NATIONAL INSTITUTE FOR HEALTH AND CLINICAL EXCELLENCE

Transcription

1 NATIONAL INSTITUTE FOR HEALTH AND CLINICAL EXCELLENCE MERCK SERONO LTD. SINGLE TECHNOLOGY APPRAISAL SUBMISSION: ERBITUX (CETUXIMAB) FOR THE FIRST-LINE TREATMENT OF METASTATIC COLORECTAL CANCER

2 Contents Section Page number Description of technology under assessment 3 Specification of the decision problem 6 Executive summary 8 Context 18 Clinical evidence 23 Cost effectiveness 58 Assessment of factors relevant to the NHS and other parties 117 References 123 Appendices

3 Section A Manufacturers and sponsors will be requested to submit section A in advance of the full submission (for details on timelines, see the Guide to the single technology appraisal process A (draft) Summary of Product Characteristics (SPC) for pharmaceuticals and a (draft) technical manual for devices should be provided (see appendix 1, section 9.1). Description of technology under assessment 1.1 Give the brand name, approved name and, where appropriate, therapeutic class. For devices please provide details of any different versions of the same device. Brand name: Erbitux Approved name: cetuximab Therapeutic class: Antineoplastic agents, monoclonal antibodies 1.2 Does the technology have a UK marketing authorisation/ce marking for the indications detailed in this submission? If so, please give the date on which authorisation was received. If not, please state current UK regulatory status, with relevant dates (for example, date of application and/or expected approval dates). No: This is subject to CHMP/EMEA opinion. We anticipate a CHMP opinion by Q2/Q if no major objections occur. 1.3 What are the (anticipated) indication(s) in the UK? For devices, please provide the (anticipated) CE marking, including the indication for use Scientific assessment by the CHMP/EMEA is ongoing; the Company is not in the position to anticipate any decision on the indication by the EMEA. However at the current time the licence is anticipated to include the following cetuximab in combination with chemotherapy is indicated for the treatment of patients with EGFRexpressing metastatic colorectal cancer with KRAS (Kirsten rat sarcoma) wild-type tumour 1.4 To what extent is the technology currently being used in the NHS for the proposed indication? Include details of use in ongoing clinical trials. If the technology has not been launched, please supply the anticipated date of availability in the UK It is our understanding that the technology is not currently used in the NHS for the proposed indication (first-line metastatic colorectal cancer) outside of clinical trials. However it is licensed in the UK as follows in metastatic colorectal cancer and locally advanced squamous cell carcinoma of the head and neck; - 3 -

4 Erbitux in combination with irinotecan is indicated for the treatment of patients with epidermal growth factor receptor (EGFR)-expressing metastatic colorectal cancer after failure of irinotecan-including cytotoxic therapy. Erbitux in combination with radiation therapy is indicated for the treatment of patients with locally advanced squamous cell cancer of the head and neck. NICE have appraised Erbitux in both aforementioned indications and published guidance for the treatment of patients with metastatic colorectal cancer after failure of irinotecan in Technology Appraisal Guidance 118. The appraisal of Erbitux in the treatment of locally advanced squamous cell carcinoma of the head and neck is ongoing. 1.5 Does the technology have regulatory approval outside the UK? If so, please provide details. Yes, as of February 2008, cetuximab is approved in 68 countries for the treatment of irinotecan-resistant metastatic colorectal cancer and 61 countries for the treatment of locally advanced squamous cell carcinoma of the head and neck (SCCHN). 1.6 Is the technology subject to any other form of health technology assessment in the UK? If so, what is the timescale for completion? We plan to make a submission for cetuximab for the 1 st line treatment of metastatic colorectal cancer to the Scottish Medicines Consortium (SMC) by Q3/Q depending on CHMP/EMEA opinion. The NICE appraisal of cetuximab for the treatment of locally advanced head and neck cancer is ongoing. Guidance is expected in June For pharmaceuticals, what formulation(s) (for example, ampoule, vial, sustained-release tablet, strength(s) and pack size(s) will be available? The following vial sizes are available: 100mg/20ml vial and 500mg/100ml vial. 1.8 What is the proposed course of treatment? For pharmaceuticals, list the dose, dosing frequency, length of course and anticipated frequency of repeat courses of treatment. It is expected that cetuximab plus 5-fluorouracil/folinic acid/ irinotecan (FOLFIRI) or cetuximab plus 5-fluorouracil/folinic acid/ oxaliplatin (FOLFOX) will be used as detailed in the following regimen: 400 mg cetuximab per m 2 body surface area day 1. All subsequent weekly doses of cetuximab are 250 mg/m 2. Cetuximab treatment is continued until progression of the underlying disease. The FOLFIRI and FOLFOX regimens are given in Tables A1 and A What is the acquisition cost of the technology (excluding VAT)? For devices, provide the list price and average selling price. If the unit cost of the technology is not yet known, please provide details of the anticipated unit cost, including the range of possible unit costs /20ml (100mg)vial /100ml (500mg) vial - 4 -

5 1.10 What is the setting for the use of the technology? Treatment for patients with metastatic colorectal cancer who are unsuitable for complete resection and have not received previous chemotherapy in the metastatic setting For patients being treated with this technology, are there any other aspects that need to be taken into account? For example, are there additional tests or investigations needed for selection, or particular administration requirements, or is there a need for monitoring of patients over and above usual clinical practice for this condition? What other therapies, if any, are likely to be administered at the same time as the intervention as part of a course of treatment? It is expected that tumour samples will need to be tested for EGFR-expression and KRAS status prior to selection for treatment. Premedication: Prior to the first infusion, patients must receive pre-medication with an antihistamine. This pre-medication is recommended prior to all subsequent infusions. Close monitoring is required during the infusion and for at least 1 hour after the end of the infusion. It is likely that cetuximab will be given concurrently with the FOLFIRI regimen in this setting as follows; Table A1 FOLFIRI regimen FOLFIRI Chemotherapy given every 14 days (1 hour after cetuximab infusion) Irinotecan 30 to 90-min infusion on day 1 Folinic acid 120-min infusion on day mg/m² 400 mg/m² (racemic) or 200 mg/m² (L-form) before 5-FU bolus injection 5-fluorouracil day mg/m² bolus on day 1 followed by a 46-hour continuous infusion of 2400 mg/m² - 5 -

6 Alternatively cetuximab may be given with the FOLFOX regimen as described below; Table A2 FOLFOX regimen FOLFOX chemotherapy given every 14 days Oxaliplatin Infusion over 120 minutes Folinic acid Infusion over 120 minutes Day 1 Day 2 (1 hour after cetuximab infusion) 85 mg/m mg/m² 200 mg/m² 5-FU Bolus (over 2 4 minutes) followed by a 22-hour infusion 400 mg/m² bolus, followed by 22- hour continuous infusion of 600 mg/m² 400 mg/m² bolus, followed by 22- hour continuous infusion of 600 mg/m² Population 2 Statement of the decision problem In this section the manufacturer or sponsor should specify the decision problem that the submission addresses. The decision problem should be derived from the final scope issued by NICE and should state the key parameters that the information in the Evidence Submission will address. Table A3 Decision problem Final scope issued by NICE People with untreated metastatic colorectal cancer Decision problem addressed in the submission Subject to CHMP/EMEA decision on the indication, the population is anticipated to be Patients with untreated epidermal growth factor receptor (EGFR)- expressing metastatic colorectal cancer with KRAS wild type status Intervention Cetuximab in combination with chemotherapy Cetuximab in combination with fluoropyrimidine-based chemotherapy with irinotecan or oxaliplatin-containing regimens. Comparator(s) oxaliplatin-including chemotherapy regimens irinotecan-including chemotherapy regimens 5-fluorouracil and folinic acid (5-FU/FA) (including oral analogues; capecitabine and tegafur with uracil) FOLFIRI and FOLFOX are considered appropriate comparators to cetuximab plus chemotherapy as the population most likely to receive intensive treatment in the first line setting is the same. i.e. fitter patients. Although 5-FU single agent therapy is detailed in the scope for this appraisal, it is apparent that the characteristics of patients considered suitable for 1 st -line combination chemotherapy differs from the population who are likely to receive fluoropyrimidine monotherapy

7 Outcomes Economic Analysis Special considerations and other issues Survival Progression-free survival Time to treatment failure Tumour response rate Resection rates of metastases Health-related quality of life Adverse effects of treatment The reference case stipulates that the cost effectiveness of treatments should be expressed in terms of incremental cost per quality-adjusted life year. The time horizon for the economic evaluation should be based on a time period over which costs and benefits can reasonably be expected to accrue. Costs will be considered from an NHS and Personal Social Services perspective. It is anticipated that individuals receiving interventions first-line may receive subsequent treatment within or outside clinical trials. They may also undergo subsequent surgical interventions following first line treatment. Any confounding that results from this should be taken into account if possible particularly with respect to choice of endpoint. If evidence allows the appraisal should consider the following subgroups: patients for whom metastases may become resectable following chemotherapy patients who are KRAS mutation negative If evidence allows the appraisal should consider the use of continuation rules based on tumour response. Overall Survival Progression-free survival Tumour response rate Disease control rate Resection rates of metastases Health-related quality of life Adverse effects of treatment Economic Analysis will report incremental cost per QALY and life years gained. Subgroup analyses include Patients for whom metastases may become resectable following treatment with cetuximab combined with chemotherapy. Patients whose metastatic disease is confined to the liver. Patients who are KRAS wildtype status

8 Section B 3 Executive summary Please provide an executive summary that summarises the key sections of the submission. All statements should be directly relevant to the decision problem, be evidence-based and clearly reference the relevant section of the submission. The summary should cover the following items. The UK approved name, brand name, marketing status and principal pharmacological action of the proposed drug Approved name: cetuximab Brand name: Erbitux Marketing status: Subject to EMEA approval we anticipate marketing authorisation for the first line treatment of metastatic colorectal cancer in July/August Mechanism of action The Epidermal Growth Factor Receptor (EGFR) is a commonly expressed transmembrane glycoprotein belonging to the tyrosine kinase growth factor receptor family. EGFR is expressed widely in normal human body tissues, and is over expressed in many types of tumour. As a transmembrane glycoprotein, the extracellular domain of the EGFR is a ligand-binding site for Transforming Growth Factor alpha (TGFα), Epidermal Growth Factor (EGF) and other factors. Upon ligand binding, the intracellular domain of the EGFR is activated, thereby triggering cellular mechanisms that regulate cell growth, propensity to tumour cell invasion and angiogenesis [Yarden 1988, Baselga 2001]. In vitro analysis using cells that express high numbers of EGFR and produce ligands for these receptors has shown that the EGFR may be activated through an autocrine pathway, thereby leading to the proliferation of cells in culture [Van de Vijver et al 1991.] Cetuximab, a chimerised antibody of the IgG1 subclass, was originally derived from a mouse myeloma cell line. The chimerisation process resulted in an antibody with binding affinity to EGFR greater than the natural ligand EGF [Goldstein et al 1995] Cetuximab blocks binding of EGF and TGFα to the EGFR and inhibits ligand-induced activation of this receptor. Cetuximab also stimulates EGFR internalisation, effectively removing the receptor from the cell surface for interaction with ligands. [Waksal et al 1999]. Cetuximab also induces antibody dependent cell cytotoxicity (ADCC) [Roda et al 2007] See Figure B1 overleaf; - 8 -

9 Figure B1 The formulation(s), strength(s), pack size(s), maximum quantity(ies), anticipated frequency of any repeat courses of treatment and acquisition cost (see section 1.9).price. The following vial sizes are available: 100mg cetuximab/20ml vial and 500mg cetuximab/100ml vial. It is expected that cetuximab (plus FOLFIRI or FOLFOX) will be used as detailed in the following regimen: 400 mg cetuximab per m 2 body surface area on day 1. All subsequent weekly doses are 250 mg/m 2 each. Cetuximab treatment is continued until progression of the underlying disease /20ml vial /100ml vial Cetuximab is available to purchase as single vials. The indication(s) and any restriction(s). At the current time the licence is anticipated to include the following cetuximab in combination with chemotherapy is indicated for the 1 st line treatment of patients with EGFR-expressing metastatic colorectal cancer with KRAS wild-type status. The recommended course of treatment. 400 mg cetuximab per m 2 body surface area on day 1. Then, subsequent weekly doses of 250 mg/m 2. Cetuximab treatment should be continued until progression of the underlying disease

10 The main comparator(s). 5-fluorouracil/folinic acid/irinotecan (FOLFIRI) alone and 5-fluorouracil/folinic acid/oxaliplatin (FOLFOX) alone. Regimens based on 5-fluorouracil alone are not considered an appropriate comparator as clinical practice reserves such treatment for patients who may not be able to tolerate combination chemotherapy which is different from the population considered for cetuximab. [Personal Communication] Whether the key clinical evidence in the submission comes from head to head randomised trials (RCTs), from an indirect comparison of two sets of randomised trials involving a common comparator (for example, placebo or other active therapy), or from non-randomised studies. The main clinical results of the randomised trials and any relevant non RCTs The key clinical evidence is generated from two randomised controlled trials. The CRYSTAL study (EMR ) and OPUS study (EMR ). For further details please see sections 5.4 for efficacy and 5.7 for safety. The presence of mutations of the Kirsten rat sarcoma (KRAS) gene has been shown to affect the outcome of patients with colorectal cancer. In a large collaborative database, the presence of KRAS mutations was examined in samples from 3439 tumours from patients with CRC and the mutation was detected in 1197 cases, giving a prevalence of approximately 35%. In a multivariate analysis it was also shown that specific mutations were associated with a poorer outcome with regards to progression free survival and overall survival. [Andreyev, 2001] Studies have shown that the presence of KRAS mutations is a negative predictor of outcome in patients treated with EGFR inhibitors. [Tabernero 2008, Lievre 2008] Analyses based on KRAS status were not prospectively planned in the original protocols for the studies submitted in the regulatory submission because, when they were designed, no conclusive evidence for potentially predictive biomarkers was available. The key studies had not selected patients on the basis of KRAS status. Tumour blocks collected during the trial were available for approximately 50% of the trial population, allowing KRAS status to be analysed in these patients. As part of the regulatory process, the EMEA requested that Merck Serono perform a retrospective analysis of the patient population with wild-type KRAS tumour from the pivotal studies for cetuximab plus chemotherapy in the 1 st line treatment of metastatic colorectal cancer. Therefore the retrospective analysis based on the KRAS subset forms the main evidence for the decision problem. An overview of the key pivotal studies in the Full Analysis Set (FAS) and KRAS-wildtype populations is described below. The full data for all populations is listed in appendix

11 CRYSTAL STUDY (EMR ) The CRYSTAL study was a randomised, controlled, multicentre phase III study comparing 5-FU/FA plus irinotecan (FOLFIRI) plus cetuximab versus 5-FU/FA plus irinotecan (FOLFIRI) as first-line treatment for epidermal growth factor receptor-expressing metastatic colorectal cancer. The primary objective of the study was to assess whether the progression-free survival (PFS) time under 5-FU/FA plus irinotecan plus cetuximab is longer than that under 5-FU/FA plus irinotecan as first-line treatment for EGFR-expressing mcrc. Full analysis set (FAS) A total of 1217 patients were randomised to receive cetuximab plus FOLFIRI or FOLFIRI alone. The ITT population consisted of 599 patients in each arm. PFS This study met its primary endpoint by demonstrating superior PFS time (assessed by an Independent Review Committee) for cetuximab in combination with FOLFIRI compared to FOLFIRI alone (p=0.0479), with a hazard ratio (HR) of 0.85 (95% CI: 0.726, 0.998). A 15% risk reduction for progression was observed for subjects treated with cetuximab plus FOLFIRI compared to FOLFIRI alone. Surgery with curative intent In a pre-planned analysis surgery with curative intent was determined. In the cetuximab plus FOLFIRI group surgical resection of liver metastases was 6.0% compared to 2.5% in the FOLFIRI alone group. There was a statistically significant difference in the number of subjects with no residual tumour after resection, in favour of cetuximab plus FOLFIRI, with p= The odds ratio was 3.01 (95% CI: 1.39, 6.49) indicating that the relative chance for this beneficial surgery of curative intent was 3 times higher for subjects in the cetuximab plus FOLFIRI group in comparison to subjects in the FOLFIRI alone group. Safety In general, the safety profile observed in the FAS was comparable with the known safety profile of cetuximab. The data indicate that the addition of cetuximab to FOLFIRI increased the incidence of diarrhoea and palmar-plantar erythrodysaesthesia syndrome; however, diarrhoea and palmar-plantar erythrodysaesthesia syndrome were manageable using standard available treatments

12 The results below represent the KRAS wild-type sub population alone in the CRYSTAL study which is pertinent to the decision problem. Kaplan-Meier plot of Progression-free survival in study EMR (CRYSTAL), (KRAS wild-type population only) Figure B2 Table B1 Variable (n=348) KRAS wild type Response Rate (%) (95% CI) Disease control rate %) (95% CI) Progression-free survival Hazard Ratio 95% CI Median (months) 95% CI Curative metastatic surgery rate (%) Cetuximab plus FOLFIRI n= FOLFIRI alone n= P value (1.7) 2 (1.1)

13 A subgroup analysis was performed for those patients whose metastatic disease was confined to the liver. Within this group, both treatment arms had better response rates and longer progression-free survival (PFS) than the whole study population. In addition the efficacy advantage of the addition of cetuximab to the standard FOLFIRI chemotherapy was more marked than in the general population, suggesting additional benefit for cetuximab within this sub-group. These results are presented in Table B2 below Patients of KRAS wild-type status with metastatic disease confined to the liver Table B2 Variable (n=67) KRAS wild type Response Rate (%) (95% CI) Disease control rate %) (95% CI) Progression-free survival Hazard Ratio 95% CI Median (months) 95% CI Safety Cetuximab plus FOLFIRI alone FOLFIRI n= 35 n= , P value The majority of adverse events in the KRAS wild type population are in line with the existing product labelling for cetuximab or 5-FU/FA plus irinotecan. The following AEs occurred more frequently with cetuximab + FOLFIRI vs FOLFIRI (difference between treatment groups 5%) in the KRAS wild type population only: neutropenia, constipation, dyspepsia, dyspnoea, dysgeusia, injection site reaction, erythema, hypotension, hypertrichosis and cheilitis. These findings are considered as not clinically relevant due to the low proportion of subjects in the KRAS wild type population compared to the safety population (FAS). Quality of life In the CRYSTAL (EMR ) study, quality of life (QoL) was assessed in both treatment groups using the QLQ-C30 (version 3.0) and the EuroQuol (EQ-5D) questionnaires. The questionnaires were assessed in a longitudinal design at randomisation, during treatment and at the final tumour assessment. In the KRAS wild type population, some statistically significant differences were observed between the two treatment groups in favour of the FOLFIRI alone treatment arm. However these differences were small and using the methodology described by Osoba et al [1998] may not be defined as clinically meaningful

14 OPUS STUDY EMR The OPUS study was an open, randomised, controlled, multicentre phase II study comparing 5- FU/FA plus oxaliplatin (FOLFOX-4) plus cetuximab versus 5-FU/FA plus oxaliplatin as first-line treatment for EGRF-expressing metastatic colorectal cancer (mcrc). The primary objective of the study was to assess whether the rate of best confirmed response of 5FU/FA plus oxaliplatin (FOLFOX-4) plus cetuximab was superior to 5- FU/FA plus oxaliplatin as first line treatment for EGFR-expressing mcrc. Full Analysis Set The ITT population consisted of 168 patients in the cetuximab plus FOLFOX-4 arm and 168 patients in the FOLFOX-4 alone arm. The best overall response rate under cetuximab plus FOLFOX-4 was 45.6% vs 36% with FOLFOX-4 alone. The chance for a best overall response of either CR or PR increased by 50% (stratified odds ratio=1.516) in the cetuximab plus FOLFOX-4 group, which was not significant at the 5% level (stratified CMH test: p=0.064).therefore the study failed its primary endpoint as assessed in the full analysis set. Results of secondary analyses supported these findings with a stratified odds ratio of In the primary analysis, the chance of a best overall response of either CR or PR in subjects with a baseline ECOG of 0 and 1 was 65% statistically significantly higher under cetuximab plus FOLFOX-4 than under FOLFOX-4 alone at the 5% level (p=0.032), with absolute rates of 49.0% and 36.8% respectively. Safety profile FAS The safety profile of cetuximab in combination with FOLFOX-4 was generally manageable, and was generally consistent with the known safety profile

15 The results below represent the KRAS wild-type sub population alone in the OPUS study which is pertinent to the decision problem. Kaplan-Meier plot of Progression-free survival in study EMR (OPUS), (KRAS wild-type population only) Figure B3 Table B3 Variable (n=134) KRAS wild type Response Rate (%) (95% CI) Disease control rate %) (95% CI) Progression-free survival Hazard Ratio 95% CI Median (months) 95% CI Curative metastatic surgery n (%) Cetuximab plus FOLFOX-4 FOLFOX-4 alone n= 61 n= (9.8) (4.1) P value

16 These data show a statistically significant benefit in terms of efficacy for the addition of cetuximab to standard first line chemotherapy in patients with metastatic colorectal cancer (mcrc) who have KRAS wild-type tumours. The data demonstrate that with the addition of cetuximab, there is a statistically significant increase in tumour response rate and progression free survival in patients in this study. The data for overall survival are not yet mature for these studies and will have to be interpreted carefully as it has not been possible to control for subsequent interventions following the initial treatment allocation. Patients received a mix of second line chemotherapies and even third line interventions. Safety in the KRAS wild-type population The frequencies of most AEs in both treatment groups were comparable in the KRAS wild type population and safety (overall) population. The majority of AEs are in line with the existing product labelling for cetuximab. The frequency of palmar-plantar erythrodysesthesia syndrome was higher in the cetuximab + FOLFOX vs FOLFOX groups, in the KRAS wild type population 13.1 vs 4.1% (8 vs 3) subjects. Summary The addition of cetuximab to either FOLFOX or FOLFIRI regimens has demonstrated a statistically significant benefit in terms of progression free survival and tumour response in patients with EGFR-expressing, KRAS wild-type metastatic colorectal cancer (mcrc). These two measures are surrogate markers for overall survival times which are of direct relevance to patients with mcrc. In the CRYSTAL study a subgroup analysis was performed for those patients whose metastatic disease was confined to the liver. Within this group, both treatment arms had better response rates and longer median progression-free survival (PFS) than the whole study population. Although in both studies there was a numerically greater rate of curative resection of liver metastases in the groups that received cetuximab in addition to standard chemotherapy, the rates were very low. Actual rates in UK practice are approx 10-20%. [Personal Communication] Therefore in our economic evaluation we intend to assess the impact of resection rates which better reflect actual resection rates in UK practice. The side effects in the two key studies were generally as expected for the separate therapies and were generally manageable. Regarding quality of life in the KRAS wild type population, the CRYSTAL study demonstrated some statistically significant differences were observed between the two treatment groups in favour of the FOLFIRI alone treatment arm. However these differences were small and may not be defined as clinically meaningful

17 In relation to the economic evaluation, details of: the type of economic evaluation and justification for the approach used the pivotal assumptions underlying the model/analysis the incremental ratios from the evaluation. A de novo economic evaluation is presented in this submission. The economic evaluation compares the costs and health outcomes of patients with EGFRexpressing metastatic colorectal cancer with KRAS wild-type status and of two treatment strategies: 1. cetuximab in combination with fluoropyrimidine and irinotecan containing regimens compared to a fluoropyrimidine and irinotecan containing regimen alone 2. cetuximab in combination with fluoropyrimidine and oxaliplatin containing regimens compared to a fluoropyrimidine and oxaliplatin containing regimen alone This economic evaluation will focus on the expected licensed population and the following characteristics: Good performance status Suitable for irinotecan or oxaliplatin containing chemotherapy Liver metastases EXCLUDING patients who are resectable at presentation The economic evaluation simulates the disease progression and survival of patients with mcrc throughout first and subsequent lines of therapy, including longer-term survival after successful curative surgery. The CRYSTAL and OPUS clinical trials both enrolled patients who were determined as having unresectable liver metastases. This economic evaluation will assess the impact of cetuximab plus FOLFIRI /FOLFOX upon rates of potentially curative resection among patients who became resectable during treatment for first line mcrc. Rates of resection will be modelled using the correlation between response rates to treatment with possible resection rates shown in Folprecht et al [2005] which may better reflect resection rates observed in UK clinical practice. Cost effectiveness results are presented in Table S1 below. Table S1: Base Case Incremental Cost Effectiveness ratios Curative-intent resection Incremental costs scenarios and effects ICER Cetuximab + Incremental costs 19,473 FOLFIRI Incremental QALYs ,287 QALY Cetuximab + Incremental costs 20,242 FOLFOX Incremental QALYs ,245 QALY Scenario subgroup analyses found that the ICER defined was primarily sensitive to two key areas: o Modification of the rate of curative liver resection rate assumed o The choice of survival analyses methodology employed

18 4 Context In this background section the manufacturer or sponsor should summarise and contextualise the evidence relating to the decision problem. The information provided will not be formally reviewed by the Evidence Review Group. 4.1 Please provide a brief overview of the disease/condition for which the technology is being used. Provide details of the treatment pathway and current treatment options at each stage. Epidemiology of colorectal cancer (CRC). Colorectal cancer is the third most common cancer in the UK, with approximately 35,000 new cases registered per year (2005 figures). It is the second most common cause of death from cancer in the UK with nearly 16,000 deaths per year ( ) [ONS 2008]. The population of patients with metastatic colorectal cancer includes both those who present with metastatic disease and those who develop metastatic disease after surgery. Estimates of the proportion presenting with metastatic colorectal cancer ranges from 20% to 55% of new cases. Of those who have undergone surgery for complete excision, of their colorectal cancer, approximately 50% will eventually develop advanced disease and distant metastases, typically within 2 years of initial diagnosis. The 5-year survival rate for metastatic colorectal cancer is 12% [NICE TA 118] There has been an evolution in understanding of the biology of the disease with respect to specific mutations that occur in its pathogenesis. The presence of mutations of the Kirsten rat sarcoma (KRAS) gene has also been shown to affect the outcome of patients with colorectal cancer. In a large collaborative database, the presence of KRAS mutations was examined in samples from 3439 tumours from patients with CRC and the mutation was detected in 1197 cases, giving a prevalence of approximately 35%. In a multivariate analysis it was also shown that specific mutations were associated with a poorer outcome with regards to progression free survival and overall survival. [Andreyev 2001] Current treatment options for patients with metastatic CRC. Historically 5-FU/FA has been a standard first-line treatment for metastatic CRC with response rates of approx 21% and median survival (mos) time of approx 11.7 months [Thirion et al, 2004] The oral 5-FU drugs capecitabine and Uftoral (oral tegafur-uracil) modulated by folinic acid (FA) showed similar mpfs and mos times in comparison to bolus 5- FU/FA. [Hoff 2001, van Cutsem 2001, Carmichael 2002, Douillard 2002] Major advances in terms of efficacy have been achieved with the introduction of the approved and widely accepted combination regimens of 5-FU with FA in combination with either irinotecan or oxaliplatin. These regimens demonstrated significant superiority over 5-FU/FA alone in terms of response rate and median time to progression (mttp) or median progesssion free survival (mpfs) time (see Table B4) in the first line setting, but only the irinotecanbased regimens showed a significantly improved mos time over 5-FU/FA alone [Saltz 2000, Douillard 2000, De Gramont 2000]

19 Table B4 Key studies for investigational drug Treatment comparator Treatment vs comparator ORR % mpfs/ttp months mos months (p-value) irinotecan [Douillard 2000] Irinotecan + infusional 5-FU/FA infusional 5- FU/FA 35 vs. 22 (ITT) (p<0.005) 6.7 vs 4.4* (p<0.001) 17.4 vs 14.1 (p 0.031) Irinotecan [Saltz 2000] Irinotecan + bolus 5- FU/FA bolus 5- FU/FA 39 vs 21 (p<0.001) 7.0 vs 4.3 (P=0.004) 14.8 vs 12.6 (P=0.04) Oxaliplatin [De Gramont 2000] FOLFOX infusional 5- FU/FA 51 vs 22 (p=0.0001) 8.2 vs 6.0 (p=0.0003) 16.2 vs 14.7 (p=0.12) * Median time to progression FA=folinic acid, FOLFOX=5-FU/FA plus oxaliplatin, 5-FU=5-fluorouracil, mos= median overall survival time, mpfs=median progression-free survival time, ORR=objective response rate The therapeutic value of irinotecan vs oxaliplatin both in combination with an identical intermittent infusional 5-FU/FA regimen was compared in a randomised phase III study with crossover in case of progressive disease [Tournigand et al 2004]. In the firstline setting, both regimens showed comparable efficacy in terms of response rate and mpfs time. The mos time was 21.5 months in subjects allocated to FOLFIRI followed by FOLFOX in the case of progressive disease and 20.6 months allocated to FOLFOX followed by FOLFIRI in the case of progressive disease. (range 17.7 to 24.6 p= 0.99). In summary, similar benefit was derived from combination therapy with either FOLFIRI or FOLFOX regimens used as 1 st and 2 nd line treatment and the order in which they were used made no difference to overall survival. As a result of this data, in the UK fluoropyrimidine-based therapies with irinotecan or with oxaliplatin is the current standard of care for subjects with previously untreated metastatic CRC. In those patients for whom intensive combination chemotherapy with FOLFIRI or FOLFOX is not a suitable option then single-agent 5-FU, IV or oral, is an alternative. [Personal communication]. Therefore although 5-FU single agent therapy is detailed in the scope for this appraisal, it is apparent that the characteristics of patients considered suitable for 1 st - line combination chemotherapy differs from the population who are likely to receive fluoropyrimidine monotherapy. Given this it is inappropriate to use fluoropyrimidine monotherapy as a comparator to cetuximab in combination with chemotherapy. The guidance in the NICE TA 93 gives an option of FOLFIRI or FOLFOX as 1 st line therapy in metastatic colorectal cancer. [NICE TA 93]. Irinotecan or oxaliplatin, within their licensed indications, are recommended as treatment options as follows: irinotecan in combination with 5-FU/FA as first-line therapy or irinotecan alone in subsequent therapy or oxaliplatin in combination with 5-FU/FA as first-line or subsequent therapy

20 With advances in surgical technique, it is now possible in some patients to achieve complete resection of metastatic disease. This is especially the case when metastatic disease is confined to the liver. A recent review of 30 published studies of resection of hepatic metastases estimated that 20-30% of such metastases are potentially resectable [Simmonds 2006]. In one series where there was no residual metastatic disease following resection, the median survival of the 77 operated patients was 48 months (95%CI 25-71), with a five-year survival rate of 50% (38-61). [Giacchetti et al 1999]. In a review of published trials [Folprecht et al 2005] a strong correlation was found between response rates and resection rates in studies including patients with isolated liver metastases (p = 0.002). It also demonstrated a correlation with response rate and the resectability of liver metastases in patients whose metastatic disease was not confined to the liver. The authors noted that the 5-year overall survival in patients with macroscopically complete resection of hepatic metastases following neoadjuvant chemotherapy was comparable to series of patients with initially resectable liver metastases. 4.2 What was the rationale for the development of the new technology? Epidermal growth factor receptor (EGFR) spans the plasma membrane of healthy cells and with its ligands, plays an important role in the development and functioning of normal tissues. EGFR signaling modulates key cellular processes such as proliferation, angiogenesis and and survival. [Baselga 2001] EGFR is expressed in approx 80% of colorectal cancers [Cunningham 2004, Saltz, 2004] Cetuximab demonstrates high affinity EGFR binding and blockade results in antitumour activity [Goldstein et al 1995]. Thus, EGFR is strongly implicated as a protein pathologically associated with CRC. As cetuximab effectively targets EGFR, blocking associated signalling, it was hypothesised that cetuximab would be effective in the treatment of CRC, hence the development of cetuximab in this disease area. It has been hypothesised that activating KRAS mutations could be responsible for acquired activation of the Ras/mitogen-activated protein kinase pathway independently of the ligand-induced activation of the EGFR and, therefore, induces a resistance to treatments whose mode of action is through EGFR inhibition. [Lièvre et al 2008, panitumumab SmPC] Studies have shown that the presence of KRAS mutations is a negative predictor of outcome in patients treated with EGFR inhibitors. [Tabernero 2008, Lievre 2008, panitumumab, SmPC]. Analyses based on KRAS status were not prospectively planned in the original protocols for the studies submitted in the regulatory submission because, when they were designed, no conclusive evidence for potentially predictive biomarkers was available. The key studies had not selected patients on the basis of KRAS status. Tumour blocks collected during the trial were available for approximately 50% of the trial population, allowing KRAS status to be analysed in these patients

21 As part of the regulatory process, the EMEA requested that Merck Serono perform a retrospective analysis of the patient population with wild-type KRAS tumour from the pivotal studies for cetuximab plus chemotherapy in the 1 st line treatment of metastatic colorectal cancer. This data forms the evidence for the decision problem. 4.3 What is the principle mechanism of action of the technology? Mechanism of action The Epidermal Growth Factor Receptor (EGFR) is a commonly expressed transmembrane glycoprotein belonging to the tyrosine kinase growth factor receptor family. EGFR is expressed widely in normal human body tissues, and is over expressed in many types of tumour. As a transmembrane glycoprotein, the extracellular domain of the EGFR is a ligand-binding site for Transforming Growth Factor alpha (TGFα), Epidermal Growth Factor (EGF) and other factors. Upon ligand binding, the intracellular domain of the EGFR is activated, thereby triggering cellular mechanisms that regulate cell growth, propensity to tumour cell invasion and angiogenesis [Yarden 1988, Baselga 2001]. In vitro analysis using cells that express high numbers of EGFR and produce ligands for these receptors has shown that the EGFR may be activated through an autocrine pathway, thereby leading to the proliferation of cells in culture [Van de Vijver et al 1991.] Cetuximab, a chimerised antibody of the IgG1 subclass, was originally derived from a mouse myeloma cell line. The chimerisation process resulted in an antibody with binding affinity to EGFR greater than the natural ligand EGF [Goldstein et al 1995] Cetuximab blocks binding of EGF and TGFα to the EGFR and inhibits ligand-induced activation of this receptor. Cetuximab also stimulates EGFR internalisation, effectively removing the receptor from the cell surface for interaction with ligands. [Waksal 1999]. Cetuximab also induces antibody dependent cell cytotoxicity (ADCC) [Roda et al 2007] See Figure B1; Figure B1-21 -

22 4.4 What is the suggested place for this technology with respect to treatments currently available for managing the disease/condition? Cetuximab plus FOLFIRI or cetuximab plus FOLFOX is expected to be used in patients fit enough to receive combination chemotherapy in patients who present with metastatic colorectal cancer, expressing EGFR and with wild type KRAS, who are unsuitable for surgery for liver metastases and have received no previous chemotherapy for metastatic disease. 4.5 Describe any issues relating to current clinical practice, including any variations or uncertainty about best practice. UK market research data [A+A tracker Data on File] has shown that the three most commonly used 1 st -line regimens in metastatic colorectal cancer are single, agent 5- FU (oral or i.v.) or combination regimens of FOLFIRI or FOLFOX. Saunders [et al 2006] in a review article noted that if oxaliplatin is used more widely in the adjuvant setting as seems likely given the data from recent trials, the use of irinotecan is likely to increase in the future. This option is further supported in NICE guidance TA100 which recommended the use of FOLFOX as a treatment option in adjuvant colorectal cancer. It is our understanding that irinotecan is likely to become generic within the next few years, which may also lead to increased use within the NHS as the acquisition price falls. The treatment of the population considered appropriate for single agent fluoropyrimidine is discussed in section Provide details of any relevant guidelines or protocols. NICE issued guidelines on improving outcomes in colorectal cancer in May [ European Society for Medical Oncology Colon Cancer: Clinical recommendations for diagnosis, adjuvant treatment and follow-up National Comprehensive Cancer Network (NCCN) Clinical practice guidelines in oncology (Colon Cancer), V.I

23 5 Clinical Evidence 5.1 Identification of studies Exact details of the search strategy are provided in appendix 2, section 9.2. The scope of the search aimed to: Identify a relevant evidence base for the use of first-line treatments in metastatic colorectal cancer in the KRAS wild-type population. Identify studies on the use of cetuximab in combination with FOLFIRI in the first line treatment of metastatic colorectal cancer in patients with (KRAS) wild-type tumour. Identify studies on the use of cetuximab in combination with FOLFOX in the first line treatment of metastatic colorectal cancer in patients with KRAS wildtype tumour. Identify studies on the use of 5-fluorouracil plus folinic acid in the first line treatment of metastatic colorectal cancer in patients with KRAS wild type tumour. A range of sources were used to identify key clinical trial evidence for each of the main comparators. Abstracts were reviewed for all trials and if it was not clear if the trial met inclusion/exclusion criteria the full text article was then reviewed. Two independent reviewers were involved in the selection of studies for inclusion in the clinical evidence section. They selected studies independently and gave rationale for the inclusion and exclusion to the other reviewer. The studies included in the clinical evidence section have been agreed upon by both reviewers Complete list of RCTs Study Design/ subject population (mcrc) EMR Randomised, (CRYSTAL) controlled trial in patients with previously untreated nonresectable mcrc expressing KRAS wild-type tumour Phase Study treatments Full Analysis Set(N) III Cetuximab + FOLFIRI vs FOLFIRI EMR (OPUS) Randomised, controlled trial in patients with previously untreated nonresectable mcrc, expressing KRAS wild-type tumour II Cetuximab + FOLFOX-4 vs FOLFOX

24 5.2.2 Inclusion criteria Randomised controlled trials. Studies on the use of cetuximab in combination with FOLFIRI in the first line treatment of metastatic colorectal cancer in patients with KRAS wild-type tumour. Studies on the use of cetuximab in combination with FOLFOX in the first line treatment of metastatic colorectal cancer in patients with KRAS wild-type tumour. Studies on the use of 5-fluorouracil plus folinic acid in the first line treatment of metastatic colorectal cancer in patients with KRAS wild-type tumour. Human studies only were considered. Exclusion criteria Studies which involved patients who received previous treatment in the metastatic colorectal cancer setting were not considered. Papers published in a language other than English were not considered. Letters and editorials were not considered. Review articles and conference summaries were excluded. Animal studies/preclinical data were not considered List of relevant RCTs Study Design/ subject population (mcrc) EMR Randomised, (CRYSTAL) controlled trial in patients with previously untreated nonresectable mcrc expressing KRAS wild-type tumour Phase Study treatments Full Analysis Set(N) III Cetuximab + FOLFIRI vs FOLFIRI EMR (OPUS) Randomised, controlled trial in patients with previously untreated nonresectable mcrc, expressing KRAS wild-type tumour II Cetuximab + FOLFOX-4 vs FOLFOX

25 5.2.4 List of relevant non-randomised controlled trials Provide details of any non-randomised controlled trials that are considered relevant to the decision problem. Provide justification for their inclusion. No non-randomised trials have been included as RCT evidence to support the decision problem has been supplied Ongoing studies Provide details of relevant ongoing studies from which additional evidence is likely to be available in the next 12 months. The COIN trial as detailed below may report preliminary results within the next 12 months. A three-arm randomised controlled trial comparing Arm A - continuous chemotherapy oxaliplatin + fluoropyrimidine (OxFP). (Control). Arm B - continuous chemotherapy (OxFp) + cetuximab. Arm C - intermittent chemotherapy (OxFp). The primary endpoint is overall survival. Secondary endpoints include progressionfree survival, failure-free survival, time of disease control, response, toxicity, quality of life and cost effectiveness

26 A Flow diagram of the number of studies included and excluded at each stage should be provided as per the QUORUM statement

27 Internal database search 2 studies met inclusion criteria Medline + Embase + Current contents + Cochrane library + Conference searches revealed 34 published articles which were reviewed for potential inclusion 34 Trials were excluded for the following reasons Not a relevant population 12 Not a trial of treatment of interest 12 Not an RCT 4 Review article 6 2 studies were included in the systematic review

28 5.3 Summary of methodology of relevant RCTs As a minimum, the summary should include information on the following aspects of the RCT, but the list is not exhaustive. Items 2 to 14 of the CONSORT checklist should be provided, as well as a CONSORT flow diagram of patient numbers ( The methodology should not be submitted in confidence without prior agreement with NICE. Where there is more than one RCT, the information should be tabulated Methods Describe the RCT design (for example, duration, degree and method of blinding, and randomisation) and interventions. The 2 studies which support the decision problem are as follows: Table B6 Study EMR (CRYSTAL) Design/ subject population (mcrc) Randomised, controlled trial in patients with previously untreated nonresectable mcrc Phase Study treatments Full Analysis Set(N) III Cetuximab + FOLFIRI vs FOLFIRI EMR (OPUS) Randomised, controlled trial in patients with previously untreated nonresectable mcrc II Cetuximab + FOLFOX-4 vs FOLFOX Further details of these two studies are presented overleaf:

29 (CRYSTAL) EMR A open-labelled, randomised, controlled, multicentre phase III study comparing 5- FU/FA plus irinotecan (FOLFIRI) plus cetuximab versus 5- FU/FA plus irinotecan (FOLFIRI) as first-line therapy for EGFR-expressing mcrc. Recruitment ran from Aug 2004 to Nov The study design is summarised below; Prescreening EGFR Assessment Screening/ baseline Randomisa -tion Study treatment 8-weekly evaluation until progressive disease Group A Cetuximab + FOLFIRI Group B FOLFIRI No study treatment Minimum of 6 weeks after end of treatment unless 2 nd line anticancer therapy planned. 3-monthly follow-up Survival status Survival status Final Tumour assessment visit End of study Visit Groups A and B both received the same chemotherapy regimen every 14 days. The regimen was based on the simplified de Gramont regimen plus irinotecan (modified FOLFIRI regimen) and details of which are given in table B7: Chemotherapy regimen every 14 days in groups A and B Table B7 Order of administration Drug Dose First Irinotecan 30 to 90-min infusion on day mg/m² Second Folinic acid 120-min infusion on day mg/m² (racemic) or 200 mg/m² (L-form) before 5-FU bolus injection Third 5-Fluorouracil day mg/m² bolus on day 1 followed by a 46-hour continuous infusion of 2400 mg/m² In addition to the chemotherapy regimen given in Table B7 the subjects in group A were to receive cetuximab every 7 days. Cetuximab was to be administered (1 hour) before chemotherapy according to the dosage regimen in the following table:

30 Table B8 Cetuximab regimen for subjects in group A Cetuximab every 7 days First infusion All subsequent infusions Cetuximab 400 mg/m² intravenous infusion over 120 min 250 mg/m² intravenous infusion over 60 min The ideal treatment cycle in this study lasted 14 days and was determined by the chemotherapy dosage interval. Treatment cycles were defined as follows: Group A: 1 treatment cycle consisted of dosing with cetuximab, irinotecan and 5- FU/FA on day 1 of the cycle, and of dosing with cetuximab on day 8 of the cycle, with follow-up through day 14 of the cycle. Group B: 1 treatment cycle consisted of dosing with irinotecan and 5-FU/FA on day 1 of the cycle, with follow-up through day 14 of the cycle. Treatment duration The planned treatment duration per subject was until demonstration of progressive disease by computed tomography (CT) or magnetic resonance imaging (MRI), occurrence of unacceptable adverse events (AEs), or withdrawal of consent. (OPUS) EMR Open-labelled, randomised, controlled, multicentre phase II study comparing 5- FU/FA plus oxaliplatin (FOLFOX-4) plus cetuximab versus 5-FU/FA plus oxaliplatin as first-line treatment for EGRF-expressing metastatic colorectal cancer. Recruitment ran from July 2005 to March The trial is summarised below; Prescreening EGFR Assessment Screening/ baseline Randomisa -tion On-Study Period (active treatment or post treatment) including 8-Weekly Evaluation Visits until PD Group A Cetuximab + FOLFOX Group B FOLFOX No study treatment Minimum of 6 weeks after the final tumour assessment visit unless other anticancer therapy planned. 3-monthly follow-up Survival status Survival status Final Tumour assessment visit End of study Visit Groups A and B both received the same chemotherapy regimen every 14 days as detailed in Table B9:

31 Chemotherapy regimen every 2 weeks for subjects in both treatment groups. Table B9 Chemotherapy every 2 weeks Oxaliplatin Infusion over 120 minutes Folinic acid Infusion over 120 minutes Day 1 (1 hour after cetuximab Day 2 infusion) 85 mg/m mg/m² 200 mg/m² 5-FU Bolus (over 2 4 minutes) followed by a 22-hour infusion 400 mg/m² bolus, followed by 22- hour continuous infusion of 600 mg/m² 400 mg/m² bolus, followed by 22- hour continuous infusion of 600 mg/m² In addition to the chemotherapy regimen given in Table B9 the subjects in group A received cetuximab every 7 days. Cetuximab was to be administered (1 hour) before chemotherapy according to the dosage regimen in the following table: Cetuximab regimen for subjects in group A Table B10 Cetuximab every 7 days First infusion All subsequent infusions Cetuximab 400 mg/m² intravenous infusion over 120 min 250 mg/m² intravenous infusion over 60 min Treatment duration Treatment in Groups A and B was continued until PD was diagnosed using computed tomography (CT) or magnetic resonance imaging (MRI) scans, unacceptable toxicity occurred, or consent was withdrawn. Blinding for CRYSTAL and OPUS studies Both of the randomised, controlled studies were open-label. Blinding was not a practical option as most subjects treated with cetuximab experience skin reactions and thus can be readily identified. However primary endpoints were determined by a blinded review of the source data by an Independent Review Committee (IRC) in both studies. KRAS Status Retrospective analyses were undertaken in both the CRYSTAL and OPUS studies to assess the effect of the interventions in patients who were of KRAS wild-type status only. This was not a prior outcome measure of either of the studies but was an important and clinically relevant analysis undertaken as a requirement of the ongoing regulatory submission. As it is anticipated that the licence for cetuximab will be restricted to KRAS wild type patients, only data from that subgroup are directly relevant to the decision problem. Therefore this submission will focus on this population in the discussion of efficacy data

32 5.3.2 Participants Provide details of the inclusion and exclusion criteria, and describe the patient characteristics at baseline. Highlight any differences between study groups. Inclusion criteria Table B11 CRYSTAL EMR Signed, written informed consent Effective contraception for both male and female subjects if conception was possible 18 years Presence of histologically confirmed adenocarcinoma of the colon or rectum First occurrence of metastatic disease (not curatively resectable) EGFR-expressing disease Life expectancy of at least 12 weeks Presence of at least 1 bidimensionally measurable index lesion not in an irradiated area ECOG PS 2 at study entry Laboratory levels: white blood cell count 3x10 9 /L; neutrophils 1.5x10 9 /L; platelet count 100x10 9 /L; haemoglobin 5.6 mmol/l (9 g/dl) ASAT and ALAT 2.5 x upper reference range, or 5 x upper reference range in case of liver metastasis Total bilirubin < 1.5 x upper reference range serum creatinine 1.5 x upper reference range Recovery from relevant toxicity to previous treatment before study entry OPUS EMR Signed, written informed consent Effective contraception for both male and female subjects if conception was possible 18 years Presence of histologically confirmed adenocarcinoma of the colon or rectum First occurrence of metastatic disease (not curatively resectable) EGFR-expressing disease Life expectancy of at least 12 weeks Presence of at least 1 bidimensionally measurable index lesion not in an irradiated area ECOG PS 2 at study entry Laboratory levels: white blood cell count 3x10 9 /L; neutrophils 1.5x10 9 /L; platelet count 100x10 9 /L; haemoglobin 6.21 mmol/l (10 g/dl) ASAT and ALAT 2.5 x upper reference range, or 5 x upper reference range in case of liver metastasis serum creatinine 1.5 x upper reference range Recovery from relevant toxicity to previous treatment before study entry Differences between studies are shown in bold

33 Exclusion criteria Table B11 CRYSTAL EMR pregnancy Brain metastasis and/or leptomeningeal disease (known or suspected) Previous exposure to EGFR-targeting therapy Previous irinotecan-based chemotherapy OPUS EMR Pregnancy Brain metastasis and/or leptomeningeal disease (known or suspected) Previous exposure to EGFR-targeting therapy Previous oxaliplatin-based chemotherapy Previous chemotherapy for CRC except Previous chemotherapy for CRC except adjuvant treatment with PD documented adjuvant treatment with PD documented >6 months after the end of adjuvant >6 months after the end of adjuvant treatment treatment Radiotherapy, surgery (excluding prior Radiotherapy, surgery (excluding prior diagnostic biopsy) or any investigational diagnostic biopsy) or any investigational drug in the 30 days before randomisation drug in the 30 days before randomisation Concurrent chronic systemic immune Concurrent chronic systemic immune therapy or hormone therapy not indicated therapy or hormone therapy not indicated in the study protocol except physiologic in the study protocol except physiologic replacement replacement Clinically relevant coronary artery Clinically relevant coronary artery disease, history of myocardial infarction disease, history of myocardial infarction in the last 12 months, or high risk of in the last 12 months, or high risk of uncontrolled arrhythmia uncontrolled arrhythmia Acute or sub-acute intestinal occlusion or Acute or sub-acute intestinal occlusion or history of inflammatory bowel disease history of inflammatory bowel disease Previous malignancy other than CRC in Previous malignancy other than CRC in the last 5 years, except basal cell cancer the last 5 years, except basal cell cancer of the skin or pre-invasive cancer of the of the skin or pre-invasive cancer of the cervix cervix Known alcohol or drug abuse Known alcohol or drug abuse Significant disease which, in the Significant disease which, in the investigator s opinion, would exclude the investigator s opinion, would exclude the subject from the study subject from the study Peripheral neuropathy >grade 1 Known hypersensitivity reaction to any of Known hypersensitivity reaction to any of the components of study treatments the components of study treatments Participation in another clinical study Participation in another clinical study within the 30 days before randomisation within the 30 days before randomisation Legal incapacity or limited legal capacity Legal incapacity or limited legal capacity Medical or psychological conditions that Medical or psychological conditions that would not permit the subject to complete would not permit the subject to complete the study or sign informed consent the study or sign informed consent Differences between studies are shown in bold

34 Baseline characteristics for the KRAS wild-type subgroup and the FAS in the CRYSTAL and OPUS studies are presented below. The data for other sets are presented in Appendix 4. Baseline characteristics in CRYSTAL study EMR Table B12 Characteristic KRAS wild type patients n=348 Full Analysis Set n=1198 Cetuximab plus FOLFIRI FOLFIRI alone Cetuximab plus FOLFIRI FOLFIRI alone Gender, n (%) Age, years Age, n (%) Ethnic origin,n (%) ECOG status, n (%) Duration of mcrc (months) Location of tumour, n (%) Prior adjuvant chemotherapy, n (%) Male Female Median Range <65 years 65 years Caucasian Black Asian other/missing Median Range Colon Rectum Both Missing No Yes n= (61.0) 67 (39) (63.4) 63 (36.6) 161 (93.6) 1 (0.6) 6 (3.5) 4 (2.3) 99 (57.6) 66 (38.4) 7 (4.1) (55.2) 76 (44.2) 1 (0.6) (75.6) 42(24.4) n= (54.5) 80 (45.5) (68.2) 56 (31.8) 172 (97.7) 1 (0.6) 2 (1.1) 1 (0.6) 104 (59.1) 65 (36.9) 7 (4.0) (58.5) 72 (40.9) 1 (0.6) (80.7) 34 (19.3) n= (61.6) 230 (38.4) (62.4) 224 (37.4) 513 (85.6) 2 (0.3) 75 (12.5) 9 (1.5) 330 (55.1) 248 (41.4) 21 (3.5) (59.8) 228 (38.1) 12 (2.0) 1 (0.2) 472 (78.8) 127 (21.2) n= (59.4) 243 (40.6) (62.9) 222 (37.1) 514 (85.8) 5 (0.8) 75 (12.5) 5 (0.8) 318 (53.1) 260 (43.4) 21 (3.5) (60.4) 225 (37.6) 11 (1.8) 1 (0.2) 483 (80.6) 116 (19.4) The baseline characteristics are balanced between the two treatment arms with respect to age, ethnicity, severity of disease and prior adjuvant chemotherapy. The population is also similar to the final analysis set of the original trial (KRAS not evaluable, KRAS mutant and KRAS wild type combined set) with the exception of a slightly lesser representation of patients classified as Asian. The populations are comparable to the population of patients in the UK anticipated to be treated with cetuximab in combination with FOLFIRI. This combination chemotherapy is only expected to be appropriate for fitter patients and inevitably this

35 will result in the target population having a lower mean age and higher performance status than the average for the whole metastatic colorectal cancer population. Baseline characteristics in OPUS STUDY EMR Table B14 Characteristic KRAS wild type patients n=134 Cetuximab FOLFOX plus FOLFOX alone Full Analysis Set n=337 Cetuximab plus FOLFOX FOLFOX alone Gender, n (%) Age, years Age, n (%) Male Female Median Range <65 years 65 years Ethnic origin,n (%) Caucasian ECOG status, n (%) Duration of mcrc (months) Location of tumour, n (%) Prior adjuvant chemotherapy, n (%) other Median Range Colon Rectum Both No Yes n= (49.2) 31 (50.8) (63.9) 22 (36.1) 61 (100.0) 0 (0) 19 (31.1) 37 (60.7) 5 (8.2) (50.8) 30 (49.2) 0 53 (86.9) 8 (13.1) n= (60.3) 29 (39.7) (63.0) 27 (37.0) 72 (98.6) 1 (1.4) 27 (37.0) 37 (50.7) 9 (12.3) (56.2) 32 (43.8) 0 61 (83.6) 12 (16.4) n= (52.7) 80 (47.3) (56.8) 73 (43.2) 169 (100) 0 (0) 65 (38.5) 89 (52.7) 15 (8.9) (54.4) 75 (44.4) 2 (1.2) 148 (87.6) 21 (12.4) n= (54.8) 76 (45.2) (64.9) 59 (35.1) 167 (99.4) 1 (0.6) 74 (44.0) 77 (45.8) 17 (10.1) (53.0) 79 (47.0) (78.0) 37 (22.0) The baseline characteristics are balanced between the two treatment arms with respect to age, ethnicity, severity of disease and prior adjuvant or neoadjuvant chemotherapy. The population is also similar to the final analysis set of the original trial (KRAS not evaluable, KRAS mutant and KRAS wild type combined set). The populations are comparable to the population of patients in the UK anticipated to be treated with cetuximab in combination with FOLFOX. This combination chemotherapy is only expected to be appropriate for fitter patients and inevitably this will result in the target population having a lower mean age and higher performance status than the average for the whole metastatic colorectal cancer population

36 5.3.3 Patient numbers CRYSTAL STUDY (EMR ) -Note that only the KRAS Wild-type population is presented here after screening Subjects screened for EGFR expression N=2020 (EGFR-expressing N=1541) KRAS wild type (n=349) Randomised (n= 348) Allocated to cetuximab plus FOLFIRI (n= 172) Did not receive allocated cetuximab plus FOLFIRI n= 0) Allocation Allocated to FOLFIRI alone (n= 176) Did not receive allocated FOLFIRI n= 0) Lost to follow-up (n= 1) Discontinued cetuximab plus FOLFIRI (n= 115) Lost to follow-up (n= 1) Discontinued FOLFIRI (n= 130) Adverse Event 11 (6.4) Death 7 (4.0) Protocol non-compliance 1 (0.6) Withdrew consent 4 (2.3) Progressive disease 75 (43.4) Symptomatic deterioration 5 (2.9) Other 12 (6.9) Follow-up Adverse Event 5 (2.8) Death 4 (2.3) Protocol non-compliance 2 (1.1) Withdrew consent 6 (3.4) Progressive disease 103 (58.5) Symptomatic deterioration 4 (2.3) Other 17 (9.7) Analysed n= 172 (ITT Population) Excluded from analysis (n=0) Analysis Analysed n= (ITT Population) Excluded from analysis (n= 0)

37 OPUS STUDY EMR Note that only the KRAS Wild-type population is presented here after screening Subjects screened for EGFR expression N=606 (EGFR-expressing N=428) KRAS wild type (n=139) Randomised (n=139) Allocated to cetuximab plus FOLFOX-4 (n=64) Did not receive allocated cetuximab plus FOLFOX n=3 (reasons) Allocation Allocated to FOLFOX-4 alone (n=75) Did not receive allocated FOLFOX n= 2) (reasons) Lost to follow-up (n= 0) Discontinued cetuximab plus FOLFOX-4 (n= 48) Adverse Event 6 (9.8) Death 1 (1.6) Protocol non-compliance 2 (3.3) Withdrew consent 6 (9.8) Progressive disease 29 (47.5) Symptomatic deterioration 1 (1.6) Other 3 (4.9) Follow-up Lost to follow-up (n= 0) Discontinued FOLFOX-4 give reasons Discontinued FO(n= LFOX- 67) 4 give reasons (n= 67) Adverse Event 8 (11.0) Adverse Death Event 85 (11.0) (6.8) Death Protocol non-compliance 25 (2.7) (6.8) Protocol Withdrew non-comp consent liance 42 (5.5) (2.7) Withdrew Progressive consent disease 44 (5.5) (60.3) Progressive Symptomatic disease deterioration 244 (2.7) (60.3) Symptomatic Other deterioration 2 (2.7) Other 2 (2.7) Analysed n= 61 (ITT population) Excluded from analysis (n= 0) Analysis Analysed n= 73 (ITT Population) Excluded from analysis (n= 0)

38 5.3.4 Outcomes Table B15 CRYSTAL EMR Primary outcome measure Progression Free Survival (PFS) time of cetuximab plus FOLFIRI compared with FOLFIRI alone for first-line treatment for EGFR-expressing mcrc. OPUS EMR Primary outcome measure Response rate of cetuximab plus FOLFOX-4 compared with FOLFOX-4 as first-line treatment for EGFR expressing mcrc. Secondary outcome measures Secondary outcome measures Progression-free survival (PFS) time. Overall survival time Overall survival time Response rate (modified World Health Organization [WHO] criteria) Disease control rate Disease control rate. Duration of response Duration of response. Quality of life population pharmacokinetic analysis Safety Safety Amendment to protocol to include Documentation of surgery with curative Curative metastatic surgery. intent Differences between studies are shown in bold Outcome measures Assessment of progression and tumour response were performed by a blinded independent review committee according to a pre-specified IRC charter. Only subjects with measurable metastatic disease expressing EGFR could be enrolled in this study. Measurable disease required the presence of at least 1 bidimensionally measurable lesion. Lesions were measured in 2 dimensions with the size estimated by the cross product of the longest diameter and the perpendicular diameter. The sum of the products of diameters (SOPD) of all index (major) lesions were calculated and reported for each time-point. Imaging of the chest (CT), abdomen and pelvis (CT or MRI) was performed at screening (baseline) for eligibility and also to establish a baseline tumour assessment. All baseline evaluations had to be performed no longer than 21 days for the CRYSTAL study and 28 days before randomisation for the OPUS study. CT scans of the chest, and CT or MRI scans of the abdomen and pelvis were required at baseline. Thereafter, at each imaging time-point, an abdominal and pelvic CT or MRI was required. All subjects were to have their best response on-study classified by the investigator using modified WHO criteria. In addition, the independent review committee (IRC) conducted an independent blinded review of the images of all subjects based on a separate charter, which outlined details of the review process. The primary confirmatory analysis of PFS time was based on the assessment of the IRC (independent review committee)

39 Progression free survival (PFS) PFS, rather than OS, was selected as the primary endpoint for the CRYSTAL study as PFS measures the direct impact of the study treatment and is not affected by follow-up treatments. This position was confirmed during scientific exchanges with the EMEA. For the CRYSTAL study the primary target variable was the PFS time. All dates related to progressive disease (PD) used in the calculation of the PFS time (progression date, last known tumour assessment) were determined independently by the IRC. A subject s PFS time was defined as: the time in months from randomisation until PD was first observed or death occurred due to any cause within 60 days of the last tumour assessment or randomisation. In subjects without a progression date or death date more than 60 days after the last tumour assessment or randomisation, the PFS time was censored on the date of last tumour assessment or randomisation. Subjects who did not receive study treatment and who had neither progressed nor died were censored on the day of randomisation (day 1). Overall survival was defined as: the time (in months) from subject randomisation to the date of his or her death. If a subject had not died, survival time was censored at the last date the subject was known to be alive. Tumour response Response was defined as follows in the Table B16 below: Table B16 Complete response (CR) Partial response (PR) Stable disease (SD) Progressive disease (PD) Disappearance of all index lesions A 50% or more decrease in the SOPD of index lesions compared to the baseline SOPD, with no evidence of PD Neither sufficient decrease to qualify for PR nor sufficient increase to qualify for PD A 25% or more increase in the SOPD of index lesions, compared to the smallest SOPD recorded for the study period (nadir SOPD) Best overall response: This was defined as the proportion of subjects having achieved confirmed CR or PR as best overall response according to radiological assessments prior to surgery with curative intent, when applicable. Duration of response (for subjects with best response of CR or PR) was defined as the time from first assessment of CR or PR until the first date of PD or death within 60 days of the last tumour assessment or randomisation, whichever was first. If a subject had not progressed or the death date was more than 60 days after the last tumour assessment or randomisation, the duration of response was censored on the date of last known tumour assessment. Time to response (for subjects with best response of CR or PR) was defined as the time from the first dosing day of study treatment until the date PR or CR was confirmed. Disease control rate was defined as the proportion of subjects having achieved confirmed CR, PR or SD as best overall response according to radiological assessments prior to surgery with curative intent, when applicable

40 Rate of surgery with curative intent (with complete resection of all lesions [R0]) was defined as the number of subjects with any resection of metastasis of curative intent and all lesions completely resected to R0, divided by all subjects qualifying for the ITT population. QoL was determined using the EORTC QLQ-C30 (version 3.0) In addition to the EORTC QLQ-C30, QoL for subjects in the United Kingdom (UK) was also assessed using the EQ-5D Statistical analysis and definition of study groups CRYSTAL EMR Primary Hypothesis The primary confirmatory analysis tested the superiority of treatment with cetuximab plus FOLFIRI over FOLFIRI treatment alone in terms of PFS time on the ITT population. The primary analysis was to be undertaken for the per protocol population only if the per protocol population were to include less than 90% of the ITT-population. As this condition was not met, no analysis of the per protocol population was performed. Secondary endpoints were to compare the groups for differences in overall survival time, response rate, disease control rate, duration of response, quality of life and safety. Power of study and determination of sample size The study required 633 events (occurrence of documented PD) in order to have 80% power for the statistical test chosen to detect a difference between treatment groups of at least 2 months, anticipating a median PFS time of 8 months under 5-FU/FA plus Irinotecan (FOLFIRI group) and 10 months under 5-FU/FA plus irinotecan plus cetuximab (cetuximab plus FOLFIRI group). Thus, with 80% power, a hazard ratio (5-FU/FA plus irinotecan plus cetuximab to 5- FU/FA plus irinotecan) of 0.8 can be detected, representing a PFS time under 5- FU/FA plus Irinotecan plus cetuximab that is 25% longer than under 5-FU/FA plus irinotecan. Simulation studies for a difference between treatment groups of 3 months for PFS time showed a power gain to 84% subjects were to be randomised at a ratio of 1:1 to the 2 treatment groups. The study was powered (80%) to detect an increase in overall survival time from 17 months in the FOLFIRI group to 21 months in the group cetuximab plus FOLFIRI group (an improvement of 23.5%) when at least 705 deaths had been observed, which was anticipated approximately 3.5 years after study start Sensitivity analysis A sensitivity analysis was employed to ensure the robustness of the data and to identify potential imbalances in schedules for tumour assessments. Progressions and censoring dates were to be assigned to the scheduled visit time-points (IRC data set) according to the following rules:

41 Table B17 Situation No tumour assessment at baseline No tumour response assessment after start of treatment before EoS No IRC confirmed PD before EoS Death within 60 days after date of randomisation Death within 60 days after date of randomisation Death within 60 days after date of last tumour assessment Date of event/ censored Yes Date of death No No Date of randomisation Yes Yes Date of death No No Date of randomisation Yes Yes Date of death No No IRC confirmed PD between scheduled visits before EoS New anti-cancer treatment started Date of last scheduled tumour response assessment visit Date of last scheduled tumour response assessment visit Date of last scheduled tumour response assessment visit Censored Yes No Yes Power calculations The study required 633 events (occurrence of documented PD) in order to have 80% power for the statistical test chosen to detect a difference between treatment groups of at least 2 months, anticipating a median PFS time of 8 months under 5-FU/FA plus Irinotecan (group B) and 10 months under 5-FU/FA plus irinotecan plus cetuximab (group A). Thus, with 80% power, a hazard ratio (5-FU/FA plus irinotecan plus cetuximab to 5-FU/FA plus irinotecan) of 0.8 can be detected, representing a PFS time under 5-FU/FA plus Irinotecan plus cetuximab that is 25% longer than under 5- FU/FA plus irinotecan. Statistical Methods Efficacy The primary confirmatory analysis tested the superiority of treatment with cetuximab plus FOLFIRI over FOLFIRI treatment alone in terms of PFS time (based on IRC data). Superiority was tested using a log-rank test (Univariate Cox Model) stratified by ECOG performance status (0 1, 2) and region, with a 2-sided significance level α=0.05. Safety Safety analyses were performed as according to the as-treated principle. Adverse events were summarised comprising any events which had an onset date on or after the first day of study therapy and up to the End-of-study visit. No formal statistical comparisons were planned. Subgroup analysis The following subgroup analyses were pre-planned: Age, gender, ethnic origin, region, EGOC status, EGFR status, liver metastases only leucocytes, ALP, involved disease sites, prior adjuvant chemotherapy, longest diameter of major lesions, QOL status at baseline, LDH Analysis by KRAS status was carried out post-hoc

42 OPUS EMR Primary Hypothesis The primary confirmatory analysis tested the equality of 5-FU/FA plus oxaliplatin plus cetuximab versus 5-FU/FA plus oxaliplatin with respect to the confirmed and IRC reviewed response. Evaluations on a per protocol population were planned if there were differences of more than 10% in key variables. This was not the case, and the per-protocol evaluation was therefore not performed. The Secondary endpoints were Curative metastatic surgery, Duration of response, Disease control rate, Progression-free survival (PFS) time, Overall survival (OS) time, Safety and to perform a population pharmacokinetic analysis. Data handling conventions For the purpose of efficacy evaluation, in case of a missing day in a date that has the month and year otherwise, the day was imputed as the 15th of the month. In the instance of a missing day and month in a date that has only the year, the day and month was imputed as the midpoint between the last known alive date and the end of that year. Safety evaluations were based upon study subjects as treated (safety population). For the purpose of calculating the duration of an AE, for example, an incomplete onset date was imputed as the 1st if the day of the month is missing, and as January 1st if the day and the month were both missing. If the date calculated was before the first cetuximab administration or if the entire onset date was missing (i.e. not even the year is given), the first administration date was used to replace the onset date, unless the observed month or year was unequivocally before the first administration of study treatment. Incomplete resolution dates was imputed as the last day of the month if the day was missing. Power calculations The ITT population was used for efficacy analysis, based on the IRC assessment of best overall response. The following hypotheses (primary confirmatory analysis) were tested in the ITT population using a confirmatory hypothesis test (significance level α=0.05): H0: The stratified odds ratio (OddsR) comparing the best overall response with versus without cetuximab equals unity (i.e. OR=1) versus the alternative hypothesis: HA: The stratified OddsR comparing the best overall response with versus without cetuximab does not equal unity (i.e. OR 1) A stratified 2-sided Mantel-Haenszel test of OddsR=1 in 2x2 tables over the two ECOG PS strata (0 1, 2) was used with calculation of the corresponding 95% CI for OddsR to test the primary hypothesis. The best overall response results were further explored in logistic regression models with treatment group (with vs without cetuximab) as covariate, adjusted for ECOG PS (0 1, 2), and other potential prognostic factors. Kaplan-Meier methods will be used to estimate the distribution of duration of response, progression-free survival time, and overall survival time

43 Sample size: Assuming a difference in rate of best overall response of at least 20% between the two treatments, i.e. an approximate 70% response rate with cetuximab and 50% response rate without cetuximab for the ECOG PS 0 1 stratum, 66% and 45% for the ECOG PS 2 stratum, the common Mantel-Haenszel OddsR over the strata was expected to be Approx. 146 subjects per group including 10% for dropouts were therefore calculated as necessary to detect a significant best overall response of at least 2.33 at level α=0.05 with a power of 90%. Safety Safety analyses were performed according to the as-treated principle. AEs were summarised comprising any AEs with an onset date on or after the first day of study therapy and up to the End of Study visit. No formal statistical comparisons were planned. Subgroup analysis The following subgroup analyses were pre-planned: Age, gender, ethnic origin, region, EGOC status, EGFR status, number of metastases/organs involved, on-study skin reaction and on-study acne like rash. Analysis by KRAS status was carried out post-hoc Critical appraisal of relevant RCTs Each RCT should be critically appraised. If there is more than one RCT, tabulate the responses, highlighting any commercial in confidence data. The critical appraisal will be validated by the Evidence Review Group. The following are suggested criteria for critical appraisal, but the list is not exhaustive. How was allocation concealed? Both randomised, controlled studies were open-label. A central, stratified, permutedblock randomisation procedure was used to balance prognostic factors between treatment groups and to minimise the predictability of treatment allocation in this open-label study. What randomisation technique was used? The same randomisation techniques were used for the CRYSTAL and OPUS studies as follows; After completion of all the pre-screening and screening evaluations, all eligible subjects were randomly assigned to 1 of the 2 treatment groups in a 1:1 ratio. To enrol a patient a phone call was made to the central randomisation centre. All subjects were assigned a 4-digit identification number (0001, 0002, etc.) at the pre-screening visit. Subjects with EGFR-expressing tumours who were to receive study medication retained this number throughout the study. Subject numbers were unique (i.e. reallocation of subject numbers was not permitted). Regions were used as strata prior to randomisation of the first subject)

44 Was a justification of the sample size provided? Yes, the sample size was calculated to provide the statistical power necessary to show the difference in treatment groups as described in section Was follow-up adequate? Trial outcome was followed up in all both treatment groups for both trials. In both studies follow-up evaluations After the End of Study visit, in all subjects every 3 months to collect information on progression status, further anti-cancer treatment and overall survival time. Were the individuals undertaking the outcomes assessment aware of allocation? Blinding of allocation to the investigators was not a practical option as most subjects treated with cetuximab experience skin reactions and thus can be readily identified as having received cetuximab treatment. However all endpoints that were used as part of the analysis were determined by the independent review committee (IRC) who remained blinded to treatment allocation. Was the design parallel-group or crossover? Indicate for each crossover trial whether a carry-over effect is likely. Both studies were parallel group studies. Was the RCT conducted in the UK (or were one or more centres of the multinational RCT located in the UK)? If not, where was the RCT conducted, and is clinical practice likely to differ from UK practice? The CRYSTAL study (EMR ) was a multinational trial with several centres located in the UK. The OPUS study did not have any centres in the UK. It was conducted in Austria, Belgium, France, Germany, Greece, Israel, Italy, Poland, Portugal, Romania, Russian federation, Spain and The Ukraine. The use of cetuximab FOLFIRI and FOLFOX regimens are widely used throughout western Europe. Whilst there is some variation in practice between these countries practice does not vary sufficiently to prevent the findings of the study being applicable to the UK. One exception to this is the relatively low rate of resection of metastatic liver disease seen in these studies. In UK practice, under the guidance of the multidisciplinary team, it would be expected that much higher rates of metastatic surgery with curative intent would be attempted.[personal communication] How do the included in the RCT participants compare with patients who are likely to receive the intervention in the UK? Consider factors known to affect outcomes in the main indication, such as demographics, epidemiology, disease severity, setting. The populations are comparable to the population of patients in the UK anticipated to be treated with cetuximab in combination with FOLFIRI or FOLFOX. The gender split shows an overall higher number of males than females and this is in keeping with UK statistics that show the ratio of genders within colorectal cancer registrations to be 45% female and 55% male. [ONS 2005]

45 The average age of the patients in the study is lower than the reported average for all colorectal cancer patients. (median = in study and years in one published UK registry [Green et al 2007]. Data on comparison of performance status and co-morbidities is not robustly available for comparison but patients with significant co-morbidities or a ECOG performance status worse than 2 were excluded from entering the study. However this form of combination chemotherapy is usually given to fitter patients and inevitably this will result in the target population having a lower mean age and higher performance status than the average for the whole metastatic colorectal cancer population. [Personal Communication] The CRYSTAL and OPUS studies are therefore representative of the UK metastatic colorectal cancer population to be considered as part of the decision problem. For pharmaceuticals, what dosage regimens were used in the RCT? Are they within those detailed in the Summary of Product Characteristics? The cetuximab regimen used was 400 mg cetuximab per m 2 body surface area on day 1. Then, subsequent weekly doses of 250 mg/m 2. Cetuximab treatment is continued until progression of the underlying disease. This is consistent with the dosage regimen in the current Summary of product characteristics for cetuximab. The regimens for FOLFIRI and FOLFOX-4 are consistent with the summary of characteristics for 5-fluorouracil, folinic acid and oxaliplatin. Were the study groups comparable? The baseline characteristics are balanced between the two treatment arms with respect to age, ethnicity, severity of disease and prior adjuvant or neoadjuvant chemotherapy. The population is also similar to the final analysis set of the original trial (KRAS not evaluable, KRAS mutant and KRAS wild type combined set) with the exception of a slightly lesser representation of patients classified as Asian in study EMR Were the statistical analyses used appropriate? Yes, they are the standard analyses undertaken for this type of research. Was an intention-to-treat analysis undertaken? ITT analyses were undertaken for both studies Were there any confounding factors that may attenuate the interpretation of the results of the RCT(s)? The results are from an initially unplanned post-hoc sub-analysis of the original study data in response to a request from regulatory agencies. However, the treatment groups are balanced and representative of the characteristics of the full data set

46 5.4 Results of the relevant comparative RCTs Efficacy Efficacy results for the KRAS wild-type subgroup and the FAS in the CRYSTAL and OPUS studies are presented below. The data for other sets are presented in Appendix 4. CRYSTAL STUDY (EMR ) Full analysis set (FAS) A total of 1217 patients were randomised to receive cetuximab plus FOLFIRI or FOLFIRI alone. The ITT population consisted of 599 patients in each arm. This study met its primary endpoint by demonstrating superior PFS time (assessed by an Independent Review Committee) for cetuximab in combination with FOLFIRI compared to FOLFIRI alone (p=0.0479), with a hazard ratio (HR) of 0.85 (95% CI: 0.726, 0.998). A 15% risk reduction for progression was observed for subjects treated with cetuximab plus FOLFIRI compared to FOLFIRI alone. In a pre-planned analysis surgery with curative intent was determined. In the cetuximab plus FOLFIRI group surgical resection of liver metastases was 6.0% compared to 2.5% in the FOLFIRI alone group. There was a statistically significant difference in the number of subjects with no residual tumour after resection, in favour of cetuximab plus FOLFIRI, with p= The odds ratio was 3.01 (95% CI: 1.39, 6.49) indicating that the relative chance for this beneficial surgery of curative intent was 3 times higher for subjects in the cetuximab plus FOLFIRI group in comparison to subjects in the FOLFIRI alone group

47 The results below represent the KRAS wild-type sub population alone in the CRYSTAL study which is pertinent to the decision problem. Kaplan-Meier plot of Progression-free survival in study EMR (CRYSTAL), (KRAS wild-type population only) Figure B4 Table B18 Variable (n=348) KRAS wild type Response Rate (%) (95% CI) Disease control rate %) (95% CI) Progression-free survival Hazard Ratio 95% CI Median (months) 95% CI Curative metastatic surgery rate (%) Cetuximab plus FOLFIRI n= FOLFIRI alone n= P value (1.7) 2 (1.1) A subgroup analysis was performed for those patients whose metastatic disease was confined to the liver. Within this group, both treatment arms had better response rates and longer progression-free survival (PFS) than the whole study population. In addition the efficacy advantage of the addition of cetuximab to the standard FOLFIRI

48 chemotherapy was more marked than in the general population, suggesting additional benefit for cetuximab within this sub-group. These results are presented in Table B19 below Patients of KRAS wild-type status with metastatic disease confined to the liver Table B19 Variable (n=67) KRAS wild type Response Rate (%) (95% CI) Disease control rate %) (95% CI) Progression-free survival Hazard Ratio 95% CI Median (months) 95% CI Cetuximab plus FOLFIRI alone FOLFIRI n= 35 n= , P value Quality of life In the CRYSTAL (EMR ) study, quality of life (QoL) was assessed in both treatment groups using the QLQ-C30 (version 3.0) and the EuroQuol (EQ-5D) questionnaires. The questionnaires were assessed in a longitudinal design at randomisation, during treatment and at the final tumour assessment. In the KRAS wild type population, some statistically significant differences were observed between the two treatment groups in favour of the FOLFIRI alone treatment arm. However these differences were small and using the methodology described by [Osoba et al 1998] may not be defined as clinically meaningful. OPUS STUDY EMR Full Analysis Set The ITT population consisted of 168 patients in the cetuximab plus FOLFOX-4 arm and 168 patients in the FOLFOX-4 alone arm. The best overall response rate under cetuximab plus FOLFOX-4 was 45.6% vs 36% with FOLFOX-4 alone. The chance for a best overall response of either CR or PR increased by 50% (stratified odds ratio=1.516) in the cetuximab plus FOLFOX-4 group, which was not significant at the 5% level (stratified CMH test: p=0.064).therefore the study failed its primary endpoint as assessed in the full analysis set. Results of secondary analyses supported these findings with a stratified odds ratio of In the primary analysis, the chance of a best overall response of either CR or PR in subjects with a baseline ECOG of 0 and 1 was 65% statistically significantly higher under cetuximab plus FOLFOX-4 than under FOLFOX-4 alone at the 5% level (p=0.032), with absolute rates of 49.0% and 36.8% respectively

49 The results below represent the KRAS wild-type sub population alone in the OPUS study which is pertinent to the decision problem. Kaplan-Meier plot of Progression-free survival in study EMR (OPUS), (KRAS wild-type population only) Figure B5 Table B20 Variable (n=134) KRAS wild type Response Rate (%) (95% CI) Disease control rate %) (95% CI) Progression-free survival Hazard Ratio 95% CI Median (months) 95% CI Curative metastatic surgery n (%) Cetuximab plus FOLFOX-4 FOLFOX-4 alone n= 61 n= (9.8) (4.1) P value

50 These data show a statistically significant benefit in terms of efficacy for the addition of cetuximab to standard first line chemotherapy in patients with metastatic colorectal cancer (mcrc) who have KRAS wild-type tumours. The data demonstrate that with the addition of cetuximab, there is a statistically significant increase in tumour response rate and progression free survival in patients in this study. The data for overall survival are not yet mature for these studies and will have to be interpreted carefully as it has not been possible to control for subsequent interventions following the initial treatment allocation. Patients received a mix of second line chemotherapies and even third line interventions. Discussion of efficacy data These data show a statistically significant benefit in terms of efficacy for the addition of cetuximab to standard first line chemotherapy in patients with metastatic colorectal cancer (mcrc) who had KRAS wild-type expressing tumours. The data demonstrate that with the addition of cetuximab, there is a statistically significant increase in tumour response rate and progression free survival in patients on study. The data for overall survival are not yet mature for these studies and will have to be interpreted carefully as it has not been possible to control for subsequent interventions following the initial treatment allocation. Patients will receive a mix of second line chemotherapies and third line interventions. Therefore the surrogate endpoints of tumour response and progression free survival are the best current data to examine the effect of addition of cetuximab to standard first line chemotherapy compared to standard chemotherapy alone. A subgroup analysis was performed for those patients whose metastatic disease was confined to the liver. Within this group, both treatment arms had better response rates and longer progression-free survival (PFS) than the whole study population. In addition the efficacy advantage of the addition of cetuximab to the standard FOLFIRI chemotherapy was more marked than in the general population, suggesting additional benefit within this sub-group. Although in both studies there was a numerically greater rate of curative metastatic surgery in the groups that received cetuximab in addition to standard chemotherapy, the rates were very low and in fact considerably lower than would be expected in routine practice [Personal communication] The analyses were performed post-hoc on a subset of the initially conceived population for these studies as a result of dialogue with and instruction from regulatory agencies. Therefore the original power of the study has been diluted. However it is clear that there is still a statistically significant benefit to patients from the addition of cetuximab to both FOLFIRI and FOLFOX in KRAS wild-type populations. Quality of life In the CRYSTAL study (EMR ) quality of life (QoL) was assessed in both treatment groups using the QLQ-C30 (version 3.0) and the EuroQuol (EQ-5D) questionnaires. The questionnaires were assessed in a longitudinal design at randomisation, during treatment and at the final tumour assessment

51 As part of the re-analysis of the data to concentrate solely on the population of patients with KRAS wild-type tumours, a further analysis of the QoL data was undertaken. QLQ-C30 Table B21 below presents demographic and baseline characteristics for all subjects in the Evaluable for QLQ-C30 population: KRAS wild-type subgroup. These characteristics were used as factors in the longitudinal QoL models. The characteristics of the two treatment groups were similar with regard to age, ECOG performance status, number of involved sites and presence of liver metastases, however there was a higher proportion of females in the FOLFIRI alone treatment group. The median (range) age was 60 (22 79) years. Approximately 59% of subjects were fully active (i.e. had ECOG score 0), 13% of subjects had more than 2 involved tumor sites and 19% had liver metastases. Compared with the patient characteristics in the full ITT population the KRAS wild-type subgroup appears to have slightly better patient characteristics, i.e. younger population, better ECOG performance status, fewer sites involved and less liver metastases. Of the 348 subjects in the ITT population: KRAS wild-type subgroup, 330 subjects (161 in the Cetuximab + FOLFIRI arm and 169 in the FOLFIRI alone arm) completed at least one evaluable QLQ-C30 questionnaire and were thus included in the Evaluable for QLQ-C30 population: KRAS wild-type subgroup. In total, 1519 questionnaires were completed. The proportion of evaluable questionnaires was 79.4% in the Cetuximab + FOLFIRI arm and 78.8% in the FOLFIRI alone treatment group. In total, 1201 questionnaires were included in the primary QoL analyses. Demographic and baseline characteristics, Evaluable for QLQ-C30 Population: KRAS wild-type subgroup Table B21 Characteristic Cetuximab + FOLFIRI (N=161) N (%) FOLFIRI (N=169) N (%) Gender Male 102 (63.4) 92 (54.4) Female 59 (36.7) 77 (45.6) Age (years) Median Range ECOG PS 0 96 (59.63) 100 (59.17) 1 60 (37.27) 62 (36.69) 2 5 ( 3.11) 7 ( 4.14) Number of (86.96) 143 (84.62) Sites Involved >2 19 (11.80) 24 (14.20) Missing 2 ( 1.24) 2 ( 1.18) Liver Metastases Yes 34 (21.12) 30 (17.75) No 127 (78.88) 139 (82.25) In the initial analysis of all patients some statistically significant differences in QoL scores were observed between the two treatment groups (cetuximab + FOLFIRI and

52 FOLFIRI alone) in favour of the FOLFIRI alone treatment arm. In general these differences were small and not considered to be clinically meaningful using the methodology as described by Osoba et al In this analysis there were no statistically significant differences between reported QoL between the two groups apart from in three specific measures; Change in baseline to worst physical functioning score: The mean changes were and in the cetuximab + FOLFIRI and FOLFIRI alone treatment groups. This suggested that the addition of cetuximab resulted in poorer outcomes in this aspect of QoL. However this was not supported by any other analysis of the physical functioning scale. Dyspnoea: Subjects in the cetuximab + FOLFIRI arm reported significantly worse scores for dyspnoea than in the FOLFIRI alone treatment group (p=0.0460). This was supported by the adverse event data. Nausea/vomiting: Subjects in the FOLFIRI alone arm reported more nausea/vomiting symptoms than for the cetuximab + FOLFIRI (p=0.0554) and this was also supported by the adverse event data. In summary, some statistically significant differences were observed between the two treatment groups in favour of the FOLFIRI alone treatment arm. However these differences were small and may not be defined as clinically meaningful. EQ-5D Table B22 presents the compliance with completion of EQ-5D questionnaires by planned assessment time windows and treatment group. Only 37 patients completed evaluable baseline questionnaires. Due to the small number of patients contributing to the EQ-5D population no formal statistical analyses were performed. Table B22: Analysis of compliance for EQ-5D questionnaires by protocol planned assessment (Evaluable for EQ-5D population) Visit N1: in TW Cetuximab / Folfiri (N=25) N2: with at least one questionnaire(2) Rate: N2/N1(%) N1: in TW Folfiri (N=17) N2: with at least one questionnaire(2) Rate: N2/N1(%) Baseline Week Week Week Week Week Final Tumour Assessment

53 In order to calculate a utility value that is representative for patients receiving 1st line chemotherapy (+/- cetuximab) treatment for metastatic colorectal cancer, a summary utility value was calculated for all patients by pooling all values at each visit. This resulted in a utility value of 0.77 (SD 0.22; N=128). This utility was maintained at a steady state throughout the course of the study demonstrating that there was no decrease or increase in quality of life over time for patients who remained on treatment. This is shown in figure B7. Figure B7: CRYSTAL UK EQ-5D data from baseline to week Utility Baseline Time (Weeks) cetuximab plus FOLFIRI FOLFIRI 5.5 Meta-analysis A meta-analysis was not undertaken as the two studies relevant to the decision problem had different primary endpoints and different comparators therefore was not considered relevant. 5.6 Indirect/mixed treatment comparisons No Indirect/mixed treatment comparisons have been included as RCT evidence to support the decision problem has been supplied

54 5.7 Safety CRYSTAL STUDY EMR Safety Full Analysis Set Population The safety profile observed in the FAS was comparable with the known safety profile of cetuximab. The data indicate that the addition of cetuximab to FOLFIRI increased the incidence of diarrhoea and palmar-plantar erythrodysaesthesia syndrome; however, diarrhoea and palmar-plantar erythrodysaesthesia syndrome were manageable using standard available treatments. Safety profile KRAS wild-type population Adverse Events occurring with >10% frequency in Cetuximab plus FOLFIRI vs FOLFIRI group. KRAS wild type population only Table B23 % of subjects Adverse event Cetuximab plus FOLFIRI n= 173 FOLFIRI alone n= 176 Diarrhoea Neutropenia Rash Fatigue Dry Skin Stomatitis Anorexia Dermatitis acneiform Pyrexia Paronychia Constipation Skin Fissures Conjunctivitis Weight decreased Dyspepsia Palmar-plantar erythrodysesthesia syndrome Acne Pruritis Mucosal inflammation Dyspnoea Dysgeusia Injection site reaction Hypokalaemia Nail Disorder Erythema Hypotension Hypomagnesaemia Skin toxicity Hypertrichosis Exoliative rash Cheilitis

55 Grade 3 or 4 AEs in 5% subjects in the KRAS wild type population Table B24 Number (%) of subjects Adverse event Cetuximab plus FOLFIRI n= 173 Any 135 (78.0) 89 (50.6) Neutropaenia 44 (25.4) 29 (16.5) Diarrhoea 30 (17.3) 16 (9.1) Leukopaenia 12 (6.9) 8 (4.5) Rash 12 (6.9) 0 Paronychia 11 (6.4) 0 Dermatitis Acnieform 10 (5.8) 0 Hypokalaemia 10 (5.8) 4 (2.3) Pulmonary Embolism 9 (5.2) 6 (3.4) Vomiting 8 (4.6) 5 (2.8) FOLFIRI alone n= 176 The majority of adverse events in the KRAS wild type population are in line with the existing product labelling for cetuximab or 5-FU/FA plus irinotecan. The following AEs occurred more frequently with cetuximab + FOLFIRI vs FOLFIRI (difference between treatment groups 5%) in the KRAS wild type population only: neutropenia, constipation, dyspepsia, dyspnoea, dysgeusia, injection site reaction, erythema, hypotension, hypertrichosis and cheilitis. These findings are considered as not clinically relevant due to the low proportion of subjects in the KRAS wild type population compared to the safety population (FAS). OPUS STUDY EMR Safety profile FAS population The safety profile of cetuximab in combination with FOLFOX-4 was generally manageable, and was generally consistent with the known safety profile. Safety profile KRAS wild-type population Adverse Events occurring with >10% frequency in Cetuximab plus FOLFOX vs FOLFOX group. KRAS wild type population only Table B25 % of subjects Cetuximab plus FOLFOX-4 alone Adverse event FOLFOX-4 n= 61 n= 73 Rash Diarrhoea Fatigue Vomiting Stomatitis Dermatitis acnieform Dry skin Skin fissures Mucosal inflammation Acne Hypomagnesaemia

56 Grade 3 or 4 AEs in 5% subjects in the KRAS wild type population Table B26 Number (%) of subjects Cetuximab plus FOLFOX-4 alone Adverse event FOLFOX-4 n= 61 n= 73 Any 51 (83.6) 46 (63.0) Neutropaenia 25 (41.0) 24 (32.9) Diarrhoea 7 (11.5) 4 (5.5) Rash 7 (11.5) 0 Leukopaenia 5 (8.2) 4 (5.5) Peripheral sensory 3 (4.9) 6 (8.2) neuropathy The frequencies of most AEs in both treatment groups were comparable in the KRAS wild type population and safety (overall) population. The majority of AEs are in line with the existing product labelling for cetuximab. The frequency of palmar-plantar erythrodysesthesia syndrome was higher in the cetuximab + FOLFOX vs FOLFOX groups, in the KRAS wild type population 13.1 vs 4.1% (8 vs 3) subjects. Safety of the addition of cetuximab to chemotherapy The addition of cetuximab to standard first-line chemotherapeutic regimens in the treatment of mcrc does result in an increase in adverse events as would be expected. The adverse events are particularly the skin reactions which are well described with cetuximab. This class effect with EGFR inhibitors is generally well manageable [Segaert et al, 2005] and may be indicative of a response to cetuximab [Cunningham et al 2004]. 5.8 Non-RCT evidence In the absence of valid RCT evidence, evidence from other study designs will be considered, with reference to the inherent limitation inferred by the study design. The level of detail provided should be the same as for RCTs and where possible more than one independent source of data should be examined to explore the validity of any conclusions. Inferences about relative treatment effects drawn from observational evidence will necessarily be more circumspect from those from RCTs Summary of methodology of relevant non-rcts n/a Critical appraisal of relevant non-rcts n/a Results of the relevant non- RCTs n/a

57 5.9 Interpretation of clinical evidence Provide a brief statement of the relevance of the evidence base to the decision problem. Include a discussion of the relevance of the outcomes assessed in clinical trials to the clinical benefits experienced by patients in practice. The evidence base presented in these two RCTs demonstrates a statistically significant benefit in terms of progression free survival and tumour response from the addition of cetuximab to standard first-line chemotherapeutic regimens in patients with EGFR-expressing, KRAS wild-type, metastatic colorectal cancer (mcrc).. These two measures are surrogate markers for overall survival times which are of direct relevance to patients with mcrc. The increase in progression free survival in these studies over standard chemotherapeutic regimens is between 7-14% with the addition of cetuximab. When the sub-group of patients whose metastatic disease is confined to the liver is considered, this increase is over 53%. It is not possible to calculate from these studies what the ultimate survival benefit will be as the overall survival data are not yet mature and there will be a confounding effect from subsequent interventions in the second and third line settings for these patients. The higher response rates seen with the addition of cetuximab to chemotherapy may also be reflected in a higher curative surgery rate for metastatic disease in patients treated in the UK. This is because the multidisciplinary team approach used in the UK may result in a higher surgery rate. There is a correlation between response rate and metastatic surgery rates [Folprecht et al, 2005] and so the higher response rates seen with the addition of cetuximab to standard chemotherapy would be expected to result in higher curative metastatic surgery rates with a corresponding increase in survival times Identify any factors that may influence the applicability of study results to patients in routine clinical practice; for example, how the technology was used in the trial, issues relating to the conduct of the trial compared with clinical practice, or the choice of eligible patients. State any criteria that would be used in clinical practice to select suitable patients based on the evidence submitted. What proportion of the evidence base is for the dose(s) given in the Summary of Product Characteristics? The populations treated in these studies were restricted to metastatic colorectal cancer patients fit enough to receive intensive combination chemotherapy with FOLFIRI or FOLFOX. Patients with significant co-morbidities and poor performance status were specifically excluded from the study. Therefore it is not possible to predict the outcome of less fit patients when treated with these regimens. However, combination chemotherapies have morbidities associated with the interventions used and so will usually be restricted to patients who are fit enough to manage them. Therefore it is likely that only patients with ECOG performance status of 0-1 would be considered for the combination regimen of cetuximab combined with either FOLFOX or FOLFIRI. Therefore the results of these studies are representative of the outcomes to be expected in clinical practice

58 In the studies all interventions (cetuximab, oxaliplatin, irinotecan and 5-FU/FA) were given according to the doses given in their respective Summaries of Product Characteristics. 6 Cost effectiveness 6.1 Published cost-effectiveness evaluations Identification of studies Describe the strategies used to retrieve relevant cost-effectiveness studies from the published literature and from unpublished data held by the manufacturer or sponsor. The methods used should be justified with reference to the decision problem. Sufficient detail should be provided to enable the methods to be reproduced, and the rationale for any inclusion and exclusion criteria used should be provided. The search strategy used should be provided in appendix 3, section 9.3. The strategy was designed to retrieve cost effectiveness studies which were relevant to the decision problem of cetuximab (Erbitux) for the treatment of first line metastatic colorectal cancer (mcrc). OVID advanced search was used for both Embase and Medline, searches were carried out on Monday 17th March The criterion utilised are presented below. All searches were limited to English, human, abstract,full text (Medline was also limited to exclude reviews): Metastatic colorectal cancer Erbitux OR Cetuximab These two terms were combined together and searched with each of the following health economic search criteria: Cost effectiveness analysis Cost benefit analysis QALY Cost effectiveness Quality of life All search terms were mapped to subject heading. The Centre for Reviews and Dissemination (CRD) website was also searched ( ID=538291&D=1&H=11&E=2&SearchFor=cetuximab&DB=) using cetuximab as a search term

59 The pre-specified exclusion criterion included articles focused upon areas not particular to the decision problem in hand: Screening programs Chemotherapy in the adjuvant setting Chemotherapy in mcrc which was second line or later Studies which included other disease sites Surgical studies Review articles Studies which focused on just one aspect of treatment These searches resulted in the following findings: Medline (see Table H1 below) o 4 studies, all of which related to second line or later therapy in mcrc so were not considered relevant Embase (see Table H2 below) o 25 studies of which: 6 were second line or later therapy in mcrc 2 included other tumour sites 15 were review papers CRD database (see Table H3 below) o 14 studies of which: 9 were second line or later therapy in mcrc 5 were horizon scanning document Therefore, this review did not identify any economic analysis in the first line treatment of mcrc with cetuximab. Details of the search strategies utilised for Medline, Embase and the CRD database can be found in Table H1, H2 and H3 respectively

60 Table H1: Search strategy used for Medline # Searches Medline (R) 1950 to present carried out Monday 17/3/08 Results 1 Colorectal Neoplasms/ or metastatic colorectal cancer.mp. 32, limit 1 to (abstracts and english language and full text and humans) 5, (Erbitux or cetuximab).mp. [mp=title, original title, abstract, name of substance word, subject heading word] 1, limit 3 to (abstracts and english language and full text and humans) and Cost-Benefit Analysis/ or "Costs and Cost Analysis"/ or cost effectiveness analysis.mp. or Quality-Adjusted Life Years/ 79, limit 6 to (abstracts and english language and full text and humans) 5, cost benefit analysis.mp. or Cost-Benefit Analysis/ 43, limit 8 to (abstracts and english language and full text and humans) 3, QALY.mp. or Quality-Adjusted Life Years/ 4, limit 10 to (abstracts and english language and full text and humans) cost effectiveness.mp. or Cost-Benefit Analysis/ 50, limit 12 to (abstracts and english language and full text and humans) 4, quality of life.mp. or "Quality of Life"/ 100, limit 14 to (abstracts and english language and full text and humans) 12, and and and and and

61 Table H2: Search strategy used for Embase # Searches Embase 1980 to 2008 week 11, carried out Monday 17/3 Results 1 metastatic colorectal cancer.mp. or Colorectal Cancer/ 32, limit 1 to (full text and abstracts and human and english language) 4, (Erbitux or cetuximab).mp. [mp=title, abstract, subject headings, heading word, drug trade name, original title, device manufacturer, drug manufacturer name] 4 limit 3 to (full text and abstracts and human and english language) and , cost effectiveness analysis.mp. or "Cost Effectiveness Analysis"/ 53, limit 6 to (full text and abstracts and human and english language) 5, Cost benefit analysis.mp. or "Cost Benefit Analysis"/ 28, limit 8 to (full text and abstracts and human and english language) 2, QALY.mp. or Quality Adjusted Life Year/ 4, limit 10 to (full text and abstracts and human and english language) cost effectiveness.mp. or "Cost Effectiveness Analysis"/ 57, limit 12 to (full text and abstracts and human and english language) 5, quality of life.mp. or "Quality of Life"/ 98, limit 14 to (full text and abstracts and human and english language) 12, and and and and and Table H3: Search strategy used for CRD database # Searches CRD database carried out Monday 17/3 by database: Results NHS EED 2 DARE 1 HTA 11 Details of the exclusion criteria and incidence of articles utilised for Medline, Embase and the CRD database can be found in Table H

62 Table H4: Exclusion criteria and incidence of articles Medline 17/3/08 Criteria Number of articles excluded 2 nd line or later 4 Incl other diseases 0 Focus on one area eg sequential therapy 0 A review 0 None of the articles identified through Medline were relevant Embase 17/3/08 Criteria Number of articles excluded 2 nd line or later 6 Incl other diseases 2 Focus on just one aspect of treatment 2 A review 14 Other (comparison of use across European countries) 1 None of the articles identified through Embase were relevant CRD database 17/3/08 NHS EED 2 found in the second line or later setting DARE 1 found in the second line or later setting HTA 11 found of which 5 were horizon scanning documents and the remaining 6 were in the second line or later setting None of the articles identified through the CRD database were relevant Description of identified studies Provide a brief overview of each study, stating the aims, methods, results and relevance to decision-making in England and Wales. Each study s results should be interpreted in light of a critical appraisal of its methodology. Where studies have been identified and not included, justification for this should be provided. No studies were identified which addressed the decision problem

63 6.2 De novo economic evaluation(s) In the absence of a relevant published economic evaluation, manufacturers or sponsors should submit their own economic evaluation. When estimating cost effectiveness, particular emphasis should be given to adhering to the reference case (see the NICE document Guide to the methods of technology appraisal ). Reasons for deviating from the reference case should be clearly explained. Particularly important features of the reference case include those listed in the table below. Attribute Reference case Section in Guide to the methods of technology appraisal Comparator(s) The comparator that has been specified in the decision problem Perspective costs NHS and Personal Social Services Perspective benefits All health effects on individuals Form of economic evaluation Time horizon Synthesis of evidence Outcome measure Health states for QALY measurement Benefit valuation Source of preference data Discount rate Equity Sensitivity analysis Cost-effectiveness analysis Sufficient to capture differences in costs and outcomes Systematic review Quality-adjusted life 5.5 years (QALYs) Described using a 5.5 standardised and validated instrument Time trade-off or 5.5 standard gamble Sample of public 5.5 Health benefits and costs both 3.5% No additional weighting to QALYs Probabilistic sensitivity analysis

64 Introduction A de novo economic evaluation is presented in this submission. The economic evaluation compares the costs and health outcomes of patients with EGFRexpressing metastatic colorectal cancer with KRAS wild-type status and of two treatment strategies: o cetuximab in combination with fluoropyrimidine and irinotecan containing regimens compared to a fleuropyrimidine and irinotecan containing regimen alone o cetuximab in combination with fluoropyrimidine and oxaliplatin containing regimens compared to a fluoropyrimidine and oxaliplatin containing regimen alone From hereon in fluoropyrimidine and irinotecan or oxaliplatin containing regimens will be referred to as FOLFIRI or FOLFOX respectively. As the understanding of the influence of KRAS mutations on the response of tumours to EGFR inhibitors has developed, the EMEA have asked for Merck Serono to present data for cetuximab in the population of patients with wild-type KRAS. This has necessitated a re-evaluation of the data from the pivotal studies and re-submission of the regulatory dossier in line with the new results. It is anticipated that the licence will be restricted to the patient population with EGFR expressing, wild-type KRAS status and this subset forms the main evidence for the decision problem. This economic evaluation will focus on the expected licensed population and the following characteristics: o Good performance status o Suitable for irinotecan or oxaliplatin containing chemotherapy o Liver metastases EXCLUDING patients who are resectable at presentation The CRYSTAL and OPUS clinical trials both enrolled patients who were determined as having unresectable liver metastases. This economic evaluation will assess the impact of cetuximab plus FOLFIRI /FOLFOX upon rates of potentially curative resection among patients who became resectable during treatment for first line mcrc. Rates of resection will be modelled as per observed in the CRYSTAL and OPUS clinical trials and in Folprecht et al [2005] which may better reflect rates observed in UK clinical practice. The economic evaluation simulates the disease progression and survival of patients with mcrc throughout first and subsequent lines of therapy, including longer-term survival after successful curative surgery. The modelling of subsequent lines of treatment was used because neither the CRYSTAL or OPUS trials have not as yet generated mature overall survival data. Survival analysis has been utilised as a means of estimating the survival benefits of cetuximab from CRYSTAL and OPUS clinical trials for 1 st -line progression and death before progression. The basic model structure is identical for both comparisons. Please note a model manual can be found in appendix H

65 6.2.1 Technology How is the technology (assumed to be) used within the economic evaluation? For example, give indications, and list concomitant treatments, doses, frequency and duration of use. The description should also include assumptions about continuation and cessation of the technology. The economic evaluation compares the costs and health outcomes of patients with EGFR-expressing metastatic colorectal cancer with KRAS wild-type status and of two treatment strategies: 1. cetuximab in combination with fluoropyrimidine and irinotecan containing regimens compared to a fluoropyrimidine and irinotecan containing regimen alone 2. cetuximab in combination with fluoropyrimidine and oxaliplatin containing regimens compared to a fluoropyrimidine and oxaliplatin containing regimen alone From hereon in fluoropyrimidine and irinotecan or oxaliplatin containing regimens will be referred to as FOLFIRI or FOLFOX respectively. This is included for simplicity even though these abbreviations may cover a range of different administration techniques and dosages for 5-fluorouracil, folinic acid and irinotecan/ oxaliplatin regimens. It is assumed that the treatment regimens are as set out in the trial protocol and administered as recorded in the trial dataset [CRYSTAL and OPUS]. See also section in this submission. Cetuximab was administered as follows: Initial loading dose of 400 mg/m 2 on day 1 Subsequent doses of 250 mg/m 2 once per week until progression of the underlying disease. Cetuximab may be given with either of the regimens listed in tables H5 and H6 below. Table H5: Irinotecan containing (FOLFIRI) treatment regimen FOLFIRI Chemotherapy given every 14 days (1 hour after cetuximab infusion) Irinotecan 30 to 90-min infusion on day 1 Folinic acid 120-min infusion on day 1 5-Fluorouracil day mg/m² 400 mg/m² (racemic) or 200 mg/m² (L-form) before 5-FU bolus injection 400 mg/m² bolus on day 1 followed by a 46-hour continuous infusion of 2400 mg/m²

66 Table H6: Oxaliplatin containing (FOLFOX) treatment regimen Day 1 FOLFOX chemotherapy given (1 hour after cetuximab every 14 days infusion) Day 2 Oxaliplatin Infusion over 120 minutes 85 mg/m 2 - Folinic acid Infusion over 120 minutes 200 mg/m² 200 mg/m² 5-FU Bolus (over 2 4 minutes) followed by a 22-hour infusion 400 mg/m² bolus, followed by 22- hour continuous infusion of 600 mg/m² 400 mg/m² bolus, followed by 22-hour continuous infusion of 600 mg/m² Treatment in the model also takes into account the effects of subsequent lines of treatment. In the second-line setting the treatment regimen in the model is either FOLFOX or FOLFIRI as described in Tournigand et al [2004] dependant upon which first line treatment is utilised. Hence, if FOLFIRI is used in the first line, then FOLFOX will be used in the second line and vice versa. The specific FOLFOX 6 treatment regimen utilised in Tournigand et al [2004] is not exactly the same as FOLFOX commonly administered in the UK. However outcomes can be assumed to be similar for the different FOLFOX regimens and costs and outcomes of this treatment regimen are applied consistently for both arms of the model. In the third line setting patients receive Best Supportive Care (BSC), and these costs and outcomes are as described in the comparator arm of a third line clinical trial of cetuximab. This study was published by Jonker et al [2007] although in particular utility data is pending separate publication (academic in confidence) and is not reported in this paper. For simplicity, from hereon data from this study will be referred to as Jonker et al. Please see Table H7 below for further information

67 Table H7: Treatments in different lines within the economic model 1 (based on CRYSTAL study) Cetuximab + FOLFIRI 1st line before progression Cetuximab and FOLFIRI administered as in section FOLFIRI alone FOLFIRI administered as in section nd line, upon progression from 1st line For both arms: FOLFOX6 regimen. This regimen a 2-week schedule of administration as follows: Oxaliplatin, 100 mg/m2 Folinic acid 400 mg/m 2 in 2 hours in day 1 followed by fluorouracil 400 mg/m 2 bolus and g/m 2 in 46 hours 3 rd line Identical for both arms. The volume and cost of therapy can be different due to the different timing to progression to 2 nd line and due to patients receiving surgery with curative intent in the first line setting Duration of treatment Upon progression, or until curative surgery is performed. Some treatment with trial therapy is added after curative surgery, as observed in CRYSTAL Until progression to 3 rd line, the number of cycles of treatment in 2 nd line is estimated from literature Best Supportive Care Best Supportive Care Until death Treatment in the period following curative surgery Following successful curative surgery Following unsuccessful surgery Identical in both arms in the model Identical in both arms in the model Until death Until death Comment Since survival and treatment has been simulated the number of therapy cycles in the model was adjusted to achieve a similar cost of therapy in the model as in the CRYSTAL and OPUS trials. Volumes and types of services used are from estimates in Tournigand et al [2004] Crude monthly cost of best supportive care was estimated from Tappenden et al (2007) Patients move into a bucket health state which models survival and death based upon Adam et al. Patients who experience recurrence of disease enter 2 nd and 3 rd lines of therapy

68 6.2.2 Patients What group(s) of patients is/are included in the economic evaluation? Do they reflect the licensed indication? If not, how and why are there differences? What are the implications of this for the relevance of the evidence base to the specification of the decision problem? It is anticipated that the licence for cetuximab will be restricted to the patient population with EGFR expressing, wild-type KRAS status and this subset forms the main evidence for the decision problem. This economic evaluation will further restrict the intended target population as per the following characteristics: Good performance status Suitable for irinotecan or oxaliplatin containing chemotherapy Liver metastases EXCLUDING patients who are resectable at presentation i.e. those who will have more than 30% FRL (Future Remnant Liver) with clear margins and no invasion of the liver pedacle, inferior vena cava or the three hepatic veins. [see appendix H4 for further description of criteria for liver resection] Liver surgery can be performed on patients with metastatic colorectal cancer to excise metastatic disease. In patients who undergo surgery resulting in no residual disease, their prognosis is significantly improved and in some studies it approximates to the prognosis of patients without metastatic disease who have complete excision of their primary tumour. Indeed some patients are cured of their metastatic colorectal cancer. This resection of liver metastases can be performed at the same time as, or soon after surgical resection of the primary tumour (primary resection). Alternatively, additional treatment with chemotherapy or portal vein occlusion can be used to make the metastases more amenable to surgery and then resection is performed (secondary resection). There are describable characteristics that will determine whether metastases are resectable. With advances in surgical technique, the only absolute contraindications to surgery are defined in the European Consensus paper by Nordlinger et al (2007): The excision required would be so extensive as to leave less than 30% of the liver tissue behind. Involved coeliac lymph nodes are present or where there is evidence of disease outside of the liver that cannot be cleared There is invasion of the two branches of the liver pedicle or of the inferior vena cava or invasion of the three hepatic veins (although some centres have proposed that some of these may be resectable with the use of vascular reconstructive techniques using grafts) Of those patients where it is technically possible to resect metastases, not all may be suitable for resection due to prognostic indicators that are associated with poorer outcome, reducing the possible benefit of surgery. It is with these prognostic factors that there is debate within the medical community and no consensus exists. Two approaches to secondary resection exist:

69 1. The use of prognostic factors to select a population of patients who have liver metastases that are initially unresectable but that may become resectable following a course of chemotherapy 2. The observation that some patients receiving chemotherapy which is primarily intended to be palliative leads to the identification of patients who accidentally become amenable to resection The population of patients treated in the CRYSTAL and OPUS trials are representative of population 2 as defined above. Patients with multiple metastatic sites had a lower response rate in the CRYSTAL and OPUS trials than those whose metastatic disease was confined to the liver and the chance of patients in this group becoming amenable to metastatic surgery is much less. Therefore we are specifying that the population of patients pertinent to the decision problem are those with metastatic disease confined to the liver. In this group a number of patients will receive benefit in a palliative sense from treatment and a subgroup will receive additional benefit as their metastatic disease will respond to the extent that they may undergo secondary resection and their prognosis may significantly improve. It is important to note that the sample of patients with KRAS wild type and metastases confined to the liver was too small in both studies for formal subgroup analysis based on progression-free survival times (Crystal N=35 Cetuximab/FOLFIRI, N=32 FOLFIRI; OPUS: N=20 Cetuximab/FOLFOX and N=18 FOLFOX ). Therefore the PFS from all patients with KRAS wild type was used in the economic model as a proxy. Since the ITT datasets from both studies have shown that the incremental clinical benefit for patients with metastases confined to the liver was greater compared to the group with multiple sites of metastases, the use of PFS data from all KRAS wild type patients to represent KRAS wild type patients with metastases confined to the liver can be considered conservative. This group of patients is targeted in this economic evaluation in order to be pragmatic regarding the generalisability of clinical data available and to demonstrate both the economic value of bringing patients to the point of curative resection from a diagnosis of unresectability and the life years gained from such treatment. Table H8 and H9 presents demographic characteristics of the cohort of patients in the KRAS wild type subgroup of the CRYSTAL and OPUS trials respectively. Table H8: Demographic Characteristics of CRYSTAL KRAS Characteristic KRAS wild type patients n=348 Cetuximab plus FOLFIRI FOLFIRI alone Gender, n (%) Age, years ECOG status, n (%) Male Female Median Range n= (61.0) 67 (39) (57.6) 66 (38.4) 7 (4.1) n= (54.5) 80 (45.5) (59.1) 65 (36.9) 7 (4.0)

70 Table H9: Demographic Characteristics of OPUS KRAS Characteristic KRAS wild type patients n=134 Cetuximab plus FOLFOX alone FOLFOX Gender, n (%) Male Female Age, Median years Range Age, n (%) <65 years 65 years ECOG 0 status, n 1 (%) 2 n= (49.2) 31 (50.8) (63.9) 22 (36.1) 19 (31.1) 37 (60.7) 5 (8.2) n= (60.3) 29 (39.7) (63.0) 27 (37.0) 27 (37.0) 37 (50.7) 9 (12.3) The demographic characteristics of patients in the KRAS analysis for both CRYSTAL and OPUS clinical trials are similar to the patient population appropriate to this decision problem. Patients in the CRYSTAL trial were aged approximately 61 years with an ECOG status of mostly 0 or 1. In OPUS the median age was 59 with approximately 90% of patients having an ECOG score of 0 or 1. This is representative of patients in the UK as shown in market research data commissioned by Merck Serono and carried out by A+A market research. More details on A+A market research including methodology can be found in Appendix H1. Information pertinent to this section are presented in Table H10 below: Table H10: A+A Market research data, Demographic characteristics of UK mcrc patients treated in the first line setting (December 2007). Characteristic Irinotecan based treatment Oxaliplatin based treatment N (ITT) Fluoropyrimidine based monotherapy 170 (152 = oral fluoropyrimidine) Rectal cancer 17 (27%) 81 (28%) 31 (18%) Colon cancer 44 (71%) 209 (71%) 138 (81%) No answer Gender (n, %) Male 42 (67.7%) 181 (61.6%) 101 (59.1%) Female 20 (32.3%) 113 (38.4%) 69 (40.4%) Mean Age (S.D.) 63 yrs (8.4) 62 yrs (9.8) 71 yrs (8.8) Range ECOG performance status 0 19 (31%) 99 (34%) 19 (11%) 1 38 (61%) 165 (56%) 100 (59%) >1 4 (7%) 23 (8%) 50 (29%) No answer

71 This research suggests that patients in the UK treated with irinotecan and oxaliplatin based treatment in the first line setting are younger (62yrs,63yrs versus 71yrs respectively) and have a better performance status (ECOG of 0 or 1, 92%, 90% versus 70% respectively) than those treated with fluoropyrimidine monotherapy. Patient choice; adverse event profile and contraindications to treatment would also be taken into account when choosing between combination therapy and fluoropyrimidine monotherapy. It is also of interest that there is no significant difference between the profiles of 1 st line patients treated by oxaliplatin or by irinotecan. This information would suggest that fluoropyrimidine monotherapy regimens are not considered an appropriate comparator in this appraisal as clinical practice reserves such treatment for patients who are older or have a lower performance status. These patients may not be able to tolerate combination chemotherapy and therefore are different from the population considered in the decision problem. This economic evaluation also models beyond the first line setting for second and third line. Table H11 and H12 below present the demographic characteristics of the patients in these sections of the model. It is assumed that patients in the model who enter the 2nd and 3rd lines of treatments have similar characteristics to those shown in Tournigand [2004] and Jonker et al [2007]. Table H11: Demographic characteristics of subjects in 2 nd line of the model Characteristic FOLFIRI FOLFOX6 Mean age 61 Metastatic site Liver 87% Lung 31% Other 39% Number of sites 1 >=2 Source: Tournigand et al [2004] 59% 41% Table H12: Demographic characteristics of subjects in 3rd line of the model Characteristic Best Supportive Care Arm Median age 63.6 Median BSA 1.84 Source: Jonker et al [2007] Was the analysis carried out for any subgroups of patients? If so, how were these subgroups identified, what clinical information is there to support the biological plausibility of this approach, and how was the statistical analysis undertaken? As part of the regulatory process, the EMEA requested that Merck Serono perform a retrospective analysis of the patient population with wild-type KRAS tumour from the pivotal studies for cetuximab plus chemotherapy in the 1 st line treatment of metastatic colorectal cancer

72 Therefore the retrospective analysis based on the KRAS subset forms the main evidence for regulatory approval. At the current time the licence is anticipated to include the following: cetuximab in combination with chemotherapy is indicated for the treatment of patients with EGFR-expressing metastatic colorectal cancer with KRAS (Kirsten rat sarcoma) wild-type status. In addition to this subgroup defined by biomarker status, section has already defined a further subgroup based on liver metastases that will form the primary target population. The rates of resection employed in the base case were defined using the response rates observed in the subgroup of patients with metastatic disease confined to the liver and deriving a resection rate following treatment that could be applied to the population of patients with metastatic disease confined to the liver. Data described by Folprecht et al [2005] has been utilised to correlate response rate to resection rate. This paper reviewed resection rates in published studies and observed that there was a relationship between observed response rates and observed resection rates in treated populations. The relationship differed depending on whether the trials described all patients with unresectable metastatic colorectal cancer (unselected) or a subgroup of patients who had unresectable disease with metastases confined to the liver (selected). In the publication, Folprecht et al presented this correlation graphically (Figure H1 below) and it is this graph which will be used to model resection rates in the base case and scenarios in the model. Figure H1: Correlation of response rates to resection (Folprecht et al)

73 Were any obvious subgroups not considered? If so, which ones, and why were they not considered? Subgroups not considered in this economic evaluation included patients either not expected to be included in the expected marketing authorisation or groups of patients for whom sufficient information was not collected in the CRYSTAL or OPUS clinical trials. Patients with KRAS mutant status have not been evaluated in this submission. This subgroup was not considered because it is expected that the EMEA/CHMP will exclude this group from the marketing authorisation. Patients with poor performance status (ECOG >2) were also not considered because there are not sufficient patient numbers in the CRYSTAL and OPUS clinical trials to make conclusions. Also such patients would not generally be considered for aggressive combination chemotherapy At what points do patients enter and exit the evaluation? Do these points differ between treatment regimens? If so, how and why? Patients enter the economic evaluation at the point of diagnosis of mcrc. They exit at the point of death. Further lines of treatment are also modelled: Second line chemotherapy Best supportive care Surgical resection of metastases The model was developed in TreeAge using a Master and clone technique therefore entry and exit points are identical for both regimens in the model Comparator technology What comparator(s) was/were used and why was it/were they chosen? The choice of comparator should be consistent with the summary of the decision problem (Section A). Irinotecan and oxaliplatin based combinations will be addressed in this economic evaluation. Information is presented comparing the use of cetuximab plus FOLFIRI to FOLFIRI alone (based on the CRYSTAL trial KRAS data) and cetuximab plus FOLFOX to FOLFOX alone (based on the OPUS trial KRAS data). This is the most appropriate approach based on current NICE guidance for combination treatment in first line Metastatic colorectal cancer. In particular TAG 93 in August 2005 recommended the use of irinotecan and oxaliplatin, within their licensed indications as treatment options for people with advanced colorectal cancer in combination with 5-fluorouracil and folinic acid as first-line therapy, or irinotecan alone in subsequent therapy. Also in TAG 93 the Appraisal Committee concluded that it would not differentiate between these two drugs in terms of clinical effectiveness [section 4.3.5] and given the different side-effect profiles of the drugs [section ], the Committee accepted the importance of having a range of drug treatment options

74 Regimens based on fluoropyrimidine monotherapy (including IV 5FU regimens and oral 5FU analogues) alone are not considered an appropriate comparator in this appraisal as clinical practice reserves such treatment for patients who may not be able to tolerate combination chemotherapy which is different from the population considered for cetuximab. Please refer to section where A+A market research data demonstrates how patients treated with fluoropyrimidine alone are both older and have poorer performance status than patients defined for this decision problem Study perspective If the perspective of the study did not reflect NICE s reference case, provide further details and a justification for the approach chosen. The perspective of the economic evaluation matches that of the NICE Reference Case. Costs are estimated from the perspective of the NHS and all relevant disease and treatment health effects to the individual are captured via quality-adjusted life years (QALYs) Time horizon What time horizon was used in the analysis, and what was the justification for this choice? The cycle length of a Markov state in the model is one week therefore the costs and effects within the economic model are calculated on a per week basis. The Markov model is designed as a life-time model, the model runs over 1,200 cycles of one week per cycle which equates to just over 23 years and stops when the average age of the patients reached 80. This was felt to be an appropriate timescale to capture the majority of benefit in this patient group and was felt to be consistent with probable survival in a group of patients who enter the model at 60 years of age. Treatment in the first line setting increasingly affects mortality as the potential for curative surgery increases if the metastases are downsized sufficiently

75 6.2.6 Framework The purpose of this section is to provide details of the framework of the analysis. Section a) below relates to model-based evaluations, and section b) below relates to evaluations conducted alongside clinical trials. Please complete the section(s) relevant to the analysis. a) Model-based evaluations Please provide the following. A description of the model type. A schematic of the model. For models based on health states, direction(s) of travel should be indicated on the schematic on all transition pathways. A list of all variables that includes their value, range (distribution) and source. A separate list of all assumptions and a justification for each assumption. A life-time, semi-markov, deterministic model was developed in TreeAge Pro 2006/2007 software (TreeAge Software Inc., Williamstown, USA) to simulate the disease progression and survival in a cohort of patients with mcrc throughout 1 st and subsequent lines of therapy (2 nd and 3 rd lines), including longer-term survival after successful curative surgery. The model explores the impact of curative-intent surgical resection of liver metastases on cost-effectiveness. The economic model simulates a sequence of treatment lines using the data available from literature and from a third-line cetuximab trial (Jonker et al 2007). The modelling of subsequent lines of treatment was used because the CRYSTAL trial has not yet generated mature overall survival data. In order to model the short and long-term outcomes of treatment in patients within the cost effectiveness model, survival analyses have been undertaken as a means of estimating the survival benefits of cetuximab, as follows: The first line is based on the results of CRYSTAL and OPUS for 1 st -line progression and death before progression The second line is based on the results from Tournigand et al [2004] (for 2 nd line progression-free survival) The third line is based on the results from the Jonker et al 3 rd -line trial (for 3 rd line progression-free survival) Figure H2 presents a simplified illustration of the model structure. Patients enter the model at the beginning of the trial period, i.e. appropriate for 1 st line therapy. Patients can then either stay in this health state or move to four alternatives: Successful curative resection Unsuccessful surgery with curative intent Second line therapy (after disease progression) Dead (through Death before progression)

76 If a patient is appropriate for resection of liver metastases with curative intent then this surgery can be successful or unsuccessful. After successful curative surgery, patients are considered tumour-free and will a have longer estimated mean life expectancy of 4.76 years, with the observed median survival time of 3.23 years (as estimated from Adam et al [2004]). Patients who move from the resection health states go to either: Patients with, successful curative resection move to either a survive following surgery state or die state with a background risk similar to the general population Or Patients with unsuccessful curative resection move to receive further lines of chemotherapy and flow back into the second line health state of the model (not depicted on graphic) and continue until death. For patients with successful curative surgery the survive following curative surgery state should be regarded as a bucket state. During this state the probabilities of progression and death are modelled using the survival curves from Adam et al [2004]. Once a patient enters the second line therapy health state then they can either stay in this health state or move to 2 alternatives: Third line therapy (after disease progression) Dead (through Death before further progression) Once a patient enters the third line therapy health state then they can either stay in this health state or die (move to health state dead)

77 Figure H2: Simplified Model Schematic Potentially curative resection of liver metastases 1 st line therapy (starts here) Death before Progression Successful resection Un successful resection Progression 2 nd line therapy Dead Further progression Death before Further progression 3 rd line therapy Death The main health benefits will be expected to be accrued in the 1 st line treatment setting due to: The advantage of cetuximab-based therapy in extending progression-free survival In the period post curative resection of liver metastases. Patients observed in the CRYSTAL and OPUS trials both reported increased response rates due to the addition of cetuximab to the relevant comparators. This enables increased rates of resection of the initially unresectable liver metastases and to increased survival in those patients who achieve cure. Disease progression over each of these different settings is modelled over first and subsequent lines of treatment. Further details can be found in section A list of all variables that includes their value, range (distribution) and source. Key modelling assumptions are presented below in Table H

78 Table H13: Key Modelling assumptions Section of the model Description of the state in the model Outcomes and data sources Service use and costs 1st line treatment This is the treatment stage where the biggest incremental benefit is expected due to the addition of cetuximab to the treatment regimen. Based on either the KRAS wild type ITT CRYSTAL or OPUS datasets. Patients remain in this state until: they die before progression progress into 2 nd line or are referred for curative intent liver surgery Death before progression and progression into second line are derived from survival analysis of the KRAS wild type subgroup of either the CRYSTAL or OPUS study data. Death before progression and progression into second line are described by a conditional probability of progression after deaths are excluded. All patients start 1st line treatment and receive either cetuximab + FOLFIRI or FOLFIRI alone for the CRYSTAL evaluation. For the OPUS evaluation they will receive either cetuximab plus FOLFOX or FOLFOX alone While patients stay progression-free their health-related utility of life is assumed constant at For both evaluations we have used the cost of branded oxaliplatin as 3.3 per mg and not the generic price. The variation for cost per QALY between this price and the generic list price is 0.09% for the CRYSTAL evaluation and 0.11% for the OPUS evaluation. Hence the branded oxaliplatin price has been utilised in the model This utility is derived from CRYSTAL UK based EQ 5D data. No utility data is available for OPUS At month 3 (13 weeks in the model) UK clinical opinion would indicate some patients can be referred for curativeintent resection of liver metastases. Upon referral to curative intent surgery, patients can die in the postoperative period, and in some patients surgery can fail to achieve curative intent i.e. those who are not considered to have an R0 resection. The model assumes that those patients for whom liver resection failed to achieve cure return to the progression-free state. The model will then treat them the same as those who started in first line and they may then die before progression or progress to second and then 3rd lines of treatment. No second surgery with curative intent is allowed in the model. The three months cut off was advised by a Merck Serono Advisory Board held in December 2007 (see Appendix H4). Postoperative death was 0% in the CRYSTAL study and set to 0 in the model. However in other studies postoperative death up to 2.6% was observed. Curative intent surgery Longer-term survival after successful curative surgery of liver metastases. In this section of the model patients either stay alive or die, with life mean life expectancy of 4.76 years as estimated by parametric approximation of the overall survival curve from the Adam et al [2004] study. Service use is based on estimates of the treatment costs in progressed patients. The proportion of progressionfree patients at each point in time is estimated from PFS published by Adam et al [2004]. Rate of surgical failure. The proportion of patients where complete excision of the tumour is not possible. 28% of patients in the cetuximab plus FOLFIRI arm of the CRYSTAL study failed surgery. This rate is applied consistently to both arms of the model and applied in both CRYSTAL and OPUS evaluations

79 Section of the model Description of the state in the model Outcomes and data sources Service use and costs 2nd line treatment on progression from the 1st line Patients move to this health state after disease progression in the first line health state Patients remain in this health state until: they die before progression progress into 3rd line Progression-free survival in 2nd line is derived from the PFS curves published in Tournigand et al [2004], regardless of the time of progression from the first line. It is assumed that the health related utility weight in the second line is lower than the utility weight in the 1st line (0.77) and higher than the utility weight in the 3rd line (0.68) and is represented by their average (0.73). Probability of [further] progression in the 2nd line is identical for both arms in the model. A slight difference in health gains in 2nd line can be observed due to the different mean times to progression in the cetuximab + FOLFIRI and FOLFIRI alone arms and the effect of discounting after year 1. A systematic review of literature has been conducted in order to ascertain the availability of evidence supporting these treatment regimens for second-line mcrc (see the literature review report in appendix H). This review produced only 3 relevant papers. Two of these were considered to be not appropriate. The clinical trial by Tournigand et al (2004) was considered to fit the search criteria and forms the basis of the second line segment of the model. The main critique is the use of FOLFOX-6 It can be assumed that the FOLFOX 6 regimen has similar outcomes to the use of UK oxaliplatin based combinations. This is validated by reported results as follows: 4.6 months PFS reported by Rothenberg (2003); 4 months reported by Scartozzi (2005) for secondline FOLFOX4. This is not dissimilar to the 4.2 months reported for second-line FOLFOX6 by Tournigand et al (2004). 3rd line treatment, on progression from 2nd line This is the final treatment line in the model. Patients remain in this health state until: Death before progression Death after progression The probability of death is derived from the results of Jonker et al comparing treatment with cetuximab + best supportive care to best supportive care alone. Similar to 2nd line therapy, the risk of death on progression into second line does not depend on the time the patient progressed from the 2nd line. It is assumed that patients would receive best supportive care, with no systemic therapy. The crude monthly cost of best supportive care from this study was applied in the model Why was this particular type of model used? This particular choice of model was used to make the full use of the clinical data available. At the time of submission to NICE neither the CRYSTAL or OPUS trials have, as yet, generated mature overall survival data. Overall survival is extrapolated from the available progression free survival curves using survival analyses

80 Additionally, it was decided to utilise a Markov model design to adequately capture the costs, benefits and adverse events associated with the treatment of mcrc What was the justification for the chosen structure? How was the course of the disease/condition represented? Please state why any possible other structures were rejected. The modelling structure was designed to reflect UK clinical practice where patients can receive up to three lines of treatment including best supportive care. Hence it is necessary to model such further lines of treatment in order to capture the expected full overall survival benefits. This particular modelling structure and health states employed are similar to that reported in the NICE assessment of irinotecan and oxaliplatin (TAG 93) What were the sources of information used to develop and inform the structure of the model? Several sources of information were used to develop the model. Firstly, an in-depth literature review was carried out to search for publication of other economic evaluations in the first line setting of mcrc. This included a review of all NICE published Technical Assessment Reports (TAR) for review of methods employed in the appraisal of mcrc treatments. Additionally consultation with individual clinical experts and a full advisory board were utilised to test assumptions and validate the approach. Finally, an independent Health Economics expert was consulted to validate the modelling techniques employed Does the model structure reflect all essential features of the condition that are relevant to the decision problem? If not, why not? The model structure considers disease progression, the possibility of curative liver resection and subsequent lines of treatment. Therefore all essential features that are relevant to the decision problem are considered For discrete time models, what was the model s cycle length, and why was this length chosen? Does this length reflect a minimum time over which the pathology or symptoms of a disease could differ? If not, why not? The cycle length of the model is one week. This cycle length was chosen to accurately capture the costs and benefits associated with treatment of mcrc. Routine assessment of patients with mcrc varies from centre to centre including the the frequency of scans. (see Appendix H)

81 Was a half-cycle correction used in the model? If not, why not? The cycle length in the model is one week. It was felt that this would accurately capture any costs and benefits accrued Are costs and clinical outcomes extrapolated beyond the trial follow-up period(s)? If so, what are the assumptions that underpin this extrapolation and how are they justified? In particular, what assumption was used about the longer-term difference in effectiveness between the technology and its comparator? Costs and outcomes are extrapolated beyond the trial period using survival analyses. In the base case CRYSTAL evaluation a Weibull method was employed. This is presented below in Figure H3. Figure H3: CRYSTAL evaluation, Simultaneous Weibull model "Real" progression in 1st line (both arms - joint Weibull) Time to progression and Hazard Function Probability of PFS Hazard Time (weeks) Survivor function (Cetux) Survivor function (Ctx) K-Meier-PFS-Cetux K-M-PFS-CTx Hazard function (cetux) Hazard function (ctx) The Weibull method was selected compared to other techniques examined it most closely modelled best fit to both arms of data. Further rationale for selection of the Weibull method can be found in Appendix H3. In the base case OPUS evaluation a lognormal method was employed. This is presented below in Figure H

82 Figure H4: OPUS evaluation, log normal model Lognormal parametric model for PFS in 1st line OPUS study 0.8 % of progresion free Cetux (Lognormal) Cetux (K-M) CTx (Lognormal) CTx (K-M) time (weeks) Further rationale for selection of the method of survival analyses can be found in Appendix H3. b) Non-model-based economic evaluations Was the evaluation based on patient-level economic data from a clinical trial or trials? The evaluation is model based and not based on patient level economic data from a clinical trial. Hence the remainder of this section is not completed Provide details of the clinical trial, including the rationale for its selection. Not applicable Were data complete for all patients included in the trial? If not, what were the methods employed for dealing with missing data for costs and health outcomes? Not applicable Were all relevant economic data collected for all patients in the trial? If some data (for example, resource-use or health-related utility data) were collected for a subgroup of patients in the trial, was this subgroup prespecified and how was it identified? How do the baseline characteristics and effectiveness results of the subgroup differ from those of the full trial population? How were the data extrapolated to a full trial sample? Not applicable

83 Are costs and clinical outcomes extrapolated beyond the trial follow-up period(s)? If so, what are the assumptions that underpin this extrapolation and how are they justified? In particular, what assumption was used about any longer-term differences in effectiveness between the technology and its comparator? Not applicable Clinical evidence Where relevant, answers to the following questions should be derived from, and consistent with, the clinical evidence section of the submission (section 5). Crossreferences should be provided. If alternative sources of evidence have been used, the method of identification, selection and synthesis should be provided and a justification for the approach provided How was the baseline risk of disease progression estimated? Also state which treatment strategy represents the baseline. Disease progression was based on the wild type KRAS sub group analyses of the CRYSTAL and OPUS clinical trials. To date neither trial has generated mature overall survival data, hence survival analyses are employed. The treatment strategy which represents the baseline is FOLFIRI from the CRYSTAL evaluation and FOLFOX from the OPUS evaluation How were the relative risks of disease progression estimated? In order to model the short and long-term outcomes of treatment in patients within the Markov cost effectiveness model, survival analyses using the Weibull technique have been undertaken as a means of estimating the survival benefits of cetuximab, based upon data from the CRYSTAL trial (for 1 st -line progression and death before progression). Disease progression in the economic model is assessed over five separate settings/ treatments each utilising the most suitable modelling technique to simulate the data as presented in table H14 below:

84 Table H14: CRYSTAL and OPUS economic models: Extrapolation technique employed for different settings of the model Setting First line, death before progression CRYSTAL Evaluation Modelling Technique Log normal Original source of data CRYSTAL wt KRAS analyses First line time to progression Weibull CRYSTAL wt KRAS analyses Second line time to progression Log normal Tournigand et al Modelling Technique Log normal Log normal Log normal OPUS Evaluation Original source of data OPUS wt KRAS analyses OPUS wt KRAS analyses Tournigand et al Third line time to progression Weibull Jonker et al Weibull Jonker et al Post curative surgery survival Log logistic Adam et al Log logistic Adam et al PFS following curative surgery Log logistic Adam et al Log logistic Adam et al Further graphical information on the survival analyses employed can be found in Appendix H Were intermediate outcome measures linked to final outcomes (such as patient survival and quality-adjusted life years [QALYs])? If so, how was this relationship estimated, what sources of evidence were used, and what other evidence is there to support it? Intermediate measures are linked to final outcomes throughout the analyses and through each different setting. Patient cost and outcomes are accumulated on a weekly basis as per the clinical setting and patient clinical characteristics. Progression free survival data from the CRYSTAL and OPUS studies are modelled using survival analyses to provide patient survival data to feed into QALY derivation Were the health effects or adverse effects associated with the technology included in the economic evaluation? If not, would their inclusion increase or decrease the estimated cost effectiveness of this technology? The adverse effects of cetuximab are considered in both cost and outcome elements of the economic evaluation. Utility data for the first line setting are taken from the CRYSTAL trial and EQ-5D data collected in the UK. The utility figure applied for progression free survival is considered equal for both arms. This figure (0.77) is calculated by taking an average figure of all completed data on the EQ-5D over the length of the first line treatment period, and as a result adverse events are included within this

85 Was expert opinion used to estimate any clinical parameters? If so, how were the experts identified, to which variables did this apply, and what was the method of elicitation used? Expert opinion was used to guide and validate decisions of parameter values and ranges to be employed in sensitivity analyses. In particular, the rate of curative liver resection observed in the UK was assessed to understand the most appropriate resection rates from Folprecht et al that could be employed. In addition the duration of therapy deemed appropriate to assess the possibility of curative resection was three months [personal communication Mr Graeme Poston]. Further clinical parameters defined by clinical opinion can be found in Appendix H What remaining assumptions regarding clinical evidence were made? Why are they considered to be reasonable? The CRYSTAL and OPUS clinical trials consist of patients who are ineligible for curative surgery. As described in section of the clinical section, reported rates for curative resection of liver metastases were relatively low in the CRYSTAL study. This was the case for both the Full Analysis Set (FAS) and KRAS wild type populations, although there was a difference in liver resection rates between the two treatment arms in favour of the cetuximab group. Data from the KRAS wild type subgroup analyses reports greater progression free survival and response rates; however the reduced sample size does not allow assessment of the number of patients who undergo curative resection. In particular, rates of potentially curative liver surgery was low compared to rates observed in UK clinical practice in which approximately 10-20% of patients may become eligible for curative liver resection after previously being defined as not resectable [personal communication Mr Graeme Poston]. Table H15 and H17 below presents response and resection rates observed in the CRYSTAL FAS and KRAS wild type populations. Table H16 and H18 presents response and resection rates observed in the OPUS FAS and KRAS wild type populations

86 Table H15: Surgery with curative intent (Full dataset ITT Population) in the CRYSTAL study Characteristic Cetuximab + FOLFIRI (n=599) FOLFIRI alone (n=599) Response rate 46.9% 38.7% Subjects with any surgery with curative intent, n (%) No residual tumour after resection p-value (CMH test) Odds ratio Failure to achieve curative resection in those operated with curative intent p= (1.39, 6.49) 36 (6.0%) 15 (2.5%) 26 (4.3%) 9 (1.5%) 27.8% 40% Table H16: Surgery with curative intent (Full dataset ITT Population) in the OPUS study Characteristic Cetuximab + FOLFOX (n=169) FOLFOX alone (n=168) Response rate 45.6% 35.7% Subjects with any surgery with curative intent, n (%) No residual tumour after resection p-value (CMH test) Odds ratio Not calculated 6* (3.5%) 4 (2.3%) 4 (2.3%) 4 (2.3%) Failure to achieve curative resection in those operated with curative intent 33 % 0% *Differences between Tables H16 and H18 are due to clinical cut points for each analysis

87 Table 17: Surgery with curative intent (KRAS wild type Population) in the CRYSTAL study Characteristic Cetuximab + FOLFIRI (n=172) FOLFIRI alone (n=176) Response Rate 59.3% 43.2% Subjects with any surgery with curative intent, n (%) No residual tumour after resection p-value (CMH test) Odds ratio Failure to achieve curative resection in those operated with curative intent P= (0.23, 9.22) 6 (3.5%) 4 (2.3%) 3 (1.7%) 2 (1.1%) 50% 50% Table H18: Surgery with curative intent (KRAS wild type Population) in the OPUS study Characteristic Cetuximab + FOLFOX (n=61) FOLFOX alone (n=73) Response rate 60.7% 37% Subjects with any surgery with curative intent, n (%) No residual tumour after resection p-value (CMH test) Odds ratio 7* (11.5%) 3 (4.1%) 6 (9.8%) 3 (4.1%) Not Done Failure to achieve curative intent resection in those operated with curative intent 15.3 % 0% *Differences between Tables H16 and H18 are due to clinical cut points for each analysis The impact of potential curative liver resection is an important factor in the modelling because mean overall survival time in patients after curative resection is estimated at 4.76 years compared to mean survival of years in those without curative surgery. The rates of resection employed in the base case were defined using the response rates observed in the subgroup of patients with metastatic disease confined to the liver and to derive a resection rate following treatment that could be applied to the population of patients with metastatic disease confined to the liver. Data described by Folprecht et al has been used to correlate response rate to liver resection rate. This paper reviewed resection rates in published studies and observed that there was a relationship between observed response rates and observed resection rates in treated populations. The relationship differed depending on whether the trials described all patients with unresectable metastatic colorectal cancer (unselected) or a subgroup of patients who had unresectable disease with metastases confined to the liver (selected). The response rates observed in the full population of patients with KRAS wild type tumours are presented in Table H19:

88 Table H19: Response rates observed in the KRAS wild type CRYSTAL AND OPUS analyses FOLFIRI/FOLFOX RR (%) CETUXIMAB PLUS FOLFIRI/FOLFOX RR (%) CRYSTAL OPUS These results can be correlated to the full population of patients that could receive cetuximab in combination with fluoropyrimidine based chemotherapy combined with oxaliplatin or irinotecan. These trials included an unselected population and so the appropriate curve described by Folprecht et al is the lower curve as presented in Figure H5. In Figures H5 H7 presented below have been colour coded to represent data for the CRYSTAL evaluation in blue and the OPUS evaluation in red. Figure H5: Response rate observed in the KRAS wild type CRYSTAL AND OPUS analyses related to Folprecht et al. By applying the RR in the CRYSTAL and OPUS studies to this curve, gives liver resection rates as shown in Table H20. Table H20: Resection rates derived from Folprecht et al unselected population CETUXIMAB PLUS FOLFIRI/FOLFOX (%) FOLFIRI/FOLFOX (%) CRYSTAL 9 15 OPUS 7 15 The response rates observed in the population of patients with KRAS wild type tumours and metastatic disease confined to the liver were as presented in Table H

89 Table H21: Response rates observed in the KRAS wild type CRYSTAL AND OPUS analyses (liver metastases only) CETUXIMAB PLUS FOLFIRI/FOLFOX RR (%) FOLFIRI/FOLFOX RR (%) CRYSTAL OPUS* * For OPUS, data is not available for response rates in the liver only metastases population and so a conservative estimate has been made for resection using the observed resection rates in the ITT analysis of the KRAS subgroup. These results can be correlated to the population of patients, with metastatic disease confined to the liver, who could receive cetuximab in combination with fluoropyrimidine base chemotherapy combined with oxaliplatin or irinotecan. This is a selected population and so the appropriate curve described by Folprecht is the upper curve as in Figure H6: Figure H6: Response rate observed in the KRAS wild type CRYSTAL AND OPUS analyses (liver metastases only) related to Folprecht et al selected patients. Table H22 shows the corresponding resection rates when we assume the use of the selected group from Folprecht et al and use the KRAS WT response rates for CRYSTAL. Table H22: Resection rates derived from Folprecht et al selected population CETUXIMAB PLUS FOLFIRI/FOLFOX (%) FOLFIRI/FOLFOX (%) CRYSTAL OPUS

90 Alternatively These results can also be modelled assuming that the ultimate treatment population will not be as selected as in the studies considered selective by Folprecht et al. This unselected population can be modelled thus: See Fig H7and Table H23. Figure H7: Response rate observed in the KRAS wild type CRYSTAL analyses (liver metastases only) related to Folprecht et al unselected patients. Table H23 below shows liver resection rates derived from: CRYSTAL wild type KRAS population Response rate for patients with liver only metastases (77%) The unselected curve from the Folprecht graph These resection rates approximate to those seen in UK clinical practice [personal communication Graeme Poston] Table H23: Resection rates derived from Folprecht et al unselected population CONTROL resection rate % ACTIVE (+cet) resection rate % CRYSTAL OPUS Data Not available These are the resection rates determined as most appropriate for use in the base case analyses

91 6.2.8 Measurement and valuation of health effects Which health effects were measured and how was this undertaken? Health Overall survival effects include both those that have a positive impact and those with a negative impact, such as adverse events. The economic model measures cost-effectiveness primarily in terms of costs per life year gained and per QALY gained in the first line. Survival analyses were employed to assess overall survival. A number of parametric models were assessed to extrapolate the observed progression-free survival data. The detail of the parametric models can be found in appendix H2. The base case analyses for CRYSTAL utilises a Weibull technique to model overall survival while the model for OPUS uses a log normal technique to model the OS survival data. The CRYSTAL and OPUS studies both report a numerically greater rate of curative resection of liver metastases in the groups that received cetuximab plus chemotherapy. Therefore the rate of curative resection and survival has been included in the model. Health related utility weights are applied to the time lived with disease at different stage of disease progression in the Markov model. Heath-related utilities were drawn from the results of two clinical trials in the first and third line setting and estimated for the second line setting. These are presented below in Table H24. EQ-5D data was not collected in the OPUS study. Table H24: Health-related utilities Heath state 1 st line treatment, progression-free 2 nd line treatment, progression-free Utility weight Source 0.77 CRYSTAL clinical trial Comments and assumptions UK EQ-5D data based on 42 patients is derived as an average of all EQ-5D completed responses over the study period (baseline to week 40) 0.73 Estimate An estimate based on the assumption that utility in 2 nd line would be lower than in 1 st line and higher than in 3 rd line. 3 rd line 0.68 Jonker et al [2007] Utility weights are measured in the trial using the HUI. Utilities did not rapidly deteriorate over the period of 0-24 weeks in 3rd line, and therefore the average was assumed until death. Possibly lower utility weights in the terminal stage were not captured in the trial and in the model. Figure H8 below presents the UK EQ-5D data from baseline to week 40 from the CRYSTAL trial

92 Figure H8: CRYSTAL UK EQ-5D data from baseline to week Utility Baseline Time (Weeks) cetuximab plus FOLFIRI FOLFIRI Table H25 below presents the patient numbers at each time point for EQ-5D data collected in the UK. TABLE H25: Patient numbers of UK EQ-5D data collected in CRYSTAL Baseline Cetuximab plus FOLFIRI FOLFIRI This quality of life maintenance is also demonstrated in information collected with the EORTC-QLQ, disease specific instrument, and presented below in Figure H9. Figure H9: CRYSTAL EORTC QLQ data from baseline to week EPRTC QLQ Global Health Status Baseline Time (weeks) cetuximab plus FOLFIRI FOLFIRI

93 Table H26 below presents the patient numbers at each time point for the EORTC - QLQ data collected in CRYSTAL. TABLE H26: Patient numbers of EORTC - QLQ data collected in CRYSTAL Baseline Cetuximab plus FOLFIRI FOLFIRI Further detail of the selection of utility weights can be found in Appendix H5 on Utility weights literature. Utility in the period following curative surgery takes into account utility in patients free of disease and patients with recurrent disease. It was assumed that patients free of disease have health-related utility equal to that of the general population. In patients with progressive disease the utility was estimated as the weighted average of utilities in the third line (0.68) and second line (0.73). EQ-5D utility scores reported for the general population can be found in Table H27 below (Petrou et al, 2005). Table H27: EQ-5D utility scores reported for the general population Age Overall utility score > Which health effects were valued? If taken from the published literature, how and why were these values selected? What other values could have been used instead? If valued directly, how was this undertaken? The economic model measures cost-effectiveness primarily in terms of costs per life year gained and per QALY gained in the first line Were health effects measured and valued in a manner that was consistent with NICE s reference case? If not, which approach was used? Clinical evidence was based on the CRYSTAL and OPUS clinical trials and not on a systematic review as in the Reference Case. A systematic review was not necessary in this case as the CRYSTAL and OPUS trials are the only head-to-head, randomised controlled trials available which meet the decision problem. No relevant clinical evidence was knowingly omitted. In line with the Reference Case, health state utilities were valued using non-diseasespecific instruments where possible. In the first line setting EQ-5D data from 42 UK patients in the CRYSTAL clinical trial is used. EQ-5D data was only collected in the

94 UK, hence the low sample size. In the third line setting HUI data from Jonker et al [2007] is used, and in the second line setting an estimate is derived as a mid way point between the first and third line utility estimates derived. In the case of successful curative resection, patients are assigned the utility value as published for the general population derived through the EQ-5D as presented in table H28 (Petrou et al 2005). Further detail on utilities employed in the economic model can be found in section Were any health effects excluded from the analysis? If so, why were they excluded? No health effects were excluded from the model If health effects were not expressed using QALYs, what health outcome measure was used and what was the justification for this approach? Health effects were expressed as QALYs Resource identification, measurement and valuation What resources were included in the evaluation? (The list should be comprehensive and as disaggregated as possible). The categories of resources in the model include: cost of 1 st line chemotherapy drugs (cetuximab, irinotecan, folinic acid, 5- fluorouracil) cost of administration in the 1 st line cost of curative intent liver surgery cost of post-resection therapy in patients who had curative result of the liver metastases operation cost of management of adverse events in the 1 st line cost of treatments in 2 nd line cost of treatment in 3 rd line, and cost of KRAS screening How were the resources measured? The costs in the model were estimated from a health service perspective and uses UK NHS cost. Further detail can be found below in Tables H28 and H

95 Table H28: Cost Variables included in the model. Unit Unit cost, Source Detail of costs Cetuximab, 100 mg UK British National Formulary number 55 Erbitux ( Merck): Intravenous infusion, cetuximab /20ml (100mg)vial, /100ml (500mg) vial Irinotecan, 1 mg 1.30 UK BNF 55 Campto (Pfizer): Concentrate for intravenous infusion, irinotecan hydrochloride 20 mg/ml, net price 2- ml vial = 53.00; 5-mL vial = Folinic acid, 1 mg 0.39 UK BNF 55, price based on the median across various vial sizes 5-fluorouracil, 1 mg UK BNF 55, price based on the median across various vial sizes Calcium Folinate (Non-proprietary): Injection, folinic acid (as calcium salt) 3 mg/ml, net price 1-mL amp = 2.28, 10-mL amp = 4.62; 7.5 mg/ml, net price 2-mL amp = 7.80; 10 mg/ml, net price 5-mL vial = 19.41, 10-mL vial = 35.09, 30-mL vial = 94.69, 35-mL vial = Fluorouracil (Non-proprietary): Injection, fluorouracil (as sodium salt) 25 mg/ml, net price 10-mL vial = 3.20, 20-mL vial = 6.40, 100-mL vial = 32.00; 50 mg/ml, 10-mL vial = 6.40, 20-mL vial = 12.80, 50-mL vial = 32.00, 100-mL vial = Oxaliplatin, 1 mg UK BNF 55 Oxaliplatin (Non-proprietary): Injection, powder for reconstitution, oxaliplatin, net price 50-mg vial = , 100-mg vial= Eloxatin (Sanofi-Aventis): Concentrate for intravenous infusion, oxaliplatin 5 mg/ml, net price 10-mL vial = , 20-mL vial = (used in the model) Outpatient drug administration, 1 infusion Day case chemotherapy administration Estimated cost of liver resection with curative intent Cost of an adverse event in 2nd-line, outpatient visit Cost of an outpatient attendance for grade 3/4 adverse event Cost of an outpatient attendance for a serious adverse event UK NHS National Tariff, UK NHS National Tariff, ,271.2 Estimated from the UK NHS National Tariff Estimated from the UK NHS National Tariff, see detail in the Appendix on adverse events Estimated from the UK NHS National Tariff, see detail in the Appendix on adverse events Same as above TOPS-FUA. Outpatient Adult Follow Up Attendance. HRG code 370F. Medical Oncology [Attendance without Treatment] : Face to Face Total Attendances. TDC day case care. HRG F98. Chemotherapy with a Digestive System Primary Diagnosis Estimated as the average of HRGs related to liver resection weighted by the number of finished consulting episodes

96 Unit Unit cost, Source Detail of costs Cost of hospitalisation for a non-serious adverse event Cost of a serious adverse event requiring hospitalisation Cost of hospitalisation for SAE in 2nd-line model Cost of oncology outpatient attendance Cost of abdominal CTscan Cost of Liver function tests (LFT) 1,050 Based on the mapping of the types of adverse events onto the UK HRGs that can likely be assigned to them (see Appendix on adverse events) 1,170 Based on the mapping of the types of adverse events onto the UK HRGs that can likely be assigned to them (see Appendix on adverse events) 1,170.8 Same as above (assumption) UK NHS National Tariff, Used in the costing of monitoring 22.0 Used in the costing of monitoring, although the cost is small and will not noticeably affect the costeffectiveness TOPS-FUA. Outpatient Adult Follow Up Attendance. HRG code 370F. Medical Oncology [Attendance without Treatment]: Face to Face Total Attendances. Christie NHS Hospital, NHS Foundation Trust, R&D Price List Christie NHS Hospital, NHS Foundation Trust, R&D Price List Cost of KRAS test 300 Verbal communication from Medical Solutions based on ad hoc patient testing would be approximate 300. The model takes into account the cost of testing for the full population. The cost of KRAS screening takes into account the whole patient population and was estimated as 462 per patient classified as wild type KRAS status. This calculation assumes that each test costs 300 and approximately a third of patients would be detected as KRAS mutated status. Therefore the patients would not be appropriate for cetuximab plus FOLFIRI treatment, even thought the cost of the test would be incurred. Further information on the costing of adverse events can be found in Appendix 8. Table H29: Estimating the cost of liver resection HRG Code Health Care Resource Group No. of FCE (Finished Consulting Episodes) G02 Liver - Complex Procedures 1,183 7,222 G03 Liver - Very Major Procedures 1,366 3,280 G04 Liver - Major Procedures >69 or w cc 1,695 1,810 G05 Liver - Major Procedures <70 w/o cc 5, Weighted average, per operation 2,271 National Average Unit Cost

97 Were the resources measured using the same source(s) of evidence as the baseline and relative risks of disease progression? Disease progression was based on data from the CRYSTAL and OPUS clinical trials and resources were measured via outcomes reported in the clinical trial dataset Were resources used to treat the disease/condition included for all relevant years (including those following the initial treatment period)? Provide details and a justification for any assumptions that were made (for example, assumptions regarding types of subsequent treatment). Resources used to treat the disease/condition were included for all relevant years What source(s) of information were used to value the resources? The costs in the model were estimated from a health service perspective and uses UK NHS cost. Further detail can be found below in Table H29 in section What is the unit cost (excluding VAT) of the intervention(s) included in the analysis? Does this differ from the (anticipated) acquisition cost reported in section 1? The unit cost of cetuximab is as included in section 1 of the submission and as presented below: /20ml (100mg) vial /100ml (500mg) vial In the economic evaluation, results presented take into account the number of vials of cetuximab utilised per patient dose. This was calculated from an assessment of vials used in the KRAS analyses of both CRYSTAL and OPUS and an average per patient defined Were the resources measured and valued in a manner consistent with the reference case? If not, how and why do the approaches differ? Resources were measured and valued in a manner consistent with the Reference Case. The resources measured were those under the control of the NHS and they were valued from the perspective of the NHS Were resource values indexed to the current price year? In all cases, the most recently published (at the time of analysis) unit cost source was utilised, therefore it was not necessary to index costs to the current price year

98 Provide details of and a justification for any assumptions that were made in the estimation of resource measurement and valuation. All modelling assumptions are included in Table H in section Time preferences Were costs and health benefits discounted at the rates specified in NICE s reference case? The economic model applies discounting to both costs and outcomes at the annual discount rate of 3.5% Sensitivity analysis Sensitivity analysis should be used to deal with sources of main uncertainty other than that related to the precision of the parameter estimates. For technologies whose final price/acquisition cost has not been confirmed, sensitivity analysis should be conducted over a plausible range of prices Which variables were subject to sensitivity analysis? How were they varied and what was the rationale for this? Table H30 below lists the full range of sensitivity analyses conducted including the Base case values and low and high inputs. All costs are in pounds sterling

99 Table H30 Parameters in the univariate sensitivity analysis for both CRYSTAL and OPUS models Parameter Low input Base case High Input Annual discount rate for effects Utility weight progression free Utility in progressive disease (2nd line) Utility in 3rd line Cost of 1 KRAS test Annual discount rate for costs Cost of a serious adverse event requiring hospitalisation 995 1,152 1,346 Cost of an outpatient attendance for grade 3/4 adverse event Cost of the infusion procedure Mean weekly cost of BSC Cost of an outpatient attendance for a serious adverse event Cost of hospitalisation for a non-serious adverse event 893 1,051 1,208 Cost of hospitalisation for SAE in 2nd-line model 995 1,171 1,346 Cost of an adverse event in 2nd-line (outpatient visit) Cost of oncology outpatient attendance Estimated mean cost of curative liver resection 1,931 2,271 2,612 Cost of hepatic ultrasonography Cost of Chest CT scan Cost abdominal CT Cost of 1 day chemotherapy administration Was probabilistic sensitivity analysis (PSA) undertaken? If not, why not? If it was, the distributions and their sources should be clearly stated; including the derivation and value of priors. Probablistic sensitivity analyses was undertaken. Distributions, sources and derivations for the PSA can be found in Appendix H Has the uncertainty associated with structural uncertainty been investigated? To what extent could/does this type of uncertainty change the results? Structural uncertainty has been assessed and presented in Table H31 below

100 Table H31: Commentary on Structural uncertainty Structural Uncertainty Inclusion/exclusion of potentially relevant comparators. Inclusion/exclusion of potentially relevant events. Statistical models to estimate specific parameters. Clinical uncertainty or lack of clinical evidence. Comment The issue of comparators is addressed by modelling available evidence from both from the OPUS and CRYSTAL trials. As in section treatment with fluoropyrimidine monotherapy is not considered appropriate for inclusion to this decision problem. The model captures the events that correspond to the key endpoints in the trials, and includes: death before progression progression death in 3 rd line curative intent surgery The model does not explicitly simulate other events of surgery, adverse events or progression in post surgery period. Various methods for survival analysis were assessed. Considerations included clinical relevance and statistical fit. Further information can be found in Appendix H3. The primary clinical uncertainties regarding a lack of evidence lay in: The lack of actual CRYSTAL and OPUS overall survival data in the first line setting The paucity of resection data in the KRAS subgroup analyses of CRYSTAL and OPUS. In particular there is a lack of data to inform the rates of curative intent surgery actually performed Statistical analysis How were rates or probabilities based on intervals transformed into (transition) probabilities? In the statistical analysis we have calculated the shapes of the progression free or overall survival distributions. Using these functions, we calculate the transition probabilities as [1-(Tc/Tp)], where Tc is the proportion of patients who are progression-free at the current time point, and Tp is the proportion progression-free at the previous time point (i.e. 1 week earlier). A full description of derivation of variables in the model can be found in Appendix H Is there evidence that (transition) probabilities should vary over time for the condition or disease? If so, has this been included in the evaluation? If there is evidence that this is the case, but it has not been included, provide an explanation of why it has been excluded. The transition probabilities do vary over time and this has been included in the model. The time-dependent probabilities of transition are calculated from the parametric approximations of the observed progression-free and overall survival distributions

101 Validity Describe the measures that have been undertaken in order to validate and check the model. The model structure and assumptions were validated by a UK Expert Panel and an expert in economic modelling. Please refer to appendix H3 for a full description of survival analyses techniques employed. 6.3 Results Provide details of the results of the analysis. In particular, results should include, but are not limited to, the following: costs, QALYs and incremental cost per QALY disaggregated results such as life years gained, costs associated with treatment, costs associated with adverse events, and costs associated with followup/subsequent treatment a statement as to whether the results are based on a probabilistic sensitivity analysis cost-effectiveness acceptability curves scatterplots on cost-effectiveness quadrants Base-case analysis What were the results of the base-case analysis? As previously stated, the Base case results are based upon the following considerations: A patient population with: EGFR expressing, wild-type KRAS status Good performance status (ECOG 0,1) Suitable for irinotecan or oxaliplatin containing chemotherapy Liver metastases excluding those who are resectable at presentation. Liver resection rates are those estimated for cetuximab in combination with FOLFIRI (based upon CRYSTAL evaluation): 22% and 12%: assuming that the resection rates are comparable to those predicted by Folprecht et al in unselected patients using the observed response rates in the KRAS wild type with liver only metastases CRYSTAL population. Liver resection rates are those estimated as for cetuximab in combination with FOLFOX (based upon OPUS evaluation): 15% and 7%: assuming that the resection rates are comparable to those predicted by Folprecht et al in unselected patients using the observed response rates in the KRAS wild type with liver only metastases OPUS population. Liver resection rates as observed in the actual FAS for both CRYSTAL and OPUS trials are presented as scenario analyses in section

102 Cost Results In the CRYSTAL evaluation, the economic model estimated that the cetuximab plus FOLFIRI regimen is associated with an incremental cost per patient of 19,473 and a total expected mean cost per patient of 45,576 in comparison to 26,103 in the FOLFIRI arm alone. In the OPUS evaluation, the cetuximab plus FOLFOX regimen is associated with an incremental cost per patient of 20,242 and a total expected mean cost per patient of 42,084 in comparison to 21,842 in the FOLFOX alone arm. A breakdown of cost by major components including estimated costs in subsequent lines of treatment is presented in Table H32 below. Table H32: Costs in the presence of curative intent surgery CRYSTAL EVALUATION* OPUS EVALUATION* Cetuximab + FOLFIRI FOLFIRI Cetuximab + FOLFOX-4 FOLFOX-4 alone Cetuximab 17,562-17,499 - Irinotecan/ oxaliplatin 5,172 5,196 5,286 4,921 5-FU FA 3,935 3,902 4,031 3,635 Therapy admin in 1st line 3,467 3,438 3,551 3,202 Curative-intent surgery Care after successful surgery 3,474 1,888 2, AE in 1st line 1, nd line setting 6,088 6,833 4,856 5,190 3rd line setting 2,962 3,288 2,708 2,863 KRAS testing Total costs 45,576 26,103 42,084 21,842 Incremental costs 19,473 20,242 * Figures in the table are rounded Table H32 above present the cost structure for the CRYSTAL and OPUS evaluations. The main driver of the incremental cost presented is the acquisition cost of cetuximab. It can be seen that most of the cost in the model is incurred in the first line setting, and in particular is associated with the acquisition costs of cetuximab, irinotecan/ oxaliplatin and fluoropyrimidine. This is presented again below in Figure H10:

103 Figure H10: Costs in the presence of curative intent surgery (base-case for the resection scenario 1 and 2) 50,000 45,000 40,000 KRAS testing 3rd line 35,000 2nd line 30,000 AE in 1st line Cost ( ) 25,000 20,000 Care after successful surgery Curative-intent surgery 15,000 Therapy admin in 1st line 10,000 FA 5,000 0 Cetuximab + FOLFIRI FOLFIRI Cetuximab + FOLFOX-4 Treatment Regimen FOLFOX-4 alone 5-FU Irinotecan/ oxaliplatin Cetuximab Health Outcome results The CRYSTAL evaluation estimates that patients treated with cetuximab plus FOLFIRI gain on average QALYs and life years gained compared to those treated with FOLFIRI alone. Table H33 below present s incremental QALYs and life years gained. Table H33: Incremental QALY and life years gained in the CRYSTAL evaluation Outcome Cetuximab + FOLFIRI alone Incremental FOLFIRI Life Years QALYs The OPUS evaluation estimates that patients treated with cetuximab plus FOLFOX in the first line setting gain on average QALYs and life years gained compared to those treated with FOLFOX alone. Table H34 below present s incremental QALYs and life years gained. Table H34: Incremental QALY and life years gained in the OPUS evaluation Outcome Cetuximab + FOLFOX-4 FOLFOX-4 alone Incremental Life Years QALYs Table H35 below presents the QALYs gained by phase of the model

104 Table H35: QALYs gained by model phase 1 st line 2 nd line 3 rd line Postoperative Total* Cetuximab + FOLFIRI QALY ** LY FOLFIRI QALY LY Cetuximab + FOLFOX QALY ** LY FOLFOX QALY LY * Due to rounding the total presented may be different from simple addition ** The health benefits numbers are rounded to 3 decimal places Results in Table H35 demonstrate that surgical intervention increases overall and total incremental QALY and life years gained in the cetuximab + FOLFIRI arm vs. FOLFIRI alone. Model results suggest that greater QALYS can be gained in the third line versus the second line. However, in the second line of treatment QALYs are accrued only during the PFS time, whereas in the third line setting QALYs are accrued for the overall survival period. Incremental cost-effectiveness ratios result: In the CRYSTAL evaluation, cetuximab plus FOLFIRI patients are estimated to gain QALYs and have an incremental cost of 19,473 compared to FOLFIRI alone patients. This translates into the following incremental cost effectiveness ratio (ICER) of cetuximab plus FOLFIRI in comparison with FOLFIRI alone of 69,287. In the OPUS evaluation, cetuximab plus FOLFOX patients are estimated to gain QALYs and have an incremental cost of 20,242 compared to FOLFOX alone patients. This translates into the following incremental cost effectiveness ratio (ICER) of cetuximab plus FOLFOX in comparison with FOLFOX alone of 63,245. Table H36 presents the ICERs for each scenario. Table H36: Incremental cost-effectiveness ratios Curative-intent resection scenarios Incremental costs and effects ICER Incremental costs 19,473 Cetuximab + FOLFIRI Incremental QALYs ,287 QALY Incremental LYs ,347 LY Incremental costs 20,242 Cetuximab + FOLFOX Incremental QALYs ,245 QALY Incremental LYs ,027 LY

105 6.3.2 Subgroup analysis What were the results of the subgroup analysis/analyses if conducted? Incremental cost-effectiveness the use of surgery Tables H37 and H38 provide an exploratory analysis of different rates of curative intent surgery in the CRYSTAL and OPUS evaluations. Base case results assess patients with liver metastases and apply response rates observed in CRYSTAL and OPUS clinical trials to derive a resection rate from an unselected population based on Folprecht et al [2004] (as presented in Figure H6 section ). It is expected that surgery with curative intent will have a significant impact on the ICER hence the objective of this subgroup analyses is to assess scenarios in which patients defined in the base case may undergo a greater or lower rate of surgery with curative intent and the impact upon cost effectiveness. Scenario 2 presented is the base case and included for iterative purposes The following scenarios are presented in Table H37 and H38 below: Scenario 1: Resection rates as observed in the CRYSTAL (11.5% and 6.7%) and OPUS clinical trials (11.5% and 4.1%) for cetuximab plus FOLFIRI/ FOLFOX and comparator arms respectively. Scenario 2: Resection rates as presented in the base case CRYSTAL evaluation (22% and 12%) and OPUS evaluation (15% and 7%). Scenario 3: Using a similar method to scenario 2 but assuming that the resection rates are comparable to those predicted by Folprecht et al in selected patients (as presented in Figure H6 section ). This is calculated by using the observed response rates in the wt KRAS with liver only metastases population. Resection rates employed for the CRYSTAL evaluation are 55% and 35%; and 42% and 26% for the OPUS evaluation

106 Table H37: Cost effectiveness derived from different resection scenarios for the CRYSTAL evaluation. Curative-intent resection scenarios (CTx / Cetuximab + CTx) Scenario 1: 11.5% / 6.7% Incremental costs and effects Incremental costs 20,224 ICER Incremental QALYs ,815 /QALY Incremental LYs ,144 /LY Scenario 2: 22% / 12% (as in base case) Incremental costs 19,473 Incremental QALYs ,287 /QALY Incremental Lys ,347 /LY Scenario 3: 55% / 35% Incremental costs 16,893 Incremental QALYs ,540 /QALY Incremental Lys ,460 /LY Table H37 presents the impact of different rates of curative surgery and the impact upon cost effectiveness. ICERs presented are in the range of 42K - 94K per QALY gained. Table H38 Cost effectiveness derived from different resection scenarios for the OPUS evaluation. Resection scenarios Scenario 1: 11.5% / 4.1% Incremental costs and effects Incremental costs 20,553 ICER Incremental QALYs ,288 /QALY Incremental LYs ,602 /LY Scenario 2 : 15% / 7% (as in base case) Incremental costs 20,242 Incremental QALYs ,245 /QALY Incremental Lys ,027 /LY Scenario 3: 42% / 26% Incremental costs 17,945 Incremental QALYs ,065 /QALY Incremental Lys ,474 /LY Table H38 presents the impact of different rates of curative surgery and the impact upon cost effectiveness. ICERs presented are in the range of 43K - 65K per QALY gained

107 Incremental cost effectiveness the use of vial sharing It has been assumed that usage of cetuximab will be for complete vials at each visit per patient. Wastage is built into the economic model assuming a patient by patient basis. As a scenario analysis, an assumption is presented based upon vial sharing. This would be representative of usage in specialist centres with sufficient numbers of patients and organisation to ensure that patients visited for treatment on the same day to ensure the vials can be shared. Additionally, irinotecan and oxaliplatin vials are routinely shared in the NHS, and if cetuximab were to become more commonly used then such a scenario may mean that cetuximab based treatment in this area would be able to exploit greater efficiencies. Table H39 below presents ICERs assuming vial sharing is possible. All other assumptions regarding rates of curative resection are as presented in the base case analyses. Table H39: Incremental cost-effectiveness ratios assuming vial sharing is possible Incremental Curative-intent costs and ICER resection scenarios effects Incremental costs 17,941 Cetuximab + FOLFIRI Cetuximab + FOLFOX Incremental QALYs ,834 Incremental LYs ,071 Incremental costs 19,022 Incremental QALYs ,431 Incremental LYs ,070 Incremental cost effectiveness the use of different Survival Analysis methods In the base case, Weibull and lognormal methods have been employed for the CRYSTAL and OPUS evaluations respectively. Variation of the method of survival analyses employed in the model impact upon the incremental cost effectiveness. For the CRYSTAL evaluation a Weibull method was used in the base case and may underestimate the benefit of the cetuximab plus FOLFIRI arm and overestimate the FOLFIRI alone arm. Presented below is an alternative method employed to assess survival. Figure H11 below presents an extrapolation using the actual Kaplan Meier survival curve up to 31 weeks after which a Weibull method is employed

108 Figure H11: CRYSTAL evaluation, Survival analyses using a combination of Kaplan-Meier and Weibull tails 1.0 Estimates of progression-free survival in 1st line using a combination of K-Meier empirical curves with Weibull extrapolation of 'tails' % Progression-free Cetuximab + FOLFIRI FOLFIRI Time, weeks Table H40 below presents cost effectiveness results derived when the extrapolation technique as presented in Figure H11 is employed. Table H40: CRYSTAL evaluation, Cost-effectiveness analysis Survival analyses using a combination of Kaplan-Meier and Weibull tails Incremental costs and effects ICER Incremental costs 19,598 Incremental QALYs ,090.1 Incremental LYs ,256.3 For the OPUS evaluation a lognormal method was used in the base case. When a Weibull method is employed the cost effectiveness results are as presented in Table H41 below. Table H41: OPUS evaluation, Cost-effectiveness analysis using a Weibull method Incremental costs and effects ICER Incremental costs 19,823 Incremental QALYs ,333 Incremental LYs ,

109 6.3.3 Sensitivity analyses What were the main findings of the sensitivity analyses? Univariate Sensitivity Analyses Table H42 below presents univariate sensitivity analysis for the CRYSTAL evaluation. Results are presented from variables that affect the variation in ICER more than 0.02% of its original value. Table H42: Univariate sensitivity analysis (CRYSTAL EVALUATION) Variable 1. Annual discount rate for effects Low EV High EV Low Input High Input Spread % Spread 58,782 76, , % 2. Utility Weight in 1 st line 68,500 70, , % 3. Annual discount rate for costs 68,995 69, % 4. Utility in 2nd line 69,057 69, % 5. Cost of a serious adverse event requiring hospitalisation 68,976 69, , % 6. Utility in 3rd line 69,212 69, % 7. Cost of an outpatient attendance for grade 3/4 adverse event 8. Estimated mean cost of curative liver resection 9. Cost of the infusion procedure 69,158 69, % 69,162 69,411 1,930 2, % 69,179 69, % 10. Mean weekly cost of BSC 69,237 69, % 11. Cost of an outpatient attendance for a serious adverse event 69,268 69, % In addition to the univariate sensitivity analysis presented in Table H42 the impact of failure of curative intent surgery has been examined. The failure rate varies from 0% (all successful resections) to 40% which was the highest observed in the FAS Crystal population(from the FOLFIRI arm, see Table ). The results are presented in Table H43 below

110 Table H43: CRYSTAL evaluation: Univariate analysis of cure rate for those who undergo resection Rate of Surgical Resection Failure ICER Difference from base case 0% 56, % 4% 58, % 8% 59, % 12% 61, % 16% 63, % 20% 65, % 24% 67, % 28% 69, % 32% 71, % 36% 74, % 40% 76, % Table H43 highlights the impact of variation of failure of curative intent surgery. In the best case assuming a 0% failure rate, the ICER is 56,492 compared to 69,396 for the base-case ICER (28% failure rate). Table H44 below presents univariate sensitivity analysis for the OPUS evaluation. Results are presented from variables that affect the variation in ICER more than 0.02% of its original value. Table H44: Univariate sensitivity analysis (OPUS EVALUATION) Variable Low EV High EV Low Input High Input Spread % Spread 1. Annual discount rate for effects 55,927 68, , % 2. Utility in 1 st line 62,143 64, , % 3. Utility in 3 rd line 63,032 64, % 4. Utility in 2 nd line 63,024 63, % 5. Annual discount rate for costs 63,053 63, % 6. Cost of the infusion procedure 63,002 63, % 7. Estimated mean cost of curative liver resection 8. Cost of an outpatient attendance for grade 3/4 adverse event 9. Cost of a serious adverse event requiring hospitalisation 63,157 63,333 1,930 2, % 63,063 63, % 63,173 63, , % 10. Mean weekly cost of BSC 63,221 63, % 11. Cost of an outpatient attendance for a serious adverse event 63,218 63, % In addition to the univariate sensitivity analysis presented above the impact of failure of curative intent surgery has been examined

111 Cost effectiveness in the model is sensitive to this assumption. The results are presented in Table H45 below. Table H45: OPUS evaluation: Univariate analysis of cure rate for those who undergo resection Rate of Surgical Resection Failure ICER Difference from base case 0% 54, % 4% 55, % 8% 56, % 12% 57, % 16% 59, % 20% 60, % 24% 61, % 28% 63, % 32% 64, % 36% 66, % 40% 68, % Table H45 highlights the impact of variation of failure of curative intent surgery. In the best case assuming a 0% failiure rate, the ICER is 54,062 compared to 69,396 for the base-case ICER (28% failure rate). Probabilistic Sensitivity Analyses In order to address uncertainty around the observed cost and effect values utilised in the model, Probabilistic Sensitivity Analyses (PSA) for both the CRYSTAL and OPUS evaluations was undertaken. Statistics and cost effects for 2000 simulations are presented below

112 Figure H12: ICER Scatter plot for the CRYSTAL evaluation 25,000 ICE Scatterplot of Cetuximab+FOLFIRI (KRAS) vs. FOLFIRI (KRAS) Base case for Surgery (12% and 22%) and base case for PFS in 1st line 24,000 23,000 22,000 Incremental Cost 21,000 20,000 19,000 18,000 17,000 16,000 15, Incremental Effectiveness (QALY) Figure H13: Cost-effectiveness acceptability curve for the CRYSTAL evaluation Acceptability Curve for Cetuximab + FOLFIRI vs FOLFIRI alone. Base case curative intent resection Rate (12% and 22% in the CTx and Cetuximab+CTx arms) Base Case Parametric Model for PFS in 1st line (Simultaneous Weibull) Proportion Cost-Effective , , , , ,000 WTP Threshold (#/QALY) Table H46 below presents summary costs and outcomes statistics for 2000 simulations for the CRYSTAL evaluation

113 Table H46: PSA summary costs and outcomes statistics Cetuximab+FOLFIRI FOLFIRI Statistic Cost QALY Cost QALY Mean 45, , Std Dev 2, , Minimum 38, , % 41, , % 42, , Median 45, , % 48, , % 49, , Maximum 52, , Figure H14: ICER Scatter plot for the OPUS evaluation 25,000 ICE Scatterplot of Cetuximab+FOLFOX-4 (KRAS) vs. FOLFOX-4 (KRAS) 24,000 23,000 22,000 Incremental Cost 21,000 20,000 19,000 18,000 17,000 16,000 15, Incremental Effectiveness (QALY)

114 Figure H15: Cost-effectiveness acceptability curve for the OPUS evaluation Proportion Cost-Effective ,000 40,000 60,000 80, , , , , , ,000 WTP Threshold ( /QALY) Table H47 below presents summary costs and outcomes statistics for 250 simulations for the OPUS evaluation. Table H47: PSA summary costs and outcomes statistics Cetuximab+FOLFOX-4 FOLFOX-4 Statistic Cost QALY Cost QALY Mean 42, , Std Dev 1, , Minimum 36, , % 38, , % 39, , Median 42, , % 44, , % 45, , Maximum 47, , Probabilistic sensitivity analyses presented suggest that from the parameters assessed, the results of the base case are relatively robust. This would suggest that the subgroup scenario analyses presented in section may provide a better understanding of the drivers of cost effectiveness in the model

115 6.3.4 Interpretation of economic evidence Are the results from this economic evaluation consistent with the published economic literature? If not, why do the results from this evaluation differ, and why should the results in the submission be given more credence than those in the published literature? The results of the base case are consistent with the economic literature, in particular with TAG 118 and the ICER reported for bevacizumab vs IFL regimen in which the ICER was in the range of 62,857 to 93,128 [from TAR Tappenden et al 2006]. Results are also consistent with TAG 93 where NICE recommended the use of FOLFOX, and the ICER reported 68k [section 4.2.9] Is the economic evaluation relevant to all groups of patients who could potentially use the technology? The base case of the economic evaluation is relevant to those patients with mcrc eligible for treatment for first line combination chemotherapy. It is restricted to those who are of KRAS wild type and who have unresectable liver metastases which may be rendered resectable following combination treatment of cetuximab plus chemotherapy. The base case of the economic evaluation is relevant to all patients in first line mcrc who would be considered suitable for fluoropyrimidine based irinotecan/ oxaliplatin combination chemotherapy. The impact of restricting the use of cetuximab in combination with chemotherapy to specific sub populations who may be expected to derive greater benefit has been explored. In particular, those patients with liver only metastases which may be rendered fully resectable and hence potentially cured as a result of cetuximab combination therapy What are the main strengths and weaknesses of the evaluation? How might these affect the interpretation of the results? The main strengths of the economic evaluation lay in the targeted patient population, and use of EQ-5D generated utility values. The base case target patient population represents that group of patients thought most likely to derive benefit from the addition of cetuximab to standard combination chemotherapy. The utilities used are based on EQ-5D data collected in the UK derived from the UK FAS of the CRYSTAL trial. The primary weakness of the economic evaluation is that due in part to the KRAS reanalysis of the clinical trials, data on resection rates are limited. We have therefore modelled a number of scenarios using data from the literature as well as using CRYSTAL resection rates. The health states in the model, while consistent with other colorectal cancer models submitted to NICE, are not designed to reflect possible differences in the utility functions for the comparator arm vs the cetuximab arm. In addition, it is important to note that the sample of patients with KRAS wildtype and metastases confined to the liver was too small in both studies for formal subgroup analysis based on progression-free survival times (Crystal N=35 Cetuximab/FOLFIRI, N=32 FOLFIRI; OPUS: N=20 Cetuximab/FOLFOX and N=18 FOLFOX ). Therefore the PFS from all patients with KRAS wildtype was used in the economic model as a proxy. Since the ITT datasets from both studies have shown that the incremental

116 clinical benefit for patients with metastases confined to the liver was greater compared to the group with multiple sites of metastases, the use of PFS data from all KRAS wildtype patients to represent KRAS wildtype patients with metastases confined to the liver can be considered conservative What further analyses could be undertaken to enhance the robustness/completeness of the results? Uncensored data from the CRYSTAL clinical trial will be available in the near future. This data may further inform the choice of extrapolation technique presented in the base case. 7 Assessment of factors relevant to the NHS and other parties The purpose of this section is to provide an analysis of any factors relevant to the NHS and other parties that may fall outside the remit of the assessments of clinical effectiveness and cost effectiveness. This will facilitate the subsequent evaluation of the budget impact analysis. Such factors might include issues relating to service organisation and provision, resource allocation and equity, societal or ethical issues, plus any impact on patients or carers. Further examples are given in section 3.4 of the NICE document Guide to the methods of technology appraisal. 7.1 What is the estimated annual budget impact for the NHS in England and Wales? The estimated budget impact for the NHS in England and Wales is 17,179,128. In the first year of launch this consists of treating: 76 patients with cetuximab in combination with FOLFIRI 326 patients for cetuximab in combination with FOLFOX, This would rise to 25,151,961 treating a total of 589 patients across both settings in the fifth year. Table BI1 below presents total budget impact, purchase costs of cetuximab and other modelled costs including administration, treatment of adverse events and costs incurred for surgical procedures. (Please see appendix BI1 for further details of the budget impact calculation including a copy of the excel spreadsheet used to model Tappenden et al.)

117 Table BI1: Budget Impact Cetuximab purchase costs 7,042,873 7,747,160 8,521,876 9,374,064 10,311,470 Other costs of treatment 10,136,254 11,149,880 12,264,868 13,491,355 14,840,490 Total Budget Impact 17,179,128 18,897,040 20,786,744 22,865,419 25,151,961 Approximately 90% of the budget increase is the purchase cost of cetuximab with the remaining 10% mainly being curative intent surgery, care after successful surgery and adverse events in first line, with small cost additions from KRAS testing, further adverse events, imaging and procedures. Further details of incremental cost can be found in section What number of patients were assumed to be eligible? How was this figure derived? The overall incidence of colorectal cancer (CRC) in England and Wales is estimated to be 35,000 patients. From this it is estimated that 2,011 patients would be eligible for combination chemotherapy per annum. This figure was calculated using an adaptation of the CRC treatment algorithm which was derived by Tappenden et al [2006] set out in the bevacizumab/ cetuximab TAR [see appendix BI 1]. This is presented in Figure BI1 below

118 Fig BI1 Estimated Number Of Patients Assumed To Be Eligible For Combination Treatment For Unresectable Liver Metastases STEP 1: Overall incidence of colorectal cancer in England and Wales 35, 000 patients STEP 2: Total number of patients with uncurable mcrc 17, 171 patients STEP 3: Patients with Liver only metastases 8,586 patients STEP 4: Patients with Liver only metastases who are unresectable 7,298 patients STEP 5: Proportion of step 4 who may be both KRAS wild type and EGFR expressing 3,795 patients STEP 6: Patients eligible for combination chemotherapy 2,011 Table BI2 below presents calculations and assumptions presented in Figure BI1 above

119 Table BI2 description of patient number derivation shown in Fig BI1 Step % of patients How derived Source From the literature ONS data Approx 49% of patients 3 50% of those in step 2. This value is the sum total of all patients who were either diagnosed with mcrc or those who have had resection for CRC and then progressed to Metastatic disease Those patients who have liver only metastases derived from Tappenden et al, See Appendix BI1 derived from Tappenden et al See Appendix BI1 4 85% of step 3 Proportion of patients who have unresectable liver metastases 5(a) 65% of step 4 The proportion of patients who would be expected to have WT KRAS status in the general population 5(b) 80% of step 5(a) The proportion of patients who would be expected to express EGFR as observed in clinical trials 6 53% of step 5 This figure represents 10% of the total in step 4 who would receive irinotecan based treatment plus 43% who would be eligible for oxaliplatin based combinations derived from Tappenden et al, See Appendix BI1 Based on Andreyev et al, Cunningham et al, 2004 Combination chemotherapy represents all patients who would be treated in this setting within NICE guidance with either oxaliplatin or irinotecan based combinations, [A+A market data] 7.3 What assumption(s) were made about current treatment options and uptake of technologies? Based upon the assumption that cetuximab will only be used in combination with either fluoropyrimidine based combinations of both oxaliplatin and irinotecan and using data from the A+A market research, in the first line setting approximately 43% of patients will receive treatment with an oxaliplatin based combination, and about 10% of patients will receive an irinotecan based treatment. Of the total of 2,011 patients who would be eligible for combination chemotherapy it is estimated that this will equate to a maximum of 1,632 and 379 patients for oxaliplatin and irinotecan based treatments respectively

120 Table BI3: Total eligible Patient numbers cost and maximum treatment cost Type of chemo Number of KRAS/EGFR eligible patients Proportion treated in first line N= Cost per patient Cost by treatment type Total FOLFIRI 3,795 10% ,576 17,273,304 FOLFOX 3,795 43% 1,632 42,084 68,681,088 85,954,392 Table BI3 describes the number of eligible KRAS wild type patients then splits this into patient numbers (N=) and cost for both FOLFIRI patients and FOLFOX patients. The total box describes the full year cost of all treatment modalities including surgery, delivery and costs associated with treatment of adverse events if ALL eligible patients were treated in a full year. Given the conservative nature of uptake in the UK we have made some assumptions surrounding the market uptake in year s one, two, three, four and five post approval. It is assumed in these calculations that there will be no impact during the 2008 calendar year. Please see table BI1 for a break down of expected costs per annum. 7.4 What assumption(s) were made about market share (where relevant)? Data from the A+A market research tracker indicate that oxaliplatin combination treatments are most common in this setting, followed by 5FU/FA then irinotecan based combination. 7.5 Unit costs were assumed? How were these calculated? Unit costs used included list price of chemotherapies, cost of administration and of treating adverse events. Sources were BNF 55 NHS national tariff costs and UK HRGs. The unit costs used in this calculation are listed in table H9. The output from the model was used to estimate resource use for both oxaliplatin and irinotecan based combinations with cetuximab. 7.6 In addition to drug costs, consider other significant costs associated with treatment. What is the recommended treatment regime for example, what is the typical number of visits, and does treatment involve daycase or outpatient attendance? Is there a difference between recommended and observed doses? Are there likely to be any adverse events or a need for other treatments in combination with the technology? Costs for each part of the combination chemotherapy were calculated based on list prices from BNF 55. Costs for administration and management of adverse events as well as an estimate for the cost of KRAS testing are included. Cost per KRAS test was estimated at 300. The cost used in the model takes into account the number of KRAS mutant patients who would not be eligible for treatment so is higher than the cost per test. It is probable that this cost is an over estimate and would reduce if used routinely due to economies of scale

121 Table BI4 below shows the expected cost per patient derived from usage in the base case of the economic model. Table BI4: Expected cost values. CRYSTAL EVALUATION* OPUS EVALUATION* Cetuximab + FOLFIRI FOLFIRI Cetuximab + FOLFOX-4 FOLFOX-4 alone Cetuximab 17,562-17,499 - Irinotecan/ oxaliplatin 5,172 5,196 5,286 4,921 5-FU FA 3,935 3,902 4,031 3,635 Therapy admin in 1st line 3,467 3,438 3,551 3,202 Curative-intent surgery Care after successful surgery 3,474 1,888 2, AE in 1st line 1, nd line setting 6,088 6,833 4,856 5,190 3rd line setting 2,962 3,288 2,708 2,863 KRAS testing Total costs 45,576 26,103 42,084 21,842 Incremental costs 19,473 20, Were there any estimates of resource savings? If so, what were they? There are no expected resource savings through the addition of cetuximab into the treatment of mcrc. Greater efficiencies of use may be generated through potential vial sharing if cetuximab plus FOLFIRI/ FOLFOX is implemented on a wide scale. 7.8 Are there any other opportunities for resource savings or redirection of resources that it has not been possible to quantify? There are no expected opportunities for resource savings or redirection of resources that it has not been possible to quantify

122 8 References A+A Tracker Data on File Adam R. Delvart V. Pascal G. et al. Rescue surgery for unresectable colorectal liver metastases downstaged by chemotherapy: a model to predict long-term survival. Annals of Surgery. 240(4): Andreyev HJ, Norman AR, Cunningham D, et al. Kirsten ras mutations in patients with colorectal cancer: the 'RASCAL II' study. Br J Cancer. 2001;85(5): Baselga J The EGFR as a target for anticancer therapy focus on cetuximab European Journal of Cancer 37(2001) S16 S22 Carmichael J, Popiela T, Radstone D, Falk S, Borner M, Oza A, Skovsgaard T, et al. Randomized comparative study of tegafur/uracil and oral leucovorin versus parenteral fluorouracil and leucovorin in patients with previously untreated metastatic colorectal cancer. JCO 2002; 20: Cunningham D, Humblet Y, Siena S, et al. Cetuximab monotherapy and cetuximab plus irinotecan in irinotecan-refractory metastatic colorectal cancer. N Engl J Med 2004; 351: de Gramont A, Figer A, Seymour M, Homerin M, Hmissi A, Cassidy J et al Leucovorin and Fluorouracil With or Without Oxaliplatin as First-Line Treatment in Advanced Colorectal Cancer. Journal of Clinical Oncology, Vol 18, Issue : Douillard JY, Cunningham D, Roth AD, Navarro M, James RD, Karasek P, et al Irinotecan combined with fluorouracil compared with fluorouracil alone as first-line treatment for metastatic colorectal cancer: a multicentre randomised trial. Lancet Mar 25;355(9209): Douillard JY, Hoff P, Skillings J, Eisenberg P, Davidson N, Harper P et al Multicenter phase III study of uracil/tegafur and oral leucovorin versus fluorouracil and leucovorin in patients with previously untreated metastatic colorectal cancer. JCO 2002;20: Folprecht G, Grothey A, Alberts S et al Neoadjuvant treatment of unresectable colorectal liver metastases: correlation between tumour response and resection rates. Ann Oncol. 2005;16(8): Giacchetti S, Itzhaki M, Gruia G, et al Long-term survival of patients with unresectable colorectal cancer liver metastases following infusional chemotherapy with 5-fluorouracil, leucovorin, oxaliplatin and surgery. Ann Oncol. 1999;10(6): Goldstein NI, Prewett M, Zuklys K, et al. Biological efficacy of a chimeric antibody to the epidermal growth factor receptor in a human tumor xenograft model. Clin Cancer Res 1995; 1: Green J, Watson J, Roche M et al, Stage, grade and morphology of tumours of the colon and rectum recorded in the Oxford Cancer Registry, British Journal of Cancer 2007: 96,

123 Hoff, P. M., Ansari, R., Batist, G., Cox, J., Kocha, W., Kuperminc, M.,et al Comparison of oral capecitabine versus intravenous fluorouracil plus leucovorin as first-line treatment in 605 patients with metastatic colorectal cancer: results of a randomized phase III study. Journal of Clinical Oncology ; Jonker DJ. O'Callaghan CJ. Karapetis CS et al Cetuximab for the treatment of colorectal cancer. New England Journal of Medicine. 357(20):2040-8, 2007 Lièvre A, Bachet JB, Boige V, et al.kras mutations as an independent prognostic factor in patients with advanced colorectal cancer treated with cetuximab. J Clin Oncol. 2008;26(3): Nordingler B, Van Cutsem E, Rougier P, Kohne C et al Does chemotherapy prior to liver resection increase the potential for cure in patients with metastatic colorectal cancer? A report from the European Colorectal Metastases Treatment Group. European Journal of cancer 2007, 43 (14) NICE TA93 - Irinotecan, oxaliplatin and raltitrexed for the treatment of advanced colorectal cancer. &bcsi_scan_filename=ta93guidance.pdf Accessed February NICE TA118 Colorectal cancer (metastatic) - bevacizumab & cetuximab: Guidance. Accessed February ONS: Office for national statistics cancer incidence and mortality in the United Kingdom. Accessed February 2008 Osoba D, Rodrigues G, Myles J, et al.interpreting the significance of changes in healthrelated quality-of-life scores. J Clin Oncol, (1998),16, Panitumumab SmPC. Electronic medicines compendium. Personal communication G Poston 7 April 2008 Personal communication M Saunders April 2008 Roda JM, Joshi T, Butchar JP, et al. The activation of natural killer cell effector functions by cetuximab-coated, epidermal growth factor receptor positive tumor cells is enhanced by cytokines.clin Cancer Res. 2007:13(21): Rothenberg,M.L., Oza,A.M., Bigelow,R.H., Berlin,J.D., Marshall,J.L., Ramanathan,R.K., Hart,L.L., Gupta,S., Garay,C.A., Burger,B.G., Le Bail,N., Haller,D.G. J Clin Oncol, (2003) 21(11): Saltz LB, Cox JV, Blanke C, Rosen LS, Fehrenbacher L, Moore MJ, et al Irinotecan plus fluorouracil and leucovorin for metastatic colorectal cancer. Irinotecan Study Group. N Engl J Med Sep 28;343(13): Saltz L, Meropol NJ, Loehrer PJ, et al. Phase II trial of cetuximab in patients with refractory colorectal cancer that expresses the epidermal growth factor receptor. J Clin Oncol 2004; 22: Saunders M and Iveson T. Management of advanced colorectal cancer:state of the art. British Journal of Cancer (2006) 95, Scartozzi,M., Sobrero,A., Gasparini,G., et al. The role of 5-fluorouracil (5-FU) reintroduction with irinotecan or oxaliplatin in truly 5-FU-refractory advanced colorectal cancer patients. ONCOLOGY, (2005) 68(2-3):

124 Segaert S et al. Clinical signs, pathophysiology and management of skin toxicity during therapy with epidermal growth factor receptor inhibitors. Ann Of Oncol 2005; 16 (9): Simmonds PC, Primrose JN, Colquitt JL, et al Surgical resection of hepatic metastases from colorectal cancer: a systematic review of published studies. Br J Cancer Apr 10;94(7): Tabernero J, Cervantes A, Ciardiello F, Correlation of efficacy to KRAS status (wt vs. mut) in patients (pts) with metastatic colorectal cancer (mcrc), treated with weekly (q1w) and q2w schedules of cetuximab combined with FOLFIRI. Abstract 435. ASCO Gastrointestinal cancers symposium Thirion P, Michiels S, Pignon JP, Buyse M, Braud AC, Carlson RW, et al; Meta-Analysis Group in Cancer. Modulation of fluorouracil by leucovorin in patients with advanced colorectal cancer: an updated meta-analysis. J Clin Oncol ;22(18): Tournigand C, André T, Achille E, Lledo G, Flesh M, Mery-Mignard D et al FOLFIRI followed by FOLFOX6 or the reverse sequence in advanced colorectal cancer: a randomized GERCOR study. J Clin Oncol ;22(2): Van Cutsem, E., Twelves, C., Cassidy, J., Allman, D., Bajetta, E., Boyer, M. et al Oral capecitabine compared with intravenous fluorouracil plus leucovorin in patients with metastatic colorectal cancer: Results of a large phase III study. Journal of Clinical Oncology 2001; Van de Vijver MJ, Kumar R, Mendelsohn J. Ligand-induced Activation of A431 Cell Epidermal Growth Factor Receptors Occurs Primarily by an Autocrine Pathway That Acts Upon Receptors on the Surface Rather than Intracellularly. J Biol Chem 1991; 266: Waksal HW. Role of an anti-epidermal growth factor receptor in treating cancer. Cancer Metastasis Rev 1999; 18: Yarden Y, Ullrich A. Growth Factor Receptor Tyrosine Kinases. Ann Rev Biochem 1988; 57:

125 9 Appendices 9.1 Appendix 1 Summary of Product Characteristics or Technical Manual or drafts A draft of the cetuximab summary of product characteristics (SmPC) which includes the use of cetuximab in the treatment of 1 st -line metastatic colorectal cancer is not currently available. Merck Serono UK propose to send a draft SmPC to NICE at the time of CHMP opinion. 9.2 Appendix 2: search strategy for section 5 The following information should be provided The specific databases searched and the service provider used (for example, Dialog, DataStar, OVID, Silver Platter), including at least: A protocol was prepared for the literature search, detailing inclusion and exclusion criteria and search terms, search dates and data span searched. Articles were identified in electronic database searches of: MEDLINE In-Process & Other Non-Indexed Citations, Ovid MEDLINE Daily and Ovid MEDLINE were searched using OVID (Search Table 1), EMBASE (using OVID)(Search Table 2), Current Contents was searched using Datastarweb. (Search Table 3), The Cochrane Library (2008, Issue 1) Searched using the Cochrane Database of Systematic Reviews (Search Table 4), The American Society of Clinical Oncology (ASCO) Abstracts Database ( Asco.org) and European The European Journal of Cancer Supplements (ECCO) (Search Table 5) The date on which the search was conducted. Database Date of search Medline Embase Current Contents the Cochrane Library ASCO / ECCO abstracts The date span of the search. Ovid MEDLINE(R) In-Process & Other Non-Indexed Citations, Ovid MEDLINE(R) Daily and Ovid MEDLINE(R) 1950 to Present (1966 to ) EMBASE 1980 to 2008 Week 12 Current Contents Search(R): Bibliogr. Records to date

126 The Cochrane Library (2008, Issue 1) - Cochrane Database of Systematic Reviews The American Society of Clinical Oncology (ASCO) Abstracts Database ( Asco.org) and ASCO GI European The European Journal of Cancer Supplements (ECCO) 2003-present The complete search strategies used, including all the search terms: textwords (free text), subject index headings (for example, MeSH) and the relationship between the search terms (for example, Boolean). Search Table1 : OVID Medline Search Strategies Basis: Ovid MEDLINE(R) In-Process & Other Non-Indexed Citations, Ovid MEDLINE(R) Daily and Ovid MEDLINE(R) 1950 to Present Date: Search String No. Detail #1 (((Colorect* or Colon* or Rect* or Duoden* or Ile* or Jejun* or Stomach* or gastri* or gastro* or GI) adj3 (cancer* or tumor* or tumour* or neoplasm* or malignan* or carcinom*)) or mcrc or m-crc or CRC).mp. or exp Colorectal Neoplasms/ #2 (KRAS or K-RAS or K RAS or Kras1 or Kras2 or Kras1P or Ki-RAS or RASK or RASK1 or RASK2 or Kirsten RAS or K-RASwt).mp. #3 (erbitux or cetuximab or (IMC-C 22 or IMC-C225 or c-225 or c225) or (EMR or EMD )).mp. #4 ((fluorouracil or 5-fluorouracil or 5 fluorouracil or 5-fu or 5fu or fluoruracil).mp. or exp Fluorouracil/) and ((leucovorin or folinic acid or acid, folinic or folinic-acid or levoleucovorin or folinate).mp. or exp Leucovorin/) #5 FOLFIRI*.mp. or ((irinotecan or Campto or camptothec* or captosar).mp. and #4) #6 FOLFOX*.mp. or ((oxaliplatin or eloxatin).mp. and #4) Search pool A: #1 AND #2 AND #4 (n=6) (5-fluorouracil plus folinic acid) in metastatic colorectal cancer in patients with KRAS Search pool B: #1 AND #2 AND #3 and #5 (n=0) cetuximab in combination with FOLFIRI in metastatic colorectal cancer in patients with KRAS Search pool C: #1 AND #2 AND #3 and #6 (n=0) cetuximab in combination with FOLFOX in metastatic colorectal cancer in patients with KRAS

127 Search Table 2: OVID EMBASE Search Strategies Basis: EMBASE 1980 to 2008 Week 12 Date: Search String No. Detail #1 (((Colorect* or Colon* or Rect* or Duoden* or Ile* or Jejun* or Stomach* or gastri* or gastro* or GI) adj3 (cancer* or tumor* or tumour* or neoplasm* or malignan* or carcinom*)) or mcrc or m-crc or CRC).mp. or exp Colorectal Tumor/ #2 (KRAS or K-RAS or K RAS or Kras1 or Kras2 or Kras1P or Ki-RAS or RASK or RASK1 or RASK2 or Kirsten RAS or K-RASwt).mp. #3 (erbitux or cetuximab or (IMC-C 22 or IMC-C225 or c-225 or c225) or (EMR or EMD )).mp. #4 ((fluorouracil or 5-fluorouracil or 5 fluorouracil or 5-fu or 5fu or fluoruracil).mp. or exp Fluorouracil/) and ((leucovorin or folinic acid or acid, folinic or folinic-acid or levoleucovorin or folinate).mp. or exp Folinic Acid/) #5 FOLFIRI*.mp. or ((irinotecan or Campto or camptothec* or captosar).mp. and #4) #6 FOLFOX*.mp. or ((oxaliplatin or eloxatin).mp. and #4) Search pool A: #1 AND #2 AND #4 (n=60) (5-fluorouracil plus folinic acid) in metastatic colorectal cancer in patients with KRAS Search pool B: #1 AND #2 AND #3 and #5 (n=27) cetuximab in combination with FOLFIRI in metastatic colorectal cancer in patients with KRAS Search pool C: #1 AND #2 AND #3 and #6 (n=22) cetuximab in combination with FOLFOX in metastatic colorectal cancer in patients with KRAS Search Table 3: Datastar Current Contents Search Strategies Date: Basis: Current Contents Search(R): Bibliogr. Records to date (CBIB) Search String No. Detail #1 (Colorect$ OR Colon$ OR Rect$ OR Duoden$ OR Ile$ OR Jejun$ OR Stomach$ OR gastri$ OR gastro$ OR GI) ADJ (cancer$ OR tumor$ OR tumour$ OR neoplasm$ OR malignan$ OR carcinom$) OR mcrc OR m-crc OR CRC #2 KRAS OR K-RAS OR K ADJ RAS OR Kras1 OR Kras2 OR Kras1P OR Ki-RAS OR RASK OR RASK1 OR RASK2 OR Kirsten ADJ RAS OR K-RASwt #3 erbitux OR cetuximab OR IMC-C ADJ '22' OR IMC-C225 OR c-225 OR c225 OR EMR OR EMD #4 (fluorouracil OR 5-fluorouracil OR 5 ADJ fluorouracil OR 5-fu OR 5fu OR fluoruracil) AND (leucovorin OR folinic ADJ acid OR acid ADJ folinic OR folinic-acid

128 OR levoleucovorin OR folinate) #5 (FOLFIRI$ OR irinotecan OR Campto OR camptothec$ OR captosar) AND #4 #6 FOLFOX$ OR oxaliplatin OR eloxatin) and #4 Search pool A: #1 AND #2 AND #4 (n=60) (5-fluorouracil plus folinic acid) in metastatic colorectal cancer in patients with KRAS Search pool B: #1 AND #2 AND #3 and #5 (n=27) cetuximab in combination with FOLFIRI in metastatic colorectal cancer in patients with KRAS Search pool C: #1 AND #2 AND #3 and #6 (n=22) cetuximab in combination with FOLFOX in metastatic colorectal cancer in patients with KRAS Search Table 4: The Cochrane Library Search Strategies Basis: the Cochrane Library (2008, Issue 1) - Cochrane Database of Systematic Reviews Date: Search String No. Detail #1 (Colorect$ OR Colon$ OR Rect$ OR Duoden$ OR Ile$ OR Jejun$ OR Stomach$ OR gastri$ OR gastro$ OR GI) ADJ (cancer$ OR tumor$ OR tumour$ OR neoplasm$ OR malignan$ OR carcinom$) OR mcrc OR m-crc OR CRC #2 KRAS OR K-RAS OR K ADJ RAS OR Kras1 OR Kras2 OR Kras1P OR Ki-RAS OR RASK OR RASK1 OR RASK2 OR Kirsten ADJ RAS OR K-RASwt #1 AND #2 Search Table 5: Conference Search Strategies Basis: Asco abstracts from ASCO and ASCO GI , webbased content searched on 1 st of April, 2008 Date: Conference ASCO ECCO Detail KRAS or K-RAS or K RAS or Kras1 or Kras2 or Kras1P or Ki-RAS or RASK or RASK1 or RASK2 or Kirsten RAS or K-RASwt in abstract section KRAS or K-RAS or Ki-RAS in abstract section

129 9.2.5 Details of any additional searches, for example searches of company databases (include a description of each database). A search was conducted of internal company databases. The internal clinical trial information repository revealed two studies which met the inclusion criteria The inclusion and exclusion criteria. Inclusion criteria Randomised controlled trials. Studies on the use of cetuximab in combination with FOLFIRI in the first line treatment of metastatic colorectal cancer in patients with KRAS wild-type tumour. Studies on the use of cetuximab in combination with FOLFOX in the first line treatment of metastatic colorectal cancer in patients with KRAS wild-type tumour. Studies on the use of 5-fluorouracil plus folinic acid in the first line treatment of metastatic colorectal cancer in patients with KRAS wild-type tumour. Human studies only were considered. Exclusion criteria Studies which involved patients who received previous treatment in the metastatic colorectal cancer setting were not considered. Papers published in a language other than English were not considered. Letters and editorials were not considered. Review articles and conference summaries were excluded. Animal studies/preclinical data were not considered The data abstraction strategy. After the selection of relevant trials, data was extracted in a structured form to capture, the study number/author, the design of the trial and the eligible patient population, the phase of the trial, details of the study interventions and the sample size

130 Appendix 3: search strategy for section 6 The following information should be provided The specific databases searched and the service provider used (for example, Dialog, DataStar, OVID, Silver Platter), including at least: Medline Embase Medline (R) In-Process Health Economic Evaluation Database NHS Economic Evaluation Database (NHS EED). We used OVID to search Medline, Embase and Medline R. NHS EED, DARE and HTA database were searched using the CRD databases portal The date on which the search was conducted. Monday 17 th March The date span of the search. The searches were conducted to the present for all databases and from: 1980 for Embase 1950 for Medline and Medline (R) The complete search strategies used, including all the search terms: textwords (free text), subject index headings (for example, MeSH) and the relationship between the search terms (for example, Boolean). Searches were conducted for the following terms in Embase and Medline: Metastatic colorectal cancer; Erbitux OR cetuximab; cost effectiveness analysis; QALY; cost effectiveness; quality of life; value in health. All searches were limited to English language and humans, abstract and full text only. The CRD database was searched for cetuximab only Details of any additional searches, for example searches of company databases (include a description of each database). Not Applicable

131 Appendix 4 Appendix 4 baseline demographics and efficacy results from other groups not directly relevant to the decision problem from the OPUS and CRYSTAL studies. Baseline demographic variables for the FAS and KRAS populations of study EMR Baseline demographic variables for the FAS and KRAS populations of study EMR

132 PFS for the FAS and KRAS populations of studies EMR and EMR

133 Tumour Response for the FAS and KRAS populations of studies EMR and EMR