as follows: age at first birth ≥ 30 years, 6%; menarche at age < 12, 20%; ≥ 1 breast biopsies, 15%; history of breast carcinoma in mother or sister, 8% (data not shown).

The distribution of risk estimates according to age and hysterectomy status is shown in Table 1. Overall, 18.6% of the population met the definition of high risk (≥ 1.66% 5-year risk), and 37% of women ages 60–70 years were at high risk. Because the presence of a uterus influences the benefit-risk ratio for tamoxifen use, risk categories are shown according to whether or not women underwent hysterectomy. Approximately 37% of women overall reported undergoing a hysterectomy; and, among women ages 60–70 years, 42% had undergone a hysterectomy. Forty percent of women who met the definition of high risk did not have a uterus. Among those who were high risk, 35%, 42%, and 40% of women ages 40–49 years, 50–59 years, and 60–70 years, respectively, did not have a uterus.

Table 1. Distribution of the 5-Year Risk Estimates of Invasive Breast Carcinoma by Age Category and Hysterectomy Status among Participants in the CLUE II Community Cohort
All categoriesFive-yr risk categories: no. of patients (%)
Total< 1.66%1.66–3.49%3.50–3.99%4.0–6.49%≥ 6.5%
  • a

    One hundred eighty-two women were missing data concerning hysterectomy status.

All study participants60484924 (81.4)1027 (17.0)31 (0.5)48 (0.8)18 (0.3)
By age      
 40–49 yrs19181832 (95.5)80 (4.2)2 (0.1)4 (0.2)0 (0.0)
 50–59 yrs19431705 (87.8)224 (11.5)5 (0.3)7 (0.4)2 (0.1)
 60–70 yrs21871387 (63.4)723 (33.0)24 (1.2)37 (1.7)16 (0.7)
With uterusa      
 40–49 yrs14081353 (96.1)50 (3.6)2 (0.1)3 (0.2)0 (0.0)
 50–59 yrs1084950 (87.6)129 (11.9)1 (0.1)4 (0.4)0 (0.0)
 60–70 yrs1205749 (62.1)416 (34.5)12 (1.0)21 (1.7)7 (0.6)
Without uterusa      
 40–49 yrs477446 (93.5)30 (6.3)0 (0.0)1 (0.2)0 (0.0)
 50–59 yrs811714 (88.0)89 (11.0)4 (0.5)2 (0.2)2 (0.2)
 60–70 yrs881575 (65.3)275 (31.2)10 (1.1)14 (1.6)7 (0.8)

Based on our findings, a provider would need to screen 5 or 6 women between the ages of 40 years and 70 years to find 1 woman who was at high enough risk to be eligible for counseling about breast carcinoma chemoprevention. An even higher number of women would need to be screened to find one woman who was both at high risk and had a favorable enough benefit-risk index for tamoxifen. Among women ages 40–49 years, 26 women with a uterus and 15 women without a uterus would need to be screened to find 1 women with a favorable enough benefit-risk index for chemoprevention. The number needed to be screened increased with age; and among, women ages 60–70 years, 142 women with a uterus and 28 women without a uterus would need to be screened to find 1 woman who was eligible for chemoprevention who also would have a favorable benefit-risk index (data not shown).

The age distribution of the 6066 women in the CLUE II cohort who were included in the analysis of the incidence of adverse health events was as follows: 40–49 years (30.5%), 50–59 years (31%), 60–69 years (28%), and ≥ 70 years (10.5%). The age distribution of women in the placebo arm of the BCPT was 35–39 years (2.8%), 40–49 years (36.5%), 50–59 years (30.6%), 60–69 years (24.1%), and ≥ 70 years (6.0%) (data not shown). The annual incidence of adverse health events among women in the CLUE II cohort and participants in the BCPT are shown in Table 2. In the tamoxifen compared to the placebo arm of the BCPT, women had a higher incidence rate of thromboembolic disease and a lower incidence rate of fractures. Compared with incidence rates from the BCPT placebo arm, the incidence of hip, spine and Colles fractures, stroke, TIA, and DVT were higher in the CLUE II cohort.

Table 2. Annual Incidence of Adverse Health Events among Participants in the CLUE II Community Cohort and the Breast Cancer Prevention Study
Type of eventNo. of eventsPopulation at riskIncidence per 1000 women CLUE II cohortIncidence per 1000 women: BCPT placebo/tamoxifena
  • BCPT: Breast Cancer Prevention Trial.

  • a

    Incidence rates per 1000 women in the Breast Cancer Prevention Trial were calculated using person-years at risk.

  • b

    Women who underwent hysterectomy were excluded from the population at risk.

Fractures    
 Hip9253671.710.84/0.46
 Spine9752351.851.18/0.88
 Colles18351053.580.88/0.54
Stroke12753432.370.92/1.45
Transient ischemic attack11252542.130.96/0.73
Deep vein thrombosis11451982.190.84/1.34
Cataracts1189519422.8921.72/24.82
Endometrial carcinomab3537670.920.91/2.53

Chemoprevention with tamoxifen is recommended for 5 years. Table 3 shows the number of women needed to treat with tamoxifen for 5 years to prevent or cause each of the health events associated with tamoxifen. Because the baseline incidence of fractures, including hip, spine, and Colles, was higher in the CLUE II community cohort compared with the placebo arm of the BCPT, the numbers of women needed to treat with tamoxifen to prevent a fracture was lower than would be expected based on the BCPT data. Similarly, the overall baseline incidence of stroke and DVT was higher in the community cohort; therefore, fewer women needed to be treated with tamoxifen (n = 143 women) to cause harm through a stroke than would be expected based on the BCPT data (n = 377 women) from a healthier population of women. The number of women needed to treat to cause endometrial cancer or cataract was similar between the BCPT and the CLUE II cohort populations.

Table 3. The Number of Women Needed to Treat with Tamoxifen for 5 Years to Prevent or Cause One Additional Health Outcome in the Breast Cancer Prevention Trial and in the CLUE II Community Cohort
Health eventBCPT relative risk ratioaBCPT/CLUE II
No. needed to treat to preventNo. needed to treat to harm
  • BCPT: Breast Cancer Prevention Trial.

  • a

    Relative risk ratio observed in the Breast Cancer Prevention Trial.

Fractures   
 Hip0.55526/259
 Spine0.74666/415
 Colles0.61588/143
Invasive breast carcinoma0.5160/79
Stroke1.59377/143
Deep vein thrombosis1.60400/152
Endometrial carcinoma2.53123/142
Developed cataracts1.1464/62

Among the 1124 women with a 5-year risk of developing an invasive breast carcinoma ≥ 1.66%, we projected that 29 invasive breast carcinoma cases would develop over a 5-year period (data not shown). Among these high-risk women, a baseline annual breast carcinoma incidence of 5.16 per 1000 women would be expected. The 5-year incidence of invasive breast carcinoma among high-risk women in the CLUE II cohort was similar to that observed in the placebo arm of the BCPT (6.76 per 1000 women); therefore, the number of women needed to treat with tamoxifen to prevent 1 additional invasive breast carcinoma also was similar between the 2 populations of women.

DISCUSSION

CLUE II is a well established cohort of women from Western Maryland who have been followed prospectively since 1989 with substantial and detailed information collected on epidemiological and health events. In this community-based cohort, we showed that 18% of white women between age 40 years and 70 years had a 5-year risk of developing invasive breast carcinoma ≥ 1.66% and could be considered eligible for chemoprevention counseling using the BCPT eligibility criteria. Although all women age ≥ 60 years were eligible for the BCPT, only 37% of the current study participants ages 60–70 years had a 5-year risk ≥ 1.66%. The results from this community-based cohort study are similar to the results estimated using the 2000 National Health Interview Survey, in which 18.7% of white women and 5.7% of black women were eligible to receive tamoxifen in the general United States population.5 However, the Nurses Health Study found that 33% of women ages 45–74 had a 5-year risk ≥ 1.67%, suggesting women from the Nurses Health Study have a higher risk profile than women in our community cohort and in the general population.13

Because the risk of potential adverse health events associated with tamoxifen increases with age, older women require a higher estimated breast carcinoma risk to derive an overall benefit from chemoprevention with agents such as tamoxifen. For example, according to benefit-risk indices calculated according to the model of Gail et al., women ages 40–49 years with a 5-year risk of breast cancer ≥ 2.0% were expected to have a favorable net benefit from tamoxifen for chemoprevention.4 In comparison, women ages 60–70 years with a uterus needed a risk of invasive breast carcinoma ≥ 6.5% to have a benefit-risk index favoring the use of tamoxifen, although the benefit was not statistically significant.4 For women without a uterus in this age group, a statistically significant, favorable benefit-risk index was evident at a 5-year risk prediction of 3.5%.4 In the current study, among women ages 60–70 years, only 0.7% had a 5-year risk of breast carcinoma ≥ 6.5%, and 3.6% had a 5-year risk of breast carcinoma ≥ 3.5%.

To our knowledge, limited information is available regarding the acceptance of tamoxifen for chemoprevention among clinicians. One report from a breast carcinoma risk evaluation clinic in a university setting found that 63% of high-risk women were offered tamoxifen for chemoprevention.14 We suspect that fewer high-risk women in the community setting are offered tamoxifen. A barrier to implementing counseling for breast carcinoma chemoprevention may be the relative difficulty in identifying women in a busy primary care setting who are at increased risk of breast carcinoma and who also are likely to have an overall benefit from chemoprevention. In the current study, 26 women ages 40–49 years and 142 women ages 60–70 years with a uterus would need to be screened using the risk-assessment model of Gail et al. to find 1 women who would be eligible to receive counseling about chemoprevention and who also would have a favorable benefit-risk index for treatment with tamoxifen. Methods to aid clinicians in identifying and counseling women who are at increased risk of breast carcinoma are needed, especially given the potential benefits of tamoxifen in breast carcinoma risk reduction and the cost-effectiveness of treatment.15

Women who were recruited for participation in the BCPT trial were ineligible if they had a life expectancy < 10 years; if they had a prior history of breast carcinoma; if they had taken estrogen or progesterone, oral contraceptives, or androgens for ≤ 3 months prior to randomization; and if they had a history of DVT or pulmonary embolism.2 We observed a higher incidence of adverse health events, including fractures, thromboembolic events, and cataracts, than reported in the placebo arm of the BCPT, as expected in an unselected, community-based setting. The higher baseline risk of adverse events highlights the difference in health status between women in the community and women who participate in clinical trials who generally are likely to be healthier. The higher baseline risk of thromboembolic disease in the community cohort indicates that fewer women in the community would need to be treated with tamoxifen to cause harm with a thromboembolic event. However, because tamoxifen has a beneficial effect in reducing the risk of fractures, a lower number of women in the community cohort also would need to be treated with tamoxifen to prevent fractures.

The net benefit of treatment with tamoxifen is influenced heavily by whether or not a woman has a uterus. The proportion of women undergoing hysterectomies will vary widely in the community. In the ongoing STAR trial, in which the objective is to determine whether raloxifene is equivalent or better than tamoxifen in reducing the incidence of breast carcinoma among postmenopausal women, > 57% of participants enrolled have undergone a hysterectomy, and 58% have Gail et al. model risk scores ≥ 3.0%.3 In the current community cohort, 42% of women age ≥ 50 years had undergone a hysterectomy and, only 2.2% had a Gail et al. model risk of breast carcinoma ≥ 3.5%. Therefore, in the community cohort with a lower hysterectomy rate, fewer high-risk women will have a favorable net benefit index for chemoprevention due to the risk of endometrial carcinoma.

To our knowledge, the impact of tamoxifen as a prevention agent on mortality from breast carcinoma has not been evaluated to date, making it difficult to compare with interventions such as early detection with mammograms. Based on results from the BCPT, 16 cases of breast carcinoma would be prevented among 1000 women ages 50–59 years who are treated with tamoxifen for 5 years. The question remains how many of these women would have died from breast carcinoma in the absence of chemoprevention, because most women with breast carcinoma will be long-term survivors. If we assume that 25% of these women would have gone on to die from breast carcinoma, then tamoxifen would have prevented 4 deaths among 1000 women who had been treated for 5 years. This can be compared with the benefits of screening for breast carcinoma, for which it is estimated that 4 deaths would be prevented among 1000 women ages 50–59 years who undergo annual screening mammogram for 10 years.16

Baseline rates of adverse health events were found to differ between women in the community and participants in the BCPT. For clinicians who are assessing a women's individual risk of breast carcinoma, care must be taken in directly extrapolating risk-benefit ratios from trial results to women in the general population. Benefit-risk indices for tamoxifen as determined by Gail et al. can be used as a guide for identifying subgroups of women who are expected to have a favorable benefit from tamoxifen. In counseling about chemoprevention with tamoxifen, estimates of the patient's risk of benefit and harm with treatment should be incorporated with an assessment of her baseline comorbidities and preferences for care.

Results from the ongoing STAR trial and other chemoprevention trials using the aromatase inhibitors will provide additional information on the potential benefits as well as risks associated with breast carcinoma chemoprevention. We estimate that 18% of white women ages 40–70 years in a community setting would meet the definition of high risk to be eligible for counseling about chemoprevention. When weighing the risks and benefits, health providers should carefully consider potential differences between their patient populations and those individuals who are eligible for clinical trials.

Ancillary