Cost-effectiveness of chemoprevention of breast cancer using tamoxifen in a postmenopausal US population




Previous cost-effectiveness analyses of tamoxifen therapy account for breast cancer risk reduction during active treatment but not for its persistent protective effect after active treatment.


A detailed, continuous time, mathematical model of breast cancer and healthcare processes was used to simulate a postmenopausal population aged <55 years in a virtual trial comparing tamoxifen treatment with no treatment for lifetime follow-up. Unlike previous work, the current model of tamoxifen therapy is based on a meta-analysis of 4 randomized, placebo-controlled chemoprevention trials with breast cancer risk reduction continuing for 10 years after treatment termination. Cancer incidence and survival data were derived from Surveillance, Epidemiology and End Results statistics. Noncancer disease incidences, quality-adjusted life year (QALY) utility weights, and costs were derived from the literature.


Tamoxifen treatment (vs no treatment) saved 29 QALYs in a population of 1000 postmenopausal women aged <55 years with an additional cost of $333,000 over the population's lifetime (average cost-effectiveness ratio, $11,530 per QALY). Tamoxifen therapy, compared with no treatment, was cost saving when higher risk populations were targeted (5-year risk ≥1.66%). The cost-effectiveness results were sensitive to parameters that characterized menopausal symptoms and adverse side effects of tamoxifen.


The current results indicated that tamoxifen chemoprophylaxis for postmenopausal women aged <55 years is a cost-effective health policy that reduces breast cancer incidence and improves life expectancy. Focusing on a postmenopausal population aged <55 years minimized the threat of adverse events associated with tamoxifen. Cancer 2011. © 2011 American Cancer Society.

The US Food and Drug Administration approved tamoxifen as a chemopreventive agent for women at higher than average risk of developing breast cancer after publication of the results from the National Surgical Adjuvant Breast and Bowel Project (NSABP) P-1 trial, which demonstrated that tamoxifen prophylaxis reduced the risk of invasive breast cancer by nearly 50%.1 This reduction, however, comes with a nontrivial side-effect profile that includes elevated risks of endometrial cancer and thromboembolic events. Because of the severity of the possible side effects, few women are taking tamoxifen chemoprophylaxis2, 3 despite the large proportion of women who are eligible to receive treatment.4

Several cost-effectiveness analyses have attempted to identify specific populations who would derive the most benefit from tamoxifen treatment, generally targeting women at elevated 5-year Gail model breast cancer risk5 to balance the benefits of treatment with the potential harms of adverse events (AEs). Those studies determined that, although tamoxifen treatment unequivocally reduces breast cancer incidence, widespread use of tamoxifen is not cost-effective because of its side-effect profile, cost, and sensitivity to assumptions related to quality-adjusted life year (QALY) utilities.6-9 Those studies used side-effect relative risks derived from the NSABP clinical trial and assumed that breast cancer risk reduction was limited to the duration of active treatment (typically, 5 years). However, the results from longer follow-up trials of tamoxifen chemoprophylaxis10, 11 have demonstrated the need to revisit these assumptions.

We present a model of tamoxifen that accounts for the effects of tamoxifen on estrogen receptor (ER)-positive breast cancer, endometrial cancer, thromboembolic events, and cataracts. Breast cancer risk reduction occurs during active tamoxifen treatment and also persists for 10 years after active treatment, as demonstrated in long follow-up trials. The side-effect profile of tamoxifen is modeled based on a meta-analysis of the 4 chemoprevention clinical trials.1, 10-12

Menopausal side effects and fear of serious AEs play a decisive role in determining patient adherence to tamoxifen therapy. In the current study, we focus our attention on postmenopausal women aged <55 years at the start of tamoxifen therapy, thereby minimizing the most serious menopause-like symptoms13 and treating a population at the lower range of risk for AEs.1, 9

We use the Archimedes Breast Cancer Model to simulate a virtual trial in which 5 years of tamoxifen treatment were compared with no treatment. We present results for multiple 5-year breast cancer risk thresholds and risk groups.


Model of Breast Cancer and AEs

The Archimedes Breast Cancer Model is a detailed, continuous-time, mathematical model of breast cancer incidence, tumor growth, detection and spread, survival, and healthcare processes associated with breast cancer. Figure 1 is a schematic representation of the model. The risk factors in the natural history model include the commonly used variables of age, age of menarche, age of first live birth, number of previous benign breast biopsies, and family history of breast cancer. In addition to the typical breast cancer risk factors, we also account for the risk factors of race/ethnicity, age of menopause, body mass index, use of combined hormone-replacement therapy, and the presence and type of benign breast disease. A description of each component of the model is provided in Table 1. The accuracy of the model is validated against Surveillance, Epidemiology, and End Results (SEER) data from the years 1980 through 200414; the Cancer Prevention Study II Nutrition Cohort24 for breast cancer incidence and mortality; and Breast Cancer Surveillance Consortium (BCSC)17 data for mammogram screening and interval cancer rates.

Figure 1.

. This is a graphic summary of the Archimedes Breast Cancer Model.

Table 1. Breast Cancer Model Summary: Brief Description of Construction of the Archimedes Breast Cancer Model
ParameterDescriptionPrimary Reference
  1. SEER indicates Surveillance, Epidemiology, and End Results Program (National Cancer Institute); BCSC, Breast Cancer Surveillance Consortium; TNM, tumor, lymph node, metastasis; BMI, body mass index; CHRT, combined hormone replacement therapy; BBD, benign breast disease.

Breast cancer pathology
 Natural history/incidenceModeled using an age and race dependent hazard function and other risk factors (see below)SEER (SEER 200814)
 DiagnosisCalibrated to frequencies of diagnostic procedures after symptoms and positive mammogramsBritish Columbia Screening  Program (Clay 199415)
 All-cause mortalityModeled using a race-dependent hazard that accounts for all mortality not related to breast cancer and breast cancer treatmentCenters for Disease Control  and Prevention (Kung 200816)
 SurvivalDependent on race/ethnicity and size of primary tumor, number of affected axillary lymph nodes, and presence/absence of invasion and distant metastasis at the time of diagnosisSEER (SEER 200814)
 Tumor size and growthSize is assigned at time of detection by palpation according to race-dependent distributions; growth is drawn from a log-normal distribution of growth rates and assumed to slow with age, with parameters chosen to match observed detection ratesSEER (SEER 200814), BCSC  (Ballard-Barbash 199717)
 InvasionAssigned according to a cumulative distribution of sizesSEER (SEER 200814)
 Spread to axillary lymph nodesModeled assuming spread is a function of tumor size but independent of absolute tumor lifetimeSEER (SEER 200814)
 Distant metastasisDetermined at diagnosis as a function of tumor sizeSEER (SEER 200814)
 StageTNM classification systemTNM system (Sobin &  Wittekind 199718)
 Breast densityModeled as a decreasing function of ageVachon 200719
Risk factors
  Age of menarche, age   of first live birth,   family historyModeled using a version of the Gail modelSpiegelman 199420
  MenopauseModeled using a functional form of relative risk that increases with age of menopause between ages 30 60 yCollaborative Group on Hormonal  Factors in Breast Cancer 199721
  BMI, CHRTModeled in terms of effects on estrogen pathwaysEndogenous Hormones and Breast  Cancer Collaborate Group22
  Benign breast diseaseModeled to capture overall prevalence and incidence rates in the US population of BBD; BBD types include nonproliferative, proliferative without atypia, and atypical hyperplasiaHartmann 200523
  Screening mammographyDetection is based on tumor size and breast density; for a given density, detection is a monotonically increasing function of tumor size; detection rates are fit to give the correct interval cancer ratesBCSC (Ballard-Barbash 199717)
  False-positive symptoms   and mammogramsFalse symptoms assigned at random such that 10% of breast lumps are cancerous; false-positive mammograms depend on breast density and time since prior mammogramBCSC (Ballard-Barbash 199717)

Disease events occur with a probability determined from an age-dependent hazard for that event. Hazard functions for endometrial cancer, pulmonary embolism (PE), deep vein thrombosis (DVT), and cataracts are approximated from incidence by age data from Gail et al.25 Breast cancer hazard rates, including both invasive and ductal carcinoma in situ (DCIS) tumors, are estimated from incidence by age data from SEER for the years 1994 through 2004.14 The occurrence of hysterectomy is modeled using incidence data from Merrill and Feuer26 and Gail et al.25 Endometrial cancer events do not occur after hysterectomy. Simulated women who experience an AE while taking tamoxifen discontinue its use at that time. We assume the occurrence of each event is independent and model only the first event of each type for each individual.

We account for survival and mortality from breast cancer, endometrial cancer, and PE, and we assume that cataracts and DVT are nonfatal events. Survival curves for breast and endometrial cancers are estimated using relative survival and SEER data from the years 1988 through 2004.14 Relative survival for PE is computed from data reported by Kniffin et al27 and Stein et al.28

Death from breast cancer, endometrial cancer, and PE are modeled using disease-specific survival functions. Death from other causes is derived from an all-cause mortality hazard calculated from the Centers for Disease Control and Prevention 2005 mortality tables.16 Explicitly modeled disease-specific mortality rates are subtracted from the all-cause mortality rate, thus avoiding double counting of the contributions of breast cancer, endometrial cancer, and PE to the overall death rate. In simulations, each woman is assigned a death age from each modeled cause of death, and the earliest age becomes her actual time of death.

Model of Tamoxifen Therapy

The effect of tamoxifen therapy on ER-positive breast cancer is divided into 2 phases: during active treatment and after active treatment. We assume that each phase can be captured by a unique hazard function. We assume that tamoxifen reduces the risk of ER-negative breast cancer by 48% among current tamoxifen users and has no effect on ER-negative cancers based on a meta-analysis by Cuzick et al.29 During the active treatment phase, the breast cancer hazard function is multiplied by a constant hazard ratio of 0.52.

The most substantial data on the long-term effects of tamoxifen come from a meta-analysis by the Early Breast Cancer Trialists' Collaborative Group of data from approximately 30,000 women with breast cancer in adjuvant tamoxifen studies.30 By using data for contralateral breast cancer, we model the hazard ratio for the effect of tamoxifen on breast cancer after active treatment as 1 − 0.0934t, where t years indicates the total duration of tamoxifen use, thereby indicating that longer tamoxifen use (up to 5 years) results in a greater breast cancer risk reduction.1, 31, 32

Because the meta-analysis by Cuzick et al29 did not report hazard ratios for all side effects of interest, we performed our own analysis using reported trial event counts and the inverse variance method (see Table 2). Tamoxifen has a statistically significant impact on endometrial cancer, PE, and DVT in our meta-analysis. Its impact is borderline significant for cataracts and is not statistically significant for myocardial infarction, osteoporotic fracture, or stroke. Consequently, endometrial cancer, PE, DVT, and cataracts are included in the model. Because of the discrepancies in relative risk ratios for endometrial cancer1 and stroke11 as AEs, we explored their elevated risks as part of a sensitivity analysis. The effect of tamoxifen on these events is captured by multiplying the appropriate hazard function by the hazard ratio in Table 2 for the years during tamoxifen treatment. Once tamoxifen therapy is terminated, hazard rates for AEs return to age-appropriate baseline values, as observed in clinical trials.10, 11

Table 2. Meta-Analysis of Adverse Events: Adverse Event Counts From Tamoxifen Chemoprevention Trials
 NSABP (Fisher 19981)RMH (Powles 200710)Italian (Veronesi 200712)IBIS-I (Cuzick 200711) 
EventPlaceboTamoxifenPlaceboTamoxifenPlaceboTamoxifenPlaceboTamoxifenHR [95% CI]
  • NSABP indicates National Surgical Adjuvant Breast and Bowel Project; RMH, Royal Marsden Hospital; IBIS-I, first International Breast Cancer Intervention Study; HR, hazard ratio; CI, confidence interval

  • a

    All women underwent hysterectomy and were not at risk for endometrial cancer.

Endometrial cancer153614aa5112.41 [1.46-3.96]
Stroke243874058121.39 [0.92-2.11]
Pulmonary embolism618231110131.79 [1.01-3.17]
Deep vein thrombosis223524365242.05 [1.34-3.15]
Cataracts50757491NANA37381.12 [1.00-1.25]
Osteoporotic fracture1371112219NANA40450.88 [0.72-1.08]
Myocardial infarction28313655551.14 [0.75-1.74]

Tamoxifen therapy is effective only for ER-positive breast cancers, which account for 78% of all breast cancers in our model.14 We neglect the distinctions of mutations of the breast cancer 1 and 2 genes (BRCA1/2) as well as cytochrome p450 family 2, subfamily D (CYP2D) mutations in the general population.


Costs for disease events are listed in Table 3 and are adjusted to 2010 US dollars using the Consumer Price Index for medical costs.36 Costs for fractional years spent in each disease state are prorated. Cost of death and costs related to the terminal year of life are separated to account for the accumulation of costs in the final months and days of life, as reported by Hoover et al.35 Cost for the last year of life is prorated when death soon follows (eg, within 1 year of) the disease event. End-of-life and death costs are applied to all simulated individuals regardless of cause of death. An annual supply of tamoxifen costs $203.34 We assume that the cost of breast cancer risk prediction, which involves several questions about a patient's medical history, is negligible. The cost of screening mammography is neglected, because the screening protocol and compliance in each arm is identical. Future costs are discounted annually at 3%, according to convention.

Table 3. Cost Tablea
Clinical ConditionReferenceCost, $US
  • DVT indicates deep vein thrombosis; PE, pulmonary embolism.

  • a

    No costs are accrued for breast cancer after Year 10, for endometrial cancer after Year 5, or for stroke after Year 2. Stroke costs are used only in sensitivity analysis when the stroke hazard ratio is not unity. PE, DVT, and cataracts accrue cost only at the time of the event. Screening mammogram costs are neglected.

Breast cancerMelnikow 20068 
 Year 1 22,418
 Year 2 1902
 Year 3 1509
 Years 4-10 1433
Endometrial cancerMelnikow 20068 
 Year 1 17,391
 Year 2 284
 Years 3-5 190
StrokeSamsa 199933 
 Year 1 27,325
 Year 2 2786
DVTMelnikow 200686274
PEMelnikow 2006816,943
Cataracts8Melnikow 200685484
Tamoxifen, 1-y 201034203
DeathHoover 20023512,147
Last y of lifeHoover 20023539,236

Quality-Adjusted Utility Weights

Quality-adjusted utilities for disease states are listed in Table 4. All disease utilities, except for those for breast cancer and cataracts, are based on data reported by Locker et al.38 Data on the utility for cataracts were derived from Melnikow et al,7 and data on the utility for breast cancer were derived from Tosteson et al37 and Meadows et al.39 The utility for terminal year of cancer is applied to the final year of life for those whose cause of death is cancer. Similarly to Melnikow et al,7 we consider QALY assumptions both with and without the adjustment for common symptoms associated with tamoxifen therapy. QALYs (and life years) are discounted annually at 3%.

Table 4. Quality-Adjusted Life Year Utility Assumptionsa
Disease StateReferenceUtility Weight
  • a

    The base case quality-adjusted life year weights are shown for the disease states that were modeled. Separate analyses were performed with and without the utility for common tamoxifen symptoms.

Healthy 1.0
Breast cancerTosteson 200837 
 Local 0.90
 Regional 0.75
 Distant 0.60
Endometrial cancerLocker 2007380.839
Terminal year of cancerMeadows 2007390.23
Deep vein thrombosisLocker 2007380.729
Pulmonary embolismLocker 2007380.741
StrokeLocker 2007380.707
CataractsMelnikow 200870.772
Common tamoxifen symptomsLocker 2007380.959

Simulation Details

Women representative of the US population are selected at random from the National Health and Nutrition Examination Survey (NHANES) database (1999-2006),40 and information on their ages of menarche, first live birth, and other variables pertinent to breast cancer are used as inputs. Family history, which is not present in NHANES, is assigned randomly based on the distribution of the number of affected first-degree relatives with breast cancer reported in Pinsky et al.41

We simulated 1.5 million postmenopausal women aged <55 years for lifetime follow-up. The average age of menopause in this population is 47 years. Simulated women with events of breast cancer (invasive or in situ), endometrial cancer, PE, DVT, or cataracts before the start of tamoxifen therapy were excluded. Each woman was simulated in 2 trial arms: a control arm with no tamoxifen therapy and a treatment arm with perfect compliance to treatment in which women with a predicted 5-year breast cancer risk greater than a predetermined threshold take tamoxifen therapy for 5 years (with tamoxifen discontinued at the occurrence of an AE). All breast cancer risks in the model and sensitivity analyses are 5-year absolute risks.


Tamoxifen chemoprophylaxis prevents more breast cancers in the treatment arm compared with the control arm in populations of increasing 5-year breast cancer risk (see Table 5). Comparing the number of breast cancers prevented (BCPrevented) with the number of AEs caused (AECaused), there is modest benefit to receiving tamoxifen therapy for low-to-average risk populations (risk between 0% and 1.66%), as evidenced by the 2:1 ratio of BCPrevented to AECaused (Table 5). For those with higher 5-year risk (≥1.66%), the BCPrevented-to-AECaused ratio improves to 3:1. This result reinforces the established finding that increased risks of side effects associated with tamoxifen therapy are more acceptable when breast cancer risk is sufficiently high.6, 8, 9, 42, 43

Table 5. Health Outcomesa
Risk Group, %Breast Cancer Cases Prevented Per 1000 WomenBreast Cancer Deaths Prevented Per 1000 WomenAdverse Event Cases Gained Per 1000 Women
  • a

    Differences in health outcomes between the treatment and control arms of a simulated population of postmenopausal women aged <55 years at baseline. Adverse events included endometrial cancer, pulmonary embolism, and deep vein thrombosis. Cataract, which occurs in approximately 70% of the population and is largely treatable, was neglected in these counts.


Cost outcomes, however, reveal a significant reversal of established results. Tamoxifen treatment, compared with control, is cost-saving when received by postmenopausal women aged <55 years who have a breast cancer risk ≥1.66% (see Table 6). The efficiency frontier is plotted in Figure 2 for all risk thresholds explored. Upon lowering the risk threshold (from a risk ≥2% down to a risk ≥0.8%), the incremental costs associated with treatment increase along with QALYs saved, resulting in incremental cost-effectiveness ratios (ICERs) that remain well below the often-used cost-effectiveness benchmark of $50,000 per QALY44 (see Table 7). Treating all postmenopausal women regardless of risk costs $51,200 per QALY saved relative to a population with a risk ≥0.8%.

Figure 2.

This plot illustrates the efficiency frontier for all risk thresholds that were explored in the current study. Quality-adjusted life years (QALYs) saved versus increased cost between treatment and control arms are illustrated for several breast cancer risk thresholds in a fixed population of 150,000 postmenopausal women aged <55 years over the population's lifetime.

Table 6. Average Cost-Effectiveness Ratio Outcomesa
Risk Group, %Life Years Saved Per 1000 Treated WomenQALYs Saved Per 1000 Treated WomenAdditional Cost Per Treated Woman, $USCost Per Life Year Saved, $USCost Per QALY Saved (ACER), $US
  • QALYs indicates quality-adjusted life years; ACER, average cost-effectiveness ratio.

  • a

    Differences in cost and life-year outcomes between treatment and control arms are shown for postmenopausal women aged <55 years. Life years, QALYs, and costs are discounted annually by 3%.

Table 7. Incremental Cost-Effectiveness Ratio Outcomesa
Risk Group, %Incremental Cost of Treatment, $USIncremental QALYs Saved With TreatmentICER, $US
  • QALYs indicates quality-adjusted life years; ICER, incremental cost-effectiveness ratio.

  • a

    Incremental costs and QALYs saved between treatment and control arms are shown relative to the next risk threshold for a fixed a population of approximately 150,000 postmenopausal women aged <55 years over the population lifetime.


The principle differences between the current model of tamoxifen therapy and previous models6-9 are 2-fold. First, previous analyses did not consider the persistence of breast cancer risk reduction after completing tamoxifen therapy. Second, the hazard ratios for AEs associated with tamoxifen therapy in the current study are derived from a meta-analysis of 4 chemoprevention trials, whereas previous work used data from 1 trial. If we consider a model of tamoxifen with no persistent risk reduction (the feature with the most significant impact on cost-effectiveness outcomes), we observe qualitatively consistent results compared with previous analyses7: Tamoxifen treatment is dominated (additional cost but no QALY savings in the treatment arm compared with control) for average risks.

A summary of the sensitivity analyses performed for women with a risk ≥1.66% is provided in Figure 3. The average cost-effectiveness results—treatment is cost-saving compared with control for populations with breast cancer risk ≥2% and is cost-effective otherwise—remain robust to sensitivity analyses performed for various discount factors (1% to 5%) and costs (tamoxifen and cancer costs increased up to 3 times).

Figure 3.

This chart illustrates a sensitivity analysis according to absolute changes in the average cost-effectiveness ratio (ACER) for varied parameters compared with base-case parameters (see Tables 2 and 3). The results are dominated when all side-effect relative risks are elevated to the upper limit of the 95% confidence interval and when persistent breast cancer risk reduction is reduced to 0 years. QALY indicates quality-adjusted life year.

Sensitivity analysis of side-effect hazard ratios reveals a large impact on average cost-effectiveness ratio (ACER) results for the population at large but minimal impact on those with a risk ≥1.66%. Although the elevated stroke hazard ratio is not statistically significant when it is varied from 1.0 to 1.39, tamoxifen treatment is dominated at low risk thresholds (risk ≥1.25%) but remains cost-saving at a risk ≥3%. Life year savings are observed for all risk thresholds, although they are marginal (9 life years per 1000 women) for a risk ≥1.25%. When the hazard ratio for endometrial cancer is varied from 2.41 to 3.96, tamoxifen therapy is dominated compared with control for QALYs and life years for women at all risk thresholds. Finally, tamoxifen therapy is dominated for all women at any risk threshold when the upper limit of the 95% confidence interval for the hazard ratio is used for each of the side-effects modeled.

The sensitivity of QALY savings to assumptions used in assigning utilities has been explored previously.7 We produce consistent results where the inclusion of a quality-adjusted utility associated with tamoxifen-induced menopausal symptoms (eg, hot flashes, vaginal discharge) leads to expensive or dominated results (Table 8). There are noteworthy discrepancies in the literature for the utility weight of these common symptoms and for other health states, including breast cancer.7, 9, 37-39 Some of these discrepancies can be explained by the characteristics of the populations that were used to derive the weights: The presence of serious disease in a population will significantly modulate the anxieties surrounding future disease prospects.7, 38 The variability and general difficulty of assigning QALY utilities is well known, and their choice can have important consequences for cost-effectiveness analyses.7, 45

Table 8. Quality-Adjusted Life Years Saveda
 QALY Utility Assumptions
Risk Group, %Base QALY (Locker 200738)Base QALY Plus Tamoxifen Symptoms (Locker 200738)Melnikow QALY (Melnikow 20087)Melnikow QALY Plus Tamoxifen Symptoms (Melnikow 20087)
  • QALYs indicates quality-adjusted life years.

  • a

    Differences in QALYs per woman are shown between the treatment and control arms under 4 sets of utility assumptions for various 5-year breast cancer risk thresholds.


The persistent reduction in breast cancer risk after tamoxifen therapy is modeled as a 10-year effect. Kaplan-Meier curves for breast cancer incidence from long-term chemoprevention trials10, 11 are highly suggestive of an even longer duration of preventive benefit. Extending the length of breast cancer risk reduction after active treatment to 20 years and for a lifetime produces cost-saving ACER results for all women. With 20 years of treatment, the ratio of breast cancers prevented to AEs caused is 4:1 for women at any risk threshold and 6:1 if their risk ≥1.66%.


Current guidelines from the American Society of Clinical Oncologists and the US Preventive Services Task Force recommend consideration of tamoxifen to reduce breast cancer risk in women who have a predicted 5-year breast cancer risk ≥1.66%.46, 47 Our current analysis indicates that the benefits of tamoxifen chemoprevention can compensate sufficiently for its side-effect profile in a postmenopausal population aged <55 years with a risk ≥1.66%. Tamoxifen use in this population is forecast to save 85 QALYs per 1000 postmenopausal women aged <55 years with cost savings of $47,580 compared with no treatment over lifetime follow-up.

The current analysis does not address possible increased mortality for tamoxifen users who develop ER-negative breast cancer1 or the possible ineffectiveness of tamoxifen in 2% to 10% of the population with a CYP2D6 mutation that inhibits its metabolism.48 These details may change outcomes for tamoxifen efficacy at the population level, although we expect that tamoxifen will remain cost-effective compared with control for women with a risk ≥1.66%. Perfect adherence to tamoxifen in the treatment arm overestimates actual tamoxifen use.2 More realistic compliance would not change ACER or ICER values but would change total QALYs saved between the treatment and control arms. Finally, common tamoxifen symptoms, which play a decisive role in patient adherence to tamoxifen,49, 50 are not modeled explicitly but are explored in a sensitivity analysis on QALY utilities.

The reduced side-effect profile of raloxifene makes it an attractive alternative to tamoxifen, but it is less effective in reducing breast cancer risk.51 Moreover, long-term breast cancer risk reduction after active treatment, a critical benefit of tamoxifen, has yet to be observed.52, 53 These factors, in addition to the 6-fold greater cost of raloxifene,34 led to our focus on tamoxifen for the current breast cancer chemoprevention cost-effectiveness analysis.

Anastrozole, an aromatase inhibitor, has demonstrated significant chemopreventive potential. Anastrozole therapy reportedly was almost twice as effective as tamoxifen in preventing contralateral breast cancer, both at the end of 5 years of treatment and 4 years after treatment.54 The side-effect profile, which includes increased risk of fractures but lower risks of endometrial cancer and thromboembolic events, may be more acceptable. Currently, the second International Breast Cancer Intervention Study (IBIS-II) is investigating anastrozole prophylaxis in a postmenopausal population.55 If the rate of breast cancer risk reduction after active treatment persists, then anastrozole would be a worthwhile new option for preventing ER-positive breast cancers in postmenopausal women.

Identifying a population that is at greatest risk for developing ER-positive breast cancer, at lowest risk for developing AEs, and affected least by common side effects is critical to promoting tamoxifen therapy. One promising approach may be to stratify risk according to a version of the Gail score, as developed by Chlebowski et al,56 which distinguishes risk according to hormone receptor status. Such stratification would increase the number of prevented breast cancers while limiting the risks of AEs to a narrower population, making the favorable cost outcomes presented here even more compelling. More precise breast cancer risk predictors and a better understanding of the risk factors that influence the side-effect profile of tamoxifen, coupled with the yet-unknown total duration of the breast cancer risk reduction achievable with tamoxifen, suggest a promising future for the chemoprevention of breast cancer.


The authors made no disclosures.