pat hways Medtech innovation briefing Published: 25 March 2015 nice.org.uk/guidance/mib27

pat hways The Prosigna gene expression profiling assay for assessing long-term risk of breast cancer recurrence Medtech innovation briefing Published: 25 March 2015 nice.org.uk/guidance/mib27 Summary The Prosigna assay measures gene expression to provide a risk of distant recurrence of breast cancer 10 years after a woman's initial diagnosis. The evidence comprises 1 multicentre study on archived breast cancer tissue showing similar precision and reproducibility across 3 testing sites and 3 prospective-retrospective studies. These analysed data collected in 2 previous randomised controlled trials of endocrine treatment, showing that the risk of recurrence measure (generated by the Prosigna assay) helped to predict distant recurrence. The estimated cost per Prosigna assay, including capital equipment and single-use components, is 1382 excluding VAT. Page 1 of

Product summary and likely place in therapy Prosigna is a prognostic gene signature assay that measures gene expression to calculate the risk of women with early-stage breast cancer developing distant recurrence 10 years from diagnosis. Prosigna is intended to be used in conjunction with other information in treatment decisions for women with hormone receptor-positive early-stage breast cancer. Effectiveness eness and safety One multicentre study (n=43) investigated the analytical validity of the assay. Archived breast cancer tissue samples from an American tissue bank were analysed at 3 centres. Prosigna showed similar precision and reproducibility across 3 testing sites to that previously reported for centralised lab tests. The clinical validity of Prosigna was investigated in 2 prospective-retrospective studies analysing data collected in 2 previous randomised controlled trials (n=1017, n=1478). A third study (n=2137) combined the data and performed a pooled analysis from the 2 randomised control trials. All 3 studies included women with early-stage breast cancer treated with endocrine therapy only. In all 3 studies, the risk of recurrence measure helped predict distant recurrence at 10 years after diagnosis. Page 2 of

Technical factors The assay measures the expression profiles of all 50 of the genes included in the PAM50 gene signature. The Prosigna assay is performed on RNA that has been isolated from formalin-fixed, paraffin-embedded breast tumour tissue. Cost and resource use Prosigna has an NHS acquisition cost of 182,600, excluding VAT, for the capital component (the ncounter Dx Digital Analyzer). The disposable component costs 1362 (Prosigna kit) plus 13 (RNA extraction kit), excluding VAT, for each test. Based on assumptions about workflow and system lifespan, the estimated overall cost per test is 1382 excluding VAT. No evidence on cost impact and resource use was available. The analysis is done in a local molecular pathology lab using the ncounter Dx Digital Analyzer. Introduction Breast cancer is the most common cancer overall and second most common cause of cancer death in women in the UK. In 2011, 49,936 women were diagnosed and 11,643 died from breast cancer (Cancer Research UK, 2014a). Breast cancer also occurs in men, although it is very rare. Approximately 350 men are diagnosed with breast cancer in the UK annually (Cancer Research UK, 2014d). The risk of developing breast cancer depends on many factors, including the following (Cancer Research UK, 2014b): age (80% of cases occur in women aged 50 years and over) race (in England, white women have a higher risk of breast cancer than women from other ethnic groups) genetic factors (including BRCA gene mutations) Page 3 of

use of certain medications (oral contraceptives, some types of hormone replacement therapy and diethylstilbestrol use in pregnancy increase breast cancer risk) lifestyle (obesity and alcohol consumption increase the likelihood of breast cancer; breastfeeding and physical activity protect against breast cancer) ionising radiation. Women with breast cancer develop local symptoms that may include lumps in the breast or armpit, change in the size or shape of the breast or nipple, or discharge from the nipple. General symptoms such as fatigue, weight loss, anorexia and bone pain are associated with advanced metastatic disease (Cancer Research UK, 2014c). Breast cancer is diagnosed through a range of tests. These include clinical examination, imaging tests such as mammography, ultrasound, MRI, CT scans and breast biopsy. Based on the results of these tests, breast cancer is categorised into 4 stages: Early breast cancer (stages 1 and 2). This means that the cancer has not spread beyond the breast or the lymph nodes in the armpit and is only present in 3 or fewer lymph nodes. Locally advanced breast cancer (stage 3). This refers to cancer which has not spread to another part of the body, but may be larger than 5 cm in diameter, growing into the skin or muscle of the chest, or present in the lymph nodes in the armpit (more than 4 positive lymph nodes). Advanced breast cancer (stage 4). This refers to cancer which has metastasised. Breast cancer survival depends mainly on the disease stage at diagnosis and the treatment received. More than 90% of women diagnosed with early breast cancer survive for at least 5 years. In contrast, only 13% of those diagnosed with advanced disease survive for more than 5 years (Cancer Research UK, 2014a). Because there is limited awareness of the condition in men, they often present only when their symptoms become severe and the disease is at an advanced stage (NHS Choices 2013). As a result, breast cancer prognosis is not as good in men as in women. In many cases, people with breast cancer have surgery followed by radiotherapy. There are 3 categories of breast cancer surgery (Cancer Research UK, 2014e): removal of the whole breast (mastectomy) lump removal (called lumpectomy or wide local excision) removal of a quarter of the breast containing the cancer (called quadrantectomy). Page 4 of

Oncologists use a number of tools to make decisions about the risk of disease recurrence after surgery. These tools categorise prognosis as 'good' (low risk of metastases) to 'poor' (high risk of metastases) and can also indicate patients that may most benefit from adjuvant treatment with endocrine therapy and chemotherapy. The risk is calculated using a number of factors, including the following: Disease stage (size of the tumour, number of positive lymph nodes): large tumours and positive lymph nodes are associated with a poorer prognosis. Grade of cancer cells (the extent to which the appearance of the cancer cells differs from that of normal cells): a higher grade is associated with a poorer prognosis. Hormone receptor (HR) expression and HER2 status of the cancer cells: HR-positive/ HER2-negative cancer has a better prognosis. A combination of the grade of the cancer cells and the expression of hormone receptor genes. The 4 most common molecular subtypes are: Luminal A, characterised by ER-positive and low grade and associated with good prognosis Luminal B, characterised by ER-positive but often high grade and associated with worse prognosis than Luminal A basal-like, characterised by ER, PR and HER2-negative (also called triple-negative breast cancer) and associated with poor prognosis HER2-enriched, characterised by HER2-overexpression and associated with poor prognosis. Technology overview This briefing describes the regulated use of the technology for the indication specified, in the setting described, and with any other specific equipment referred to. It is the responsibility of healthcare professionals to check the regulatory status of any intended use of the technology in other indications and settings. Page 5 of

About the technology CE marking The Prosigna breast cancer prognostic gene signature assay, the ncounter Dx Prep Station and the ncounter Dx Digital Analyzer (both parts of the ncounter Dx Analysis System) were CE-marked in October 2013 as Class I medical devices under the In Vitro Diagnostic Medical Devices Directive 98/79/EC. The Prosigna assay is manufactured by NanoString Technologies. Description The Prosigna assay is an in vitro diagnostic tool that uses information from gene expression to calculate the risk of women with HR-positive early-stage breast cancer developing distant recurrence 10 years from diagnosis. The assay measures the expression profiles of genes included in the PAM50 gene signature, a set of 50 genes used for classifying the 'intrinsic' subtypes of breast cancer (Parker et al. 2009), as well as 8 housekeeping genes (for normalisation), 6 positive controls and 8 negative controls. The Prosigna assay is performed on RNA isolated from formalin-fixed paraffin-embedded (FFPE) breast tumour tissue. A pathologist examines the histology slides taken from the tumour biopsy, identifying and marking all areas containing breast cancer. A trained technician then isolates the RNA from these areas using an RNA extraction kit. For the assay to be done, tumour tissue must meet the following criteria: invasive ductal/not otherwise specified, invasive lobular, or mixed histopathological breast cancer type 10-μm thick tissue sections mounted on slides 4 mm 2 tumour size 10% tumour cellularity (the proportion of cancer cells in a sample) 125 ng RNA (12.5 ng/μl) tumour area: 100 mm 2 for 1 slide, or 20 99 mm 2 for 3 slides, or 4 19 mm 2 for 6 slides. Analysis is done using the ncounter Dx Analysis (a companion system made by NanoString Technologies). The test results, which are provided using the ncounter Analysis System Services web interface, describe: Page 6 of

the intrinsic tumour subgroup (Luminal A/B, basal-like or HER2-enriched) the risk of recurrence (ROR) score, calculated based on the PAM50 analysis combined with a proliferation score and gross tumour size (the score is then adjusted to a 1 100 scale based on regression coefficients generated from a training set of breast tumour samples; see table 1) a risk category, calculated by combining information from the gene analysis and the nodal status. Table 1 Risk score classification algorithm using Prosigna Nodal status Node-negative Prosigna score range Risk classification 0 40 Low 41 60 Intermediate 61 100 High Node-positive (1 3 nodes) 0 15 Low 16-40 Intermediate 41 100 High Node-positive ( 4 nodes) 0 100 High In addition to the ncounter Dx Analysis system, the following single-use components are needed to use Prosigna: Formalin-fixed paraffin-embedded RNA isolation kit. The manufacturer recommends the Roche FFPET RNA Isolation Kit, which has been validated specifically for use with Prosigna, but other RNA isolation kits may be used if they meet the manufacturer's specifications. A Prosigna kit, which contains the following reagents: RNA reference sample Reporter CodeSet, containing a unique DNA probe sequence for RNA hybridisation and purification Capture ProbeSet, containing a unique DNA probe sequence for RNA hybridisation and purification Codeset barcode sticker Page 7 of

Prep pack, containing hybridisation buffer, 12-well notched strip tubes and lids, prep station tips, cartridge adhesive cover and tip sheaths ncounter cartridge(s) Prep plate, containing reagents necessary for post-hybridisation processing and immobilisation of the RNA. The ncounter Dx Analysis is a non-disposable component needed before Prosigna is used. It comprises the ncounter Dx Prep Station, which immobilises the CodeSet complexes for data collection based on fluidic technologies, and the ncounter Dx Digital Analyzer, which derives data by counting fluorescent barcodes. All of the above components are designed to be used in a molecular pathology laboratory. Intended use Prosigna is intended for the long-term prognosis of women with early-stage breast cancer who meet the following criteria: postmenopausal post-surgical (mastectomy or breast-conserving) hormone receptor-positive cancer node-negative (stage 1 or 2) node-positive (stage 2 or 3a). Setting and intended user Prosigna is intended for use in secondary care settings, specifically in hospitals with the equipment and expertise for gene expression analysis. The test would be requested by a clinician involved in managing breast cancer, most likely an oncologist, and is intended to be done by professionals trained in highly complex molecular biology techniques. Current NHS options Currently, the process for predicting the risk of disease recurrence and death in people with breast cancer involves the use of tools such as the Nottingham Predictive Index (NPI) or the online prognosis algorithm Adjuvant! Online. Page 8 of

The NPI uses information about tumour grade, lymph node involvement and tumour size to predict prognosis. The NPI is most commonly used to divide people into 5 prognostic groups ranging from good to poor prognosis. The Adjuvant! Online computer programme provides estimates of the potential benefits of adjuvant endocrine therapy and chemotherapy by taking into account a number of patient and tumour characteristics. These are the patient's age, disease stage, number of positive lymph nodes, tumour grade and other comorbidities. NICE guidance on early and locally advanced breast cancer: diagnosis and treatment recommends using Adjuvant! Online to better estimate risk of recurrence and the absolute benefit of adjuvant treatment for patients with early invasive breast cancer. The guidance also recommends that adjuvant therapy should be considered for all patients with early invasive breast cancer after surgery. The decision about adjuvant therapy should be made after considering all prognostic and predictive factors and following discussion with the patient about the potential benefits and side effects of the treatment. NICE diagnostic guidance on gene expression profiling and expanded immunohistochemistry tests for guiding adjuvant chemotherapy decisions in early breast cancer management recommends using Oncotype DX to guide chemotherapy decisions in patients with ER-positive, lymph node-negative and HER2-negative early-stage breast cancer. PREDICT has recently become available to the NHS, a tool similar to NPI and Adjuvant! Online which is based on cancer registry data for women treated in the UK and includes HER2 and Ki-67 status (a histology marker of proliferation). NICE is aware of the following CE-marked technologies that appear to fulfil a similar function to Prosigna: Oncotype DX (Genomic Health) MammaPrint (Agendia) EndoPredict (Sividon Diagnostics). Costs and use of the technology Information on the cost of using the technology has been provided by the manufacturer. Page 9 of

The ncounter Dx Analysis System costs 182,600 per unit, excluding VAT. The Prosigna kit costs 1362 per test, excluding VAT. This reference price is set in US dollars ($2080) and will fluctuate in line with the exchange rate. The RNA extraction kit costs 13 per test, excluding VAT. The maintenance cost is 13,600 per year for a 5-year contract and is charged from year 2. The anticipated lifespan of the ncounter Dx Analysis System is 5 years. The manufacturer estimates that each system can process up to 30 patient samples per 8-hour work day (240 work days annually). Using a standard annuity method with an equipment lifespan of 5 years and a discount rate of 3.5% (Drummond et al. 2005), along with the assumptions/costs (including maintenance cost) listed above, the average cost is estimated to be 1382 per test. The manufacturer provides training for up to 6 people as part of the system cost. Additional training is available at a cost of $10,000 ( 6840) for 1 4 trainees. The Adjuvant! Online, NPI and PREDICT tools are all free for use by the NHS. The Oncotype DX test is recommended in NICE diagnostics guidance in the context of a confidential pricing scheme agreed with NICE. No other practical difficulties have been identified in using or adopting the technology. Likely place in therapy Prosigna is likely to be used, together with the results of other investigations, in the planning of treatment for women with HR-positive early-stage breast cancer following surgery (mastectomy or breast-conserving). Specialist commentator comments One specialist commentator stated that the likely place in therapy for Prosigna remains undefined. It is unclear which treatment decisions will be informed by the result of this assay if it were adopted into current practice, especially since it cannot be used to guide decisions about adjuvant chemotherapy. The commentator also noted that the data analysis in the available clinical evidence is retrospective from subpopulations of 2 historical trials. Page 10 of

The same commentator reflected that there is no evidence that Prosigna is a clinically helpful test for patients with stage 1 disease. This is because it has yet to be proven that the test can guide decisions about the length of endocrine therapy in this population. Such evidence would come from comparing Prosigna directly with the standard of care in a prospective randomised trial, but this has not yet been done. One commentator noted that it is unclear if this test has clinical- or cost-effectiveness in determining the length of endocrine therapy, because there is no long-term clinical evidence to support this. They noted that Prosigna determines prognosis from years 5 to 10 after diagnosis, but the additional benefits of extended endocrine therapy only appear beyond this point. One specialist commentator noted that the role of Prosigna in guiding decisions about adjuvant chemotherapy needs further clarification because the test will likely be used for this purpose in clinical practice. They said that they would like to use Prosigna to inform individual discussions with patients about the potential benefits of chemotherapy. For example, patients with a low risk of recurrence may not benefit from adjuvant chemotherapy and will consequently have endocrine therapy alone. The same specialist commentator noted that multi-parameter assays like Prosigna and Oncotype Dx may well be used in patients with stage 2 disease, along with clinical parameters, to identify those with a good prognosis who could safely avoid chemotherapy. One specialist commentator noted that although currently Prosigna is indicated only for postmenopausal women, premenopausal women with non-hereditary breast cancer could potentially benefit from the technology in the future. According to 1 specialist commentator, Prosigna can potentially help to identify which patients scheduled to have endocrine treatment only are more likely to have disease recurrence after 5 years. They felt that this may at least be of value in encouraging treatment compliance and post-treatment monitoring in high-risk patients, and in selecting candidates for inclusion in clinical trials. If there were evidence that the assay could also predict site of recurrence, it could lead to reconsidering disease management in high-risk patients. For example, although at diagnosis older women are less likely to have metastases than younger women, when they do have metastases their prognosis is poorer (Purushotham et al. 2014). One specialist commentator noted that although the results from the study by Nielsen et al. (2014) on analytical validity are reassuring, they do not provide any evidence for clinical utility which is the key question for a test with such a high initial cost. Until data on Prosigna's clinical utility, cost effectiveness and impact on quality of life are provided, the commentator felt that there is no justification for adopting it into clinical use. Another specialist commentator stated that although Page 11 of

Nielsen et al. state that the inter-laboratory assay variability was within acceptable limits, the capital equipment costs compared with the annual test ratio may limit the assay to a centralised service. Equality considerations NICE is committed to promoting equality and eliminating unlawful discrimination. We aim to comply fully with all legal obligations to: promote race and disability equality and equality of opportunity between men and women eliminate unlawful discrimination on grounds of race, disability, age, sex, gender reassignment, pregnancy and maternity (including women post-delivery), sexual orientation, and religion or belief, in the way we produce our guidance (these are protected characteristics under the Equality Act 2010). Prosigna is intended for women with breast cancer only and men are excluded from its use. Gender and cancer (a chronic condition) are both considered protected characteristics as defined by the Equality Act 2010. Women of white heritage and those aged over 50 are at increased risk of breast cancer. Gender, race and age are protected characteristics defined in the Equality Act 2010. Evidence review Clinical and technical evidence Regulatory bodies A search of the Medicines and Healthcare Products Regulatory Agency (MHRA) website revealed no manufacturer Field Safety Notices or Medical Device Alerts for this equipment. No reports of adverse events were identified from searches of the US Food and Drug Administration (FDA) Manufacturer and User Device Facility Experience (MAUDE) database. Clinical evidence 7 studies on Prosigna were identified. Three of these studies were excluded from this briefing because of overlapping patient cohorts (Sestak et al. 2013, Filipts et al. 2014) and inappropriate patient populations (Jorgensen et al. 2014). Abstracts were not considered for inclusion in the briefing. Three studies on the clinical validity of Prosigna and 1 study on the analytical validity of Prosigna were included in this briefing. Page 12 of

Analytical validity Nielsen et al. (2014) studied the analytical validity of Prosigna across 3 laboratories in North America (1 in Canada and 2 in the USA). The authors used 43 FFPE breast tumour blocks archived from the tissue bank at Washington University in St. Louis, Missouri. The study tested the precision and reproducibility of the Prosigna assay using the ncounter Dx Analysis System, following the EP05-A2 protocol (a protocol from the Clinical and Laboratory Standards Institute for the evaluation of in-vitro diagnostic devices; Chesher 2008). A total standard deviation (SD) of less than 4.3 risk of recurrence (ROR) units was defined as 'highly reproducible'. The measured SD for precision and reproducibility was less than 1 ROR unit and 2.9 ROR units respectively. Assay results were stable in the presence of moderate amounts of surrounding non-tumour tissue (less than 70% of non-tumour tissue by area). The authors concluded that the precision and reproducibility of the Prosigna assay, estimated across multiple testing sites, was similar to that previously reported for centralised lab tests. Clinical validity Dowsett et al. (2013) studied the ROR in postmenopausal women (n=1017) with ER-positive early-stage breast cancer treated with anastrozole or tamoxifen as part of a randomised controlled trial (ATAC). Patients in the trial had either anastrozole or tamoxifen as monotherapy for 5 years. The median follow-up was 10 years. The authors investigated whether ROR can add prognostic information to clinicopathologic variables and whether it can predict the risk of disease recurrence after endocrine therapy better than the recurrence score provided by Oncotype DX and the IHC4 score. According to their results, ROR added significant prognostic information beyond the clinical treatment score in all women analysed and after subdividing them into 4 groups (the clinical treatment score contains information on nodal status, tumour size, grade, age, and treatment). The ROR score derived using Prosigna provided more prognostic information than that derived using Oncotype DX, and its prognostic ability was as equal to that of IHC4 in all but the HER2-negative/ node-negative group. In addition, more patients were considered as high risk and fewer as intermediate risk than by Oncotype DX (table 3). Gnant et al. (2014) studied the ROR in postmenopausal women (n=1478) who had taken part in another randomised controlled trial (ABCSG-8). These women had HR-positive early-stage breast cancer and were randomized to have 2 years of adjuvant tamoxifen followed by 3 years of anastrozole or 5 years of adjuvant tamoxifen. The median follow-up was 10 years. The authors investigated whether ROR can add prognostic information to clinicopathologic variables. According to their results, ROR added significant prognostic information beyond the Page 13 of

clinicopathologic variables in all women analysed, and after subdividing them into 8 groups (table 4). Sestak et al. (2014) investigated the ability of Prosigna to predict the occurrence of metastases 5 years from diagnosis, also using data and tissue samples from the ABCSG-8 trial. The authors used long-term follow-up data and tissue samples from 2137 postmenopausal women with HR-positive, early-stage breast cancer from the ABCSG 8 (n=1275) and ATAC (n=862) multicentre randomised controlled trials. Only patients who had 5 years of endocrine treatment and had not developed metastases at that stage were included in the analysis. The median follow-up was 10 years. The prognostic ability of the ROR score was also compared with the information provided by the clinical treatment score. Although the clinical treatment score was the strongest prognostic factor at 5 years after diagnosis, the ROR score was the strongest at 5 to 10 years after diagnosis (table 5). Table 2 Overview of the Nielsen et al. (2014) multicentre technical study Study component Objectives/ hypotheses Study design Setting Description To validate the analytical performance of Prosigna on the ncounter Dx Digital Analyzer across multiple laboratories. Prospectively planned retrospective multicentre validation study based on the EP05-A2 protocol. 3 centres in North America (1 in Canada and 2 in USA). Page 14 of

Inclusion/ exclusion criteria Inclusion criteria: Every case should represent a unique breast cancer patient. All must be primary breast cancers. All are pathology confirmed invasive ductal or lobular carcinoma, a mixtures of these types, or classified as no special type. All are hormone-receptor positive breast cancer. All must have a recorded tumour size. FFPE blocks should be <10 years old. A minimum of 10 cases each of 100 mm 2 tumour area (1 slide per test) and 4 100 mm 2 tumour area (3 slides per test). The archived tissue samples were obtained from the tissue bank at Washington University in St. Louis, Missouri. Primary outcomes Statistical methods Precision and reproducibility. The variance components model used to characterize the sources of variability was calculated based on the formulas: ROR score = site + operator + lot + run + within-run for precision. Measurement = FFPE block + site + tissue section + error for reproducibility. Variance components were estimated using the R procedure 'lmer'. Participants 43 FFPE breast tumour blocks from tissue bank at Washington University in St. Louis, Missouri, were used in the analysis. Results Precision: SD was less than 1 ROR unit. Reproducibility: total SD was 2.9 ROR units. The ROR scores for RNA inputs at the extremes of the range were the same as those at the nominal input level. Assay results were stable in the presence of moderate amounts of surrounding non-tumour tissue (<70% by area). Page 15 of

Conclusions The precision and reproducibility of the Prosigna assay, estimated across multiple testing sites is similar (relative to the overall test range) to what was previously reported for centralized lab tests. Abbreviations: CI, confidence interval; FFPE, formalin-fixed, paraffin-embedded; ITT, intention to treat; n, number of patients; ROR, risk of recurrence; RR, relative risk; SD, standard deviation. Table 3 Overview of the Dowsett et al. (2013) prospective-retrospective e multicentre study Study component Objectives/ hypotheses Study design Setting Inclusion/ exclusion criteria Description To determine whether the ROR score can add prognostic information to clinicopathologic variables and whether it can predict the risk of disease recurrence better than Oncotype DX and the IHC4 score. Prospective-retrospective multicentre study. Patients enrolled between 1998 and 1999 in the ATAC trial. Median follow-up was 10 years. Postmenopausal women with hormone receptor-positive early-stage breast cancer with FFPE tissue samples available recruited as part of the ATAC randomised controlled trial. In addition patients must: not have had chemotherapy have Oncotype DX analysis available have sufficient residual RNA for PAM50 analysis. Primary outcomes Statistical methods Prognostic performance of distant recurrence in years 0 to 10 in all patients according to ROR. Cox proportional hazard regression models were fitted to time from random assignment to first disease recurrence. Hazard ratios, and 95% CIs were estimated to study the association between ROR score and distant recurrence after 10 years of follow-up. Page 16 of

Participants 1017 postmenopausal women with early-stage breast cancer from the ATAC randomised controlled trial. Mean (±SD) age was 64 (±8.3) years. 65% of the women had T1-stage cancer and 73% had node-negative disease. Women from the ATAC trial had either anastrozole or tamoxifen given for 5 years as monotherapy. Results ROR (including tumour size) added significant prognostic information beyond the CTS in all women analysed (LR-Δχ 2 : 33.9 p<0.001) and after subdividing them into 4 groups of node-negative (LR-Δχ 2 : 24.6 p<0.001), node-positive (LR-Δχ 2 : 9.3 p=0.002), HER2-negative (LR-Δχ 2 : 28 p<0.001), and HER2-negative/ node-positive (LR-Δχ 2 : 23.4 p<0.001). Prosigna ROR provided more prognostic information (LR-χ 2 : 99.9 p<0.001) in comparison to RS from Oncotype DX (LR-χ 2 : 38.2 p<0.001). More patients were scored as high risk and fewer as intermediate risk by ROR than by Oncotype DX. Prosigna's prognostic ability was equal to IHC4 in all but one group (HER2-/node-), where the ROR added more information (LR-Δχ 2 =22.4, p<0.001) than IHC4 (LR-Δχ 2 =13.6, p<0.001). Conclusions The ROR score added clinically meaningful prognostic information to the clinical treatment score in all patients and all subgroups at 10 years after diagnosis. Abbreviations: CI, confidence interval; CTS, clinical treatment score; FFPE, formalin-fixed, paraffin-embedded; HR, hazard ratio; ITT, intention to treat; LR-χ 2, likelihood ratio chi-squared; LR-Δχ 2, changes in likelihood ratio chi-squared; n, number of patients; RR, relative risk; ROR, risk of recurrence; SD, standard deviation. Table 4 Summary of the Gnant et al. (2014) prospective-retrospective e multicentre study Study component Objectives/ hypotheses Study design Description To determine whether the ROR score can add prognostic information to clinicopathologic variables. Prospective-retrospective multicentre study. Page 17 of

Setting Inclusion/ exclusion criteria Primary outcomes Statistical methods Patients enrolled between 1996 and 2004 in the ABCSG-8 trial. Median follow-up was 11 years. Postmenopausal women with hormone receptor-positive early-stage breast cancer with FFPE tissue samples available, recruited as part of the ABCSG-8 randomised controlled trial. Prognostic performance of distant recurrence in years 0 to 10 in all patients according to ROR over and above standard clinical variables. Cox proportional hazards models were used to assess the effects of individual prognostic factors, a combined linear predictor (clinical linear predictor), ROR score, ROR score-derived risk groups, and intrinsic subtypes; HRs with 95% CIs were also estimated to study the association between ROR score and distant recurrence at 10 years of follow-up. Participants 1478 postmenopausal women with early-stage breast cancer from the ABCSG-8 randomised controlled trial. Median age was 63 years (range: 41 79 years). 70% had T1-stage cancer and 71% had node-negative disease. The ABCSG-8 trial randomized post-menopausal women with HR-positive early-stage breast cancer to 2 years of tamoxifen followed by 3 years of anastrozole, or 5 years of tamoxifen. Results In all tested subgroups, ROR score adds statistically significant prognostic information to the clinical predictors (p<0.0001). PAM50 assigns an intrinsic subtype to all cases, and the Luminal A cohort had a significantly lower ROR at 10 years compared with Luminal B (p<0.0001). Significant and clinically relevant discrimination between low and high risk groups occurred also within all tested subgroups. Conclusions The ROR score and the ROR-based risk groups added clinically meaningful prognostic information to the clinical linear predictor in all patients and all subgroups at 10 years after diagnosis. Abbreviations: CI, confidence interval; CTS, clinical treatment score; FFPE, formalin-fixed, paraffin-embedded; HR, hazard ratio; ITT, intention to treat; n, number of patients; RR, relative risk; ROR, risk of recurrence; SD, standard deviation. Page 18 of

Table 5 Summary of the Sestak et al. (2014) prospective-retrospective e multicentre study Study component Objectives/ hypotheses Study design Setting Inclusion/ exclusion criteria Primary outcomes Statistical methods Description To determine whether the ROR score provides prognostic information in the period beginning 5 years after diagnosis. Prospective-retrospective multicentre study. Patients enrolled between 1996 and 2004 in the ABCSG-8 trial and between 1998 and 1999 in the ATAC trial. Median follow-up was 10 years. This study included postmenopausal women with hormone receptor-positive early-stage breast cancer with FFPE tissue samples available recruited as part of the ABCSG-8 and ATAC randomised controlled trials. Prognostic performance of distant recurrence in years 5 to 10 in all patients according to ROR. Kaplan Meier analysis derived from Cox proportional hazards regression models with associated 95% CIs was used to study the association between ROR score and distant recurrence after 5 years of follow-up. Participants 862 and 1275 postmenopausal women with early-stage breast cancer from the ATAC and ABCSG-8 randomised controlled trials respectively. Women from the ATAC trial had either anastrozole or tamoxifen given for 5 years as monotherapy. The ABCSG-8 trial randomized post-menopausal women with HR-positive early-stage breast cancer to either 2 years of tamoxifen followed by 3 years of anastrozole or 5 years of tamoxifen. Page 19 of

Results Women categorised into the high-risk group had 16.6% (95% CI: 13.1% to 20.9%) risk of distant recurrence in years 5 to 10. Those in the intermediate-risk group had a risk of 8.3% (95% CI 6.1% to 11.2%), and those in the low-risk group had a risk of 2.4% (95% CI 1.6% to 3.5%). The ROR score was significantly prognostic by itself (univariable HR: 2.69, 95% CI 2.12 to 3.43, p<0.001) and when it was added to the CTS (bivariable HR: 2.07, 95% CI 1.63 to 2.64, p<0.001) in years 5 to 10 after diagnosis. In the node-negative subgroup the ROR score added more prognostic information in bivariable analysis (HR: 2.11, 95% CI 1.48 to 3.00, p<0.001) compared with the CTS (HR: 1.56, 95% CI 1.15 to 2.12, p>0.001). In the node-negative/human epidermal growth factor receptor 2-negative subgroup, the ROR score added more prognostic information in bivariable analysis (HR: 2.41, 95% CI 1.65 to 3.50, p<0.001) compared with the CTS (HR: 1.65, 95% CI: 1.21 to 2.24, p>0.001). Conclusions The ROR score added clinically meaningful prognostic information to the clinical treatment score in all patients and all subgroups in years 5 to 10 after diagnosis. Abbreviations: CI, confidence interval; CTS, clinical treatment score; FFPE, formalin-fixed, paraffin-embedded; HR, hazard ratio; n, number of patients; ROR, risk of recurrence; RR, relative risk; SD, standard deviation. Recent and ongoing studies Four ongoing or in-development trials on Prosigna were identified in the preparation of this briefing: NCT01974856: a prospective observational study based in Germany investigating the effect of the test on the choice of treatment in patients with early stage breast cancer, funded by the manufacturer. This is an ongoing study with estimated completion date in April 2015. NCT01899079: a prospective observational study based in Spain investigating the effect of the test on the choice of treatment in patients with early stage breast cancer, funded by the manufacturer. This is an ongoing study which was originally planned for completion in July 2014. NCT02213042: a multicentre randomised controlled trial based in Spain evaluating the changes in multiple biomarkers (including Prosigna) identified based on pre-treatment biopsy Page 20 of

and disease progression biopsy. This study is sponsored by GlaxoSmithKline. This is an ongoing study with estimated completion date in August 2018. ISRCTN42400492: a multicentre randomised controlled trial (referred to as OPTIMA) based in the UK to validate the use of a multi-parameter assay including Prosigna for selection of chemotherapy in hormone-sensitive early breast cancer. The preliminary phase of this study is ongoing with estimated completion date in April 2015. The main phase of this study is due to start in 2015 and is predicted to run for approximately 7 years. Costs and resource consequences If Prosigna were to be adopted in the NHS, it is likely to be used in patients with stage 1 breast cancer. Cancer Research UK statistics report about 50,285 new cases of breast cancer in the UK in 2011; data on staging (for 1 English region) suggest that 41% of new diagnoses in women are stage 1, and approximately 70% have hormone-positive cancer (Cancer Research UK, 2014a). The estimated eligible population is, therefore, about 14,431 patients per year. The proportion of women with stage 2 and 3a breast cancer likely to be offered Prosigna if it were adopted, is smaller, because in current practice all patients with node-positive disease would be offered chemotherapy (unless they are unfit). According to the manufacturer, Prosigna is currently available at 1 NHS Hospital, the Royal Marsden NHS Foundation Trust. It will be available from additional NHS sites in 2015 as part of the planned OPTIMA trial. No published evidence on resource consequences of Prosigna was identified in the systematic review of evidence. Strengths and limitations of the evidence The evidence for clinical validity for Prosigna is based on the prospective-retrospective analysis of archived material from 2 multicentre randomised controlled trials (ATAC and ABCSG-8). These trials had long-term follow-up for a median of 10 years and included women with early-stage breast cancer at low and high risk of developing metastases. The gold standard for validating a prognostic score would be in a prospective clinical trial, in which the test itself is the intervention under investigation. However, specimens collected in the context of randomised clinical trials and stored for future use are the best source of specimens for retrospective technical and prognostic studies (Simon et al. 2009, McShane 2012). In addition, Nielsen et al. (2014) used a validated protocol for assessing the precision and reproducibility of the Prosigna assay. The experimental design was based on Clinical Laboratory Page 21 of

and Standards Institute guidelines for the evaluation of the precision of in vitro diagnostic devices, outlined in EP05-A2. In addition, the study was carried out in 3 independent laboratories (rather than satellite sites where systems are very similar). This provides strong evidence that the assay is robust as long as the extracted RNA quality meets minimum standards. In the material from the ATAC study, the Prosigna assay was used on samples which had previously had tumour micro-dissection and RNA extraction for a study analysing Oncotype DX. As a result, extrapolating from these data to real-world clinical settings could be challenging (Zanotti et al. 2014). In addition, none of the women analysed by Gnant et al. (2014), Sestak et al. (2014) and Dowsett et al. (2013) had chemotherapy as part of their initial treatment. As a result, the prognostic value of the ROR score in a chemotherapy-treated population is unknown. Sestak et al. (2014) combined data previously analysed by Dowsett et al. (2013) and Gnant et al. (2014). As a result, there is some overlap between their patient cohort and those of the other studies. Dowsett et al. (2013) and Sestak et al. (2014) used the clinical treatment score as a comparator rather than the online tools Adjuvant! Online and PREDICT, or the NPI, which are standard practice in the UK. Similarly, Gnant et al. (2014) used a combination score of clinicopathologic parameters as the comparator for Prosigna. Such indices are always incomplete because they may not include all parameters used by clinicians in other health systems to aid clinical decision-making. Lastly, all included studies received financial support or disclosed competing interests from the manufacturer, and this introduces the potential for bias in the reporting of outcomes. Relevance to NICE guidance programmes NICE has issued the following guidance: Early and locally advanced breast cancer: Diagnosis and treatment (2009) NICE guideline CG80 Breast cancer (2011) NICE quality standard 12 Early and locally advanced breast cancer (2011) NICE pathway Familial breast cancer (2011) NICE pathway Intraoperative tests (RD-100i OSNA system and Metasin test) for detecting sentinel lymph node metastases in breast cancer (2013) NICE diagnostics guidance 8 Page 22 of

Gene expression profiling and expanded immunohistochemistry tests for guiding adjuvant chemotherapy decisions in early breast cancer management: MammaPrint, Oncotype DX, IHC4 and Mammostrat (2013) NICE diagnostics guidance 10 Advanced breast cancer (update): Diagnosis and treatment (2014) NICE guideline CG81 References Chesher D (2008) Evaluating Assay Precision. The Clinical Biochemist Reviews (Suppl 1): S23 6 Cancer Research UK (2014a) Breast cancer incidence statistics (accessed December 2014) Cancer Research UK (2014b) Definite breast cancer risks (accessed December 2014) Cancer Research UK (2014c) Breast cancer symptoms (accessed December 2014) Cancer Research UK (2014d) Breast cancer in men (accessed January 2015) Cancer Research UK (2014e) Types of treatment for breast cancer (accessed January 2015) Cuzick J, Dowsett M, Pineda S et al. (2011) Prognostic value of a combined estrogen receptor, progesterone receptor, Ki-67, and human epidermal growth factor receptor 2 immunohistochemical score and comparison with the Genomic Health recurrence score in early breast cancer. Journal of Clinical Oncology (32): 4273 8 Dowsett M, Sestak I et al. (2013) Comparison of PAM50 risk of recurrence score with oncotype DX and IHC4 for predicting risk of distant recurrence after endocrine therapy. Journal of Clinical Oncology 31(22): 2783 90 Drummond M, Sculpher M, Torrance G, O'Brien BJ, Stoddart GL (2005) Methods for economic health valuation of health care programmes. 3rd ed. Oxford University Press Filipits M, Nielsen TO et al. (2014) The PAM50 risk-of-recurrence score predicts risk for late distant recurrence after endocrine therapy in postmenopausal women with endocrine-responsive early breast cancer. Clinical Cancer Research 20(5): 18 305 Gnant M, Filipits M et al. (2014) Predicting distant recurrence in receptor-positive breast cancer patients with limited clinicopathological risk: Using the PAM50 Risk of Recurrence score in 1478 Page 23 of

postmenopausal patients of the ABCSG-8 trial treated with adjuvant endocrine therapy alone. Annals of Oncology 25(2): 339 45 Jorgensen CLT, Nielsen TO et al. (2014) PAM50 breast cancer intrinsic subtypes and effect of gemcitabine in advanced breast cancer patients. Acta Oncologica 53(6): 776 87 McShane LM (2012) Statistical challenges in the development and evaluation of marker-based clinical tests. BMC Medicine 10: 52 NHS choices (2013) Breast cancer in men (accessed January 2015) Nielsen T, Wallden B et al. (2014) Analytical validation of the PAM50-based Prosigna Breast Cancer Prognostic Gene Signature Assay and ncounter Analysis System using formalin-fixed paraffin-embedded breast tumor specimens. BMC Cancer 14: 177 Parker JS, Mullins M et al. (2009) Supervised risk predictor of breast cancer based on intrinsic subtypes. Journal of Clinical Oncology 27(8): 1160 7 Purushotham A, Shamil E, Cariati M et al. (2014) Age at diagnosis and distant metastasis in breast cancer - a surprising inverse relationship. European Journal of Cancer 50(10): 1697 705 Sestak I, Dowsett M et al. (2013) Factors predicting late recurrence for estrogen receptor-positive breast cancer. Journal of the National Cancer Institute 105(19): 1504 11 Sestak I, Cuzick J et al. (2014) Prediction of Late Distant Recurrence After 5 Years of Endocrine Treatment: A Combined Analysis of Patients From the Austrian Breast and Colorectal Cancer Study Group 8 and Arimidex, Tamoxifen Alone or in Combination Randomized Trials Using the PAM50 Risk of Recurrence Score. Journal of Clinical Oncology, epub ahead of print. Simon RM, Paik S et al. (2009) Use of archived specimens in evaluation of prognostic and predictive biomarkers. Journal of the National Cancer Institute 101(21): 1446 52 Ward S, Scope A, Rafia R, Pandor A, Harnan S, Evans P, Wyld L (2013) Gene expression profiling and expanded immunohistochemistry tests to guide the use of adjuvant chemotherapy in breast cancer management: A systematic review and cost-effectiveness analysis. Health Technology Assessment 17(44): V-302 Page 24 of

Zanotti L, Bottini A et al. (2014) Diagnostic tests based on gene expression profile in breast cancer: from background to clinical use. Tumour Biology 35(9): 8461 70. Search strategy and evidence selection Search strategy Embase 1980 to 2014 Week 48, Ovid MEDLINE(R) In-Process & Other Non-Indexed Citations and Ovid MEDLINE(R) 1946 to Present; searched 1 December 2014 For clinical evidence: 1. prosigna.mp. 2. PAM50.mp. 3. nanostring.mp. 4. single tube multiplex.mp. 5. ncounter.mp. 6. 1 or 2 or 3 or 4 or 5 7. (breast adj3 cancer).mp. [mp=title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier] 8. Breast Neoplasms 9. 7 or 8 10. 6 and 9 11. limit 10 to (english language and humans and yr="2003 -Current") The CRD database was searched using the following keywords: Page 25 of

Any field/prosigna OR Any filed/pam50 OR Any field/ncounter For economic evidence: 1. prosigna.mp. 2. PAM50.mp. 3. nanostring.mp. 4. single tube multiplex.mp. 5. ncounter.mp. 6. 1 or 2 or 3 or 4 or 5 7. (breast adj3 cancer).mp. 8. Breast Neoplasms/ 9. 7 or 8 10. cost$.mp. [mp=ti, ab, sh, hw, tn, ot, dm, mf, dv, kw, nm, kf, px, rx, ui] 11. economic$.mp. [mp=ti, ab, sh, hw, tn, ot, dm, mf, dv, kw, nm, kf, px, rx, ui] 12. 10 or 11 13. 6 and 9 and 12 14. from 13 keep 1-15 15. limit 14 to yr="2003 -Current" 16. limit 15 to english language Page 26 of

17. remove duplicates from 16 Cochrane Database of Systematic Reviews: Issue 12 of 12, December 2014. Cochrane Central Register of Controlled Trials: Issue 11 of 12, November 2014. Database of Abstracts of Reviews of Effect: Issue 4 of 4, October 2014. Health Technology Assessment Database: Issue 4 of 4, October 2014. NHS Economic Evaluation Database: Issue 4 of 4, October 2014. 1. prosigna 2. pam50 3. ncounter 4. 1 or 2 or 3 5. cost$ 6. economic$ 7. 5 or 6 8. 4 and 7 Evidence selection Evidence was selected based on the following criteria: Population: women with early breast cancer Intervention: Prosigna gene expression profiling test Comparator: Adjuvant! Online, Nottingham Prognostic Index Outcomes: analytical validity and clinical validity. For the clinical evidence, a total of 254 publications were retrieved. After excluding abstracts, case studies, editorials, letters, reviews, animal studies, and non-english language studies, 52 were Page 27 of

selected to be read in full. Four publications were included in this briefing after the following studies were excluded for the reasons outlined below: Sestak et al. (2013) and Filipits et al. (2014) due to overlapping cohorts with Sestak et al. (2014) Jorgensen et al. (2014) as the population was women with advanced breast cancer For the economic evidence, 13 publications were retrieved and 2 were selected to be read in full. None was included in this briefing. Changes after publication October 2015: Minor maintenance About this briefing Medtech innovation briefings summarise the published evidence and information available for individual medical technologies. The briefings provide information to aid local decision-making by clinicians, managers, and procurement professionals. Medtech innovation briefings aim to present information and critically review the strengths and weaknesses of the relevant evidence, but contain no recommendations and are not formal NICE guidance. Development of this briefing This briefing was developed for NICE by King's Technology Evaluation Centre (KiTEC). The interim process and methods statement sets out the process NICE uses to select topics, and how the briefings are developed, quality assured and approved for publication. Project team King's Technology Evaluation Centre (KiTEC), King's Health Partners Medical Technologies Evaluation Programme, NICE Peer reviewers and contributors Anastasia Chalkidou, Senior Health Technology Assessor, KiTEC Page 28 of

Muralikrishnan R. Kartha, Senior Health Economist, KiTEC Stephen Keevil, Director, KiTEC Cornelius Lewis, Director, KiTEC Viktoria McMillan, Centre Manager, KiTEC Specialist commentators The following specialist commentators provided comments on a draft of this briefing: Dr Cheryl Gillett, Head of Tissue Banking for Health Schools, King's College London Dr Iain MacPherson, Clinical Senior Lecturer in Medical Oncology, University of Glasgow Dr Luke Hughes-Davies, Consultant Oncologist, Cambridge University Hospitals Copyright National Institute for Health and Care Excellence 2015. All rights reserved. NICE copyright material can be downloaded for private research and study, and may be reproduced for educational and not-for-profit purposes. No reproduction by or for commercial organisations, or for commercial purposes, is allowed without the written permission of NICE. ISBN: 978-1-4731-1077-9 Page of