NATIONAL INSTITUTE FOR HEALTH AND CLINICAL EXCELLENCE. Overview

NATIONAL INSTITUTE FOR HEALTH AND CLINICAL EXCELLENCE Overview Hormonal therapies for the adjuvant treatment of early oestrogen-receptor positive breast cancer The overview is written by members of the Institute s team of technical analysts. It forms part of the information received by the Appraisal Committee members before the first committee meeting. The overview summarises the evidence and views that have been submitted by consultees and evaluated by the Assessment Group, and highlights key issues and uncertainties. A list of the sources of evidence used in the preparation of this document is given in appendix A. Information that has been submitted to the Institute in confidence has been removed from this document 1 Background 1.1 The condition Breast cancer is the uncontrolled, abnormal growth of malignant breast tissue. It is the most common type of cancer among women in the UK. The incidence of breast cancer in England and Wales in 2003 was 38,864, accounting for approximately 30% of all reported cancers in women. The age standardised incidence rate in England and Wales is 144 and 113 per 100,000 population respectively, and women have a 1 in 9 lifetime risk of developing breast cancer. Breast cancer is also the most common cause of cancer-related deaths in women. Approximately 10,500 women died from breast cancer in England in 2003. National Institute for Health and Clinical Excellence v4 Page 1 of 26

Over the past 2 decades the incidence of breast cancer has increased in the UK, but the mortality rate has decreased. This has been attributed to the national breast screening programme implemented for women aged between 50 and 64, and to improved treatment options. However, breast cancer survival rates for England and Wales are below the European average. Breast cancer incidence increases with age and around 80% of breast cancers occur in women older than 50. Factors associated with increased breast cancer risk include previous breast cancer, early menarche, late menopause, hormone replacement therapy, oral contraception, obesity and alcohol consumption. It is also thought that breast cancer risk is increased in women who have not had children, or had children late, and women who have not breast-fed. Family history and genetic predisposition also play an important role, because women who possess mutations of breast cancer susceptibility genes (BRCA1 or 2) are at a higher risk of developing breast cancer. However, the majority of breast cancers occur in women with no direct family history of the disease. Once breast cancer is diagnosed, prognosis and treatment decisions depend on the extent of the disease. This is assessed by tumour staging, based on the size and nature of the primary tumour, the involvement of the regional lymph nodes and the presence of distant metastases (table 1). Breast tissue contains receptors for the female hormones oestrogen and progesterone. These receptors allow the breast tissue to grow or change in response to changing levels of those hormones. Approximately two thirds of women diagnosed with breast cancer have hormone-receptor-positive (HR+) tumours, which are more likely to respond to hormone therapy. HR+ tumours also tend to grow less aggressively, resulting in a better prognosis. National Institute for Health and Clinical Excellence v4 Page 2 of 26

Table 1 UICC a staging of breast cancer based on TNM b classification Ductal carcinoma in situ (DCIS) Early invasive breast cancer Locally advanced disease Metastatic breast cancer Stage Tumour size Regional lymph nodes Distant metastases 0 Negative Not detectable I < 2 cm Negative Not detectable II 2 5 cm (or < 5 cm) Negative (or positive) Not detectable III > 5 cm Positive Not detectable IV any Positive or negative a Union Internationale Contre le Cancer ; b Tumour Node Metastases 5-year survival Not reported 84% 71% 48% Detectable 18% The main risk for women who have had early breast cancer is recurrence of the cancer. The risk of recurrence is approximately 8% in the first year, rising to 13% in the second year and gradually declines thereafter to an annual risk of 2% after 10 years (data taken from assessment report figure 1). However, for women with bigger and/or node-positive tumours, these risks are much higher. Consultees from the professional organisations have commented that the published 5-year survival rates (table 1) may overestimate the current risk of recurrence because, with screening, increased patient awareness and better treatments, the recurrence and death rates have fallen rapidly over the past 15 years. 1.2 Current management Current treatment options for early breast cancer depend on disease characteristics (such as stage and hormone receptor [HR] status of the tumour), on patient characteristics (such as age and menopausal status) and on personal preferences. National Institute for Health and Clinical Excellence v4 Page 3 of 26

Treatment can be divided into surgical treatment and adjuvant treatment after surgical removal of the primary cancer. The purpose of surgery is to control the disease locally (within the breast and axillary lymph nodes) and to determine the prognostic characteristics of the primary cancer. This can be followed by adjuvant radiotherapy and/or chemotherapy. The aim of adjuvant treatment is to prevent recurrence. Even with early detection and successful surgery to remove the primary tumour and the local lymph nodes, the disease recurs in up to 45% of patients in the following 15 years, and approximately 35% of women die of breast cancer over this period. Hormonal therapy Oestrogen is the main hormone involved in the development and growth of HR+ breast cancer. In premenopausal women oestrogens are directly produced in the ovaries until production declines during the menopause. After the menopause oestrogens are still produced (to a lesser extent) in the peripheral tissues, by the enzyme aromatase converting androgens secreted from the adrenal cortex into oestrogens. The aim of hormonal therapy is to deprive the tumour cells of the proliferative stimulus provided by oestrogen. This can be achieved by blocking the binding of oestrogen to its receptor in the nucleus of responsive cells, as with tamoxifen. Five years of tamoxifen therapy reduces the recurrence rate to 33% and the mortality rate to 26% 15 years after the initial diagnosis. It is beneficial irrespective of age, menopausal status or previous use of chemotherapy. For postmenopausal women tamoxifen also provides protection against bone fractures and lowers serum cholesterol levels. However, long-term use of tamoxifen can be associated with vaginal bleeding, endometrial thickening, increased risk for endometrial cancer, thromboembolic events and ocular degeneration. Some of these adverse events associated with tamoxifen relate to the partial agonist effect of the drug. Resistance to tamoxifen has also been described. National Institute for Health and Clinical Excellence v4 Page 4 of 26

In 2002, the NICE cancer service guidance, Improving outcomes in breast cancer summarised contemporary service provision for diagnosis, treatment and follow-up of patients with early breast cancer. In the UK, 5 years of tamoxifen therapy has become standard adjuvant treatment for postmenopausal women with early HR+ breast cancer. In 2005, the Scottish Intercollegiate Guideline Network recommended in its breast cancer guidelines that aromatase inhibitors should be used in postmenopausal women if there are relative contraindications to the use of, or intolerance of, tamoxifen. Furthermore, the guidelines recommended switching to aromatase inhibitors in postmenopausal women after either 2 3 years or 5 years of tamoxifen therapy. 2 The technologies Aromatase is expressed in non-ovarian tissues such as muscle and fat in both premenopausal and postmenopausal women. Aromatase inhibitors act by blocking the conversion of androgens to oestrogens in the peripheral tissues in postmenopausal women, thereby reducing the circulating levels of oestrogens. Three aromatase inhibitors are available (table 2). Table 2 Summary description of technologies Generic name Anastrozole Letrozole Exemestane Proprietary name Arimidex Femara Aromasin Manufacturer AstraZeneca Novartis Pharmaceuticals Pfizer Dose 1 mg/day 2.5 mg/day 25 mg/day Acquisition cost excluding VAT ( British national formulary edition 50) 68.56/28 days 83.16/28 days 88.80/30 days 266.40/90 days Anastrozole is a reversible, non-steroidal aromatase inhibitor and has been shown to suppress oestradiol by more 80%. It is licensed in the UK for adjuvant treatment of postmenopausal women with HR+ early invasive breast National Institute for Health and Clinical Excellence v4 Page 5 of 26

cancer, and for this indication treatment should be given for 5 years. It is also licensed for the treatment of advanced breast cancer in postmenopausal women. Letrozole is a reversible, non-steroidal aromatase inhibitor and has been shown to suppress oestradiol by 75 to 80%. It is licensed in the UK for various indications in the advanced breast cancer and neo-adjuvant setting. In the adjuvant setting it is licensed in the UK for the following circumstances. Adjuvant treatment of postmenopausal women with HR+ invasive early breast cancer (treatment should continue for 5 years or until tumour relapse occurs, whichever comes first). Treatment of early invasive breast cancer in postmenopausal women with HR+ invasive early breast cancer who have already received standard adjuvant tamoxifen therapy for 5 years. Treatment with letrozole should continue for 3 years or until tumour relapse occurs, whichever comes first. Exemestane is an irreversible steroidal aromatase inhibitor and has been shown to suppress oestradiol by more than 90%. It is licensed in the UK for the adjuvant treatment of postmenopausal women with oestrogen receptor positive invasive early breast cancer, following 2 3 years of initial adjuvant tamoxifen therapy. For this indication, treatment with exemestane should be given for the remainder of the 5-year adjuvant period. It is also licensed for the treatment of advanced breast cancer in women with natural or induced postmenopausal status whose disease has progressed following antioestrogen therapy. Because aromatase inhibitors reduce circulating oestrogen levels, a decrease in bone mineral density can be anticipated. A warning has therefore been issued in the summaries of product characteristics of all three aromatase inhibitors that women should have their bone mineral density formally assessed by bone densitometry, for instance DEXA scanning at the start of treatment and at regular intervals thereafter. Treatment or prophylaxis for osteoporosis should be initiated according to local clinical practice and carefully monitored. National Institute for Health and Clinical Excellence v4 Page 6 of 26

An additional requirement to check plasma lipid levels has also been included in the summaries of product characteristics for anastrozole. 3 The evidence This appraisal covers three slightly different indications, based on the marketing authorisations for the drugs under appraisal (table 3). A fourth indication, planned sequence (that is, planning to give 2 3 years aromatase inhibitor therapy following 2 3 years of tamoxifen therapy), is currently not included in any licence and will not be discussed in this overview. Anastrozole is not licensed in the unplanned switch or extended adjuvant settings, but the results from the Assessment Group s analyses for those unlicensed indications are included in this overview for information. Table 3 Definition of indications under appraisal Indication Primary adjuvant Unplanned Switching Extended Adjuvant Definition 5 years of aromatase inhibitor instead of tamoxifen Unplanned 2 3 years of aromatase inhibitor after 2 3 years of tamoxifen Aromatase inhibitor after standard therapy with tamoxifen Drugs licensed for the indication Anastrozole, Letrozole Exemestane Letrozole 3.1 Clinical effectiveness The Assessment Group identified seven prospective randomised controlled trials that met the inclusion criteria (figure 4 in assessment report, page 31). Baseline characteristics were well balanced between treatment arms within the included studies. There was some variation between studies in the distribution of prognostic factors, the methods of randomisation and concealment and blinding, intention to treat (ITT) analysis and follow up (table 4). In three studies (GABG, IES and MA-17) event-driven analyses led to trials being stopped early because of the benefit observed and the median followup of each was shorter than the term of treatment. National Institute for Health and Clinical Excellence v4 Page 7 of 26

Table 4 Characteristics of included studies Primary adjuvant Anastrozole Letrozole Study name ATAC BIG 1-98 Number of participants 9366 (6241 in relevant arms) 8010 Average age 64 61 Months of follow up 68 26 and 36 % node negative 61% 58% % HR+ 84% 99.8% Prior chemotherapy 21% 25% Blinding Adequate Unclear ITT analysis Available in addition to analysis in Minor exclusions HR+ women Unplanned switching Study name Anastrozole a Anastrozole a Exemestane GABG (combined analysis of ABCSG-8 and ARNO-95) b a ITA Number of participants 3224 448 4742 Average age 62 63 64 Months of follow up 28 36 31 % node negative 74% 0% 51% % HR+ 98% 91% (anastrazole) 86% (tamoxifen) 81% Prior chemotherapy 0% 67% 32% Blinding Unclear Open label Adequate ITT analysis Different randomisation in the sub-studies Minor exclusions IES Minor exclusions a Anastrozole in not licensed in the unplanned switch indication. b The GABG analysis is based on a planned combined analysis of two studies, ABCSG-8 and ARNO-95, that randomised at different times. Extended adjuvant Letrozole Anastrozole Study name MA 17 ABCSG-6a Number of 856 5187 participants 62 Not reported Months of follow up 28 Not reported % node negative 50% 66 and 69% % HR+ 98% 95% Prior chemotherapy 45% Not reported Blinding Probably adequate Unclear ITT analysis Minor exclusions Minor exclusions A meta-analysis of GABG and ITA has also been carried out; this is referred to as the Jonat meta-analysis (available as conference presentation only). The Assessment Group and consultees from the professional organisations raised National Institute for Health and Clinical Excellence v4 Page 8 of 26

concerns that there is insufficient information available on the methodology used to determine whether the Jonat meta-analysis is appropriate. 3.1.1 Clinical outcomes The primary outcome in all studies was disease-free survival. However, this was defined in slightly different ways across the studies. Secondary outcomes included overall survival, breast-cancer-related survival, breast cancer recurrence, loco-regional recurrence, contralateral breast cancer, and distant recurrence. Primary adjuvant Both anastrozole and letrozole significantly increased disease-free survival. This was also found for the HR+ subgroup with anastrozole (table 5). Of the secondary outcomes, overall survival or breast-cancer-related survival were not significantly increased by either drug. Breast cancer recurrence was statistically significantly reduced by both drugs. Distant recurrences were significantly reduced with letrozole and with anastrozole in the ITT population. Event rates were generally lower in the study with letrozole than with anastrozole probably because of the shorter follow up in the study. Unplanned switch Both exemestane and anastrozole (not licensed for this indication) significantly increased disease-free survival (table 6), but for anastrozole this was reported only in the Jonat meta-analysis. Of the secondary outcomes, overall survival was increased by both drugs, but statistically significantly so for anastrozole only in the Jonat meta-analysis. No hazard ratios for breastcancer-related survival were reported. Breast cancer recurrence was statistically significantly reduced by both drugs. Distant recurrences were statistically significantly reduced with anastrozole in one study but in the smaller study (ITA) with a worse prognosis the hazard ratio did not reach statistical significance. Event rates were generally higher in the small anastrozole study (ITA) because of the inclusion of patients with a worse prognosis. National Institute for Health and Clinical Excellence v4 Page 9 of 26

Table 5 Outcomes for the primary adjuvant indication Event rate (%) Aromatase inhibitor Control Hazard ratio (95% confidence interval [CI]) Absolute risk reduction (ARR) (%) Disease-free survival (figures refer to mortality) Anastrozole (ATAC) 18.6 21.0 0.87 (0.78 0.97) 2.4 Anastrozole (ATAC HR+) 16.2 19.1 0.83 (0.73 0.94) 2.9 Letrozole (BIG 1-98 ) 8.8 10.7 0.81 (0.70 0.93) 1.9 Overall survival (figures refer to mortality) Anastrozole (ATAC) 13.3 13.6 0.97 (0.85 1.12) 0.3 Anastrozole (ATAC HR+) 11.3 11.6 0.97 (0.83 1.14) 0.3 Letrozole (BIG 1-98 ) 4.1 4.8 0.86 (0.70 1.06) 0.6 Breast-cancer-related survival (figures refer to mortality) Anastrozole (ATAC) 7.6 8.6 0.88 (0.74 1.05) 1.0 Anastrozole (ATAC HR+) 5.8 6.6 0.87 (0.70 1.09) 0.8 Letrozole (BIG 1-98 ) 2.8 3.8 NR 1.1 Breast cancer recurrence Anastrozole (ATAC) 13.0 16.1 0.79 (0.70 0.90) 3.1 Anastrozole (ATAC HR+) 10.8 14.2 0.74 (0.64 0.87) 5.2 Letrozole (BIG 1-98 ) 5.7 7.7 0.72 (0.61 0.88) 2.0 Loco-regional recurrence Anastrozole (ATAC) 2.69 3.24 NR 0.6 Letrozole (BIG 1-98 ) 0.85 1.22 NR 0.4 Contralateral breast cancer Anastrozole (ATAC) 1.1 1.9 0.58 (0.38 0.88) 0.8 Anastrozole (ATAC HR+ve) 1.0 2.0 0.47 (0.29 0.75) 1.0 Letrozole (BIG 1-98 ) 0.4 0.7 NR 0.3 Distant recurrence Anastrozole (ATAC) 10.5 12.1 0.86 (0.74 0.99) 1.6 Anastrozole (ATAC HR+) 8.6 10.2 0.84 (0.70 1.00) 1.6 Letrozole (BIG 1-98 ) 4.4 5.8 0.73 (0.60 0.88) 1.4 NR not reported National Institute for Health and Clinical Excellence v4 Page 10 of 26

Table 6 Outcomes for the unplanned switch indication Event rate (%) Aromatase inhibitor Disease-free survival (figures refer to mortality) Control Hazard ratio (95% CI) Absolute risk reduction (ARR) Anastrozole (GABG) NR NR NR NR Anastrozole (ITA) NR NR NR NR Anastrozole (Jonat ) NR NR 0.59 (0.48 0.74) NR Exemestane (IES) 11.0 14.9 0.73 (0.62 0.86) 3.8 Overall survival (figures refer to mortality) Anastrozole (GABG) 2.8 3.7 NR 0.9 Anastrozole (ITA) 1.8 4.4 NR 2.7 Anastrozole (Jonat ) 3.3 4.5 0.71 (0.52 0.98) 1.2 Exemestane (IES) 6.4 7.9 0.83 (0.67 1.02) 1.5 Breast-cancer-related survival (figures refer to mortality) Anastrozole (GABG) 1.5 1.9 NR 0.4 Anastrozole (ITA) 1.8 3.1 NR 1.3 Exemestane (IES) 4.0 5.2 NR 1.2 Breast cancer recurrence Anastrozole (GABG) 4.2 7.0 0.59 (0.44 0.81) 2.7 Anastrozole (ITA) 5.8 15.1 NR 9.3 Anastrozole (Jonat ) 4.6 8.0 0.55 (NR) 3.4 Exemestane (IES) 8.7 12.2 0.70 (0.58 0.83) 3.5 Loco-regional recurrence Anastrozole (GABG) 1.1 1.5 NR 0.4 Anastrozole (ITA) 0.9 5.8 0.15 (0.03 0.65) 4.9 Anastrozole (Jonat ) 0.95 1.8 NR 0.8 Exemestane (IES) 1.8 2.4 NR 0.5 Contralateral breast cancer Anastrozole (GABG) 0.7 1.0 NR 0.3 Anastrozole (ITA) 0.4 0.9 NR 0.4 Anastrozole (Jonat ) 0.7 1.1 NR 0.6 Exemestane (IES) 0.5 1.1 0.50 (0.26 0.97) 0.6 Distant recurrence Anastrozole (GABG) 2.4 4.4 0.54 (0.37 0.80) 2.0 Anastrozole (ITA) 4.5 8.4 0.49 (0.22 1.05) 4.0 Anastrozole (Jonat ) 2.9 5.1 0.61 (NR) 2.2 Exemestane (IES) 6.4 8.8 NR 2.5 Extended adjuvant The effect of letrozole and anastrozole (not licensed for this indication) on disease-free survival in the extended adjuvant setting was not reported (table 7). Of the secondary outcomes, letrozole s effect on overall survival and breast-cancer-related survival was not statistically significant (no data for anastrozole). Letrozole was found to increase overall survival significantly National Institute for Health and Clinical Excellence v4 Page 11 of 26

when data only for women whose disease had been node-positive were analysed (Hazard ratio [HR] = 0.61, 95% confidence interval [CI] 0.38 to 0.98). Breast cancer recurrence was statistically significantly reduced by both drugs. The event rates for loco-regional recurrence and contralateral breast cancer were very low, and no statistically significant effect was reported. Letrozole reduced the risk of distant recurrences significantly Table 7 Outcomes for the extended adjuvant indication Event rate (%) Aromatase inhibitor Disease-free survival (figures refer to mortality) Control HR (95% CI) Absolute risk reduction (ARR) Letrozole (MA-17) NR NR NR NR Anastrozole (ABCSG 6a) NR NR NR NR Overall survival (figures refer to mortality) Letrozole (MA-17) 2.0 2.4 0.82 (0.57 1.19) 0.4 Breast-cancer-related survival (figures refer to mortality) Letrozole (MA-17) 0.6 0.9 NR 0.2 Anastrozole (ABCSG 6a) NR NR NR NR Breast cancer recurrence Letrozole (MA-17) 3.6 6.0 0.58 (0.45 0.76) 2.4 Anastrozole (ABCSG 6a) 7.8 11.9 0.64 (0.41 0.99) 4.2 Loco-regional recurrence Letrozole (MA-17) 0.7 1.3 NR 0.6 Anastrozole (ABCSG 6a) 2.6 3.2 NR 0.6 Contralateral breast cancer Letrozole (MA-17) 0.7 1.1 0.63 (0.18 2.21) 0.4 Anastrozole (ABCSG 6a) 1.6 2.1 NR 0.6 Distant recurrence Letrozole (MA-17) 2.2 3.6 0.60 (0.43 0.84) 1.4 Anastrozole (ABCSG 6a) 4.1 7.5 NR 3.3 3.1.2 Adverse events The aromatase inhibitors and tamoxifen have different adverse-effect profiles. Tamoxifen is associated with small but statistically significant increases in endometrial cancer, thromboembolic events and stroke. Some consultees noted in their comments on the assessment report that the hysterectomy rate was also significantly higher in patients treated with tamoxifen. National Institute for Health and Clinical Excellence v4 Page 12 of 26

The data on aromatase inhibitors show a trend towards increases in osteoporosis. The professional organisations expressed the view that this effect represents a combination of the loss from the protective effect of tamoxifen and an additional loss as a result of oestrogen deprivation. A statistically increased risk of fracture was seen for anastrozole in the primary adjuvant and unplanned switch settings, and with letrozole in the primary adjuvant setting. A summary of fracture rates reported in the trials is shown in table 8. Hip fractures were reported only in the primary adjuvant study with anastrozole and the extended adjuvant study with letrozole. In both studies, hip fracture risk was numerically, but not statistically significantly, higher in the aromatase arms, but the numbers of hip fractures in the trials was low. The Assessment Group points to data that indicate that the statistical significance of the increase in osteoporosis increases with follow-up time (assessment report pages ii and 84). Patient organisations point out in their comments on the assessment report that subsequent analyses of the ATAC study have demonstrated that the risk of fractures, although significantly higher for women on anastrozole compared with women on tamoxifen, did remain constant and no time-related increase in risk was observed. Table 8 Fracture rates Indication Drug Fracture rate (%) Aromatase Inhibitor Control Primary adjuvant Anastrozole (ATAC) 11* 7.7 Letrozole (BIG 1-98) 5.7* 4 Unplanned switch Anastrozole (GABG) 2.1* 1 Anastrozole (ITA) 0.9 0.9 Exemestane (IES) 3.1 2.3 Extended adjuvant Letrozole (MA.17) 5.3 4.6 * statistically significant difference to control arm Hypercholesterolaemia was increased by letrozole in the primary adjuvant study, but not in the extended adjuvant study. There may be a small increased risk of cardiac disease with aromatase inhibitors. None of the studies were designed to specifically evaluate cardiac morbidity; however, in several studies there was a numerical increase in the National Institute for Health and Clinical Excellence v4 Page 13 of 26

number of patients experiencing cardiac adverse events, including myocardial infarction and cardiac failure. Only the primary adjuvant study with letrozole showed a significant effect for grade 3 5 cardiac events and for cardiac failure, favouring tamoxifen. Tamoxifen, however, is known to decrease cholesterol levels and possibly the risk of cardiac events in women of this age. It is therefore difficult to assess any potential effect on cardiovascular events fully, because these were not consistently reported in the trials. The Assessment Group requested standardised information on cardiovascular toxicity from the manufacturers, but this was provided for only two trials. For reasons of consistency, the Assessment Group could not use this additional information. 3.1.3 Quality of life In their submissions to this appraisal, patient organisations expressed the view that the less severe menopausal symptoms observed with an aromatase inhibitor, such as hot flushes and vaginal bleeding, were very important to the quality of life in women undergoing cancer treatment. However, in the clinical studies there was no significant difference in overall health-related quality of life between tamoxifen treatment and the primary adjuvant anastrozole, the extended adjuvant letrozole and the unplanned switch strategies. 3.1.4 Summary of clinical evidence Based on individual trials, there is consistent evidence that aromatase inhibitors improve disease-free survival in the primary adjuvant and the unplanned switch settings, and breast cancer recurrence in all settings, with absolute risk reductions (ARRs) for disease-free survival of between 1.9 and 3.8%. There is evidence from some trials that aromatase inhibitors improve the risk of loco-regional or distant recurrence or contralateral breast cancer in all three settings, but this is not consistent between trials. However, none of the individual studies have established an advantage in terms of either allcause mortality or quality of life. National Institute for Health and Clinical Excellence v4 Page 14 of 26

3.2 Cost effectiveness The Assessment Group reviewed the literature and the submitted economic evidence, and generated its own economic model. 3.2.1 Economic literature One cost-effectiveness study of anastrozole from the US was identified, which did not include endometrial cancer as an outcome. The study therefore favoured tamoxifen, but included a 25% reduction in benefit for anastrozole because of the effect on hip fracture. This resulted in an incremental cost per QALY of $76,000 over a 20-year horizon, or $202,000 over an 8-year horizon. 3.2.2 Submitted economic evidence Economic models were submitted by AstraZeneca (anastrozole, primary adjuvant and unplanned switch [the latter indication is unlicensed]), Novartis (letrozole, primary adjuvant and extended adjuvant), and Pfizer (exemestane, unplanned switch). Each of the submitted models were Markov models with 1000 simulations, a mean patient age of 61 64 years, a 25-year or lifetime horizon and contained broadly similar health states and model structures. The differences in the models included the assumptions on the duration of benefit, the source of the utility data, and the inclusion of adverse events. The individual models were based on the data from the respective relevant clinical studies (table 4). The results from the manufacturers economic models are shown in table 9 (base case) and are discussed in sections 3.2.4 and 3.2.5 (sensitivity analyses). 3.2.3 The ScHARR Model The Assessment Group developed a probabilistic Markov model that used trial evidence for all three treatment strategies and drugs separately, and five submodels were developed: anastrozole (primary adjuvant and unplanned switch, the latter being an unlicensed indication), letrozole (primary adjuvant and extended adjuvant), and exemestane (unplanned switch). Resource use and utilities were taken from the respective trial data where available or from National Institute for Health and Clinical Excellence v4 Page 15 of 26

published literature. The model uses an annual cycle length and is run until patients reach 100 years of age. The comparator was 5-year tamoxifen therapy immediately following surgery, with a proportion of patients also receiving chemotherapy. Figure 1 illustrates the health states modelled, and the possible transitions, with disease-free survival being the starting health state. Figure 1 Structure of the Assessment Group model Loco-regional or contralateral Relapse Remission Disease-free Survival Metastatic disease Death from other causes Death from Breast cancer Details of the model assumptions are explained fully in the assessment report on page 120. Costs and utilities for the health states in the model are shown in the assessment report in tables 30, 31, 32 (pages 125, 127, 129) and table 35 (page 142). If patients in the model do not relapse for 15 years, they are assumed to have the same risk of progression as the general population. After patients relapse in the model, the type of treatment received, the cost of treatment and survival are the same in both arms. Also, their survival is assumed to be independent of the time of relapse. This assumption means that survival for patients may be slightly overestimated. An important feature of the model is the assumption for the duration of benefit after the 5-year treatment period: In the base case analysis benefits of National Institute for Health and Clinical Excellence v4 Page 16 of 26

aromatase inhibitors in terms of relative risk gradually decline over the next 10-year period, to the extent that by year 15 the number of patients in disease-free survival is the same in both arms. This is considered to be a conservative assumption. An alternative assumption was tested in a sensitivity analysis (called benefits maintained ). In the benefits maintained scenario the rate of recurrence is the same for both arms after the treatment period, based on a long-term follow up study on tamoxifen, thereby preserving the benefits of aromatase inhibitors. In their comments on the assessment report, many consultees criticised the base case assumptions as inappropriate, because these imply that, during the 10 years following treatment, patients who had received aromatase inhibitors perform worse than those who had received tamoxifen. However, consultees from the professional organisations stated that there is insufficient information available to predict whether benefit will be maintained or not and stated that any interpretation of post-treatment data should be very cautious and should take care to avoid undue extrapolation. Modelling of adverse events Generally, only adverse events that were statistically different between the treatment arms were included in the model, apart from fracture rates for aromatase inhibitors. Because of the potential long-term risks of hip and other fractures in postmenopausal women it was considered prudent to take this risk into account. Tamoxifen is associated with higher rates of vaginal bleeding, endometrial cancer, venous thromboembolic events and ischaemic cerebrovascular disorders, and this has been included in the model. Hypercholesterolaemia was excluded as an adverse event because of the current uncertainty in the evidence base. Cardiovascular events were not modelled. The costs of treating adverse events were added to the cost of the health state. Some adverse events are assumed to cause a reduction in quality of life (such as vaginal bleeding) and others are also associated with a risk of mortality (such as hip fractures and endometrial cancer). National Institute for Health and Clinical Excellence v4 Page 17 of 26

The impact of fracture on the cost effectiveness was modelled separately using the Assessment Group s model developed for the appraisal on bisphosphonates, raloxifene, strontium ranelate and teriparatide for the prevention of osteoporotic fractures. The base case includes a differential effect on fracture between treatment arms extending 5 years beyond the treatment period, with the relative risk for fracture being constant during the time of endocrine therapy, and declining gradually to control level over the next 5 years (details of the fracture modelling in assessment report on pages 139 140). Costs and utilities used for adverse events are shown in the assessment report in tables 33 and 36, pages 135 and 143). 3.2.4 Base case results from the Assessment Group s and the manufacturers models (table 9) Generally, treatment with aromatase inhibitors resulted in increased drug costs and slightly decreased follow-up costs (for example, recurrence) compared with tamoxifen. Adverse events had a very minor contribution to the costs (assessment report figures 6 10, pages 146 150). Primary adjuvant: in the Assessment Group s model 5-year treatment with anastrozole or letrozole resulted in an incremental cost per QALY of 21,592 and 22,761 respectively when compared with 5-year tamoxifen therapy. The corresponding results from the submission models were lower: 7610 for anastrozole and 10,286 for letrozole. Unplanned switch: in the Assessment Group s model 2 3-year treatment with anastrozole (not licensed for this indication) or exemestane, after 2 3- year tamoxifen therapy, resulted in an incremental cost per QALY of 13,766 and 15,861 respectively when compared with 5-year tamoxifen therapy. The corresponding results from the submission models were lower: 2446 for anastrozole and ***** for exemestane. Because of the shorter follow up in the letrozole study underpinning this modelling, the results for letrozole could be associated with a higher uncertainty than the results for anastrozole. National Institute for Health and Clinical Excellence v4 Page 18 of 26

Extended adjuvant: in the Assessment Groups model 5-year treatment with letrozole after 5-year tamoxifen therapy resulted in an incremental cost per QALY of 9248 when compared with 5-year tamoxifen therapy followed by placebo. The corresponding result from the submission model was 7725. Table 9 Incremental cost per QALY vs tamoxifen Anastrozole (assessment report submission) Primary adjuvant (5 years instead of tamoxifen) 21,592 7610 Unplanned switching (2 3 years after 2 3 years of tamoxifen) ( 13,766 a ) 2446 Extended adjuvant (after 5 years of tamoxifen) a Letrozole (assessment report submission) Exemestane (assessment report submission) a not licensed 22,761 10,286 a 15,861 ***** a 9248 7725 a 3.2.5 Sensitivity analyses The Assessment Group carried out a number of deterministic sensitivity analyses, which are shown in the figures below. The model results were sensitive to the assumptions on the duration of benefit, with incremental cost effectiveness ratios (ICERs) more than halved when the benefits maintained assumption was used. National Institute for Health and Clinical Excellence v4 Page 19 of 26

Sensitivity Analysis Primary Adjuvant Anastrazole Letrozole cost per QALY 40,000 30,000 20,000 10,000 0 Basecase Benefits maintained Discount rate : 3.5% both 10 year time horizon Cost of recurrence doubled Sensitivity analysis Sensitivity analysis - Unplanned Switch Anastrazole Exemestane cost per QALY 20,000 15,000 10,000 5,000 0 Basecase Benefits maintained Discount rate : 3.5% costs & benefits 10 year analysis Cost of recurrence doubled Sensitivity analysis cost per QALY 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0 Basecase Sensitivity analysis - Extended Adjuvant Letrozole Benefits maintained Discount rate : 3.5% costs & benefits Sensitivity Analysis 10 year analysis Cost of recurrence doubled National Institute for Health and Clinical Excellence v4 Page 20 of 26

The Assessment Group provided a further sensitivity analysis associated with the assumptions on fracture effects (described in the assessment report, page 155). When fracture prediction was based on bone mineral density loss found in the ATAC study (7%), which indicated a doubling of the fracture risk, the ICER increased from 20,750 (base case) to 26,042. When fracture risk was assumed to be constant over 10 years the ICER was 28,250. This latter ICER would increase to 31,075 if patients who were identified as having low bone mineral density were treated with bisphosphonates. However, when the benefits maintained assumption on the duration of the breast cancer benefit was used, this latter figure would be 13,656. 1 The manufacturer s submission for anastrozole in the primary adjuvant setting provided sensitivity analyses on the effect of the duration of benefit. When the benefit was assumed to extend for a further 5 years after the 5-year treatment period (base case) the ICER was 7610 in the HR+ trial population (table 9). When the benefit was assumed to extend for only 1 year after the 5-year treatment period the ICER increased to 12,463 per QALY. When the benefit was assumed to extend over the entire time horizon of the model (25 years) the ICER decreased to 5,331. When the time horizon was reduced to 10 years, the ICER increased to *******. In the unplanned switch setting, the manufacturer s model showed an ICER of ********when the benefit was assumed to extend for only 1 year after the 5- year treatment period. For a 10-year time horizon, the ICERs were ********and ********for scenarios with benefits extending 5 years beyond the treatment period and 1 year beyond the treatment period respectively. The manufacturer s model for letrozole showed that age at the start of treatment has a major impact, with the ICER being 23,700 for women aged 71, compared with 10,286 for women aged 61 (primary adjuvant setting). The higher ICER for the older patient group was also confirmed in the manufacturer s model for the extended adjuvant setting and in the Assessment Group s sensitivity analysis. 1 Calculation by technical team at NICE based on data in assessment report. National Institute for Health and Clinical Excellence v4 Page 21 of 26

The manufacturer s model for letrozole in the extended adjuvant treatment showed that the ICERs for node-negative and node-positive patients were 11,784 and 5373 respectively. The results from the probabilistic sensitivity analyses carried out by the Assessment Group, using the conservative assumption on the duration of benefit, are summarised in Table 10 (full CEACs in assessment report, pages 158 162). Table 10 Probability of being cost effective (estimated from figures 13, 14, 17, 18 and 20 in the assessment report) Probability of cost effectiveness (%) Willingness to pay 20,000 per QALY Willingness to pay 30,000 per QALY Primary adjuvant Anastrozole 35 62 Letrozole 36 62 Unplanned switch Anastrozole a 65 80 Exemestane 55 85 Extended adjuvant Letrozole more than 95 more than 95 a not licensed for this indication 3.2.6 Subgroups Many consultees, either in their submissions or in their comments on the assessment report raised the issue of subgroups. Analyses of higher- and lower-risk populations in the trials did not identify any subgroups that show a major and convincing deviation from the overall hazard ratio in each study, apart from the ATAC study, in which women who were positive for oestrogen receptors but not for progesterone receptors appeared to derive a greater benefit from anastrozole than women who were positive for both receptors. However, this observation has not been convincingly replicated in other studies. Consultees from the professional organisations suggested that it is highly appropriate to consider the use of conventional prognostic variables (node status, grade and tumour size) to stratify patients into groups who will derive greater than average or less than average additional advantage from National Institute for Health and Clinical Excellence v4 Page 22 of 26

treatment with an aromatase inhibitor compared to tamoxifen. These consultees raised the point that it should be possible to identify groups for whom an aromatase inhibitor provides a more cost effective intervention than treating all women who may benefit. Some professional organisations took the view that for women with breast cancer at low risk of recurrence 5 years (for example, T1, Grade I or II, node negative) tamoxifen remains an appropriate treatment. 4 Issues for consideration Should aromatase inhibitors be viewed as a class even though exemestane is a steroidal irreversible inhibitor, while letrozole and anastrozole are reversible inhibitors? Will the observed short-term improvement in disease-free survival be maintained and translate into consistent overall survival benefits? Which assumption on the duration of benefit should be used in the judgement of cost effectiveness? Is it necessary to consider subgroups with different prognostic factors? If so, what are the most appropriate risk factors and biological characteristics to select high-, intermediate- and low-risk groups to enable the best use of adjuvant aromatase inhibitors? Are there any issues over the potential long-term adverse event rates with aromatase inhibitors? If tamoxifen and aromatase inhibitors are given in sequence, what is the optimal sequence? Some consultees have suggested that aromatase inhibitors should be given first, because the risk of recurrence in an HR+ women is greatest in the 2nd and 3rd year following initial diagnosis. Is there sufficient evidence to consider the optimal length of treatment with an aromatase inhibitor, as noted in the scope for this appraisal? If so, what is the optimal length of treatment? National Institute for Health and Clinical Excellence v4 Page 23 of 26

5 Authors Elisabeth George, Helen Chung, Louise Longworth NICE Appraisal Team April 2006 National Institute for Health and Clinical Excellence v4 Page 24 of 26

Appendix A: Sources of evidence considered in the preparation of the overview A The assessment report: Danny Hind, Sue Ward, Enrico De Nigris, et al. (ScHARR, University of Sheffield) Hormonal therapies for early breast cancer: systematic review and economic evaluation, January 2006 B Submissions from the following organisations: I Manufacturers/sponsors: AstraZeneca Novartis Pfizer II Professional/specialist and patient/carer groups: Joint submission by Breakthrough Breast Cancer, Breast Cancer Care and Cancer BACUP Joint submission by the National Cancer Research Institute, Breast Cancer Study Group and the British Breast Group Joint submission by the Association of Cancer Physicians, the Royal College of Radiologists, the Joint Collegiate Council on Clinical Oncology and the Royal College of Physicians C Comments on the assessment report from the following organisations: I Manufacturers/sponsors: AstraZeneca Novartis Pfizer II Professional/specialist and patient/carer groups: Guideline Development Group NHS Quality Improvement Scotland Breakthrough Breast Cancer National Institute for Health and Clinical Excellence v4 Page 25 of 26

British Oncology Pharmacists Association & Cancer Network Pharmacists Forum Royal College of Nursing Royal College of Pathologists Royal College of Radiologists, the Joint Collegiate Council on Clinical Oncology, the Royal College of Physicians, the National Cancer Research Institute Breast Cancer Group and British Breast Group Society for Endocrinology National Institute for Health and Clinical Excellence v4 Page 26 of 26