3 The manufacturer's submission
3.1 The manufacturer's approach to the decision problem was in line with the NICE scope for the population, intervention, outcomes and the economic evaluation. The manufacturer's submission focussed on a comparison of erlotinib with gefitinib for first-line treatment of locally advanced or metastatic EGFR-TK mutation-positive NSCLC. The manufacturer's submission did not include pemetrexed plus cisplatin or carboplatin as a comparator because of the declining use in clinical practice of this combination for first-line treatment of locally advanced or metastatic EGFR-TK mutation-positive NSCLC and the absence of suitable data for comparison in this population.
3.2 The manufacturer identified two randomised controlled trials (EURTAC and OPTIMAL) that compared erlotinib with platinum doublet chemotherapy as first-line treatment for patients with locally advanced or metastatic EGFR-TK mutation-positive NSCLC. The manufacturer based its evidence submission on the EURTAC trial with the OPTIMAL trial as supporting evidence. No studies were identified that compared erlotinib directly with gefitinib in this patient population, and so the manufacturer presented an indirect treatment comparison to assess the relative effectiveness of erlotinib and gefitinib.
3.3 The EURTAC trial was a European-based, open-label, phase III, randomised trial of first-line erlotinib treatment compared with platinum doublet chemotherapy for patients with stage IIIb or stage IV NSCLC and EGFR-TK mutation-positive tumours. The trial included 173 randomised patients and was conducted in 42 centres in Spain, France and Italy. Patients were screened for EGFR-TK mutations and those with EGFR-TK mutation-positive tumours were randomised to receive either 150 mg of erlotinib orally once a day or one of the following standard platinum-based chemotherapy regimens: cisplatin or carboplatin plus docetaxel; cisplatin or carboplatin plus gemcitabine. In the randomisation, patients were stratified according to Eastern Cooperative Oncology Group (ECOG) status (either ECOG=0, or ECOG=1 or 2) and the mutation type (deletion in exon 19 or mutation in exon 21 L858R). Treatment continued until disease progression, unacceptable adverse reactions, death, or until four chemotherapy cycles were completed. Following disease progression, patients were allowed to cross over in either direction, if clinically appropriate.
3.4 The primary outcome examined in the EURTAC trial was the length of progression-free survival. This was assessed as the time from randomisation to the first occurrence of progressive disease or death from any cause. Secondary outcomes included overall survival, best overall response, disease control, health-related quality of life and safety. Best overall response was defined in terms of the number of patients with either a complete or partial response (as defined by the Response Evaluation Criteria in Solid Tumours [RECIST] version 1 criteria) and disease control included patients with either a complete or partial response and those with stable disease for at least 6 weeks.
3.5 The manufacturer's submission described the results of the intention-to-treat analysis for all randomised patients. The median and 95% confidence limits of progression-free and overall survival between the erlotinib and the platinum doublet chemotherapy arms were obtained from the Kaplan–Meier estimate of the survival function. A two-sided log-rank test was used to assess the difference in outcomes between the two treatment arms. A Cox proportional hazards model was used to estimate the hazard ratio and 95% confidence intervals.
3.6 The EURTAC trial included 153 patients at the time of the interim analysis and 173 at the updated analysis. For the updated analysis there were 86 patients in the erlotinib arm and 87 in the platinum doublet chemotherapy arm. Data for progression-free survival and overall survival from the EURTAC trial are still being collected. Both the interim and updated analyses showed that progression-free survival was statistically significantly longer for patients treated with erlotinib than for patients treated with platinum doublet chemotherapy. In the updated analysis the median progression-free survival in the platinum doublet chemotherapy arm was 5.2 months compared with 9.7 months in the erlotinib arm. The risk of disease progression or death was statistically significantly reduced (by 63%, HR 0.37, 95% CI 0.25 to 0.54, p<0.0001) for patients in the erlotinib arm. In the updated analysis the manufacturer reported overall survival results for 69 (40%) events. The median overall survival was 19.5 months in the platinum doublet chemotherapy arm and 19.3 months in the erlotinib arm (hazard ratio 1.04 [95% CI 0.65 to 1.68], p=0.8702). More patients in the platinum doublet chemotherapy arm received second and further-line treatments than patients in the erlotinib arm (77% [n=67] compared with 45% [n=39]). In the platinum doublet chemotherapy arm, 66 of the 67 patients received at least one treatment with either erlotinib or gefitinib. In the updated analysis, the best overall response (as defined in section 3.4) was statistically significantly greater in the erlotinib arm than the platinum doublet chemotherapy arm (58.1% [95% CI 47.0% to 68.7%] compared with 14.9% [95% CI 8.2% to 24.2%], p<0.0001).
3.7 The manufacturer submitted the results of the OPTIMAL trial, which was carried out in 22 centres in China, as additional evidence. The OPTIMAL trial was a multicentre, open-label, phase III, randomised trial of first-line erlotinib treatment compared with platinum doublet chemotherapy for chemotherapy-naive patients with stage IIIb or stage IV NSCLC whose tumours were EGFR-TK mutation-positive. Patients were randomised (n=165) to receive either 150 mg of erlotinib orally once daily or gemcitabine plus cisplatin chemotherapy. Treatment continued until disease progression, unacceptable adverse reactions or death, or until four chemotherapy cycles were completed. Following disease progression, patients were allowed to cross over in either direction, if clinically appropriate.
3.8 In the most recent analysis from the OPTIMAL trial, progression-free survival was statistically significantly longer in patients treated with erlotinib than in patients treated with platinum doublet chemotherapy. The median progression-free survival in the platinum doublet chemotherapy arm was 4.6 months (95% CI 4.21 to 5.42) compared with 13.7 months (95% CI 10.58 to 15.28) in the erlotinib arm. The risk of progression or death was statistically significantly reduced (by 84%, HR 0.16; 95% CI 0.10 to 0.26, p<0.0001) for patients in the erlotinib arm. The overall survival data from the OPTIMAL trial were not presented because too few deaths had been recorded at the time of the analysis.
3.9 The manufacturer did not perform a meta-analysis of progression-free survival from the EURTAC and OPTIMAL trials because heterogeneity between the treatment effects was identified using an assessment of heterogeneity recommended by the Cochrane Collaboration. The manufacturer noted that factors possibly contributing to the heterogeneity included: the different ethnicity of the patients in the trials; better adherence in the OPTIMAL trial and poorer efficacy of the comparator in the OPTIMAL trial.
3.10 A systematic review identified four randomised controlled trials comparing gefitinib with various doublet chemotherapy regimens in East Asian populations (IPASS, First-SIGNAL, WJTOG3405 and NEJGSG002). The data from the gefitinib trials were pooled by assuming that the doublet chemotherapy was of equal efficacy in each of the four trials (Ku et al. 2011). Across the four studies, the estimated hazard ratio for median progression-free survival was 0.45 (95% CI 0.38 to 0.55, p<0.001).
3.11 For the indirect comparison of erlotinib with gefitinib the manufacturer assumed that the platinum doublet chemotherapy arms of the EURTAC and OPTIMAL trials could be linked to the gefitinib meta-analysis using platinum doublet chemotherapy as the anchor point. From an assessment of the similarities and differences between the studies, the manufacturer concluded that ethnicity is the key factor for the differences and so a robust indirect comparison should involve studies based in an East Asian population. The manufacturer presented results from four possible indirect comparisons of the two erlotinib trials and combinations of them against the gefitinib meta-analysis. In the indirect comparisons the hazard ratio for median progression-free survival varied between 0.36 (95% CI 0.22 to 0.59) and 0.82 (95% CI 0.54 to 1.26) depending on the combination of studies chosen. In the manufacturer's view the hazard ratio for progression-free survival from the indirect comparison of EURTAC with the gefitinib meta-analysis (hazard ratio 0.82 [95% CI 0.54 to 1.26]) was the most appropriate estimate of the clinical effectiveness of erlotinib compared with gefitinib in patients with EGFR-TK mutation-positive NSCLC in England and Wales.
3.12 The manufacturer stated that there were insufficient data on health-related quality of life collected in the EURTAC trial for any analysis to be done. In the OPTIMAL trial quality of life was assessed using the Functional Assessment of Cancer Therapy-Lung (FACT-L) questionnaire and the Trial Outcome Index. Results were presented from 128 (83.2%) patients and demonstrated that approximately 70% of patients receiving first-line erlotinib experienced significant, clinically relevant improvements in quality of life compared with 30% of patients receiving platinum doublet chemotherapy across all FACT-L scales measured.
3.13 The incidence and nature of adverse reactions to erlotinib in the EURTAC and OPTIMAL trials were consistent with previously collected data on the use of erlotinib for first-line maintenance treatment and relapsed NSCLC. The manufacturer noted the longer duration of active treatment with erlotinib compared with chemotherapy and that the extended treatment period may also have increased the number of adverse reactions reported. In the EURTAC trial, patients in the erlotinib arm had a typical treatment duration of 9–10 months before progression or unacceptable adverse reactions, whereas patients in the chemotherapy arm received a maximum of four cycles over approximately 3 months. Most of the reported adverse reactions in both arms were grade 1 or grade 2 (432/527 events [82.0%] in the chemotherapy arm and 621/681 events [91.2%] in the erlotinib arm). Fewer patients experienced grade 3 or 4 events in the erlotinib arm (31 patients [41.3%]) than in the chemotherapy arm (49 patients [66.2%]).
3.14 In the EURTAC trial low grade skin reactions and diarrhoea were the most commonly reported adverse reactions in patients who received erlotinib. Skin reactions were mainly mild or moderate, with 5% of patients experiencing grade 3 rash and 1% experiencing dry skin. No grade 4 skin reactions were reported. Diarrhoea was also mainly mild or moderate, with 4% of patients experiencing grade 3 diarrhoea.
3.15 The manufacturer presented a de novo economic analysis that assessed the cost effectiveness of erlotinib compared with gefitinib for the first-line treatment of EGFR mutation-positive NSCLC. In line with the NICE reference case, outcomes were expressed in terms of quality-adjusted life years (QALYs), an NHS and personal social services perspective was adopted, and costs and benefits were discounted at 3.5%. The treatments compared in the model were first-line erlotinib (one 150-mg tablet daily until disease progression) or gefitinib (one 250-mg tablet daily until disease progression). No second-line treatments were included because the second-line treatment options were considered identical for both erlotinib and gefitinib. The manufacturer presented a semi-Markov economic model with three health states: progression-free survival, progressed disease and death. The model had a 10-year time horizon and a cycle length of 1 month.
3.16 The manufacturer considered that the EURTAC study was more representative of the outcomes expected in UK clinical practice than the OPTIMAL study and so the clinical data in the model were derived from the EURTAC trial and the indirect comparison of erlotinib (EURTAC trial) and gefitinib (Ku et al. 2011). An area under the curve approach was used to calculate the proportion of patients in the progression-free survival health state each month. For erlotinib, the estimated survival curve for the progression-free state was based on the observed EURTAC data up to month 16 and was then extrapolated assuming an exponential distribution. For gefitinib, the progression-free survival curve was derived by transforming the erlotinib survival curve using the hazard ratio for progression-free survival (HR 0.82) from the indirect comparison of erlotinib (EURTAC trial) and gefitinib (Ku et al. 2011). The same transition probabilities, derived from the EURTAC data, were used for both erlotinib and gefitinib for the transition between the progression-free survival health state and death and between the progressed disease health state and death.
3.17 Utilities in the model were based on values from the study of Nafees et al. (2008). These utility values were estimated using the standard gamble approach with 105 members of the UK general public who were asked to value health-state descriptions of patients receiving second-line chemotherapy for NSCLC. These values have been used in four previous NICE appraisals of drugs for NSCLC (Pemexetred for the first-line treatment of non-small-cell lung cancer [NICE technology appraisal guidance 181], Pemexetred for the maintenance treatment of non-small-cell lung cancer [NICE technology appraisal guidance 190], Gefitinib for the first-line treatment of locally advanced or metastatic non-small-cell lung cancer [NICE technology appraisal guidance 192], Erlotinib monotherapy for the maintenance treatment of non-small-cell lung cancer, [NICE technology appraisal guidance 227]). The utility values for the progression-free survival health state were treatment dependent and were calculated from the response rate and the incidence of adverse reactions (grade 3 or 4 rash; grade 3 or 4 diarrhoea). The utility value for the progression-free health state for patients receiving erlotinib (0.661) was based on the response rate in the EURTAC trial (58.10%). The value for patients receiving gefitinib (0.656) was based on a gefitinib response rate (28.23%) which was estimated indirectly by applying the relative response from the gefitinib meta-analysis to the chemotherapy response rate observed in the EURTAC trial (14.9%). The utility decrement value for progressed disease (−0.1798 relative to the progression-free survival stable disease baseline value of 0.6532) was taken from the study of Nafees et al. (2008) and assumed that the choice of first-line treatment had no influence on the utility patients experienced post progression.
3.18 The manufacturer included costs associated with drug acquisition and administration, best supportive care, terminal care, monitoring and adverse reactions in the economic model. These were estimated from a range of secondary sources such as reference costs, BNF and submissions for previous NICE technology appraisals. The monthly cost of erlotinib with the list price (see section 2.3) was £1631.53 based on a daily dose of 150 mg. The manufacturer also presented analyses based on the erlotinib drug cost with the earlier 14.5% discount and with the revised patient access scheme. Under the terms of the gefitinib patient access scheme approved by the Department of Health, there is a single fixed cost of £12,200 per patient when the third monthly pack of gefitinib is supplied. In the base-case analysis, the proportion of patients for whom the £12,200 payment was made was derived by applying the hazard ratio for progression-free survival from the indirect comparison of erlotinib and gefitinib (HR 0.82) to the 'time to last dose' curve generated from the EURTAC data. This results in approximately 76% patients incurring the fixed cost for gefitinib. No administration cost for the erlotinib patient access scheme was included in the economic model because it is a simple discount. For the gefitinib patient access scheme, the manufacturer assumed that the administration cost includes a one-off £70 cost (patient registration, invoicing and query management) and an ongoing monthly cost of £35 (completion of request pack and payment reconciliation).
3.19 Results from the manufacturer's base-case analyses (including the discount under the patient access scheme as revised in 2012) for erlotinib compared with gefitinib show an incremental cost-effectiveness ratio (ICER) of £21,874 per QALY gained. From deterministic sensitivity analyses for a range of parameters, the manufacturer identified the main factors affecting the cost effectiveness as the hazard ratio for progression-free survival for gefitinib and the proportion of patients for whom the gefitinib patient access scheme payment was needed. Varying the hazard ratio for progression-free survival from the indirect comparison from 0.36 to 0.58 resulted in an ICER between £15,712 and £16,552 per QALY gained. When the proportion of patients incurring the fixed charge for gefitinib was varied from 85% to 100%, the ICER was always less than £10,066 per QALY gained. The manufacturer also presented a probabilistic sensitivity analysis which resulted in an ICER of £25,791 per QALY gained for erlotinib compared with gefitinib. There was a 36% probability of erlotinib being cost effective if the maximum acceptable ICER was £20,000 per QALY gained; the probability was 63% if the maximum acceptable ICER was £30,000 per QALY gained.
3.20 The ERG stated that without consideration of pemetrexed in combination with another drug (doublet chemotherapy) as a comparator, the evidence presented in the manufacturer's submission was incomplete and did not allow a full evaluation of erlotinib as set out in the decision problem. The ERG considered pemetrexed-based doublet chemotherapy a valid comparator because almost all patients whose tumours are EGFR-TK mutation-positive have non-squamous lung cancer. In addition some of these will be treated with pemetrexed-based doublet chemotherapy in hospitals that do not routinely test for EGFR and also in situations when delaying treatment to await EGFR-TK status would be detrimental to the patient's health. The ERG stated that the difference in efficacy between pemetrexed and gefitinib has become clearer since the publication of Gefitinib for the first-line treatment of locally advanced or metastatic non-small-cell lung cancer (NICE technology appraisal guidance 192). The ERG stated that pemetrexed is the only first-line treatment for patients with non-squamous cell lung cancer which has demonstrated a statistically significant gain in overall survival when compared with third-generation chemotherapy. Recently published updates to a randomised controlled trial of gefitinib have reported no overall survival gain for gefitinib compared with third-generation chemotherapy.
3.21 In the ERG's view the EURTAC trial was well-designed and suitably powered to demonstrate its primary objective. It considered the inclusion and exclusion criteria to be reasonable and the baseline characteristics of patients in EURTAC trial to reflect patients in UK clinical practice who would be considered eligible for treatment with an EGFR-TK inhibitor. The ERG was unable to comment definitively on the quality of the supporting evidence from the OPTIMAL trial because the clinical study report was not made available.
3.22 The ERG considered that the use of conventional proportional hazards methods to estimate hazard ratios in either the gefitinib or erlotinib trials compared with any other drug is problematic. The assumption of proportional hazards was not tested by the manufacturer. The ERG presented plots of the hazard rates for gefitinib and erlotinib and comparators, which suggested an assumption of proportional hazards was not valid. A comparison of the cumulative hazards for each of the six trials of a tyrosine kinase inhibitor (either gefitinib or erlotinib) compared with platinum doublet chemotherapy showed two separate phases. During the first 4 months of treatment (corresponding approximately to the period of standard chemotherapy), there is very little difference in hazards between intervention and comparator arms. However, in the following 2–3 months the slopes of the lines in all trial arms increase, but with the comparator arms diverging rapidly from the erlotinib or gefitinib arms. A more appropriate method of estimating the relative efficacy involves treating these two time periods as separate phases (equivalent to active therapy followed by observation/maintenance therapy) and deriving separate hazard ratios for each phase (using a landmark analysis for the second phase). In the ERG's view, relative efficacy should be estimated using this approach and the estimates obtained explored in a revised economic model.
3.23 The ERG highlighted that the manufacturer identified heterogeneity between the EURTAC and OPTIMAL trials by comparing the median progression-free survival. In the ERG's view the heterogeneity identified by the manufacturer is simply a consequence of using this outcome measure. A comparison of the Kaplan–Meier curves for progression-free survival from the two trials shows close correspondence in the comparator arms. The two erlotinib arms follow very similar trends although they are slightly separated. Crucially, across successive time periods the gradients of the cumulative hazard curves are very similar. The ERG concluded that the balance of evidence favours including results from both EURTAC and OPTIMAL trials in any indirect comparison.
3.24 The ERG was not convinced that any of the four options for the indirect comparison described by the manufacturer are appropriate. It believed that data from the EURTAC and OPTIMAL trials should be pooled and that revised relative efficacy measures be used (see section 3.22). From an analysis of the progression-free survival and the cumulative hazard curves, the ERG showed that after 12 months the results for patients in the IPASS trial (gefitinib compared with doublet chemotherapy) diverge from the other gefitinib trials. The ERG recommended that a sensitivity analysis that excludes the IPASS data should be undertaken as part of the indirect comparison.
3.25 The ERG was only able to offer a limited critique of the cost-effectiveness results submitted by the manufacturer because of its concerns about the structure of the model. In the ERG's view pemetrexed plus cisplatin should be included as a comparator and there was also an argument for including the four third-generation platinum doublets (docetaxel, gemcitabine, paclitaxel and vinorelbine) in a full evaluation as in Gefitinib for the first-line treatment of locally advanced or metastatic non-small-cell lung cancer (NICE technology appraisal guidance 192). In the ERG's view the omission of all comparators other than gefitinib has resulted in a simple model structure without a robust, multi-way economic comparison that would most likely have reduced the probability of erlotinib appearing as the most cost-effective option.
3.26 The ERG highlighted that the current model yielded an overall survival benefit for patients with EGFR-TK mutation-positive NSCLC receiving first-line erlotinib compared with those receiving gefitinib, which has not been demonstrated by the published evidence from randomised controlled trials. The submitted model does not include any data on overall survival and after disease progression all surviving patients are assumed to follow the same post-progression course and incur the same costs. The direct consequence of the simple model structure is that most of the estimated difference in progression-free survival between patients receiving gefitinib and those receiving erlotinib is preserved by a common post-progression phase, which translates into a similar difference in overall survival.
3.27 Additional evidence was provided by the manufacturer in response to NICE's request in the appraisal consultation document for an updated economic model and analyses. The updated model included, as requested, an assumption of equal progression-free survival and equal utilities for the progression-free survival health state for the two treatments (erlotinib and gefitinib). The manufacturer provided analyses exploring the sensitivity of the cost-effectiveness results to varying the proportion of patients (equally for erlotinib and gefitinib) in the progression-free survival health state at day 60 (for whom the fixed charge for gefitinib is incurred under the patient access scheme). The proportion in the base case was 80%, which was the proportion of patients still receiving erlotinib at the start of the third month of the EURTAC trial. In the sensitivity analyses, the proportion was varied, equally for erlotinib and gefitinib, from the base case to 100%. The costs in the model were not modified. Results from the updated model showed that erlotinib becomes more cost effective than gefitinib when at least 91% of patients incur the gefitinib fixed charge. In the manufacturer's view the proportion of patients incurring the gefitinib fixed charge at day 60 is likely to be more than 90%, as demonstrated in each of the four gefitinib trials. Also data from recent market surveys in Europe and the UK indicate that at least 95% of patients receiving gefitinib have 60 or more days of treatment (and thus incur the fixed charge).
3.28 The ERG explored the analyses using the manufacturer's updated model and confirmed that the Committee's requests specified in the appraisal consultation document had been implemented. However, the ERG noted that the manufacturer assumed the same rates of adverse reactions for the two treatments (erlotinib and gefitinib) when calculating the utility value. The ERG showed that when the same utility value is used for both treatments and the different rates of adverse reactions are retained in the updated model, there is a small additional cost of £5.24 per patient for erlotinib treatment. The ERG was concerned that the cost of administering the gefitinib patient access scheme (a mean cost of £438 per patient over the treatment period) was overstated in the manufacturer's model. The ERG assumed that pack ordering and reconciliation would be needed only once a year and estimated a mean cost for administering the gefitinib patient access scheme of between £111 and £118 per patient. The ERG explored the impact of updating the manufacturer's model with these costs and included the adverse reaction rates for each treatment. The ERG's results demonstrated that erlotinib is cost effective compared with gefitinib when 95% or more of patients receiving gefitinib incur the fixed charge for gefitinib.
3.29 Full details of all the evidence are in the manufacturer's submission and the ERG report, which are available from www.nice.org.uk/guidance/TA258