Chemoprevention: Drug pricing and mortality

The case of tamoxifen

Authors


  • The National Cancer Institute provided review of the article but had no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; or preparation or approval of the article.

Abstract

BACKGROUND.

Tamoxifen is a prototypic cancer chemopreventive agent, yet clinical trials have not evaluated its effect on mortality or the impact of drug pricing on its cost-effectiveness.

METHODS.

A state-transition Markov model for a hypothetical cohort of women age 50 years was used to evaluate the effects of tamoxifen on mortality and tamoxifen price on cost-effectiveness. Incidence and mortality rates for breast and endometrial cancers were derived from Surveillance, Epidemiology and End Results statistics, and noncancer outcomes were obtained from published studies. Relative risks of outcomes were derived from the National Surgical Adjuvant Breast and Bowel Project P-1 trial. Costs were based on Medicare reimbursements.

RESULTS.

Projected overall mortality for women at 1.67% 5-year breast cancer risk showed little difference with or without tamoxifen, resulting in a cost-effectiveness ratio of $1,335,690 per life-year saved as a result of tamoxifen use. Adjusting for the differential impact of estrogen receptor-negative cancers, tamoxifen increased mortality for women with a uterus until the 5-year breast cancer risk reached ≥2.1%. Assigning the Canadian price for tamoxifen dramatically reduced the incremental cost (to $123,780 per life-year saved). At that price, the use of tamoxifen was less costly and more effective for women with 5-year breast cancer risks >4%.

CONCLUSIONS.

Tamoxifen may increase mortality in women at the lower end of the “high-risk” range for breast cancer. If prices in the U.S. approximated Canadian prices, then tamoxifen use for breast cancer risk reduction in women with a 5-year risk >3% could be a reasonable strategy to reduce the incidence of breast cancer. Because they are used by many unaffected individuals, the price of chemopreventive agents has a major influence on their cost-effectiveness. Cancer 2006. © 2006 American Cancer Society.

Tamoxifen, which is a prototype chemoprevention agent, was approved by the U.S. Food and Drug Administration (FDA) for breast cancer risk reduction in 1998 for women who had a 5-year risk of breast cancer ≥1.67%. This recommendation was based on the results of the National Surgical Adjuvant Breast and Bowel Project (NSABP) P-1 trial, in which women who took tamoxifen had a 49% reduction in the incidence of breast cancer over 5 years.1 The U.S. Preventive Services Task Force2 and the Canadian Task Force on Preventive Health Care3 have advised physicians to counsel women who have a 5-year estimated breast cancer risk of ≥1.67% concerning the potential benefits and harms of tamoxifen.

Despite its beneficial effects on breast cancer risk, tamoxifen has been associated with serious adverse effects. The NSABP P-1 trial documented increased risks of venous thromboembolism, endometrial cancer, and cataracts that required surgery; a borderline increase in the risk of stroke; and a borderline decrease in the risk of osteoporotic fractures.3 Risks of adverse outcomes were greater among women age >49 years, the same group of women generally at higher risk of breast cancer.4 Studies of tamoxifen for breast cancer risk reduction conducted in Europe have revealed no risk reductions5, 6 or smaller7 risk reductions than the NSABP P-1 trial. A metaanalysis of all tamoxifen risk-reduction trials estimated a 38% reduction in breast cancer risk.8

The effect of tamoxifen on mortality when used for risk reduction has not been well defined because of the need for many years of follow-up. The metaanalysis did not detect any effect on mortality, concluding that follow-up data were not sufficient to judge tamoxifen's effect on breast cancer deaths.8 In all trials that reported a risk reduction, tamoxifen reduced the risk of estrogen receptor (ER)-positive breast cancers and had essentially no effect on ER-negative breast cancers,3, 7, 8 which imply a poorer prognosis.9 Because of this, women with breast cancer that is not prevented by tamoxifen are likely to have higher than average mortality. How the greater proportion of ER-negative cancers that occur in women who take tamoxifen for risk reduction may affect their breast cancer prognosis and the resulting potential effect of tamoxifen on mortality have not been explored to date.

Clearly, the price of tamoxifen also will influence its cost-effectiveness. The impact of high drug prices on the U.S. health care system and on individuals has received considerable attention from the media, the public, and Congress.10–13 Several states and cities in the U.S., as well as many individual purchasers, have opted to obtain medications from Canada, where prices often are lower.10 These price differentials are particularly striking in the case of tamoxifen; the average wholesale price (AWP) in the U.S. is 7 times higher than typical Canadian Internet pharmacy prices.14

We used Markov modeling to explore the effect of tamoxifen for breast cancer risk reduction on health outcomes, mortality, and cost per life-year saved. Potential differences in breast cancer prognosis based on different proportions of hormone treatment-responsive cancers among women receiving and not receiving tamoxifen were examined. We adopted a rigorous approach to cost estimation and focused on the impact of different prices of tamoxifen covering both U.S. and Canadian pricing.

MATERIALS AND METHODS

This study was exempted from review by the University of California–Davis Office for Human Research Protection.

Study Design

We adopted the payer's perspective, considering the effects of receiving tamoxifen for 5 years among women age ≥ 50 years with a 5-year breast cancer risk ranging from 1% to 5%, compared with a strategy of usual care without tamoxifen. We projected health outcomes, life expectancy, and costs15 by using a woman age 50 years with a 5-year breast cancer risk of 1.67%, the threshold for risk reduction therapy used in the NSABP P-1 trial and approved by the FDA,3, 16 as the base case.

Model

We developed a state transition Markov Model17, 18 in DATA version 4.0 (TreeAge Software) to track hypothetical cohorts of women who did or did not take tamoxifen for breast cancer risk reduction. The base-case model followed women age 50 years until death or age 100 years in annual cycles and included outcomes described in the NSABP P-1 trial: reduced risks of invasive breast cancer, ductal carcinoma in situ (DCIS), and osteoporotic fractures and increased risks of endometrial cancer, deep venous thrombosis (DVT), pulmonary embolism, and cataracts that required surgery. We considered mortality from breast cancer, endometrial cancer, pulmonary embolism, hip fracture, and age-specific death from all other causes. Mortality rates for DCIS, cataracts, DVT, and nonhip fractures were assumed to be negligible. Breast and endometrial cancer were modeled as “tunnel” states of 10 years' and 5 years' duration, respectively, based on an examination of the survival curves for these cancers generated by the Surveillance, Epidemiology, and End Results (SEER) Program.20 The survival analysis, based on the American Joint Committee on Cancer staging system (3rd edition) was conducted for women age 50 years and older who had a malignant or in situ cancer diagnosed between 1990 and 1999.

Probabilities

Participants in randomized trials are a highly selected group with better health status than the general population.19 To create a model more generalizable to the U.S. population, baseline, age-specific incidence rates for outcomes were obtained from SEER data from 1990 to 1999 for endometrial cancers,20 adjusting the denominator for the rate to reflect only those women who had not had a hysterectomy, from California hospital-discharge data for venous thromboembolism21 and from population-based published rates for osteoporotic fractures22, 23 and cataracts.24 The 5-year incidence for invasive breast cancer was modeled in the base-case analysis at 1.67%, which was the eligibility threshold for the NSABP P-1 trial, and varied from 1% to 5%. The incidence of DCIS was tied to the invasive breast cancer rate based on the relative proportion of DCIS to invasive breast cancer found in the SEER data.20

Mortality rates for were obtained from SEER from 1990 to 1999 for breast and endometrial cancer,20 from the published literature for hip fracture25–30 and pulmonary embolism,31–34 and from the National Center for Health Statistics for death from all other causes.35 For the base-case analysis, relative risks of invasive breast cancer, DCIS, osteoporotic fractures, endometrial cancer, pulmonary embolism, DVT, and cataracts that required surgery for women age 50 years and older on tamoxifen were taken from the NSABP P-1 published trial results.3 Table 1 shows model probabilities, relative risks, and costs.

Table 1. Model Variables and Sources
VariableMeasureSource
  • SEER indicates Surveillance, Epidemiology, and End Results; ER, estrogen receptor; RR, risk ratio; 95% CI, 95% confidence interval; $US, U.S. dollars; CP, clinical pathways; AWP, average wholesale price; HMO, health maintenance organization.

  • *

    Calculated from hip fracture rates (Melton et al., 198822) and proportions (Siris et al., 200123).

  • Overall mortality rates were based on SEER; proportion of ER-negative cancers (39.5%) from tamoxifen trials overview (Cuzick et al., 20038) was applied to SEER rates to adjust projected mortality for ER status.

  • It was assumed that treatment costs for breast cancer occurred during the receipt of tamoxifen were not to include treatment with tamoxifen.

  • §

    Unpublished data.

Incidence rates (per 100,000 women per year)
 Endometrial cancer SEER, 200320
  Ages 50–54 y75 
  Ages 55–59 y108 
  Ages 60–64 y139 
  Ages 65–69 y153 
  Ages 70–74 y155 
  Ages 75–79 y152 
  Ages 80–84 y144 
  Age ≥85 y103 
 Pulmonary embolism White et al., 200321
  Ages 50–54 y20 
  Ages 55–64 y41 
  Ages 65–74 y95 
  Ages 75–84 y139 
  Age ≥85 y174 
 Deep venous thrombosis White et al., 200321
  Ages 50–54 y41 
  Ages 55–64 y82 
  Ages 65–74 y190 
  Ages 75–84 y277 
  Age ≥85 y347 
 Cataracts Leske and Sperduto, 198324
  Ages 55–64 y182 
  Age ≥65 y1220 
 Osteoporotic fracture Melton et al., 198822; Siris et al., 200123
  Vertebral fracture*
   Age ≥50 y14.8 
  Wrist fracture*
   Age ≥50 y44.3 
  Hip fracture
   Ages 45–54 y27.9 
   Ages 55–64 y95.5 
   Ages 65–74 y317.7 
   Ages 75–84 y1310.5 
   Age ≥85 y2683.9 
 Disability requiring long term care after hip fracture Braithwaite et al., 200325
  Age 80 y13,000 
Overall mortality rates (per 100,000 cases per year)
 Breast cancer: Age ≥50 y (ER adjustment) SEER, 200320
  Year 12440 (2000) 
  Year 22780 (3600) 
  Year 32870 (3730) 
  Year 42430 (2810) 
  Year 52180 (2700) 
  Year 62200 (2320) 
  Year 71940 (2300) 
  Year 81550 (1670) 
  Year 91720 (1740) 
  Year 10570 (700) 
 Endometrial cancer: Age ≥50 y SEER, 200320
  Year 13390 
  Year 22560 
  Year 31560 
  Year 4890 
  Year 5890 
 Hip fracture (adults)20,000Braithwaite et al., 200325; Fransen et al., 200226; Hannan et al., 200127; Heikkinen et al., 200128; Wolinsky et al., 199729; Parker et al., 200030
 Pulmonary embolism (adults)16,000Carson et al., 199231; Janke et al., 200032; Kniffin et al., 199433; Anderson et al., 199134
Outcomes: RR (95% CI)
 Breast cancer
  Age 50–59 y0.49 (0.27–0.74)Fisher et al., 19981
  Age ≥60 y0.45 (0.29–0.81)Fisher et al., 19981
 Deep venous thrombosis1.71 (0.85–3.58)Fisher et al., 19981
 Osteoporotic fracture0.79 (0.60–1.05)Fisher et al., 19981
 Cataracts1.57 (1.16–2.14)Fisher et al., 19981
 Pulmonary embolism3.19 (1.12–11.15) 
 Endometrial cancer4.01 (1.70–10.90)Fisher et al., 19981
Clinical condition: costs ($US)
 Breast cancer on tamoxifen
  Year 1$14,830CP
  Year 2$1048 
  Year 3$797 
  Year 4$746 
  Years 5–10$746 
 Breast cancer after no tamoxifen
  Year 1$14,958 
  Year 2$1269CP
  Year 3$1007 
  Year 4$956 
  Years 5–10$956 
 Endometrial cancer
  Year 1$11,604 
  Year 2$190 
  Year 3$127CP
  Year 4$127 
  Year 5$127 
 Hip fracture, acute nursing home care$14,313CP
 Hip fracture, home care$362CP
 Posthip fracture disability$8529CP
 Cataracts$3659CP
 Deep venous thrombosis$4186CP
 Osteoporotic fractures
  Wrist$212 
  Spinal compression$189CP
  Hip$9502 
 Pulmonary embolism$13,167CP
 Tamoxifen (1-y supply)
  AWP§$1212 
  HMO§$415Medi-Span, 200514
    Canadian Internet ($US)§$163 
 Last year of life Grady et al., 200443
  Ages 30–74 y$35,799 
  Ages 75–84 y$35,082 
  Age ≥85 y$26,389 
    

For the sensitivity analyses, relative risks were varied 1 at a time to the extremes of the 95% confidence intervals obtained from the NSABP P-1 trial, then substituted with relative risks from a recently published metaanalysis of tamoxifen.8 To examine the potential effect on overall prognosis of the higher proportion of ER-negative breast cancers among women who receive tamoxifen, we applied the proportion of ER-negative cancers found in the metaanalysis8 to the model and estimated breast cancer mortality for women with ER-positive and ER-negative cancers based on SEER data.20

Costs

To estimate direct medical care costs associated with inpatient, outpatient, and nursing facility care, detailed clinical pathways were defined by the physician authors (J.M. and J.N.) for each outcome state. Each pathway was specified for a typical patient as a comprehensive series of 1) Diagnosis-Related Group (DRG) codes for hospital charges reimbursed under the prospective payment system; 2) Current Procedural Terminology codes for laboratory tests, anesthesiology, and all other professional services that were not covered by the DRG reimbursement; and 3) drugs administered or prescribed. The Medicare average reimbursement for hospital care was calculated by dividing the total Medicare reimbursements for that DRG by the associated number of discharges using data from the Medicare Provider Analysis and Review.36

Costs were assigned to laboratory procedures on the basis of the midpoint for the Clinical Diagnostic Laboratory Fee Schedules.37 Costs for anesthesiology were based on the sum of national average time units and base units multiplied by the average anesthesia conversion factor.38 For all other professional services, costs were based on total transitioned facility relative value units multiplied by a conversion factor.39

Prices for drugs coded under Healthcare Common Procedure Coding System “J-codes” and reimbursed by Medicare (prior to Part D) were obtained from the Medicare Single Drug Pricer File.39 Drugs not covered by Medicare were assigned costs on the basis of reported AWP.14 Generic forms of tamoxifen recently have become available in the U.S. We investigated tamoxifen pricing at local pharmacies and at U.S. and Canadian Internet pharmacies and obtained reimbursement rates for tamoxifen by MediCal and several large health maintenance organizations (HMOs). Alternative price estimates for the model were based on averages of 1) California HMO reimbursements and 2) Canadian Internet prices converted to U.S. dollars.

Age-specific net costs for the last year of life were estimated from Medicare reimbursements.40 All costs were based on reimbursements or payment schedules for 2000, and all other costs were adjusted to constant Year 2000 dollars using the Medical Care Component of the Consumer Price Index. Future costs and life-years were discounted at 3% throughout the model. Table 1 shows costs for each clinical condition included in the model and a range of tamoxifen prices.

Analyses

Monte Carlo simulations41 were used to estimate tamoxifen-related deaths and mortality from other causes (excluding breast and endometrial cancer deaths). Mortality for theoretical cohorts of 10,000 women, with 5000 using tamoxifen, was estimated from 1000 simulations for 10-year and 50-year intervals. Simulations were repeated with breast cancer mortality adjusted for differences in mortality among women with ER-negative and ER-positive breast cancer.

Cost-effectiveness analyses were conducted separately for women with and without a uterus, because women who have undergone hysterectomy are no longer at risk for endometrial cancer. Our base-case analysis estimated the incremental cost-effectiveness ratio of 5-years of tamoxifen use by a woman age 50 years with a 5-year risk of invasive breast cancer of 1.67%. In sensitivity analyses, models were run for women starting tamoxifen at age 60 years and at age 70 years.

RESULTS

Tamoxifen-related outcome-specific and other all-cause mortality estimates from the Monte Carlo simulation are shown in Figure 1 (for a 10-year time horizon) and Table 2 (for a 50-year time horizon) for 5-year breast cancer risks of 1.67%, 3%, and 5%. Both benefits and harms of tamoxifen are more evident after 10 years but tend to equalize with the lifetime time horizon. At a 5-year breast cancer risk of 1.67%, the projected 50-year mortality from tamoxifen-related outcomes essentially is equal between the 2 strategies; whereas, in the 10-year time horizon, a few more deaths were projected for women receiving tamoxifen. Adjusting the model for the higher proportion of ER-negative breast cancers occurring in tamoxifen users, the number of projected breast cancer deaths for the tamoxifen cohort increased from 8 deaths to 10 deaths over 10 years and from 67 deaths to 75 deaths over the 50-year projection. For higher breast cancer risks, the benefits of tamoxifen in reducing breast cancer mortality overtake mortality from tamoxifen-related adverse events.

Figure 1.

Projected mean tamoxifen-related and other (all cause) 10-year mortality was based on 1000 Monte Carlo simulations of 3 hypothetical cohorts of 10,000 women (5000 women receiving tamoxifen for 5 years at age 50 years and 5000 women not receiving tamoxifen). Mortality is shown for cohorts with 5-year breast cancer risks of 1.67%, 3%, and 5%. UCA indicates endometrial cancer; PE, pulmonary embolism; Hip Fx, hip fracture.

Table 2. Projected 50-Year Mortality from Tamoxifen-Related Outcomes Based on 1000 Monte Carlo Simulations of a Cohort of 5000 Women who Started Tamoxifen for 5 years at Age 50 Years and a Cohort of 5000 Women who Did Not Start Tamoxifen*
Tamoxifen-related outcomesProjected 50-Year Mortality: Mean (Standard Deviation)
5-year Breast cancer Risk score = 1.67%5-year Breast cancer Risk score = 3.0%5-year Breast cancer Risk score = 5.0%
TamoxifenNo tamoxifenTamoxifenNo tamoxifenTamoxifenNo tamoxifen
Breast cancer67 (8.1)82 (8.6)131 (11.1)143 (11.4)211 (14.8)229 (15.3)
Endometrial cancer33 (5.7)22 (4.7)32 (5.4)22 (4.6)32 (5.6)21 (4.4)
Hip fracture135 (11.2)136 (11.2)131 (11.3)132 (11.4)125 (11.0)125 (11.1)
Pulmonary embolism28 (4.9)24 (4.5)28 (5.3)24 (4.9)26 (4.9)22 (4.7)
Total no. of deaths263264322321394397

Estimated total and incremental costs and effects and incremental cost-effectiveness ratios from the Markov model are shown in Table 3 for the base case for women with and without a uterus. That analysis indicated a negligible increase in life expectancy for women at the FDA threshold of a 1.67% 5-year risk at extremely high cost. The incremental cost per year of life saved was $1,335,690, and the mean life expectancy was projected to increase by 1.6 days. For a woman age 50 years with her uterus and with a 1% 5-year breast cancer risk, tamoxifen was dominated (more costly and less effective). At a 5% 5-year risk, the incremental cost-effectiveness was $64,778 per life-year saved, with a projected improvement in mean life-expectancy of 26.7 days (Fig. 2). The incremental cost-effectiveness ratio was more favorable for women who had a prior hysterectomy, reaching $177,116 per life-year saved at a 5-year breast cancer risk of 1.67% and with a mean projected increase in life-expectancy of 11.5 days. For these women, at 5% risk the incremental cost-effectiveness ratio was $46,954 per life-year saved, with a projected improvement in mean life expectancy of 36.5 days. For women age ≥60 years with a uterus, tamoxifen was dominated except among women with a very high breast cancer risk: at a 5% 5-year risk, the incremental cost per life-year saved was $143,945.

Figure 2.

This sensitivity analysis shows 5-year breast cancer risk and cost per life-year saved for tamoxifen among women age 50 years with and without a uterus for the base case analysis and with the analysis adjusted for the worse prognosis of estrogen receptor-negative tumors.

Table 3. Base-Case Cost-Effectiveness Analysis: 5-Year Breast Cancer Risk of 1.67%
InterventionLife-YearsIncremental effect (Days)Cost (US$)Incremental cost (US$)Incremental cost per Life-Year saved (US$)
  1. US$ indicates U.S. dollars; AWP, average wholesale price; HMO, health maintenance organization.

For women with a uterus
 No tamoxifen19.6924 17,695  
 Tamoxifen prophylaxis
  At AWP of $1212/year19.69671.623,37856831,335,690
  At HMO reimbursement rate of $415/year19.69671.619,4601765414,916
  At Canadian Internet price of $163 (US)/year19.69671.618,221527123,780
For women without a uterus
 No tamoxifen19.7250 17,411  
 Tamoxifen prophylaxis
  At AWP of $1212/year19.756411.522,9635552177,116
  At HMO reimbursement rate of $415/year19.756411.519,014160351,146
  At Canadian Internet price of $163 (US)/year19.756411.517,76635511,315

Figure 2 presents the effects of adjusting for the higher proportion of ER-negative breast cancers arising in women taking tamoxifen on the projected cost per life-year saved for women with and without a uterus over a 1% to 5% 5-year risk of breast cancer. Accounting for the higher projected mortality of ER-negative breast cancers, life expectancy for women on tamoxifen is shortened, and tamoxifen was dominated for women with a uterus until the 5-year breast cancer risk reached 2.1% or more.

In threshold analyses in which relative risks from the NSABP P-1 trial ranged through their 95% confidence intervals, tamoxifen was dominated at a 5-year breast cancer risk of 1.67% if the relative risk of invasive breast cancer increased from 0.51 to 0.55, if the relative risk of endometrial cancer rose from 4.01 to 4.50, or if the relative risk for pulmonary embolism rose from 3.19 to 4.11. The use of pooled relative risks from the trial metaanalysis8 in place of estimates from the NSABP P-1 trial resulted in a lower incremental cost-effectiveness ratios for tamoxifen: $154,622 per life-year saved (compared with $160,615) at a 5-year breast cancer risk of 3% for women with a uterus. When the projected benefits of tamoxifen were limited to 5 years' duration, tamoxifen was dominated until the 5-year breast cancer risk reached 3%.

The model was highly sensitive to the price of tamoxifen. A survey of U.S. Internet and Sacramento-area retail pharmacy prices ranged from $216 to $1560 for a 1-year supply. The effects of changes in the price of tamoxifen on the projected cost per life-year saved are shown for women with and without a uterus in Figure 3. When average HMO reimbursement ($415 per year) was used as the price for tamoxifen in place of the AWP, the incremental cost per life-year saved was $414,916 at 1.67% 5-year breast cancer risk and $42,180 per life-year saved at a 3% risk. Applying the average Canadian Internet pharmacy price of $163 per year for tamoxifen to the model at the 1.67% 5-year risk reduced the incremental cost per life-year saved for women with a uterus to $123,780. For higher 5-year breast cancer risks (≥4%), tamoxifen was dominant (i.e., less costly and more effective).

Figure 3.

This sensitivity analysis shows the price of tamoxifen on the cost per life-year saved for women age 50 years with a 5-year breast cancer risk of 1.67%. HMO indicates health maintenance organization; AWP, average wholesale price; USD, U.S. dollars.

The tamoxifen price to achieve an incremental cost-effectiveness ratio of $100,000 per life-year saved for women with a uterus and a 5-year breast cancer risk of 1.67% was $144 per year. This price is lower than any U.S. price identified but higher than the Ontario formulary price for tamoxifen ($126 per year).42

DISCUSSION

Our analysis indicates that tamoxifen for breast cancer risk reduction is unlikely to have an important effect on overall mortality for women at a 5-year breast cancer risk of 1.67%. The difference in projected lifetime outcomes from taking or not taking tamoxifen for women with a uterus with a 5-year risk of 1.67% is small; consequently, the associated cost-effectiveness ratio is very high. When the effect of the poorer prognosis of the ER-negative breast cancers that occur among women taking tamoxifen on their breast cancer mortality is accounted for, tamoxifen is projected to reduce life expectancy slightly until women reach a minimum 2.1% 5-year breast cancer risk. For women who have undergone a hysterectomy, conversely, the risk of endometrial cancer from tamoxifen is eliminated; thus, tamoxifen becomes a more cost-effective approach to breast cancer prevention. Raloxifene has recently been shown to be comparable to tamoxifen in reducing the risk of invasive breast cancer, and its use was associated with fewer thromboembolic events, fewer cataracts, possibly fewer endometrial cancers, and possibly more ductal carcinoma-in-situ.44 Given these findings, if it were priced similarly to tamoxifen, it is likely that raloxifene would have a more favorable cost-effectiveness ratio than tamoxifen, closer to that found in our analysis for women who had a prior hysterectomy.

The current findings are highly sensitive to the price of tamoxifen. For women at breast cancer risks ≥3%, tamoxifen at Canadian prices both may increase life expectancy and may reduce overall costs. For health policy makers in the U.S., this illustrates the potential effect of negotiating pharmaceutical prices at a national level and the risk of using the AWP to define the negotiated price, which is done by most Medicaid plans.45 The range of tamoxifen prices in the U.S. and their deviation from the AWP is striking. Policies recommending medications for widespread prevention of cancer or other diseases may be more cost-effective for larger segments of the population in countries where the regulation of pharmaceutical pricing is related to acquisition costs.

Regardless of the price of tamoxifen, the projected benefits of tamoxifen for women at or near the threshold risk for breast cancer (1.67%) are very small or nonexistent. Other analytical models of tamoxifen have reached similar conclusions, finding that a 5-year breast cancer risk of 3% to 4% typically is required to show improved life expectancy from tamoxifen.46–48 In general, it appears that tamoxifen has greatest benefit for women ages 40 to 49 years; however, even in this age range, a 5-year breast cancer risk of 3.4% is needed for tamoxifen to be considered cost effective.49 Among women ages 50 to 59 years who were enrolled in the NSABP P-1 trial, the mean increase in life expectancy was estimated at 40 days in 1 analysis with an incremental cost of $82,784 per life-year saved.50 It is noteworthy that 75% of women enrolled in the trial had an estimated 5-year breast cancer risk >2%, and 17% had a risk >5%. Thirty-seven percent of women had undergone a previous hysterectomy. To our knowledge, no previously published analysis has accounted for the difference in prognosis for women with ER-negative cancers.

Our evaluation is limited by lack of consideration of the effects of tamoxifen on quality of life or costs of time lost from work. Additional work is needed to examine the impact of tamoxifen use on these important outcomes. We also did not consider the potential for increased risk of stroke from tamoxifen, because this risk was not statistically significant in the NSABP P-1 trial.1

The results of our analysis raise questions about the impact of tamoxifen for breast cancer risk reduction on short-term and longer term mortality. We have identified profound effects of tamoxifen price on cost-effectiveness. Our analysis is consistent with the conclusions of Freedman et al., who reported that a relatively small proportion of women who potentially may be eligible to take tamoxifen based on the current recommended risk threshold would experience a net benefit.51 When the lower survival rates from the ER-negative cancers that were not prevented by tamoxifen were considered in our analysis, tamoxifen appeared to be less effective and more costly than usual care at the 1.67% threshold risk. For the smaller group of women with a very high estimated breast cancer risk (>3% over 5 years), and particularly for those who have undergone hysterectomy, there is a potential benefit from tamoxifen. In addition to reducing cancer incidence, chemoprevention strategies require careful attention to side effects, mortality, and costs. To be effective, agents must be targeted to those with sufficient risk of the outcome to sustain the substantial potential benefit.

Acknowledgements

We thank Della Fletcher for expert coding of clinical pathways; Mairin Rooney for graphics design; Michael DeGregorio, PharmD, for his insights on tamoxifen pharmacology; Robert Mowers, PharmD, for information and advice on United States tamoxifen pricing and reimbursement; and Joel Lexchin, MD, for information on Canadian tamoxifen pricing.

Ancillary