Aflibercept in combination with irinotecan and fluorouracil-based therapy for treating metastatic colorectal cancer that has progressed following prior oxaliplatin-based chemotherapy

The Committee considered the evidence on the clinical effectiveness of aflibercept, noting that it was derived from the VELOUR trial. The Committee agreed that the VELOUR trial was of good quality and directly relevant to the decision problem; however, it considered that the trial had limitations. The Committee, echoing comments from a patient expert, would have liked the manufacturer to have collected and presented trial data relating to health-related quality of life. The Committee would also have liked the manufacturer to have followed and presented event data for all patients after the end of the trial as defined. The Committee discussed the initial Evidence Review Group's (ERG) concern that patients in the VELOUR trial appeared to have been fitter and younger than those seen in UK clinical practice. The Committee also heard from clinical specialists that the disease and demographic characteristics seen in patients in the VELOUR trial differed from those for patients treated in UK clinical practice; however, evidence from the VELOUR trial showed that response to treatment does not vary across patient groups. The Committee was aware that the studies provided by the manufacturer in response to consultation suggested that patients receiving second-line treatment for metastatic colorectal cancer in the UK were somewhat older than those in the VELOUR trial. The Committee agreed that the patient population in the VELOUR trial was otherwise reasonably representative of patients seen in the UK and therefore concluded that the results of the VELOUR trial are generalisable to UK clinical practice.

The Committee discussed the results for overall survival, the primary end point of the VELOUR trial. The Committee noted that, in its submission and in response to clarification requests by the ERG, the manufacturer produced a range of estimates for the difference in overall survival between the aflibercept and placebo groups of the trial. The Committee considered that the data observed directly from the trial were sufficiently mature at the study cut-off date to establish median overall survival, and agreed that the difference in median overall survival of 1.44 months reflects a statistically significant but clinically small benefit. The Committee noted that the restricted mean difference of 1.92 months (based on the unlikely and conservative assumption that all remaining patients die immediately at the end of the trial) was higher than the median. The Committee therefore concluded that the difference in median overall survival is likely to underestimate the mean survival benefit of aflibercept.

The Committee discussed the estimated mean survival benefit of 4.7 months derived from extrapolation, in light of the trial data. The Committee noted that, to estimate this benefit, the manufacturer extrapolated the survival curves from a trial with a median follow-up of just under 2 years up to 15 years. The Committee noted the comments received in the ACD consultation, and agreed to explain its concerns over the survival extrapolation in more detail. The Committee noted the marked difference between the estimated mean survival benefit of 4.7 months and that of 1.44 months based on median values of overall survival. The Committee understood that the manufacturer considered the extrapolated mean value to represent the magnitude of the clinical benefit of aflibercept better than the median because there were a few patients who experienced a sustained survival benefit from treatment with aflibercept. Although the Committee agreed that a small proportion of patients, with as yet undefined characteristics, appeared to derive greater benefit from aflibercept than most patients in the trial, it agreed that extrapolating the survival curves over 15 years could result in highly uncertain estimates for overall survival. The Committee therefore discussed whether the manufacturer's estimates of mean overall survival were robust. It noted that, during the trial's maximum 3-year follow-up period, 66% of patients in the aflibercept group and 75% of those in the placebo group had died, and a proportion had been censored (data academic in confidence), reducing the number of patients at risk of dying to 47 patients (3.8%) at 30 months and 1 patient (0.1%) at 36 months. The Committee heard from the ERG that, because of this, the hazard ratios for overall survival for patients with follow-up nearing 36 months had wide confidence intervals, reflecting imprecise and uncertain estimates. The Committee noted that, at 36 months' follow-up, the Kaplan–Meier curves indicated that approximately 17% of patients randomised to aflibercept were alive but that this 17% represented only 1 patient remaining uncensored and at risk of dying, which considerably increased the uncertainty around the long-term effect on survival. The Committee noted that the cut-off date for the trial was 07 February 2011, and that the manufacturer was aware that some patients from the trial were still alive. The Committee met 18 months after this date, but no further data to support the manufacturer's extrapolation were available. The Committee appreciated that estimating mean overall survival often requires extrapolating beyond a trial period, but considered that the manufacturer's extrapolation of overall survival from a population with very few patients at risk of dying after 30 months' follow-up, over a further 12 years, was associated with great uncertainty (see section 4.24).

The Committee discussed the mean survival benefit of 4.7 months in light of the different parametric functions used by the manufacturer to estimate overall survival. The Committee was aware that, to estimate this benefit, the manufacturer used the log-logistic function, which it considered to provide the best fit to the observed data, and extrapolated the survival curves over 15 years. The Committee noted that the estimates using other parametric functions ranged from 3.0 months (with the Weibull function) to 5.3 months (with the log-normal function). The Committee discussed which extrapolation period could be considered appropriate to estimate mean overall survival, in view of the life expectancy of patients with metastatic colorectal cancer in clinical practice. The Committee was aware that extrapolation periods should reflect the time in which all patients will have died, but that a longer than 5-year survival for patients with metastatic colorectal cancer is very unusual. It was also aware that, with surgical resection of liver metastases, survival can increase, but that a very small proportion of patients in the VELOUR trial had surgical resection of liver metastases (data designated as academic in confidence), and the Committee was not presented with information about their survival. The Committee also considered survival statistics from the US cancer registry Surveillance, Epidemiology, and End Results (SEER), which showed that 6.9% of patients with metastatic colorectal cancer survive for 5 years. However, because this registry included patients who had received multiple lines of therapy, including surgical resections of tumours and therapies that may not have been considered established NHS practice, the Committee did not consider the data from the SEER registry to be a reasonable proxy for the life expectancy of the population specified in the marketing authorisation of aflibercept. The Committee agreed that a shorter extrapolation period better reflected the natural history of the disease at this stage, and yet accounted for patients who derived greater benefit from aflibercept than most patients in the VELOUR trial. It considered the robustness of the mean overall survival benefit, obtained using the log-logistic function, of 3 months (5 years extrapolation time), 4.7 months (15 years extrapolation time) and 6.6 months (without truncating the survival curves). The Committee was concerned that the log-logistic function had a very 'heavy tail' (that is, a high probability of getting large values at the end of the time horizon) compared with other parametric functions, and that this is likely to have led to an overestimate of the survival benefit of aflibercept. The Committee was also concerned that the manufacturer did not characterise the uncertainty around any of the estimates. In summary, the Committee concluded that, because of the uncertainties around the survival extrapolation, the actual trial data and the life expectancy of patients at this stage of the disease, extrapolating overall survival with the log-logistic function over 15 years did not provide a plausible mean overall survival benefit.

The Committee considered the relationship between progression-free survival and overall survival from the VELOUR trial. The Committee was aware that the manufacturer used disease progression assessed by an independent review committee in its base case. The difference in median progression-free survival between aflibercept and placebo using this methods was 2.23 months, which was a higher value than when disease progression was determined by investigator assessment (1.74 months) and higher than the mean progression-free survival (see section 3.25).The Committee considered the shapes of the Kaplan–Meier curves (reflecting the trial data) for overall survival and for progression-free survival. It noted that the curves continued to diverge during the trial period for overall survival, whereas, for progression-free survival, the curves initially diverged but then converged at around 12 months, reflecting almost the same rate of progression for patients randomised to aflibercept or placebo from that time onward. The Committee heard from the clinical specialists that, because the overall survival curves continued to separate for both patients who had or had not stopped treatment, the survival curves might reflect a disease-modifying effect in that aflibercept might have altered the natural course of the disease whereby, despite the disease progressing, patients lived longer even after treatment stopped (that is, survival post disease progression was increased more than progression-free survival). The Committee discussed how the disease-modifying effect could be explained clinically. It heard that aflibercept may have delayed death by shrinking tumours, and so extended the period before the tumour grew again. However, the Committee was not presented with evidence that tumours had shrunk, and was aware that the disease had progressed in all patients during the trial. The Committee agreed that there was no robust evidence to make firm conclusions about the likely cause of the different shapes of the overall survival and progression-free survival curves, and that the magnitude of progression-free survival depended on the method used to calculate it.

The Committee considered that the subgroup analysis presented by the manufacturer for patients with liver metastases only compared with metastases not confined to the liver. The Committee noted that, in this subgroup, there was a statistically significant interaction test at the 10% level. The Committee was aware that the 10% significance level was less specific (that is, a higher chance of a positive finding) than the more conventional 5% level. The Committee agreed that there is no evidence to suggest that aflibercept would be more effective in patients with liver metastases only than in patients with metastases confined to other organs. The Committee was aware that patients with liver metastases only are more likely to be considered for surgical resection of the metastases and possibly live longer than those with widespread metastases. The Committee therefore discussed whether aflibercept can make liver metastases operable in patients with metastatic colorectal cancer. It noted that only a very small minority of patients in the VELOUR trial proceeded to have surgical resection of liver metastases after treatment with aflibercept. The Committee heard from the clinical specialist that, in approximately 20% to 30% of patients who have surgery to remove liver metastases, metastatic colorectal cancer can be cured. The Committee, however, was not presented with evidence about rates of resection and cure with aflibercept in the subgroup of patients with liver metastases only. The Committee also considered that resecting liver metastases to achieve a cure was more appropriate in the first-line setting than in the second-line setting. Furthermore, it heard from the manufacturer that the modelling of the subgroup did not include the costs of surgical resection. The Committee concluded that it would be appropriate to include this cost and that including it is likely to affect the incremental cost-effectiveness ratio (ICER). The Committee agreed that, given the lack of evidence, aflibercept cannot be considered an effective treatment option to make liver metastases resectable. The Committee therefore concluded that this subgroup should not be considered further.

The Committee considered the subgroup that excluded patients who had received oxaliplatin-based therapy in the adjuvant setting and whose disease had relapsed within the following 6 months. The Committee heard from the clinical specialists that, in clinical practice, patients in this subgroup would not be treated differently from the overall trial population. In addition, the Committee noted that the manufacturer acknowledged that the analysis for this subgroup was planned after the trial results had been compiled (post hoc), and that the test for interaction did not show that the treatment effect in this subgroup differed from the effect in the rest of the trial population. The Committee therefore concluded that it did not need to consider further the subgroup that excluded patients whose disease had relapsed 6 months or less after starting oxaliplatin-based adjuvant therapy.

The Committee discussed the adverse events associated with aflibercept. The Committee noted that more patients in the aflibercept group (27%) stopped treatment because of adverse events than in the placebo group (12%). The Committee also noted that adding aflibercept to FOLFIRI increased the adverse events typically associated with FOLFIRI, most notably neutropenia, although it heard from clinical specialists that neutropenia would not routinely be treated in clinical practice. The Committee heard from the manufacturer that the dose of FOLFIRI used in the trial was higher than the dose that is routinely used in clinical practice and might have caused some of the adverse events. The Committee was also aware that aflibercept increased the risk of hypertension, as would other anti-vascular endothelial growth factor therapies. The Committee concluded that treatment with aflibercept plus FOLFIRI was associated with a considerable burden of adverse effects, but that, being a new treatment, less is known about its adverse effects profile than for other available treatments.

The Committee considered the structure of the model submitted by the manufacturer, and how it captured the main aspects of the condition. The Committee noted that the manufacturer chose to split the stable-disease health state into 2 sub-states to capture the costs and health benefits for patients with stable disease who either receive second-line treatment with aflibercept plus FOLFIRI or FOLFIRI alone, or who have stable disease but have stopped second-line treatment for reasons other than disease progression. The Committee heard from the ERG that the manufacturer applied the same utility value to the 2 sub-states of the stable-disease health state. It further heard that the acquisition and administration costs of second-line treatments in the model did not depend on the proportion of patients in each state, and that they were calculated outside the model. The Committee noted that the costs and quality-adjusted life years (QALYs) in the stable-disease health state were not specific to the 2 sub-states ('on second-line treatment' and 'post second-line treatment'). The Committee concluded that overall the model adhered to the NICE reference case for assessing cost effectiveness.

The Committee discussed whether or not the 15-year time horizon used by the manufacturer in the model was appropriate. The Committee appreciated that that the choice of the time horizon is a sensitive parameter in the model given the uncertainty associated with extrapolating overall survival. The Committee was aware that the time horizon should be sufficiently long to capture all the costs and health benefits in the full population (that is, a lifetime horizon should be used). The Committee therefore concluded that a time horizon of 15 years was, in principle, appropriate because all patients are likely to have died by 15 years; however, the Committee agreed that, when the time horizon is much longer than the trial duration, and the life expectancy of most patients, it is particularly important to explore the assumptions underlying how overall survival is extrapolated.

The Committee discussed the manufacturer's assumptions for extrapolating overall survival in the model, and the alternative assumptions considered by the ERG in its exploratory analyses. The Committee noted that the manufacturer assumed that the survival benefit from treatment with aflibercept plus FOLFIRI increases relative to treatment with FOLFIRI alone until around 12 months after starting treatment, and then decreases over the 15-year time horizon, but that the hazard ratio never reaches 1.0 (that is, a patient previously randomised to aflibercept will always have a lower risk of dying, even if not receiving aflibercept, relative to a patient previously randomised to placebo). The Committee noted that the ERG explored 2 alternative scenarios:

• the first assumed that the risk of death becomes the same in both treatment groups at the point at which the trial ends and continues to be the same for the remainder of the time horizon period (that is, the hazard ratio becomes 1.0 after 3 years)

• the second assumed that the overall survival curves for aflibercept plus FOLFIRI and FOLFIRI alone converge over the time horizon (that is, the hazard ratio gradually increases from the end of the trial until the survival curves come together, then the hazard ratio becomes 1.0 thereafter).
  
  The Committee understood that, in the ERG's second scenario, patients receiving aflibercept plus FOLFIRI need to have a higher risk of death than patients receiving FOLFIRI alone (that is, the hazard ratio may be greater than 1.0) for the curves to converge. The Committee considered that the manufacturer's assumption that the treatment benefit continues beyond the trial period and until 15 years is highly uncertain given that most patients had died during the 3-year follow-up period of the trial. The Committee considered that the ERG's analysis that allows the hazard ratio to become greater than 1.0 could be considered implausible. The Committee agreed that the ERG's first scenario, which assumes equal risk of death for all patients beyond the trial period (hazard ratio equals 1.0), represents an acceptable compromise between the 2 extremes of assuming continuing treatment effect (manufacturer's base case) and allowing for a reversed treatment effect (ERG's second scenario). The Committee noted that, in response to consultation, the manufacturer implemented a new scenario in its revised base case in which the hazard ratio begins to taper to 1.0 36 months after starting treatment, over a 12-month period. The Committee agreed that as a means to extrapolate overall survival both its preferred scenario (that is, the ERG's first scenario) and the manufacturer's new scenario were associated with some degree of uncertainty. In the absence of further evidence to validate the manufacturer's new approach, the Committee maintained its preference for the ERG's first scenario.

The Committee considered the estimates of health-related quality of life used in the manufacturer's model, noting that it would have preferred these data to have been collected from the VELOUR trial. The Committee was aware that the manufacturer got the utility value for the stable-disease state from the 'mCRC utilities study' and revised it after consultation to a value derived from the ASQoP study because the data from this study were new, and not because the Committee questioned the validity of the original value. The Committee noted that the ERG preferred another value from the ASQoP study for the stable-disease state but, because the difference between the manufacturer's revised value (0.78) and the ERG's preferred value (0.77) was small and likely to have a negligible impact on the ICER, the Committee concluded that either value could be considered appropriate.

The Committee discussed the appropriate utility value for the progressed-disease state in the model, noting that, because approximately three-quarters of the QALY gain in the model was accrued after disease progression, the model is highly sensitive to this parameter. The Committee, having noted the mean and median time to disease progression in the manufacturer's utility study, considered that the utility value chosen by the manufacturer for the progressed-disease state did not reflect the entire duration of progressed disease but only early progressed disease, and so was likely to be an overestimate (see section 3.28). The Committee was aware of the participation bias associated with studies of this nature. Furthermore, it heard from the ERG that the manufacturer's study was small, and produced counter-intuitive estimates in 1 subgroup analysis. The Committee was aware that, although the manufacturer stated that the data queries noted in its submission had been resolved, the manufacturer had yet to submit the study for peer-reviewed publication. The Committee was aware that, in its base case, the ERG used an alternative lower value of 0.60, which had been used in NICE's technology appraisal guidance on bevacizumab and cetuximab for the treatment of metastatic colorectal cancer, and that the ERG considered this value to represent a reasonable balance of the utility values for progressed disease used in other NICE guidance, which ranged from 0.21 to 0.69. The Committee was aware that the utility value of 0.69 used in NICE's technology appraisal guidance on cetuximab, bevacizumab and panitumumab for the treatment of metastatic colorectal cancer after first-line chemotherapy for progressed disease was based on patients who had lived long enough to receive more courses of chemotherapy than patients in the VELOUR trial, and so likely reflected patients with a better health state. The Committee agreed that no utility values for progressed disease were universally accepted as valid, but that it would be important that the utility value reflected the entire progressed-disease state. The Committee was aware that the quality of life for patients with metastatic colorectal cancer deteriorates relatively slowly other than during the last few months, when it may deteriorate faster, and that exploring a utility value of 0.3 during the last 2 months of life was a reasonable attempt by the manufacturer to address this. The Committee also agreed that adjusting the utility values for age was appropriate to reflect the natural deterioration in health-related quality of life in patients with the disease. The Committee concluded that the most plausible utility value for the progressed-disease health state would lie between the manufacturer's and the ERG's estimate.

The Committee discussed the costs of administering aflibercept plus FOLFIRI in the model, noting that the manufacturer assumed no extra cost for administering aflibercept in its original model. The Committee was aware that aflibercept would normally be prepared in a sterile compartment, and would therefore incur an extra cost; the Committee estimated that this cost is likely to be higher than the £15 used by the ERG. The Committee was also aware that the marketing authorisation for aflibercept stipulates that aflibercept should be administered over 1 hour before the infusion with FOLFIRI, but that the cost for an additional hour of infusion time (£45) was not included in the ERG base case. The Committee acknowledged that the manufacturer's revised base case accounted for the extra preparation cost and the cost for an additional infusion time for aflibercept.

The Committee noted that, in response to consultation, the manufacturer had provided data showing that the average age of patients treated in the NHS with second-line chemotherapy for metastatic colorectal cancer was 60 years in 1 study and 63 years in 3 others. The Committee agreed that the 70-year age of starting treatment, as initially assumed by the ERG in its base case, was therefore too high. It concluded that an age between 60 and 65 years is more appropriate.

The Committee discussed the costs in the model derived from the manufacturer's survey of clinical oncologists about community-based care, and personal and social care. The Committee noted that this study was small and therefore associated with uncertainty, and did not provide evidence that the oncologists in the survey were representative of practitioners in the UK. The Committee noted that the manufacturer's model incorporated median estimates from the survey because the responses from clinicians on 1 parameter (the number of visits received by a patient from a palliative care team) included an outlier, whereas the ERG argued that the mean was more appropriate. The Committee agreed that, if the sample of clinicians was appropriately homogenous and reflected similar practices, the distribution of the data collected from the survey would be largely uniform, and it would be more appropriate to use the mean rather than the median. The Committee noted that, although the manufacturer continued to use the median value in its revised base case, it presented a scenario analysis that incorporated the mean after excluding the outlier, an approach that the ERG considered appropriate. Although the Committee agreed that mean values should normally be used to estimate resource use and costs, it concluded that, in this instance, using the mean after excluding the outlier could be considered appropriate.

The Committee discussed whether aflibercept should be considered an innovative treatment. The Committee acknowledged that aflibercept represented a novel recombinant fusion protein. However, the Committee concluded that all benefits of a substantial nature relating to treatment with aflibercept plus FOLFIRI had been captured in the QALY calculation.

The Committee considered supplementary advice from NICE that should be taken into account when appraising treatments that may extend the life of patients with a short life expectancy and that are licensed for indications that affect small numbers of people with incurable illnesses. For this advice to be applied, all the following criteria must be met.

• The treatment is indicated for patients with a short life expectancy, normally less than 24 months.

• There is sufficient evidence to indicate that the treatment offers an extension to life, normally of at least an additional 3 months, compared with current NHS treatment.

• The treatment is licensed or otherwise indicated for small patient populations.
  
  In addition, when taking these criteria into account, the Committee must be persuaded that the estimates of the extension to life are robust and that the assumptions used in the reference case of the economic modelling are plausible, objective and robust.

The Committee considered the criterion for short life expectancy and the evidence for life expectancy in this group of patients. The Committee noted the overall survival estimates presented by the manufacturer from the VELOUR trial with the observed median survival in the placebo group of VELOUR of 12.1 months and the estimated mean overall survival of 18.1 months. The Committee also noted the ERG's preferred estimate of 10.5 months from the literature. The Committee concluded that patients receiving current standard NHS treatment would have an expected survival of less than 24 months from the point at which they would be considered for second-line therapy and that therefore the criterion for short life expectancy was fulfilled in this appraisal.

The Committee considered the criterion that the treatment is licensed or otherwise indicated for small patient populations. The Committee noted the manufacturer's suggestion that approximately 4000 patients in England and Wales would receive second-line treatment for metastatic colorectal cancer. The Committee was concerned that aflibercept holds a marketing authorisation for treatment of a much larger population with neovascular (wet) age-related macular degeneration, but that this was a different formulation of aflibercept marketed by another company. The Committee understood that when one technology is marketed by different companies (for different indications, using different brands), these should not be added for the purpose of establishing the cumulative population to be considered in the context of life-extending treatments at the end of life, and that therefore the criterion for a small population size was fulfilled in this appraisal.

The Committee considered the criterion that treatment offers an extension to life of normally at least an additional 3 months. The Committee noted the comments received in consultation on the ACD, and agreed to explain its concerns over the magnitude of the mean survival benefit more fully. The Committee noted that, based on the number of patients who had died during the trial (70.4%), 50% of those who received aflibercept lived for up to 1.44 months longer than people who received placebo, and acknowledged the difficulty in finding robust mean overall survival data considering the issues with the extrapolation carried out (see section 4.7). The Committee noted that, in response to consultation, the manufacturer pointed out that the original base-case model, using the Committee's preferred assumption to extrapolate overall survival (section 4.14), predicted that aflibercept would extend life by 3.4 to 3.7 months. The Committee discussed whether the estimates for mean overall survival produced by the model were robust indicators of what overall survival benefit can be seen in clinical practice, noting that all of the extrapolation assumptions were associated with great uncertainty. The Committee was aware that the longer the time horizon, the greater the influence of the 'tails' of the extrapolation curves, which define the difference in mean overall survival between the treatment arms, and to which the model is highly sensitive. The Committee agreed that, although there is a rationale for a 15-year time horizon in order to capture the very small number of patients who might have very prolonged survival, this introduced considerable uncertainty, and produced implausible results given that the extrapolation was based on a population with a small number of patients still at risk of dying beyond 30 months. Although the use of a 15-year time horizon is, in principle, appropriate, when extrapolating the relative benefit is associated with uncertainty, the Committee considered it appropriate to consider shorter time horizons as a means to explore the uncertainty. The Committee noted that, when the model time horizon was shortened to 5 years, the difference in mean overall survival decreased to 2.7 to 2.8 months. The Committee was mindful that, when there is quantitative evidence that a treatment offers a 3-month life extension, it must also be persuaded that the estimates of life extension are robust and that the assumptions used in the reference case of the economic modelling are plausible, objective and robust. The Committee agreed that, given the considerable uncertainty around extrapolating overall survival and its implementation in the modelling, it is important to take into account what has actually been observed in the trial (see section 4.6 and 4.8) and in the absence of other evidence, the Committee was not satisfied that the estimates from fitting parametric functions to Kaplan–Meier data or those produced by the model were sufficiently robust to accept that the 3-month life extension criterion is fulfilled. The Committee therefore concluded that aflibercept did not meet the criteria for an end-of-life therapy as defined by NICE.

The Committee noted that, in its response to consultation on the appraisal consultation document, the manufacturer pointed out that, in NICE's technology appraisal guidance on cetuximab, bevacizumab and panitumumab for the treatment of metastatic colorectal cancer after first-line chemotherapy, the Committee had considered a modelled survival benefit of 2.7 to 3.2 months to show sufficient evidence for a 3-month survival benefit for panitumumab. The Committee was aware that, in judging whether panitumumab met the criterion for life extension, the Committee for NICE's technology appraisal guidance on cetuximab, bevacizumab and panitumumab for the treatment of metastatic colorectal cancer after first-line chemotherapy had taken into consideration the difficulty in accommodating the cross‑over in the panitumumab trials and that the mean progression-free survival benefit for panitumumab was similar to that for cetuximab, and that the latter resulted in an overall survival benefit of 4.7 months. Therefore, it had considered that there was sufficient evidence to indicate that panitumumab offers an extension to life of approximately 3 months.

The Committee discussed the ICERs for aflibercept in combination with irinotecan and fluorouracil-based therapy for metastatic colorectal cancer based on the revised analyses provided in response to consultation. The Committee agreed that the cost-effectiveness analysis should assume equal risk of death for all patients beyond the trial period, and that the starting age of the modelled cohort should be between 60 and 65 years. The Committee noted that the manufacturer's ICER closest to these assumptions was £44,000 per QALY gained (for age 60), but would increase for the higher age bracket, if the mean value was used from the manufacturer's survey of clinical oncologists after removing the outlier and if an extrapolation function with a less heavy tail had been used. Because the manufacturer's ICERs incorporated a utility value for progressed disease deemed by the Committee to be high, the Committee considered the ICER produced by the ERG using the Committee's preferred assumptions, but which used a utility value for progressed disease of 0.6. The Committee noted that this was approximately £51,000 per QALY gained and would be higher if an extrapolation function with a less heavy tail had been used. The Committee therefore concluded that the most plausible ICER was higher than the normally acceptable maximum ICER range of £20,000 to £30,000 per QALY gained, and that aflibercept in combination with irinotecan and fluorouracil-based therapy could not be considered a cost-effective use of NHS resources for patients with metastatic colorectal cancer.

The Committee noted the comments received during consultation on the appraisal consultation document that some patients appeared to gain particular benefit ('a bimodal distribution'), and which stressed the importance of offering only certain patients aflibercept. The Committee was aware that there is currently no established method in clinical practice to identify patients with metastatic colorectal cancer who could particularly benefit from treatment, and it was not presented with evidence on how these patients could be selected for treatment with aflibercept. The Committee was aware that, as a post-authorisation commitment to the European Medicines Agency, the manufacturer initiated a biomarker program encompassing 3 studies to help select patients who may be more likely to benefit. The Committee agreed that the results of these studies would be useful for a future review of this appraisal.