- © 2002 by American Society of Clinical Oncology
Outcomes of Tamoxifen Chemoprevention for Breast Cancer in Very High-Risk Women: A Cost-Effectiveness Analysis
- Dawn Hershman,
- Vijaya Sundararajan,
- Judith S. Jacobson,
- Daniel F. Heitjan,
- Alfred I. Neugut and
- Victor R. Grann
- From the Herbert Irving Comprehensive Cancer Center and Department of Medicine, College of Physicians and Surgeons, Joseph L. Mailman School of Public Health, Columbia University, New York, NY; and Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, Australia.
- Address reprint requests to Victor R. Grann, MD, MPH, Health Outcomes Research, Herbert Irving Comprehensive Cancer Center, College of Physicians and Surgeons, Columbia University, PH18-201A, 630 West 168th St, New York, NY 10032; email: vrg2{at}columbia.edu
Abstract
PURPOSE: To estimate the effects on survival, quality-adjusted survival, and health care costs of using tamoxifen for primary prevention in subgroups of women at very high risk for breast cancer.
PATIENTS AND METHODS: A decision analysis was performed using a hypothetical cohort of women that included subgroups with atypical hyperplasia, Gail risk greater than 5, lobular carcinoma-in-situ, or two or more first-degree relatives with breast cancer. Data sources were the Breast Cancer Prevention Trial, the Surveillance, Epidemiology, and End-Results program, time trade-off preference ratings, the Group Health Cooperative of Puget Sound, and the United States Health Care Financing Administration.
RESULTS: Our model predicted that tamoxifen would prolong the average survival of cohort members initiating use at ages 35, 50, and 60 years by 70, 42, and 27 days, respectively. It would prolong survival even more for those in the higher-risk groups, especially those with atypical hyperplasia (202, 89, and 45 days). Tamoxifen use was also projected to extend quality-adjusted survival by 158, 80, and 50 days in the atypical hyperplasia group. For younger women in the highest risk groups, chemoprevention with tamoxifen was estimated to have cost savings or be cost-effective, both with and without quality adjustments.
CONCLUSION: Chemoprevention with tamoxifen may be particularly beneficial to women with atypical hyperplasia, 5-year Gail model risk greater than 5%, lobular carcinoma-in-situ, or two or more first-degree relatives with breast cancer. The benefits may be greater if tamoxifen is initiated before age 50 years rather than after and if the breast cancer risk reduction conferred by tamoxifen lasts longer than 5 years. For women with a very high risk of invasive breast cancer, chemoprevention with tamoxifen seems to be cost-effective.
IN 1998, THE NATIONAL Surgical Adjuvant Breast and Bowel Project reported on the results of its Breast Cancer Prevention Trial (BCPT) of tamoxifen among women with higher than average breast cancer risk.1 To be eligible for the BCPT, women had to have the same risk of developing breast cancer as an average 60-year-old.1,2 Tamoxifen was found to decrease the risk of invasive breast cancer by close to 50% in such women. This effect seemed to increase slightly with age; tamoxifen reduced risk by 44% in women who entered the study at 35 to 49 years, by 51% in women who entered the study at 50 to 59, and by 55% in those aged 60 or older at entry.1,3,4
Shortly after these findings were reported, tamoxifen became the first chemopreventive agent approved by the United States Food and Drug Administration for the prevention of breast cancer.5 However, few physicians seem to be prescribing tamoxifen for this purpose, perhaps because for women like most of those who participated in the BCPT, it seems unlikely to achieve significant benefits in terms of survival or quality-adjusted survival. Subset analyses from the BCPT suggested that tamoxifen might be especially beneficial for women with atypical ductal hyperplasia, lobular carcinoma-in-situ, Gail risk greater than 5, or with two or more first-degree relatives diagnosed with breast cancer.1 These subsets of women have risks for breast cancer four to seven times greater than those of women in the general population.6-8 The BCPT found that tamoxifen reduced risk by 80% among women with atypical hyperplasia and by 50% among those with Gail model risk greater than 5, lobular carcinoma-in-situ, or two or more first-degree relatives with breast cancer.1
Therefore, we have used Markov modeling to estimate the effects of tamoxifen chemoprevention on survival, quality-adjusted survival, and health care costs in hypothetical cohorts of women similar to those in the four high-risk subgroups who participated in the BCPT.
PATIENTS AND METHODS
Using Data 3.5.7 Software (Treeage Software Inc, Williamstown, MA) and assumptions derived from the results of the BCPT (Table 1), the medical literature about adverse events (Table 2), and recommendations from a group of clinical oncologists, we developed a Markov model simulating outcomes up to age 100 or death (Markov cycles).1,9-12 The model and assumptions used for this study are similar to those we developed for a prior analysis of all the women who participated in the BCPT.17 The health state outcomes were good health, good health after stopping tamoxifen, noninvasive and invasive breast cancer, hip fracture, thrombophlebitis/pulmonary emboli, endometrial cancer, cataracts, and death.
Annual Incidence Rates per 1,000 for Outcomes of the BCPT1 Used in the Markov Model
After transforming the average annual breast cancer incidence rates in the BCPT (Table 1) into transition probabilities, we modeled survival, quality-adjusted survival, and cost-effectiveness associated with starting to take tamoxifen in the first year of each of three age ranges: 35 to 49, 50 to 59, and ≥ 60 years. We based our projections of survival after a diagnosis of breast or endometrial cancer on the National Cancer Institute’s Surveillance, Epidemiology, and End Results CD-ROM data for 1973 to 1997 (Table 2).9 In accordance with the BCPT results, we assumed that 70% of women initially diagnosed with breast cancer would be node-negative and 30% would be node-positive.1 We also assumed, on the basis of published data, that estrogen receptor status would have no effect on survival after adjustment for tumor size and lymph node status.13,14 We assumed that the beneficial effects of taking tamoxifen for 5 years would last for only those 5 years.
Other Assumptions Used in the Markov Model9-12
Our assumptions regarding the known adverse side effects of tamoxifen use were as follows: that patients who developed endometrial cancer, thrombophlebitis/pulmonary emboli, or cataracts while taking tamoxifen would discontinue tamoxifen use immediately; that risk for these conditions would be elevated only during treatment; that risks for each of these conditions would be independent of one another; and that women who discontinued tamoxifen would not experience any of its health benefits thereafter. We also assumed that women who had developed one or more of these conditions would have the same risk of breast cancer as other women, and that those who developed breast cancer while taking tamoxifen would have the same mortality risk as other women with breast cancer.9 Our assumptions about mortality associated with these conditions are based on national data for these events (Table 2).
We did not include in our reference model other potential benefits or side effects of tamoxifen use, such as osteoporosis, cardiovascular disease, or stroke, because the association of these outcomes with tamoxifen use did not reach statistical significance in the BCPT (see Sensitivity Analysis, below).1
Cost Parameters
We obtained age-stratified data on cancer care costs from the Group Health Cooperative of Puget Sound (Table 3).18 All costs were adjusted for increases in the medical care component of the Consumer Price Index and are given in 1998 dollars.20 We based our estimates of the costs of treating complications, such as endometrial cancer, thrombophlebitis with pulmonary embolism, cataracts, and hip fractures (Table 4), on Medicare payments for 1998 from the Health Care Financing Administration.22,23 We based our estimates of the costs of care for those dying without cancer on an analysis of Health Care Financing Administration data by Lubitz and Riley.24 Drug cost data were obtained from the 1998 Pharmacy Fundamental Reference.5 All costs reflect the mortality assumptions shown in Table 2.
Costs Used in the Markov Model*: Costs of Care for Patients With and Without Invasive Breast Cancer†
Costs Used in the Markov Model*: Other Medical Costs for All Patients
Direct costs, such as those due to time off from work, travel, and other out-of pocket expenses, were not determined and were believed to be inconsequential compared with direct medical costs.15 Indirect costs, such as loss of earnings, were excluded from the model.
Quality-of-Life Estimates
We used preference ratings from a recent study of healthy women aged 33 to 50.25 These ratings were elicited with a time trade-off questionnaire that described three health state scenarios: use of chemoprevention, diagnosis for and treatment for invasive breast cancer, and metastatic disease (Table 5). The participants were asked how much of their life expectancy in each state they would trade not to be in that health state. To estimate the utility of a given health state on a 0 to 1 scale, we used the formula 1 − T/L, where T equals the time the respondent was willing to trade and L equals the life span assigned to that health state in the scenario.28-30 In our computer model, we calculated quality-adjusted life-years of survival by multiplying the life-years of survival associated with each health state by the preference rating of that state. We assumed that women who died with breast cancer would have metastatic disease for 3 years before death.13,31 We used utilities from the literature for complications of tamoxifen (Table 5).
Preference Ratings Used for Health States in the Markov Model
Cost-Effectiveness
To calculate the cost-effectiveness of tamoxifen chemoprevention for high-risk women in the BCPT, we developed ratios of the incremental costs of tamoxifen to incremental survival and quality-adjusted survival.16
Discounting
We examined differences in survival and quality-adjusted survival to age 100 or death with no discounting. For cost-effectiveness, we used a discount rate of 3% per year to calculate the present value of life-years saved, quality-adjusted survival, and costs.16
Sensitivity Analysis
We performed a series of studies of the effects on our outcomes: if the breast cancer risk associated with tamoxifen in the BCPT lasts for 10 or 15 years instead of 5 years; if after 2 years only 80% of women continue to take tamoxifen; if the costs of tamoxifen are higher or lower; and if tamoxifen reduces the risk of hip fractures (an end point that did not reach statistical significance in the BCPT).
RESULTS
Survival and Cost-Effectiveness in the 5-Year Reference Model
On the basis of a prior analysis of our model, daily use of tamoxifen prolongs survival by 70 days for those who initiate use at ages 35 through 49 years, by 42 days for those who initiate use at ages 50 through 59 years, and by 27 days for those who begin at ages 60 through 69 years. Similarly, the incremental quality-adjusted survival benefit decreases as the age at which tamoxifen use is initiated increases.17 The discounted cost-effectiveness ratio and the quality-adjusted cost-effectiveness ratio increase as the age of tamoxifen initiation increases (Table 6).
Cost-Effectiveness of Tamoxifen Chemoprevention, by High-Risk Subgroup
In this analysis, we investigated the benefits of tamoxifen use for different high-risk subgroups. For each of the subgroups analyzed, the model predicted a greater survival and quality-adjusted survival benefit than for the BCPT cohort overall (Figs 1 and 2⇓). This benefit was greatest for women with a history of atypical hyperplasia who started tamoxifen at ages 35, 50, and 60; their survival benefits were 202 days, 89 days, and 45 days, respectively. Women with a 5-year Gail model risk greater than 5% also enjoyed survival benefits (195, 86, and 43 days), as did those with lobular carcinoma-in-situ (162, 73, and 38 days) and those who had two or more first-degree relatives with breast cancer (99, 50, and 28 days). In these subgroups, tamoxifen use also resulted in similar, but smaller, quality-adjusted survival benefits (Fig 2).
Fig 1. Estimated mean incremental survival in days associated with taking tamoxifen, by risk group.
Fig 2. Estimated mean incremental quality-adjusted survival in days associated with taking tamoxifen, by risk group.
The cost per life-year saved and cost per quality-adjusted life-years saved were also markedly lower than for the study participants as a whole in each of the subgroups analyzed (Table 6).
Sensitivity Analysis
If the benefits of taking tamoxifen persisted up to 10 or 15 years, the survival benefit was projected to increase for each of the subgroups analyzed (Table 7). The increased survival benefit was greatest for use starting at age 35 years and decreased with age. Tamoxifen use in 35-year-olds with atypical hyperplasia or a Gail risk of ≥ 5 was estimated to be cost saving (less expensive than surveillance) in the 10-year model and even more favorable in the 15-year model.
Sensitivity Analysis of the Effects on Survival, Quality-Adjusted Survival, and Cost-Effectiveness of 10 and 15 Years of Breast Cancer Risk Reduction Due to Tamoxifen, by High-Risk Subgroup
The price of tamoxifen was an important determinant of the cost-effectiveness ratio (Table 8). Reducing the price by 40% would result in cost savings or cost-effectiveness for tamoxifen use among all the high-risk groups.
Sensitivity Analysis of the Effect of Changes in Tamoxifen Costs on Cost-Effectiveness
We performed several other sensitivity analyses using a simulated cohort similar to that of the women with atypical hyperplasia who participated in the Trial (Table 9). We selected this group because it had the highest risk of all the subgroups. The assumption that 20% of women in this group would stop taking tamoxifen after only 2 years had very little effect on survival, quality-adjusted survival, and cost-effectiveness. An additional analysis, incorporating the reduced risk of fracture observed in the tamoxifen group of the BCPT participants, found that women who took tamoxifen could expect to have longer survival and lower costs per life-year saved than untreated women (Table 9).
Sensitivity Analysis in the Atypical Hyperplasia Subgroup, Assuming That 20% of Tamoxifen Users Discontinue Use After 2 Years and That Tamoxifen Use Reduces Hip Fracture Risk
DISCUSSION
Breast cancer has remained an elusive target for primary prevention. Until recently, only screening, via mammography and clinical breast examination, had achieved modest improvements in survival of early breast cancer, at increased health care costs. In general, the effectiveness of screening increases as the prevalence of preclinical disease increases. For example, despite the controversy regarding mammographic screening for average-risk women aged 40 to 49, models suggest that mammographic screening is justified for women younger than 50 years who possess a risk equal to or greater than that of the average 50-year-old.32,33 However, mammographic screening is not primary prevention.
In 1998, tamoxifen was shown to reduce breast cancer incidence among women with above-average risk; the tamoxifen BCPT was the first study to demonstrate the efficacy of any treatment for primary prevention of breast cancer.3 We subsequently used modeling to evaluate the effects of tamoxifen on survival and quality-adjusted survival and the cost-effectiveness of chemoprevention with tamoxifen.17 The model indicated that taking tamoxifen could benefit women who initiated treatment at age 35 but not at older ages.
We have now used our model to assess the same associations among women in very high-risk groups. We have found that tamoxifen could be expected to confer modest survival benefits overall, but that 35-year-olds in the high-risk groups could be expected to have a three- to four-fold increase in survival and quality-adjusted survival. Average survival gains were as much as a year when the benefits of tamoxifen were assumed to extend from 5 years to 10 or 15 years.
In policy-oriented decision analyses, gains in life expectancy are averaged across the population receiving the intervention.34 Therefore, even when a measure greatly extends the life expectancy of the individual with the condition targeted by the intervention, it extends overall survival only slightly in a low-risk population. For example, cervical cancer screening extends the survival of women who have neoplasia detected preclinically by 25 years, but it adds only 3 months to the average survival of screenees.34,35 An intervention is generally considered worthwhile if it results in average life-expectancy gains of ≥ 2 months in the population that receives it.34
Cost-effectiveness analysis calculates the ratio of the financial cost of a treatment or preventive measure to units of effectiveness, such as life-years or quality-adjusted life-years gained.36,37 Quality adjustment for this purpose usually involves multiplying life-years gained with the treatment or preventive measure by a quantitative estimate of the associated quality of life. Quality adjustment incorporates both mortality and morbidity into a single multidimensional weighted measure.29 We based our quality adjustment on the preference ratings assigned by average-risk women to chemoprevention and other breast cancer–related health states.17,26 The Panel on Cost-Effectiveness in Health and Medicine maintains that average-risk individuals are representative of society and sensitive to both its needs and those of patients.38,39
Our model estimated that chemoprevention with tamoxifen would be cost-effective in all the subgroups analyzed who were ≤ 50 years of age when starting tamoxifen use (except for women > 50 years old with ≥ two affected first-degree relatives), and almost cost-effective for ≥ 60 year olds with atypical hyperplasia or Gail model risk greater than 5. This finding is similar to those in prior reports on the overall BCPT population, in which the cost per life-year saved was estimated to be more than $50,000 for all individuals who were ≥ 50 years old. Quality-adjusted cost-effectiveness in the overall cohort was more than $50,000 per life-year saved in all age groups.17
Other analyses have shown that tamoxifen is most beneficial for younger women with an elevated risk of breast cancer.40-42 Gail et al4 have estimated the risk-benefit ratio of prophylactic tamoxifen by calculating the risks of thrombosis and endometrial cancer and comparing them to the benefits of decreased hip fracture and breast cancer. They conclude that the use of tamoxifen should not be based on a single number, such as a projected 5-year risk of breast cancer of 1.66%, but rather on a weighing of the various risks and benefits associated with taking tamoxifen.4 Our model incorporates a similar type of risk-benefit analysis, adjusts for factors affecting quality of life, and determines the potential costs associated with the various outcomes.
The benefits of tamoxifen use diminish with increasing age because older women have higher risks of mortality from competing causes of death, such as cardiovascular disease.43 Among women in their 60s and 70s in the general population, the incidence rate of cardiovascular disease is five times higher than the rate of breast cancer.43 However, in the high-risk groups we studied, tamoxifen use prolonged survival by 1 to 3 months.
In two European trials, breast cancer incidence in the tamoxifen group did not differ from that in the placebo group.44,45 However, the participants in these studies differed from the BCPT population in risk factors for breast cancer.1
Using decision analysis and computer modeling for subgroups of women at highest risk for breast cancer, we have shown substantial benefits to both survival and quality-adjusted survival when tamoxifen is used as a chemopreventive agent, especially in young women who are at very high risk for developing breast cancer. Our findings support targeting chemoprevention to those with the highest risk.
Acknowledgments
Supported in part by grant no. CRTG-98-260-01 from the American Cancer Society, Atlanta, GA; the Sindab African American Breast Cancer Project, the Avon Breast Cancer Research and Care Program, and the Breast Cancer Alliance; and Cancer Center Support Grant no. P30 CA13696-26 from the National Cancer Institute (NCI), Bethesda, MD. D.H. was the recipient of an NCI-funded postdoctoral fellowship (T32-CA09529).
- Received November 8, 2000.
- Accepted August 6, 2001.
