A Stochastic Economic Evaluation of Letrozole versus Tamoxifen as a ...

ORIGINAL RESEARCH ARTICLE

Pharmacoeconomics 2003; 21 (7): 513-525 1170-7690/03/0007-0513/$30.00/0 © Adis Data Information BV 2003. All rights reserved.

A Stochastic Economic Evaluation of Letrozole versus Tamoxifen as a First-Line Hormonal Therapy For Advanced Breast Cancer in Postmenopausal Patients Jon Karnon1 and Trefor Jones2 1 2

School of Health and Related Research, University of Sheffield, Sheffield, England Novartis Pharmaceuticals UK, Camberley, England

Abstract

Background: Letrozole is a third-generation aromatase inhibitor that is a feasible alternative to tamoxifen as a first-line hormonal therapy for patients with advanced breast cancer. Objective: This paper presents the results of an economic evaluation comparing letrozole and tamoxifen as first-line hormonal therapies in postmenopausal women diagnosed with advanced breast cancer. Perspective: UK National Health Service. Design: A decision model (Markov process) was built describing possible patient pathways from the point of diagnosis to death. The model was populated using patient-specific clinical trial data, data from the existing literature, and expert opinion. Stochastic analyses of the model were undertaken, whereby the majority of the input parameters were described as probability distributions to represent the uncertainty about their true value. Costs were presented in year 2000 values. Results: The baseline results showed that letrozole is a cost-effective alternative to tamoxifen with a mean incremental cost per life-year gained of £2342, whilst the incremental cost increases to just over £10 000 at the 95th percentile of the cost-effectiveness range (2000 values). Conclusions: The results of the economic analysis indicate that letrozole is a cost-effective alternative first-line therapy compared with tamoxifen for postmenopausal women with advanced breast cancer, achieving additional life-years with a modest increase in costs.

Breast cancer is the most common cancer in women in the UK, with around 33 000 newly diagnosed cases and 15 000 deaths from the disease each year.[1] Most patients present with cancer limited to the breast and nodes, for which surgery followed by systemic therapies are potentially curative, although around 50% of patients diagnosed with early breast cancer will eventually progress to an advanced form

of disease.[2] Advanced breast cancer is not curable. Treatment for advanced breast cancer involves the administration of sequential systemic therapies, with a different therapy usually being administered following disease progression. The treatment aim for advanced breast cancer is to slow down or stop tumour growth for some period of time, whilst retaining an adequate quality of life for the patient.

514

Karnon & Jones

Table I. Summary of the results of the randomised controlled trial[9] of letrozole versus tamoxifen that provided the core data for the economic evaluation Description of study: Double-blind, double-dummy, randomised, multicentre, 2-arm, phase III trial comparing letrozole 2.5 mg/day versus tamoxifen 20 mg/day as first-line therapy in postmenopausal women with advanced breast cancer Study population: A total of 907 patients, randomised in 201 investigational sites in 29 countries, were available for analysis on an intent-to-treat basis Methodology: The study was randomised and double-blind, with a parallel arm design for the core phase of the study. On progression of disease or any other reason leading to discontinuation of core treatment, patients could be switched to the alternative treatment, still under double-blind conditions, provided that they remained suitable for hormonal anticancer treatment Indication and main criteria for inclusion: Postmenopausal patients with histological or cytological evidence of breast cancer presenting with locally advanced or loco-regionally recurrent disease not amenable to treatment by surgery or by radiotherapy, or with metastatic disease, were eligible for study. Patients had not been previously treated with hormonal anticancer agents for their advanced disease. Patients had to be estrogen-receptor and/or progesterone-receptor positive or with the status of both receptors unknown Duration of treatment: The randomised treatment was administered until diagnosis of progression of disease, or until other reasons (e.g. adverse events) led to discontinuation. If the patient remained suitable for hormonal anticancer therapy, treatment could be switched to the alternative, still under double-blind conditions. The median duration of the follow-up was 18 months Results – efficacy: The letrozole and tamoxifen monotherapy treatment arms were well balanced with respect to baseline demographic characteristics and extent of disease. Letrozole was superior to tamoxifen in prolonging first-line time to progression (TTP; median 9.4 vs 6.0 months, p = 0.0001) and time to treatment failure (TTF; median 9.1 vs 5.7 months, p = 0.0001), and in the rates of overall objective response and clinical benefit. For the crossover data, both TTP (median 4 vs 3.4 months) and TTF (median 3.6 vs 3.3 months) were longer in patients switching from tamoxifen to letrozole than vice versa. Over 30% of patients in both treatment arms were still alive at the end of the follow-up period so the estimation of mean survival times was difficult Results – safety: Adverse events irrespective of relationship to study treatment were reported for 90% of patients in the letrozole arm and 87% of patients in the tamoxifen arm. Serious adverse events (SAEs) considered related to study treatment were reported for 2% of patients treated with letrozole and 3% treated with tamoxifen. The most frequent treatment-related SAE was thromboembolic events reported for 1% (3 patients) on letrozole and 2% (7 patients) on tamoxifen

Though the median survival from diagnosis with advanced breast cancer is around 2 years, there is wide variation between patients.[3] Patients with hormone-sensitive cancers (estrogen-receptor [ER]/ progesterone-receptor [PgR] positive) have the best prognosis with 50–60% of ER-positive tumours responding to primary hormonal therapy with a mean response duration of between 1–2 years.[4] There are three main forms of hormonal treatment that are administered to postmenopausal women with advanced breast cancer – antiestrogens, progestogens, and aromatase inhibitors.[5] At this time, tamoxifen remains the first-line hormonal therapy for breast cancer of all stages,[6] due to its perceived equivalent efficacy and lower toxicity relative to alternative hormonal therapies.[3] Previously, progestogens were employed as the main second-line hormonal therapies, but newer more selective aromatase inhibitors have shown improved survival in this setting and are replacing progestogens as second-line therapies.[7] Indeed, the new aromatase inhibitors represent such a significant advantage over the earlier agents in terms of both © Adis Data Information BV 2003. All rights reserved.

efficacy and tolerability that they are now being considered as a feasible alternative to tamoxifen as a first-line therapy.[8] Letrozole is a third-generation aromatase inhibitor that has been compared with tamoxifen as a first-line therapy for advanced breast cancer in a randomised controlled trial (RCT).[9] Table I provides a summary of the main design elements and results of this RCT, which demonstrated that letrozole showed superior objective response rates and time to progression (TTP), whilst being at least as well tolerated as tamoxifen. It is now a requirement for the cost effectiveness of new healthcare interventions to be proven in some countries, including the UK, Canada and Australia. This is the result of an increasing recognition that decisions relating to the allocation of resources in fixed-budget healthcare systems should be informed by the associated treatment costs as well as the relative effectiveness of the alternative interventions. In a high prevalence treatment area such as breast cancer, aggregate treatment costs will rise Pharmacoeconomics 2003; 21 (7)

Letrozole vs Tamoxifen for First-Line Therapy

quickly even if a treatment at the individual level is perceived to be inexpensive. This paper reports the results of a cost-effectiveness analysis comparing alternative first-line hormonal therapies – letrozole (2.5mg daily) and tamoxifen (20mg daily) – in postmenopausal women with advanced breast cancer that is ER and/or PgR positive or of unknown receptor status. To evaluate the full impact of the alternative treatment strategies, the evaluation compared treatment costs with survival for patients receiving the alternative hormonal therapies to estimate the additional cost per life-year gained. The incremental cost per QALY is also estimated in the sensitivity analyses. Methods Data from a recent RCT comparing letrozole and tamoxifen were employed in this evaluation,[9] but the period of follow-up within the trial was insufficient to describe the lifetime patient pathways, so the trial data were supplemented with data from the literature and expert panels. These data from disparate sources were synthesised within a Markov process that described relevant events along patients’ pathways from diagnosis to death, and this was analysed using DATA decision modelling software (Treeage Software Inc., Williamstown, USA). The perspective for the evaluation was that of the UK National Health Service and the time horizon covered the full lifetime of the patients. Discount rates of 6% for resources and 1.5% for life-years were applied to all analyses, in accordance with UK Treasury guidelines.[10] All costs are presented for the year 2000. Although the primary outcome measure used in the evaluation was life-years gained, a simple analysis of the QALY differential between the two therapies was undertaken employing the reasonable assumption that the majority of the extra life-years gained are gained within the first-line hormonal therapy state. The range of utility values for patients with advanced breast cancer, without progression, has been presented as 0.59–0.8,[11] so these values were applied to the survival advantage to estimate the incremental cost per additional QALY gained. © Adis Data Information BV 2003. All rights reserved.

515

The remaining methods are described in more detail in the following three sections, which cover the structure of the decision model, the data used to populate the model, and the methods used to undertake the stochastic analysis of the model. The Model Structure

The model structure is presented in figure 1, portraying the possible pathways from the point of diagnosis with advanced breast cancer to death. The structure of the model was extended from a decision model developed to evaluate second-line hormonal therapies for advanced breast cancer.[12] In the present model, patients receive their assigned first-line hormonal therapy and continue to do so until they experience disease progression, at which point patients either die or move to an alternative treatment state. Within the first cycle of the model, patients are subject to a probability of experiencing a serious adverse event (SAE), which requires additional resources for treatment of the effects but is not explicitly linked to the patient’s prognosis. The effect of SAEs on the progression of individual patients – increasing the probability of ending first-line hormonal therapy – is included in the RCT data describing the timing and destination of patients ending first-line therapy. These data are applied to the aggregate patient cohort for each therapy, i.e. patients progressing and patients dropping out of first-line therapy without progression (7–8% of patients for both therapies) are pooled, as this strategy does not affect the second-order analysis of the model. The data sources used to inform the subsequent transition probabilities between the other states within the model account for the limited dropout rate and so there was no need to model the prognostic consequences of dropouts separately. From the first-line hormonal therapy state patients move either to an alternative hormonal therapy (maximum of three lines), first-line chemotherapy, or to an end-stage phase of the disease, where palliative pain medication, corticosteroids and radiotherapy are offered. From the chemotherapy states, patients can receive an alternative chemotherapeutic regimen (up to two lines) or move to endPharmacoeconomics 2003; 21 (7)

516

Karnon & Jones

No progression

First-line maintenance

First-line maintenance Chemotherapy 1

Start: first-line

Change of

hormone Progression

Second-line hormone

treatment

Chemotherapy 1 Second-line hormone

End-stage palliative End-stage palliative Death

No progression

Death Second-line maintenance

Second-line maintenance Chemotherapy 1

Second-line Change of

hormone Progression

treatment

Third-line hormone End-stage palliative

Death No progression

Progression

treatment

End-stage palliative

Death No progression

Progression

Change of

Observational

treatment

care (chemotherapy) End-stage palliative

Death No progression Letrozole

Observational

Chemotherapy 2 Progression

Change of

care (chemotherapy)

treatment


Death First-line maintenance Second-line maintenance Third-line maintenance Chemotherapy 1 maintenance Advanced

Chemotherapy 2 maintenance

breast cancer

Chemotherapy 1 maintenance Chemotherapy 2 Observational care (chemotherapy) End-stage palliative

Chemotherapy 2 maintenance Observational care (chemotherapy) End-stage palliative Death

as for first-line hormone as for second-line hormone as for third-line hormone as for chemotherapy 1 as for chemotherapy 2 Observational care (chemotherapy) Chemotherapy 2

Observational care (chemotherapy)


Death

Chemotherapy 2 maintenance

M

Chemotherapy 1

Death

Chemotherapy 1 maintenance Chemotherapy 2

Chemotherapy 1


Third-line maintenance

Change of

hormone

Third-line hormone

Death

Third-line maintenance Chemotherapy 1

Third-line

Chemotherapy 1

Third-line hormone

Observational care (chemotherapy) Chemotherapy 2 Third-line hormone

End-stage palliative End-stage palliative End-stage palliative

Death

Death

Death Tamoxifen

M as for letrozole

Fig. 1. Advanced breast cancer Markov model structure. M = Markov cycle.

© Adis Data Information BV 2003. All rights reserved.

Pharmacoeconomics 2003; 21 (7)


stage palliation. Patients without disease progression after 6 months of chemotherapy are observed without systemic therapy (in the health state ‘observational care’) until progression occurs, following which patients either receive more chemotherapy or third-line hormonal therapy or move to end-stage palliative care. Patients entering the end-stage palliative care state may only progress to death. Transitions between the states within the model occur at the end of each 3-month cycle, although a half-cycle correction factor is inserted to correct for the movement of the average patient midway through a cycle. The Markov process allows for the use of time-dependent data relating to the entry point to the model so the transition probabilities from the first-line hormonal therapy state describe the survival curve defined by the accompanying RCT. The remaining hormonal and chemotherapy treatment states each consist of two separate states describing the ‘first 3 months in state’ and ‘3 months onwards in state’ to enable the specification of different probabilities of progression, death and subsequent therapy in these qualitatively different time periods. The Input Data

For each state described in the model, data describing the proportion of patients progressing, the TTP, the subsequent therapy, the proportion of patients dying, and the time to death were required. In addition, data describing the resource utilisation associated with each state and the unit costs of the component resources were required. To identify these data a range of sources were employed, which are described in the following sections. First-Line Hormonal Therapy

SAEs, the proportion of patients experiencing, and the timing of, progression or death, and the choice of secondary treatment were informed using patient-specific data derived from the recent RCT of letrozole versus tamoxifen described in table I.[9] The data were analysed on an intent-to-treat approach. SAEs were defined as any event resulting in hospitalisation, a prolongation of hospitalisation, death, lasting disability or the appearance of a sec© Adis Data Information BV 2003. All rights reserved.

517

ond unrelated cancer. Of the 907 evaluable patients in the study, 11 letrozole and 15 tamoxifen recipients experienced SAEs. The majority of SAEs were costed using the relevant health-related group (HRG) and the corresponding costs from the UK NHS national schedule for reference costs,[13] or drug costs from the British National Formulary.[14] In the UK, HRGs are a nationally recognised method of classifying case-mix by categorising patients into a manageable number of groups that are clinically similar. Although HRGs have been found to differ in the extent to which the conditions included in each group consume similar amounts of resource,[15] they provide a useful costing source. Concomitant drug interventions had been recorded in the RCT, which provided information on drug resource use, but no other resource utilisation data were collected. The number and type of patient consultations and laboratory tests and procedures were assumed to be similar to patients receiving second-line hormonal therapy, which were based on a series of expert interviews described in a previous study.[16] Current UK costs were applied to the estimated resources, which were mainly costed using NHS reference costs,[13] and the annual study of UK costs of health and social care.[17] The TTP data were the most important clinical inputs for the advanced breast cancer model. Mean transition probabilities for the respective first-line hormonal therapies for each 3-month period were estimated by dividing the number of patients ending treatment during the period by the total number of patients receiving the therapies at the start of the period (minus half the number of censored patients [lost to follow-up without experiencing an outcome event] during each period). Beta distributions representing the probability of progressing in each 3-month period from the time of diagnosis were generated using the available patient-specific data from the RCT (see specification of probability distributions in section Methods for the Stochastic Analysis of the Model). From the first-line therapy state patients could die directly, or progress to a different therapy state. Table II presents the mean proportions of patients progressing to the alternative Pharmacoeconomics 2003; 21 (7)

518


Table II. Proportion of patients moving to different states following first-line hormonal therapya Months

Letrozole (453 patients)

Tamoxifen (454 patients)

proportion of patients ending firstline letrozole

1st-line CT

2nd-line HT

palliation

0–3

0.18

0.08

0.35

0.34

0.23

0.25

0.12

0.36

0.31

0.22

4–6

0.18

0.08

0.17

0.67

0.08

0.31

0.06

0.26

0.57

0.11

7–9

0.13

0.04

0.17

0.77

0.02

0.22

0.04

0.18

0.71

0.07

10–12

0.21

0.00

0.14

0.81

0.05

0.23

0.00

0.11

0.84

0.05

13–15

0.14

0.04

0.12

0.84

0.00

0.17

0.00

0.19

0.71

0.10

16–18

0.21

0.00

0.14

0.71

0.14

0.15

0.00

0.29

0.57

0.14

19–21

0.15

0.11

0.33

0.44

0.11

0.15

0.00

0.11

0.78

0.11

22–24

0.15

0.40

0.20

0.40

0.00

0.19

0.00

0.14

0.86

0.00

25–27

0.12

0.00

0.25

0.75

0.00

0.22

0.00

0.00

0.67

0.33

28–30

0.12

0.00

0.00

1.00

0.00

0.15

0.00

0.00

1.00

0.00

31–33

0.12

0.00

0.00

1.00

0.00

0.14

0.00

0.00

1.00

0.00

34–36

0.07

0.00

0.00

1.00

0.00

0.13

0.00

0.00

1.00

0.00

37–39

0.19

0.00

0.00

1.00

0.00

0.09

0.00

0.00

1.00

0.00

40–42

0.13

0.00

0.00

1.00

0.00

0.12

0.00

0.00

1.00

0.00

43–45

0.08

0.00

0.00

1.00

0.00

0.09

0.00

0.00

1.00

0.00

46–48

0.22

0.00

0.00

1.00

0.00

0.33

0.00

0.00

1.00

0.00

Source: Letrozole trial P025.[9]

CT = chemotherapy; HT = hormonal therapy.

Karnon & Jones


death

proportion of transition probabilities for patients ending first-line tamoxifen patients moving to the alternative subsequent states ending firstline tamoxifen death 1st-line CT 2nd-line HT palliation

a

transition probabilities for patients ending first-line letrozole moving to the alternative subsequent states


destination states from first-line hormonal therapy. The number of patients comprising each proportion is also provided so the reader can recreate the relevant beta distributions. Second-Line Hormonal Therapy

The RCT[9] used to provide information for the first-line transition probabilities was designed as a crossover trial, with over half of the original patients crossing over to receive the alternative hormonal therapy as a second-line intervention for advanced breast cancer. The data from this trial provided patient-level data describing progression in the first 3 months and beyond, as well as the subsequent therapies administered to these patients. Transition probabilities were estimated directly for the probability of progressing within the first 3 months of second-line therapy. The mean TTPs were estimated from the start of the fourth month on second-line therapy and converted to a transition probability to represent the probability of progressing in subsequent cycles. Less than 10% of patients were censored so the ‘areas under the curve’ provided reasonable estimates of the mean TTP for the alternative therapies. The trial data were subjected to bootstrap analysis (1000 bootstrap replications) to derive a range for each of these four transition probabilities. The mean values and the ranges provided information for triangular distributions to be used in the stochastic analysis. Resource use covering drugs (other than letrozole/tamoxifen), patient consultations, laboratory tests, and hospitalisation during the period of second-line hormonal therapy were all obtained from previously published economic evaluations of second-line letrozole.[12,16] Current UK costs[13,17] were applied to the estimated resources. Third-Line Hormonal Therapy, First- and Second-Line Chemotherapy, Observational Care and End-Stage Palliative Care

From the point of third-line hormonal therapy, the structure of the model used in this study was virtually identical to that specified in the previously published evaluations of letrozole as a second-line hormonal therapy.[12,16] The only difference being © Adis Data Information BV 2003. All rights reserved.

519

that the current model did not include an observational care state leading from third-line hormonal therapy. The data populating this latter section of the model, including information on resource utilisation, were taken directly from the previously published studies (applying current UK costs[13,17] to the resource components), which had used a modified Delphi technique to determine values for those probabilities, therapeutic choices and resource utilisation that were relevant to the UK that were not identified in the existing literature.[12] The Delphi technique consisted of two rounds of individual interviews with six physicians. In the first interview each respondent was asked to estimate values for the missing data, in the second interview each expert was provided with the average estimates of all experts and asked to review their initial estimates. The mean values of all second round estimates provided information for the missing values for the model. The therapeutic choices derived from this process included the following: the progestogen megestrol was assumed to be the most likely third-line hormonal therapy; three first-line chemotherapy regimens were identified for inclusion in the analysis; doxorubicin (adriamycin) was considered to be most often employed as second-line chemotherapy.[12] The Delphi panel also estimated that patients entering the end-stage state (palliative care) remain in that state for an average of 1.5 months before progressing to death. The transition probabilities employed in the model beyond the states describing first-line hormonal therapies are presented in table III with their respective sources. The assumed resource use associated with each state and their estimated costs are presented in table IV. Methods for the Stochastic Analysis of the Model

A probabilistic sensitivity analysis was undertaken, whereby the value of input parameters within the model were described as probability distributions.[18,19] The probabilities of disease progression and death at the end of each 3-monthly cycle were represented by probability distributions. DistribuPharmacoeconomics 2003; 21 (7)

520


Table III. Transition probabilities for states beyond first-line hormonal therapy Transition probability

Mean

Transition probability

Mean

Transition probability

Mean

Progressing in first 3 months of 2nd-line tamoxifena

0.28

Progressing in first 3 months of 3rd-line HTb

0.45

Progressing in first 3 months of 2nd-line CTb

0.39

Moving to death from 2nd-line HT

0.06

Moving to death from 3rd-line HT

0.27

Moving to death from 2nd-line CT

0.52

Moving to 1st-line CT from 2nd-line HT

0.54

Moving to 1st-line CT from 3rd-line HT

0.87

Moving to observational care from 2nd-line CT

0.54

Moving to 3rd-line HT from 2nd-line HT

0.40

Moving to end-stage palliative from 3rdline HT

0.13

Moving to end-stage palliative from 2nd-line CT

0.46

Moving to end-stage palliative from 2nd-line HT

0.06

Progressing in all subsequent cycles

0.25


0.35


0.45

Moving to death from 3rd-line HT

0.49

Moving to death from 2nd-line CT

0.58


0.03

Moving to 1st-line CT from 3rd-line HT

0.87

Moving to observational care from 2nd-line CT

0.50


0.25

Moving to end-stage palliative from 3rdline HT

0.13

Moving to end-stage palliative from 2nd-line CT

0.50


0.70

Progressing in first 3 months of 1st-line CTb

0.17

Remaining in observational carec

0.68


0.06

Moving to death from 1st-line CT

0.79

Moving to 2nd-line CT from observational care

0.21

Progressing in first 3 months of 2nd-line letrozolea

0.36

Moving to 2nd-line CT from 1st-line CT

0.68

Moving to 3rd-line HT from observational care

0.03


0.06

Moving to observational care from 1st-line CT

0.17

Moving to end-stage palliative from observational care

0.08


0.54

Moving to end-stage palliative from 1stline CT

0.16 0.26

0.40



0.06

Moving to death from 1st-line CT

0.51


0.28

Moving to 2nd-line CT from 1st-line CT

0.64


0.03

Moving to observational care from 1st-line CT

0.20


0.25

Moving to end-stage palliative from 1stline CT

0.17


0.70


0.06

a

Data source: Letrozole trial P025.[9] (Second-line letrozole refers to second-line letrozole after first-line tamoxifen).

b

Data source: literature (as described in text).

c

Data source: expert opinion (as described in text).

CT = chemotherapy; HT = hormonal therapy.

Karnon & Jones




tions were also defined for the transition probabilities for each of the subsequent therapies following disease progression during the administration of the first-line hormonal therapies. Theoretical considerations were applied to the choice of probability distribution for the different types of parameters. This involved examining the characteristics of the different types of input parameters and assigning a particular type of probability distribution with properties that matched the input parameters. Information for the distribution parameters for each probability distribution was provided by the available data for each input parameter. All the clinical data describing events during first-line hormonal therapy were proportions. Employing Bayesian methods for specifying prior distributions the beta distribution provides the most realistic representation of proportions.[20] As the relevant data represented a series of patients who had either experienced an event, or not, the method used to estimate the parameters for the beta distribution invoked the Bayesian use of a non-informative prior distribution. If no prior information is available a uniform prior can be assumed with beta distribution parameters a = 1 and b = 1. The beta distribution parameters are ‘x+a’ and ‘n–x+b’, where x is the number of observations with the event of interest and n–x the number without the event. In the health states following first-line hormonal therapy, the TTP parameters were represented by triangular distributions because information for their values was provided only by point estimates and ranges. The remaining clinical parameters were described by point estimates because less reliable data were available to estimate distribution parameters, but also because these parameters occurred further downstream and had less impact on the cost-effectiveness results. Though detailed descriptions of the resource utilisation associated with each health state were identified, it was difficult to obtain objective measures of uncertainty around the estimated cost values. Therefore, minimum and maximum plausible values for each cost estimate were defined as 75 and © Adis Data Information BV 2003. All rights reserved.

521

125% of the original point estimate, which were also represented by triangular distributions. Undertaking a probabilistic sensitivity analysis involves estimating the relevant output parameters from a model for a large number of randomly sampled sets of input parameters. Following experimentation with the model it was found that the random sampling of 5000 sets of input parameter values from the specified probability distributions provided a full definition of the output probability distributions. The estimates of costs and survival associated with each set of input parameter values were combined to establish probability distributions of the model’s outputs. These model outputs were analysed statistically to represent the uncertainty in the cost-effectiveness results in a more intuitive manner than traditional forms of sensitivity analysis. Results The cost analysis of the SAEs experienced by patients receiving the alternative first-line therapies revealed that the cost per patient experiencing an event for letrozole and tamoxifen were £1571 and £2476, respectively. The cost of remaining in each of the health states included in the model for a 3-month period is presented in table IV. These results show that the administration of chemotherapy is associated with two of the more expensive states, especially second-line chemotherapy, whilst hospitalisation increases the costs associated with palliative care. Applying these state-specific costs to the model, table V presents the mean cost-effectiveness results. The mean baseline results were calculated using the mean values from the output distributions for costs and life-years derived from the stochastic analysis of the model for both letrozole and tamoxifen. The results show that patients receiving letrozole gain an additional 0.714 life-years (discounted at 1.5%), whilst the difference in lifetime treatment costs (discounted at 6%) is £1673. The mean cost of gaining an additional life-year from the use of letrozole compared with tamoxifen was estimated as £2342. ‘Credible intervals’, the Bayesian equivalent of the ‘confidence interval’,[21] can be estimated from Pharmacoeconomics 2003; 21 (7)

522


Table IV. Costs for each health state included in the advanced breast cancer model (cost [2000 values]a per 3-month period, unless otherwise stated) Intervention

Cost (£)

Proportion of patients receiving interventions in each health state 1st-line let. HT

1st-line tam. HT

2nd-line let. HT

2nd-line tam. HT

3rd-line HT

1st-line CT

2nd-line CT

obs. care

end-stage

0.38

0.44

0.64

0.64

Medication Megestrol (per day)

1.00

Tamoxifen (per day)

0.12

1 1 1

1

Letrozole (per day)

2.97

Aminoglutethimide (per day)

0.36

1

Estrogen (per day)

2.57

0.08

Fluoxymesterone [halotestin] (per day)

0.50

0.08

0.2

Morphine (per day)

0.68

0.26

0.26

0.26

0.26

0.49

Naproxen (per day)

0.29

0.2

0.2

0.2

0.2

0.28

Codeine (per day)

5.55

0.14

0.14

0.14

0.14

CMF (per cycle)

59.57

0.4

CAF (per cycle)

120.30

0.25

CEF (per cycle)

113.20

0.23

Ondansetron (per cycle)

89.10

0.56

Dexamethasone (per cycle)

1.76

0.39

Doxorubicin (per cycle)

341.04

0.5

Docetaxel (per cycle)

1347.80

0.19

Paclitaxel (per cycle)

1269.80

0.18

0.51

Consultations Oncologist

106

0.93

0.93

0.93

0.93

0.93

1

1

0.47

0.47

General practitioner

24

0.58

0.58

0.58

0.58

0.58

0.49

0.48

0.46

0.46

Radiographer

14

0.27

0.27

0.27

0.27

0.27

0.23

0.25

0.35

0.35

Laboratory tests 9

0.91

0.91

0.91

0.91

0.91

0.91

0.91

0.18

0.18

Blood tests

2.5

0.89

0.89

0.89

0.89

0.89

0.91

0.91

0.3

0.3

Bone scintography

68.63

0.57

0.57

0.57

0.57

0.61

0.38

0.38

0

0

Ultrasound

42.46

0.19

0.19

0.19

0.19

0.42

0.52

0.53

0.01

0.01

Chest x-ray

13.66

0.44

0.44

0.44

0.44

0.34

0.11

0.11

0.01

0.01

Bone x-ray

24.58

0.27

0.27

0.27

0.27

0.25

0.23

0.24

0.01

0.01

Continued next page

Karnon & Jones


Biochemical

523

CAF = cyclophosphamide, doxorubicin (adriamycin), fluorouracil; CEF = cyclophosphamide, epirubicin, fluorouracil; CMF = cyclophosphamide, methotrexate, fluorouracil; CT = chemotherapy; HT = hormonal therapy; let.= letrozole; obs. = observational; tam. = tamoxifen.

Drug costs from British National Formulary,[14] laboratory costs from the NHS Reference Costs,[13] consultation and hospitalisation costs from the annual study of UK health and social care by Netten and Curtis.[17]

3674 1090 831 397 657 397 657 169 Palliative care

Mean cost per 3-month period (£)

334 Oncology


a

576

2890

0.45 (20.5)

0.26 (13) 0.1 (14) 0.14 (8.5) 0.11 (8) 0.1 (10) 0.04 (5) 0.04 (5) 0.04 (5)

0.02 (9) 223 General medicine

0.02 (9)

0.02 (9)

0.02 (9)

0.04 (5)

0.01 (8.5) 0.03 (9.3)

3rd-line HT 2nd-line tam. HT 2nd-line let. HT 1st-line tam. HT 1st-line let. HT Hospitalisation (days)

Intervention

Table IV. Contd

Cost (£)

Proportion of patients receiving interventions in each health state

1st-line CT

0.01 (8)

2nd-line CT

0

obs. care

0.02 (21)

end-stage


the stochastic analysis by using the relevant percentiles in the outputs as the respective credible intervals. The 5th percentile (analogous to the lower limit of the 95% confidence interval) shows letrozole dominating tamoxifen (lower costs and larger effects), whilst the 95th percentile reveals an ICER of £6068 per life-year gained. The cost-effectiveness acceptability (CEAc) curve provides a more flexible interpretation of cost effectiveness,[22] presenting the probability that letrozole is cost effective (positive net benefits) for a range of non-negative values of an additional life-year. The CEAc curve derived from the 5000 observations of costs and life-years for the two therapies in this study is presented in figure 2. The curve shows that the probability that letrozole is cost effective passes the 0.5 probability threshold at a value of a life-year of around £2500. The remaining portion of the CEAc curve shows that the probability of positive net benefits rises to over 0.9 at a value of a life-year of between £5000 and £6000. Approximating the difference in QALYs associated with the two therapies by applying a range of utility values for patients with advanced breast cancer, without progression, to the difference in lifeyears shows that the mean cost per QALY gained with letrozole is between £2927 and £3969. Discussion This paper has presented a comprehensive economic analysis of the third-generation aromatase inhibitor, letrozole, versus the currently preferred first-line hormonal therapy, tamoxifen in postmenopausal patients with advanced breast cancer that is ER and/or PgR positive or of unknown receptor status. Based on patient-specific data derived from a clinical trial, supplemented with data identified in the published literature, and expert opinion in the form of Delphi panels, the evaluation used a Markov process to structure lifetime patient pathways and to combine the data from the various sources. Our results demonstrate that letrozole is cost effective compared with tamoxifen. The model structure differed from previous economic models of advanced breast cancer. Hutton et Pharmacoeconomics 2003; 21 (7)

Karnon & Jones

al.[23] compared second-line chemotherapies for advanced breast cancer using a decision model that depicted detailed pathways, including different grades of response to treatment, treatment options for SAEs and the impact of cumulative toxicity. Such detailed modelling was necessary because the time horizon for the model was so short (median survival was 9 months) and the primary output was quality-adjusted survival. Such detail was not incorporated in the present evaluation because the model described a longer time horizon, which would have increased the complexity of the patient pathways to a greater degree than was justified by the available data. The decision model employed by Hillner et al.[24] to compare autologous bone marrow transplant with standard chemotherapy in metastatic breast cancer included separate states to describe complete and partial remission, no progression and progression. This separation was not included in the current evaluation because reliable data describing patient pathways from these states following each line of hormonal therapy and chemotherapy were not available. Nuijten et al.[12] evaluated patients from the point of second-line hormonal therapies and separated patients on the basis of whether they had experienced a SAE. In the current evaluation, data were not available to differentiate between such patients further down the pathways so the experience of SAEs was handled within the first cycle of the model. Similarly, the current model did not facilitate the movement of patients receiving hormonal therapy to an observational care state as the cost difference between such states would be negligible (due to the relatively low cost of the hormonal therapies). Table V. Model outputs of costs, life-years and incremental cost per life-year (2000 values) Outputs

Results

Letrozole cost

£11 303.07

Letrozole life-years

4.182

Tamoxifen cost

£9630.55

Tamoxifen life-years

3.468

Mean ICER

£2342

5th percentile

£948

95th percentile

£10 134


Probability letrozole cost effective

524

1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0

5

10

15

20

25

30

Value of an additional life-year (£; thousands) Fig. 2. Cost-effectiveness acceptability curve for letrozole versus tamoxifen as first-line hormonal therapy for postmenopausal women with advanced breast cancer (2000 values).

This evaluation did not explicitly account for the quality of life of patients with advanced breast cancer, which is known to be an important factor.[25] However, it was possible to approximate the differential effectiveness of the alternative therapies with respect to QALYs under the reasonable assumption that the majority of the extra life-years enabled by letrozole were gained within the first-line hormonal therapy state. The range of utility values for patients with advanced breast cancer, without progression, has been presented as 0.59–0.8.[11] Applying these values the approximate mean cost per QALY gained with letrozole is between £2927 and £3969, which reinforces the incremental cost effectiveness of letrozole. Conclusions The efficacy of letrozole as a first-line hormonal therapy for metastatic breast cancer has been tested in a clinical trial setting and has shown increased response rates and TTP. The present economic evaluation has extrapolated the available clinical trial data using published data and expert opinion to estimate the cost per life-year saved compared with the standard first-line therapy – tamoxifen. The inherent uncertainty in data available to model cost effectiveness was described using input distributions for most parameters included in the model. The mean results of the economic analysis indicate that the letrozole is a cost-effective alternative first-line Pharmacoeconomics 2003; 21 (7)


therapy compared with tamoxifen, achieving additional life-years at a mean cost of £2342. The estimated credible intervals showed that even at the 95th percentile of the cost-effectiveness range, the ICER was just over £10 000. Acknowledgements Jon Karnon was funded by Novartis Pharmaceuticals UK Plc, the manufacturers of letrozole, to undertake this study. Trefor Jones is employed by Novartis.

References 1. Cancer Research Campaign. Fact sheet 1: incidence. London: CRC Publications, 1998 2. Harris, JR, Morrow M, Bonadonna G. Cancer of the breast. In: De Vita VT, Hellman S, Rosenberg SA, editors. Cancer: principles and practice of oncology. 4th ed. Philadelphia: Lippincott Co, 1993 3. Haller DG, Fox KR, Schuchter LM. Metastatic breast cancer. In: Fowble B, Goodman RL, Glick JH, Rosato EF, editors. Breast cancer treatment: a comprehensive guide to management. St Louis: Mosby Year Book, 1991 4. Muss HB. Endocrine therapy for advanced breast cancer. Breast Cancer Res Treat 1992; 21: 15-26 5. Leonard RC, Rodger A, Dixon JM. ABC of breast diseases: metastatic breast cancer. BMJ 1994; 309 [6967]: 1501-4 6. Kuerer HM, Buzdar AU, Singletary SE. Biologic basis and evolving role of aromatase inhibitors in the management of invasive carcinoma of the breast. J Surg Oncol 2001; 77 (2): 139-47 7. Buzdar AU, Hortobagyi G. Update on endocrine therapy for breast cancer. Clin Cancer Res 1998; 4 (3): 527-34 8. Buzdar AU. Role of aromatase inhibitors in advanced breast cancer. Endocr Relat Cancer 1999; 6 (2): 219-25 9. Mouridsen H, Gershanovich M, Sun Y, et al. Superior efficacy of Letrozole versus Tamoxifen as first-line therapy for postmenopausal women with advanced breast cancer: results of a phase III study of the International Letrozole Breast Cancer Group. J Clin Oncol 2001; 19 (10): 2596-606 10. HM Treasury. Appraisal and evaluation in central government. London: HMSO, 1997 11. Earle CC, Chapman RH, Baker CS, et al. Systematic overview of cost-utility assessments in oncology. J Clin Oncol 2000; 18 (18): 3302-17


525

12. Nuijten M, Meester L, Waibel F, et al. Cost effectiveness of Letrozole in the treatment of advanced breast cancer in postmenopausal women in the UK. Pharmacoeconomics 1999; 16 (4): 379-97 13. NHS Executive. The new NHS 2000 reference costs. London: Department of Health, 2000 14. British National Formulary 40. London: Bristish Medical Association/Royal Pharmaceutical Society of Great Britain, 2000 September 15. NHSE. Costing for contracting. Costed HRGs: evaluation summary report. Leeds: NHSE, 1995 16. Nuijten M, McCormick J, Waibel F, et al. Economic evaluation of Letrozole in the treatment of advanced breast cancer in postmenopausal women in Canada. Value in Health 2000; 3 (1): 31-9 17. Netten A, Curtis L.Unit costs of health and social care 2000. PSSRU: University of Kent, 2000 18. Briggs AH. A bayesian approach to stochastic cost-effectiveness analysis. Health Econ 1999; 8 (3): 257-61 19. Felli JC, Hazen GB. Sensitivity analysis and the expected value of perfect information. Med Decis Making 1998; 18 (1): 95-109 20. Iverson GR. Bayesian statistical inference. Thousand Oaks (CA): Sage, 1984 21. Briggs AH. Handling uncertainty in cost-effectiveness models. Pharmacoeconomics 2000; 17 (5): 479-500 22. Briggs A, Fenn P. Confidence intervals or surfaces?: uncertainty on the cost-effectiveness plane. Health Econ 1998; 7 (8): 723-40 23. Hutton J, Brown R, Borowitz M, et al. A new decision model for cost-utility comparisons of chemotherapy in recurrent metastatic breast cancer. Pharmacoeconomics 1996; 9 Suppl. 2: 8-22 24. Hillner BE, Smith TJ, Desch CE. Efficacy and cost-effectiveness of autologous bone marrow transplantation in metastatic breast cancer: estimates using decision analysis while awaiting clinical trial results. JAMA 1992; 267 (15): 2055-61 25. de Haes JC, de Koning HJ, van Oortmarssen GJ, et al. The impact of a breast cancer screening programme on qualityadjusted life-years. Int J Cancer 1991; 49 (4): 538-44

Correspondence and offprints: Dr Jon Karnon, School of Health and Related Research, University of Sheffield, 30 Regent Street, Sheffield, S1 4DA, England.
