5 Outcomes

5.1 The Diagnostics Advisory Committee (section 11) considered evidence from a number of sources (section 12).

How outcomes were assessed

5.2 The assessment consisted of a systematic review of the evidence on test performance and clinical-effectiveness data for the 4 tests included in the evaluation. The outcome measures included:

  • Analytical validity, defined as the ability of the test to accurately and reliably measure the expression of mRNA or proteins by breast cancer tumour cells (that is, repeatability and reproducibility).

  • Clinical validity, defined as prognostic ability or the degree to which the test can accurately predict the risk of an outcome (for example, the risk of distant metastases in 10 years).

  • Clinical utility, defined as the ability of the test to improve clinical outcomes such as overall survival. This includes direct harms arising from the test, reclassification of risk compared with existing tools, its impact on clinical decision-making and the ability of the test to predict benefit from chemotherapy. Within the context of clinical utility, the predictive ability of a test refers to the capability of the test to accurately predict patients who will benefit most, in relative terms, from chemotherapy, that is, whether patients classified as high risk benefit more in relative terms than patients classified as low risk.

5.3 In the base-case economic analysis, the External Assessment Group used the available data on the clinical validity and clinical utility of the tests to populate the model. The risk of distant recurrence (prognosis) was computed from data on the clinical validity of the tests. The reclassification of risk by the new tests (presented as 2 Nottingham Prognostic Index [NPI] subgroups), the impact of the test results on clinical decision-making (the proportion of patients receiving adjuvant chemotherapy) and the predicted benefit of chemotherapy by risk group (reduction in the risk of distant recurrence) were based on data on the clinical utility of the tests. In all cases, the systematic review showed that data on the clinical validity of the tests were more robust than data on their clinical utility. Therefore the External Assessment Group used sensitivity analysis to explore alternative scenarios with different assumptions of the clinical utility of the tests and, in some cases alternative assumptions of clinical validity.

5.4 For 2 of the 4 tests (Oncotype DX and MammaPrint), the current systematic review updated an existing systematic review of gene expression profiling tests for breast cancer. Two previous systematic reviews (one an update of the other) reviewed the literature relating to both Oncotype DX and MammaPrint. Marchionni et al. (2008) is an exhaustive literature review of various electronic databases covering biomedical literature between 1990 and 2006. In 2010, Smartt updated this systematic review and included all relevant evidence that became available between 2007 and 2009.

5.5 The External Assessment Group undertook a systematic review of the evidence on cost effectiveness for the 4 tests. Genomic Health (Oncotype DX) submitted an economic model and Clarient submitted a report detailing an economic analysis of Mammostrat. The External Assessment Group constructed a de novo economic model. The outcomes of interest for the economic evaluation were the morbidity and mortality associated with invasive breast cancer and its treatment. These included survival and health-related quality of life, including the impact of adverse events associated with chemotherapy. The de novo economic model followed a linked evidence approach in which intermediate outcomes (results of the tests) were linked to treatment outcomes and hence quality-adjusted life year (QALY) gains. Costs and QALYs were assigned to each of the 4 tests and the comparator.

5.6 The population identified in the scope for this evaluation included assessment of the gene expression profiling and expanded immunohistochemistry tests in men with breast cancer if data were available. No such data that would allow the evaluation of these technologies in men were identified in the systematic review.

Clinical effectiveness

5.7 The terms analytical validity, clinical validity and clinical utility, used in this section, are defined in section 5.2.

5.8 Much of the clinical evidence was related to the Oncotype DX and MammaPrint tests because these tests are much further along the validation pathway than IHC4 and Mammostrat. The highest-quality evidence was reported for Oncotype DX, although limitations or gaps in the clinical data were identified for all tests. Most studies, for all tests, were retrospective in design, analysing archived tumour samples from a cohort of patients with documented information on patient characteristics and outcomes. Retrospective analyses are associated with increased bias compared with prospective randomised controlled trials. Some of the studies involved a prospective analysis of retrospective archived material from a previous randomised controlled trial. Potential issues still remain, including the effects of confounding and the possible incompleteness of some biological specimens.

5.9 Study populations were generally heterogeneous, although most of the evidence on Oncotype DX came from oestrogen receptor positive (ER+), lymph node negative (LN−) populations. Some studies included a small number of participants. Studies including larger sample sizes, in excess of 1000 samples, were available for Oncotype DX, Mammostrat and IHC4. Follow‑up was short or not reported in a number of studies. Five studies were specific to a UK population, including 3 for Oncotype DX, 1 for IHC4 and 1 for Mammostrat.

MammaPrint

5.10 A range of studies provided evidence on the prognostic ability of MammaPrint in heterogeneous populations. However, the previous systematic reviews indicated that evidence relating to the clinical validity of MammaPrint was not always conclusive nor supported the prognostic value of the test. Four studies suggested that the test could predict prognosis, 1 study of prognostic utility did not reach statistical significance and in another the methods and results were at variance with other studies. In terms of clinical utility, the previous reviews identified 1 prospective observational study (Bueno‑de‑Mesquita et al. [2007], also known as the RASTER study) demonstrating that MammaPrint had an impact on clinical decision-making when used in addition to current practice in the Netherlands (Dutch Institute for Healthcare Improvement [CBO] guidelines). Adjuvant systemic treatment was advised less often when Dutch CBO guidelines were used compared with use of MammaPrint. Therefore, Bueno‑de‑Mesquita et al. (2007) reported that the addition of MammaPrint to the standard Dutch clinical assessment of risk (modified by patient preference) in a cohort of 427 patients increased the number of patients receiving adjuvant systemic therapy by 20 (5%). At the time of the previous systematic review, follow‑up was not long enough to provide evidence of its effect on clinical end points such as distant metastasis-free survival or its utility in predicting treatment benefit; however, follow‑up data at 5 years have recently been published in the study by Drukker et al. (2013). The study reported, among other outcomes, that in the group of patients classified as being at low risk with MammaPrint and at high risk with Adjuvant! Online (of whom 76% had not received adjuvant chemotherapy), the 5‑year distant recurrence-free interval was 98.4%. The previous systematic reviews recommended that further evidence from randomised controlled trials was needed in addition to robust evidence on the prediction of benefit from chemotherapy.

5.11 The External Assessment Group identified 7 additional, non-UK-based, studies of MammaPrint. Of these 7 studies, 4 on the clinical validity of MammaPrint demonstrated that the MammaPrint score is a strong independent prognostic factor, and may provide additional value to standard clinicopathological measures. A mix of evidence exists for outcomes at 5 and 10 years. The population in all these studies was relatively small. One of the studies was of a Japanese population, and follow‑up was limited to only 5 years in 2 of the studies. For example, Mook et al. (2010) showed that in 148 women the distant metastasis-free survival at 5 years was 93% in the low-risk group and 72% in the high-risk group with an associated hazard ratio (HR) of 4.6 (95% confidence interval [CI] 1.8 to 12.0, p=0.001). The External Assessment Group did not identify any prospective studies of the impact of MammaPrint on long-term outcomes such as overall survival, but the prospective observational RASTER study published 5‑year follow‑up data after the External Assessment Group completed its assessment, and this was discussed by the Committee (see section 6.13). Six studies with data on the clinical utility of MammaPrint were identified by the External Assessment Group. Five studies reported use of MammaPrint to reclassify patients into high- and low-risk groups and compared this with the risk assigned according to current local guidance. They reported a high level of discordance between MammaPrint and current classification, although these studies did not demonstrate how this would impact on treatment decisions. For example, Bueno‑de‑Mesquita et al. (2009) compared MammaPrint risk categories and risk assessment based on Adjuvant! Online, St Gallen guidelines, NPI and Dutch CBO guidelines (2004). Discordance between MammaPrint and the other risk assessment measures was 38%, 41%, 26%, and 30% respectively. One study reported that the use of MammaPrint would result in altered treatment advice for 40% of patients based on the assumption that all patients classified as high risk would receive chemotherapy and patients classified as low risk would not receive chemotherapy. Because the study was retrospective, altered treatment advice assumed in the analysis represented potential changes and not actual changes from using the test in clinical practice.

5.12 A study of the benefit of chemotherapy according to risk group stratification by MammaPrint was identified (Knauer et al. 2010) but was omitted from the systematic review because it was based on a pooled analysis of 6 primary studies. The External Assessment Group did not include the pooled analysis in the systematic review to avoid double counting of studies already included in the review. In addition, it did not consider the findings of the pooled analysis to be robust because the authors did not reanalyse the tumour samples, it is unclear how individual patient data were combined, and there were potential issues with the statistical analyses performed (for example, although median follow‑up was 7.1 years the data were arbitrarily truncated at 5 years).

5.13 Robust evidence of clinical utility is not available for MammaPrint so it is not yet clear whether using the test will improve the use of adjuvant chemotherapy in the management of breast cancer in the UK. In summary, most studies of MammaPrint were retrospective in design, used small sample sizes and had heterogeneous patient populations and some studies included only pre-menopausal women. Moreover, no studies were conducted in the UK. The evidence for MammaPrint is based on the use of the test with fresh samples. It is not clear whether this evidence would apply if the test were used on formalin-fixed paraffin-embedded samples. Overall, the External Assessment Group considered that further robust evidence on the clinical validity and clinical utility of the test would be helpful.

Oncotype DX

5.14 Oncotype DX was reported to be furthest along the validation pathway by previous systematic reviews. In terms of clinical validity, these reviews reported evidence that the Oncotype DX recurrence score was significantly correlated with disease-free survival and overall survival. Furthermore, the recurrence score was shown to be a better predictor of distant recurrence at 10 years than traditional clinicopathological predictors. The evidence on clinical utility was limited. One study (Paik et al. 2006) demonstrated a significantly increased benefit from the use of chemotherapy in the Oncotype DX high-risk group compared with the low-risk group, although the review highlighted that the study may have potential flaws. The study indicated that this benefit difference was caused by the better prognosis without chemotherapy (and hence the reduced absolute benefit these patients would receive) and the decreased relative benefit of chemotherapy in the lower-risk groups. The specific cancers in the low-risk groups were less likely to respond to chemotherapy, independent of actual survival probability. Key gaps were identified in the evidence base related to the extent to which the test added to the management of breast cancer and the proportion of patients who would benefit from the test. The previous systematic reviews indicated that prospective confirmation of the clinical utility of Oncotype DX was needed.

5.15 The External Assessment Group identified 12 additional studies of Oncotype DX. Further larger studies now support the prognostic ability of Oncotype DX. One large-scale UK study in post-menopausal women with ER+, LN− early breast cancer found that an increase in risk score was significantly associated with an increased risk of distant recurrence. Furthermore, the evidence base has been extended to include the LN+ population. The External Assessment Group did not identify any prospective studies of the impact of Oncotype DX on long-term outcomes such as overall survival. Four studies were identified that presented further evidence on the impact of Oncotype DX on clinical decision-making. These indicated that the use of Oncotype DX leads to changes in treatment decisions for between 32% and 38% of patients. However, only 1 of these studies was performed in the UK and limitations in relation to the generalisability of the study were identified. In addition, the study only included a small sample of patients (interim results on 106 patients were available for the systematic review, the dataset for 142 patients was available for and used in the External Assessment Group's cost-effectiveness analysis). Four recent publications reported evidence that Oncotype DX predicts benefit from chemotherapy. However, only 1 of these studies on an LN+ population (Albain et al. 2010) presented new data. The other 3 publications (Tang et al. 2011, Mamounas et al. 2010 and Tang et al. 2010) reported the same trial data as Paik et al. (2006). The first evidence of improvements in quality of life and reduced patient anxiety as a result of using Oncotype DX have been reported, but the studies had small sample sizes (for example, Lo et al. [2010] included 89 patients). In summary, Oncotype DX is considered to have the most robust evidence base of the tests reviewed in this guidance, with data on the analytical validity, clinical validity and clinical utility of the test. The studies varied considerably in their size, design and patient populations. Many of the Oncotype DX studies were small and retrospective. A small number of studies were conducted in the UK. The External Assessment Group considered that further robust evidence on the clinical utility of the test would be helpful.

IHC4

5.16 No studies on analytical validity of IHC4 (based on ER, progesterone receptor [PR], human epidermal growth factor receptor 2 [HER2] and Ki‑67 in addition to classical clinical and pathological variables combined using an algorithm) were identified. Of the 4 individual tests that make up IHC4, 2 (ER and HER2) are commonly measured in the NHS. However, the quantitative assessment of ER needed for IHC4 calculations is not routinely performed. Outstanding issues around the reproducibility of detecting Ki‑67 also exist. This is noteworthy because the test is designed for local use and different local processing methods may potentially lead to different results. The External Assessment Group identified 1 study on the clinical validity of IHC4 (Cuzick, 2011), which reports that the IHC4 score is a highly significant predictor of distant recurrence. The authors validated the test in a cohort of 786 patients with ER+ cancer treated in the UK, and demonstrated that the IHC4 score was highly significantly predictive of outcome, with a hazard ratio of 4.8 (95% CI 2.2 to 10.2) for a change from the 25th to 75th percentile in a univariate analysis. This study also reported evidence comparing IHC4 against Oncotype DX. The study was rated as high quality. The External Assessment Group did not identify any prospective studies of the impact of IHC4 on long-term outcomes such as overall survival. It did not identify any published evidence on the clinical utility of IHC4 in terms of its ability to change treatment decisions or its ability to predict chemotherapy benefit. In summary, the External Assessment Group considered that the evidence base for IHC4 is currently limited to clinical validity (prognostic ability), although this evidence is considered to be relatively robust, and further evidence would be helpful on analytical validity and clinical utility.

Mammostrat

5.17 The External Assessment Group did not identify any specific studies on the analytical validity of Mammostrat, although some limited evidence on analytical validity was reported in studies of clinical validity and clinical utility. Three studies were identified that provided data to support the use of Mammostrat as an independent prognostic tool for women with ER+, tamoxifen-treated breast cancer. Although the evidence base for Mammostrat is at present relatively limited, these studies included a large sample size, appeared to be of reasonable quality, and 1 study provided data from a UK setting. The External Assessment Group did not identify any prospective studies of the impact of Mammostrat on long-term outcomes such as overall survival. In addition, clinical utility data on Mammostrat (from 1 study) suggest that the low- and high-risk groups benefit from chemotherapy, but not the intermediate-risk group. There was no published evidence on reclassification of risk groups compared with conventional means of risk classification, and no evidence on the impact of the test on clinical decision-making. Overall, the External Assessment Group considered that further evidence of analytical validity and clinical utility would be helpful.

Economic analysis

5.18 Four studies were identified as meeting the inclusion criteria of the systematic review of cost-effectiveness evidence (2 for MammaPrint and 2 for Oncotype DX). None were conducted in England. Genomic Health and Clarient also submitted economic analyses on the cost effectiveness of Oncotype DX and Mammostrat in England respectively. Several issues were highlighted in the critique of these analyses, which needed further consideration. These included assumptions about: the baseline level of chemotherapy in clinical practice; the risk of distant recurrence; patients who would be offered the test; the proportion of patients who would be offered chemotherapy after reclassification with the new test; the cost of chemotherapy and therapy used to prevent or treat associated adverse events.

5.19 The External Assessment Group constructed a de novo economic model to specifically address the decision problem for this evaluation and to estimate the cost effectiveness of the 4 tests in England.

5.20 The population assessed in the economic model was women with ER+, LN−, HER2− early breast cancer up to 75 years old at diagnosis. One analysis assumed that all women in the group received the new tests. However, the External Assessment Group's clinical specialists suggested that the new tests may be targeted at a subgroup of this population – those at intermediate risk of distant recurrence for whom the decision about whether or not to give chemotherapy is most uncertain. A subgroup analysis was performed that assumed that the new test was given only to women with an NPI score above 3.4 (used as a proxy for those women at intermediate risk of distant recurrence), based on the assumption that most women at low risk (with an NPI score below 3.4) would not be considered for chemotherapy and that there would not be many women at high risk (with an NPI score of above 5.4) within the population considered.

Clinical outcomes

5.21 Modelling was used to estimate clinical outcomes. All women in the model were assumed to be treated with endocrine therapy. A state transition model was used to simulate breast cancer outcomes for patients treated with endocrine therapy alone or with the addition of chemotherapy. Outcomes associated with breast cancer were simulated using multiple health states including recurrence-free survival, recurrence (distant and local), adverse events from chemotherapy, and death.

Costs

5.22 The costs included in the economic model were the costs of the different tests, treatment costs (endocrine therapy and chemotherapy), costs of short-term and long-term adverse events associated with chemotherapy (including the secondary prevention of short-term adverse events), costs associated with managing distant recurrence, local recurrence and terminal care.

5.23 The cost of the MammaPrint test is £2675 (this cost was used in the economic model). The Oncotype DX test costs £2580 (this cost was used in the original economic analysis, but a revised cost was used for the economic analysis conducted for the confidential revised price). IHC4 was estimated to cost £100–£200 (£150 was used in the economic model) for quantitative analysis of ER (which may need additional time compared with traditional assessment of ER status), and assessment of PR and Ki‑67 (which are not routinely collected) and running the algorithm (it was assumed that HER2 would be measured as part of standard practice). The Mammostrat test has an indicative cost between £1120 and £1620 (£1135 was used in the economic model).

Cost effectiveness

5.24 The primary analysis compared current clinical practice with treatment guided by Oncotype DX and IHC4. The systematic review of the evidence indicated most evidence for Oncotype DX compared with the other tests, and that the evidence base for Oncotype DX, in particular in relation to prognostic ability, was reasonably sound. There was less evidence for IHC4, but there was evidence relating to the performance of IHC4 compared with Oncotype DX. This evidence, with some additional assumptions when compared with the analysis of Oncotype DX, was used to model the cost effectiveness of IHC4. Additional assumptions include the reproducibility of the test and the use of risk groups as opposed to a continuous risk score. There was no evidence on the ability of IHC4 to predict benefit from chemotherapy; in the IHC4 analysis, the predicted benefit of chemotherapy was applied according to the Oncotype DX risk classification.

5.25 In addition to the primary economic analysis, further economic analyses were undertaken for Mammostrat and MammaPrint. These additional analyses were deemed exploratory by the External Assessment Group because there are significant limitations in the evidence base and the generalisability of the data to practice in England. Only data from studies conducted in the USA (Mammostrat) and The Netherlands (MammaPrint) were available to estimate the reclassification of patients using the new test. There was concern whether these data are generalisable to England and there was also uncertainty in the data. There were no studies showing the impact of these tests on the management of breast cancer in the England. In addition, the External Assessment Group reported considerable uncertainty in the data used to estimate the predicted benefit of chemotherapy by MammaPrint risk groups.

5.26 All analyses assumed that the new tests were used in addition to current practice (for IHC4 it was assumed that quantitative analysis of ER [which may need additional time compared with traditional assessment of ER status], PR and Ki‑67 is carried out in addition to current practice and data combined in an algorithm). Full details of the results can be found in section 5.6 of the diagnostics assessment report. A brief summary of the key results (including incremental cost-effectiveness ratios [ICERs]) of the base-case analysis and sensitivity analyses is presented below. Results of the 3 analyses (Oncotype DX and IHC4; Mammostrat; MammaPrint) cannot be directly compared because the data came from different studies with different patient characteristics and methodologies, and the basis for each model therefore varies significantly.

5.27 The base-case analysis modelled a hypothetical cohort of 1000 women over a lifetime horizon (100 years was used as the upper age limit). Two analyses are presented. In the first the tests were used for all women with ER+, LN−, HER2− early breast cancer aged up to and including 75 years. In the second the tests were used only for women with ER+, LN−, HER2− early breast cancer up to and including 75 years with an NPI score above 3.4 (used as a proxy for those women at intermediate risk of distant recurrence). Results are presented on a per patient basis and any differences in expected values are a result of rounding error.

Oncotype DX and IHC4

5.28 Tests used for all women with ER+, LN−, HER2− early breast cancer and Oncotype DX is assessed at the list price of £2580. In the primary economic analysis comparing Oncotype DX and IHC4 with current practice, the proportions of patients receiving chemotherapy were 19.11%, 9.57% and 14.42% respectively. The model predicted that there would be 64, 71 and 76 distant recurrences when using Oncotype DX, IHC4 or current practice respectively. Total costs and QALYs, assuming predictive benefits (that is, benefits from identifying who will benefit most from chemotherapy) based on Paik et al. (2006), are summarised in table 1.

Table 1 Per-patient costs, QALYs and ICERs in the primary economic analysis (Oncotype DX and IHC4 compared with current practice)a

Mean cost (£)

Mean QALYs

ICER – compared with current practice

Oncotype DX

£9094

13.54

£26,940b

IHC4

£6340

13.49

Dominant

Current practice

£6519

13.44

 

Abbreviations: ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life year.

a The analysis assumed the tests are used for all women with ER+, LN−, HER2− early breast cancer in England. Predictive benefit was based on Paik et al. (2006) and is based on the list price of Oncotype DX.

b Rounding error contributes to the difference from expected value.

5.29 Compared with current practice, Oncotype DX was associated with an incremental cost of £2575 and incremental QALYs of 0.1, yielding an ICER of £26,940 per QALY gained. IHC4 was £179 cheaper than current practice (cost saving), with incremental QALYs of 0.05 and was predicted to be dominant (that is, provide more QALYs at a lower cost) compared with current clinical practice. Oncotype DX, IHC4 and current practice were also compared using incremental analysis; that is, the least effective strategy was compared with the next least effective strategy that was neither dominated nor extendedly dominated. The cost-effectiveness acceptability curve showed that the probability of IHC4 being cost effective (when compared with current practice) was almost 100% if the maximum acceptable ICER was £20,000 per QALY gained. At the same maximum acceptable ICER, the probability of Oncotype DX being cost effective, when compared with current practice only, was 12.4%.

5.30 Tests used for women with ER+, LN−, HER2− early breast cancer and an NPI score above 3.4, and Oncotype DX is assessed at the list price of £2580. In the primary economic analysis the proportion of these patients predicted to receive chemotherapy was 34.72%, 26.31% and 33.60% with Oncotype DX, IHC4 and current practice respectively. The model predicted that there would be 117, 129 and 144 distant recurrences when using Oncotype DX, IHC4 or current practice respectively. Total costs and QALYs, assuming predictive benefits based on Paik et al. (2006), are summarised in table 2.

Table 2 Per patient costs, QALYs and ICERs in the primary economic analysis (Oncotype DX and IHC4 compared with current practice)a

Mean cost (£)

Mean QALYs

ICER – compared with current practice

Oncotype DX

£10,911

13.06

£9007b

IHC4

£8318

12.97

Dominant

Current practice

£8816

12.83

 

Abbreviations: ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life year.

a The analysis assumed the tests are used for women with ER+, LN−, HER2− early breast cancer and an NPI score above 3.4 in England. Predictive benefit was based on Paik et al. (2006) and is based on the list price of Oncotype DX.

b Rounding error contributes to the difference from expected value.

5.31 Compared with current practice, Oncotype DX was associated with an incremental cost of £2095 and incremental QALYs of 0.23, which resulted in an ICER of £9007 per QALY gained. IHC4 was £498 cheaper than current practice (cost saving), with incremental QALYs of 0.14 and was predicted to be dominant (that is, provide more QALYs at a lower cost) compared with current clinical practice. Oncotype DX, IHC4 and current practice were also compared using incremental analysis; that is, the least effective strategy was compared with the next least effective strategy that was neither dominated nor extendedly dominated. The cost-effectiveness acceptability curve showed that the probability of IHC4 being cost effective (when compared with current practice) was almost 100% if the maximum acceptable ICER was £20,000 per QALY gained. At the same threshold, the probability of Oncotype DX being cost effective, when compared with current practice only, was 91.6%.

5.32 Sensitivity analyses (univariate). A range of univariate sensitivity analyses were undertaken to explore the impact of varying the main model parameters. Analyses of varying the assumptions underlying the structure of the model were also performed. The ICERs for Oncotype DX compared with current clinical practice, for all women with ER+, LN−, HER2− early breast cancer and those with an NPI above 3.4, were sensitive (defined as changes in the ICER by 10% or more) to some of the assumptions made in the model. These included the time horizon modelled, the starting age of the cohort, the risk of recurrence, the proportion of patients receiving chemotherapy after reclassification with the new test, the benefit of chemotherapy in the different risk groups and the distribution of patients by NPI score. For example, the ICERs for Oncotype DX (compared with current practice) when offered to all women with ER+, LN−, HER2− early breast cancer were £91,300 (assuming 30% relative risk reduction from chemotherapy for all patients) per QALY gained and £64,900 (assuming 40% relative risk reduction from chemotherapy for all patients) per QALY gained. The ICERs for IHC4 compared with current clinical practice, for all women with ER+, LN−, HER2− early breast cancer and those with an NPI score above 3.4, were sensitive to more assumptions (such as the time spent in the distant recurrence health state, the proportion of patients receiving chemotherapy under current practice and the cost of chemotherapy), but IHC4 remained dominant compared with current practice (that is, it provided more QALYs at a lower cost) except when the cost of IHC4 was set at £400 (the resulting ICER was £1557 per QALY gained).

5.33 Following the first consultation, the manufacturer of Oncotype DX submitted a proposal to make it easier for the NHS to access the technology. The proposal makes Oncotype DX available at a revised price. The proposed price is commercial in confidence. The proposal is made for patients at an intermediate risk of distant recurrence, defined as an NPI score above 3.4 in this guidance. An External Assessment Group analysis of the proposal, using the proposal price and the assumption that Oncotype DX is validated as a prognostic tool but does not predict the benefit patients will get from chemotherapy, yielded an ICER of £22,600 per QALY gained compared with current clinical practice for patients with an NPI score above 3.4.

Mammostrat (exploratory analysis)

5.34 Test used for all women with ER+, LN−, HER2− early breast cancer. The proportion of patients receiving chemotherapy increased with the use of Mammostrat when compared with current practice (21.16% and 14.42% respectively). Current practice was associated with a mean cost of £7699 and mean QALYs of 12.86. Mammostrat was associated with a mean cost of £9040 and mean QALYs of 12.91. The ICER for Mammostrat was estimated to be £26,598 per QALY gained. However there were significant uncertainties and limitations associated with this analysis. These included uncertainty about the generalisability of the risk reclassification data to a UK population of patients with ER+, LN−, HER2− breast cancer, and the lack of evidence on the impact of the test on decision-making. In addition, the robustness of evidence on the predictive ability of the test is uncertain – clinical utility data from 1 study suggest that the low- and high-risk groups benefit from chemotherapy, but not the intermediate-risk group. The cost-effectiveness acceptability curve showed a 36.0% probability of Mammostrat being cost effective if the maximum acceptable ICER is £20,000.

5.35 Test used for women with ER+, LN−, HER2− early breast cancer and an NPI score above 3.4. The proportion of patients receiving chemotherapy increased slightly with the use of Mammostrat when compared with current practice (34.27% and 33.60% respectively). Current practice was associated with a mean cost of £9717 and mean QALYs of 12.34. Mammostrat was associated with a mean cost of £10,985 and mean QALYs of 12.29. Mammostrat was shown to be dominated by current practice. The cost-effectiveness acceptability curve showed an 18.0% probability of Mammostrat being cost effective if the maximum acceptable ICER is £20,000.

5.36 Sensitivity analyses (univariate). A range of univariate sensitivity analyses were undertaken to explore the impact of varying model parameters. When offering the test to all women with ER+, LN−, HER2− early breast cancer the ICER was very sensitive to the proportion of patients who would receive chemotherapy based on the test result. The ICER ranged between £18,879 per QALY gained to being dominated, when using the confidence intervals from the Ross et al. (2008) study for the predicted benefit of chemotherapy in terms of the reduction of distant recurrence. The ICER was not sensitive to the assumptions about utility values, management costs and the time spent in the recurrence health state. Mammostrat remained dominated under the assumptions examined in the sensitivity analysis when the test was offered to women with ER+, LN−, HER2− early breast cancer and an NPI score above 3.4.

MammaPrint (exploratory analysis)

5.37 Test used for all women with ER+, LN−, HER2− early breast cancer. The proportion of patients receiving chemotherapy increased with the use of MammaPrint when compared with current practice (44.18% and 14.42% respectively). Current practice was associated with mean costs of between £6408 and £6629, and mean QALYs of between 13.39 and 13.49. MammaPrint was associated with mean costs of between £10,017 and £10,748 and mean QALYs of between 13.47 and 13.78. Because of uncertainty around the evidence on the benefit of chemotherapy for the MammaPrint risk groups, the results for MammaPrint were presented as a range (based on the confidence interval for the benefit of chemotherapy). The ICER was estimated to be between £12,240 and £53,058 per QALY gained. Additional uncertainties include the lack of UK data in a relevant population (patients with ER+, LN−, HER2− breast cancer; particularly in relation to risk reclassification compared with UK practice), the impact of the test on clinical decision-making in the UK and reliance on data mainly from pre-menopausal populations.

5.38 Test used for women with ER+, LN−, HER2− early breast cancer and an NPI score above 3.4. The proportion of patients receiving chemotherapy increased with the use of MammaPrint when compared with current practice (90.31% and 33.60% respectively). Current practice was associated with mean costs of between £8281 and £8872 and mean QALYs of between 12.81 and 13.07. MammaPrint was associated with mean costs of between £12,278 and £14,014 and mean QALYs of between 12.99 and 13.73. Because of uncertainty around the evidence on the benefit of chemotherapy for the MammaPrint risk groups, the results for MammaPrint were presented as a range (based on the confidence interval for the benefit of chemotherapy). The ICER for MammaPrint was estimated to be between £6053 and £29,569 per QALY gained. Additional uncertainties include the lack of UK data and the reliance on data mainly from pre-menopausal populations.

5.39 Sensitivity analyses (univariate and multivariate). Given the uncertainty in the base-case analysis a limited number of sensitivity analyses were undertaken. Univariate analyses included: assuming no additional cost to the NHS for the use of fresh tissue samples and that 5% of patients classified as good prognosis and 95% of patients classified as poor prognosis received chemotherapy. A multivariate sensitivity analysis explored different values for the benefit of chemotherapy in terms of reduction in the risk of distant recurrence, assuming MammaPrint was used in all women with ER+, LN−, HER2− early breast cancer.

  • National Institute for Health and Care Excellence (NICE)