Bevacizumab in combination with capecitabine for the first-line treatment of metastatic breast cancer

Clinical effectiveness

3.1 The manufacturer conducted a literature search and identified 2 randomised controlled trials (TURANDOT and RIBBON-1) that investigate the effect of first-line bevacizumab plus capecitabine in adults with metastatic breast cancer. The TURANDOT trial was excluded because it is ongoing and no efficacy data are available. The RIBBON-1 trial was an international, multicentre, double-blind, phase III, randomised, placebo-controlled trial comparing bevacizumab plus chemotherapy with chemotherapy alone for the first-line treatment of HER2-negative, locally recurrent or metastatic breast cancer.

3.2 The RIBBON-1 trial enrolled 1237 patients to receive bevacizumab plus chemotherapy or chemotherapy plus placebo. Investigators were able to select their choice of chemotherapy before randomisation. Patients were enrolled into 2 different cohorts; in 1 cohort patients received either an anthracycline or a taxane, and in the other cohort patients received capecitabine, reflecting the choice of first-line therapy for these patients in routine clinical practice. Patients were then randomised to bevacizumab plus the chosen chemotherapy or to the chosen chemotherapy plus placebo. The manufacturer stated that only the results from the capecitabine cohort provided evidence on the use of bevacizumab in its licensed indication, in combination with capecitabine for the first-line treatment of metastatic breast cancer. The manufacturer highlighted that anthracyclines and taxanes were not considered appropriate as first-line treatment for all patients in the capecitabine cohort; about 40% of the patients had previously received taxanes and around 63% had received anthracycline therapy for early breast cancer.

3.3 In the capecitabine cohort of the RIBBON-1 trial 615 patients were randomised in a 2:1 ratio to the bevacizumab plus capecitabine arm (n=409) and the capecitabine plus placebo arm (n=206). Randomisation was stratified by the following criteria: disease-free interval (12 months or less, more than 12 months since completion of adjuvant chemotherapy or surgery if no adjuvant chemotherapy); previous adjuvant chemotherapy; and number of metastatic sites (fewer than 3, 3 or more). The dosage of bevacizumab was 15 mg/kg by intravenous infusion every 3 weeks, and the dosage of capecitabine was 1000 mg/m² orally twice daily for 2 weeks of a 3-week cycle. Treatment was continued until disease progression, unacceptable toxicity, investigator or patient decision to stop treatment, or death. Patients continued to receive capecitabine if bevacizumab was discontinued before disease progression. After disease progression, patients in either arm could move to an open-label phase consisting of treatment including bevacizumab and chemotherapy at the investigator's discretion. Patients who chose not to enter the post-progression phase and patients who discontinued treatment during the post-progression phase were followed up in a survival follow-up phase.

3.4 The primary endpoint in the trial was investigator-assessed progression-free survival according to Response Evaluation Criteria in Solid Tumors (RECIST) criteria. It was defined as the time from randomisation to first disease progression or death from any cause. Progression-free survival based on an Independent Review Committee (IRC) review of the data was considered a secondary endpoint and presented as a sensitivity analysis to support the investigator-assessed primary endpoint. Other secondary endpoints included objective response rates, defined as the percentage of patients with a complete or partial response determined on 2 consecutive assessments more than 4 weeks apart; duration of objective response, defined as the time from the first tumour assessment that supported an objective response to the time of disease progression, or death from any cause; overall survival, defined as the time from randomisation until death from any cause; and the 1-year survival rate, defined as the percentage of patients still alive 1 year after randomisation. In addition, progression-free survival and overall survival were calculated for a number of pre-specified subgroups, post hoc exploratory subgroups, and subgroups specified after the trial had begun but before the analysis was completed (for example, the subgroup of patients previously treated with a taxane, which was included in the manufacturer's economic model).

3.5 There was a statistically significant increase in the investigator assessed median progression-free survival of 2.9 months, from 5.7 months in the capecitabine plus placebo arm to 8.6 months in the bevacizumab plus capecitabine arm. The stratified hazard ratio for progression was 0.69 (95% confidence interval [CI] 0.564 to 0.840, p=0.0002). Median overall survival improved by 2.9 months, from 22.8 months with capecitabine plus placebo to 25.7 months with bevacizumab plus capecitabine. The stratified hazard ratio for death was 0.88 (95% CI 0.69 to 1.13, p=0.33), indicating a 12% improvement in overall survival with bevacizumab plus capecitabine compared with capecitabine plus placebo. However this improvement was not statistically significant. The manufacturer acknowledged that the results from the patients who crossed over to bevacizumab in the open-label post-progression phase of the trial (44.7% in the bevacizumab/capecitabine arm and 52.4% in the capecitabine/placebo arm) may have confounded overall survival results. This was because the trial was not designed to evaluate the effect of subsequent therapies.

3.6 A number of subgroup analyses for progression-free survival (16 in total) and for overall survival (24 in total) with no correction for multiple testing were presented in the manufacturer's submission. The manufacturer highlighted that bevacizumab plus capecitabine gave a progression-free survival benefit over capecitabine plus placebo in all of the pre-specified subgroups defined by stratification variables, although not all were statistically significant. The manufacturer investigated a number of additional planned and post hoc subgroups and showed that some subgroups (for example, the group previously treated with a taxane) had a greater overall survival benefit than the intention-to-treat (ITT) population of the capecitabine cohort.

3.7 The manufacturer focused on the subgroup of patients who had a previous adjuvant or neo-adjuvant taxane. This subgroup of 245 patients had an increase in median progression-free survival of 4.5 months, from 4.2 months in the capecitabine plus placebo arm to 8.7 months in the bevacizumab plus capecitabine arm. The hazard ratio for progression was 0.62 (95% CI 0.45 to 0.84). This benefit also translated into an overall survival benefit, with an increase in median overall survival of 7.9 months, from 20.5 months in the capecitabine plus placebo arm to 28.4 months in the bevacizumab plus capecitabine arm. The hazard ratio for death was 0.67 (95% CI 0.46 to 0.98). These overall survival results were based on 70 deaths in the bevacizumab plus capecitabine arm and 44 deaths in the capecitabine plus placebo arm. The manufacturer stated that patients previously treated with a taxane had worse outcomes than the patients in the ITT population, and the addition of bevacizumab increased their progression-free survival and overall survival to levels similar to or above those of the ITT population. The manufacturer presented the results of two similar metastatic breast cancer trials (the AVADO and E2100 trials) which demonstrated the same pattern of progression-free and overall survival gains from bevacizumab in patients who have previously received a taxane. The AVADO trial compared bevacizumab plus docetaxel with docetaxel plus placebo, and the E2100 trial compared bevacizumab plus paclitaxel with paclitaxel alone.

3.8 The primary safety analyses were based on all patients who received any trial treatment, defined as at least 1 full or partial dose of either trial treatment during the blinded phase of the trial. This population was referred to by the manufacturer as the safety population. The manufacturer stated that adding bevacizumab to capecitabine resulted in adverse events that were predictable based on previous use of bevacizumab, and generally manageable. Grade 3–5 adverse events were higher with bevacizumab plus capecitabine (36.6%) compared with capecitabine plus placebo (22.9%). In addition, the following adverse events were higher with bevacizumab plus capecitabine compared with capecitabine plus placebo: hypertension (10.6% compared with 1%), proteinuria (2.2% compared with 0%), sensory neuropathy (3% compared with 0.5%) and venous thromboembolic events (5% compared with 3.5%).

3.9 Health-related quality of life data were not collected in the
RIBBON-1 trial. The manufacturer stated that the most important factor causing distress among cancer patients was the fear of disease progression. Therefore a major objective of each successive line of therapy, in addition to extending overall survival, was to maintain progression-free survival for as long as possible.

3.10 The ERG stated that the literature search conducted by the manufacturer was appropriate, that all relevant studies had been identified, and that the RIBBON-1 trial on which the manufacturer's submission was based was relevant to the decision problem in its analysis. The ERG stated that the patient population in the trial was in line with the marketing authorisation for bevacizumab in combination with capecitabine. The ERG commented that the trial was well conducted, the baseline characteristics appeared to be balanced across the treatment groups, and the stratification factors were appropriate. The ERG noted that the dose for capecitabine in the trial was 1000 mg/m² rather than the licensed dose of 1250 mg/m². However, this was considered appropriate and in line with clinical practice. The ERG stated that the results from the trial could be generalised to patients in the UK.

3.11 The ERG noted that the hazard ratios for investigator- and IRC-assessed progression-free survival were almost identical, indicating that the evidence of progression-free survival benefit with bevacizumab plus capecitabine was robust. The ERG was aware that the progression-free survival benefit did not translate into a statistically significant overall survival benefit, but stated that interpreting differences in overall survival was difficult because patients from both the capecitabine plus placebo arm and the bevacizumab plus capecitabine arm were able to cross over to receive bevacizumab in the open-label phase of the trial. Other anticancer therapies were also available on progression, and in a minority of instances before progression, so bias may have been introduced.

3.12 The ERG noted the subgroup analyses conducted by the manufacturer, and commented that most increases in progression-free survival with bevacizumab plus capecitabine compared with capecitabine plus placebo were statistically significant in these subgroups. However, the only overall survival results that were statistically significant were for subgroups of patients younger than 50 years and subgroups of patients previously treated with a taxane or anthracycline as neoadjuvant or adjuvant chemotherapy. The ERG stated that the results of the subgroup analyses should be considered with caution because no statistical adjustments were performed to control for multiple testing in any of the 40 subgroups and of all outcomes, thus increasing the likelihood of significant results emerging by chance when using the usual level of significance of 5%.

3.13 The ERG agreed that there was a greater proportion of adverse events in the bevacizumab plus capecitabine arm, but that no new safety concerns were identified. The ERG also agreed that bevacizumab plus capecitabine did not lead to a clinically relevant increase in adverse events typically associated with chemotherapy, such as febrile neutropenia, neutropenia, and sensory neuropathy. The ERG stated that the difference in adverse events between the 2 arms could largely be attributed to differences in grade 3 adverse events (27% in the bevacizumab plus capecitabine arm compared with 14% in the capecitabine plus placebo arm).

3.14 Regarding the safety of bevacizumab plus capecitabine compared with capecitabine plus placebo in the subgroup of people who had previously received a taxane, the ERG stated that it was not possible to compare the proportions of patients who experienced any adverse events, any grade 3–5 adverse events, any serious adverse events or any adverse events leading to discontinuation of bevacizumab or placebo because the manufacturer did not present these data. The ERG extracted some data from the economic model, and stated that adverse events of special interest mostly appeared to be similar in frequency in the subgroup and in the overall trial population. A slightly greater proportion of patients in the subgroup reported grade 3 or higher cardiac disorders (4.4%) than in the overall safety population (2.1%). However, the ERG stated that these findings must be viewed with caution because of the small numbers of patients in this subgroup.

Cost effectiveness

3.15 In a systematic review of the literature the manufacturer found no cost-effectiveness studies comparing bevacizumab plus capecitabine with capecitabine plus placebo as first-line treatments for metastatic breast cancer. No relevant cost-effectiveness analyses were identified. The economic evaluation was based on the subgroup of patients from the RIBBON-1 trial who had previously received a taxane, and all efficacy and treatment duration parameters were derived from this subgroup. The manufacturer assumed that patients in this subgroup would probably have received an anthracycline as well. The manufacturer stated that this subgroup reflected the marketing authorisation for capecitabine. Capecitabine monotherapy has a marketing authorisation for 'the treatment of patients with locally advanced or metastatic breast cancer after failure of taxanes and an anthracycline-containing chemotherapy regimen or for whom further anthracycline therapy is not indicated'. The manufacturer acknowledged that this post hoc subgroup analysis of patients previously treated with a taxane was the main weakness of the economic evaluation.

3.16 The manufacturer developed a 3-state model. All patients enter the model in the progression-free survival health state and in each month can either progress to a 'worse' health state (that is, from progression-free survival to progressed disease or from either state to death) or remain in the same health state. The manufacturer stated that these health states were consistent with previous modelling of metastatic cancer. The progression-free survival health state is designed to capture a patient's relatively high quality of life before disease progression and the progressed disease state is designed to capture the relatively poor quality of life after disease progression. Survival data from the capecitabine plus placebo arm of the subgroup previously treated with a taxane from the RIBBON-1 trial were used to inform disease progression in the comparator arm. The treatment duration in the trial was used to determine the expected cost of treatment with each regimen in the base case. The model has a 1-month cycle length, includes a half-cycle correction and both costs and benefits are discounted at 3.5%. The time horizon was 15 years.

3.17 The proportions of patients who are progression-free in each month were taken directly from Kaplan-Meier survival curves for each treatment arm in the RIBBON-1 trial until the 12th month of treatment, after which an exponential distribution of survival time was assumed. The number of patients in each treatment arm dying from any cause while in the progression-free survival state was used to derive a constant rate and probability of mortality. The mortality rate in the progression-free survival state was assumed to be at least as great as the underlying sex- and age-related mortality in the general population.

3.18 A number of tunnel states, health states which can only be passed through in a certain order, were generated for patients with progressed disease according to the time spent in this state. The tunnel states were arranged so that each state had a progression only to death or the next temporary state. Patients who entered the progressed disease state had a probability of dying that increased each month based on an extrapolation of the survival data for patients with progressed disease. Mean overall survival was the sum of mean duration of progression-free survival and mean duration of progressed disease.

3.19 During the progressed disease phase, patients in the capecitabine cohort of the RIBBON-1 trial received a variety of different therapies. The manufacturer modelled survival in progressed disease based on adjusted analyses that aimed to 'uncross' the survival curves by excluding survival gains from patients who crossed over to bevacizumab in the open-label phase of the trial. An exponential survival distribution was assumed thereafter. The data were 'uncrossed' using a rank preserving structural failure time model to take account of the bias that may have been introduced by allowing patients from both treatment arms to receive bevacizumab after progression, potentially distorting overall survival rates in the control arm.

3.20 The manufacturer carried out a literature review to identify relevant health-related quality of life data to use in the economic evaluation. Three studies that measured utility values directly were identified and, of these, the manufacturer calculated utility values for progression-free survival and progressed disease from the results of the mixed model analysis presented by Lloyd et al. (2006). The manufacturer stated that it was most appropriate to use a base-case progression-free survival utility value that was derived from a large population, and then to adjust that base-case utility by response rate. In addition, the utility values from Lloyd et al. have been used in previous health technology appraisals for metastatic breast cancer. For patients in the progressed disease state, a health state utility value of 0.496 was incorporated in both treatment arms. For patients in the progression-free survival state a treatment-specific weighted average of the values for stable disease and treatment response, based on the reported overall response rate, was calculated: 0.784 in the bevacizumab plus capecitabine arm and 0.774 in the capecitabine plus placebo arm. The manufacturer acknowledged that the utility values reported by Lloyd et al. were not derived from patient experience, and presented a sensitivity analysis using data from Peasgood et al. (2010) to derive estimated utilities from patients valuing their own health.

3.21 The drug costs incorporated in the model for bevacizumab plus capecitabine and capecitabine were from BNF 62 (£4001.53 per month and £312.41 per month respectively). No vial sharing was assumed for bevacizumab. The manufacturer's submission assumed administration and pharmacy costs of £348.82 in the first month and £205.97 per month for subsequent months of treatment with bevacizumab and capecitabine. Administration and pharmacy costs for capecitabine alone were assumed to be £255.32 per month. The manufacturer stated that in clinical practice some patients stop treatment before disease progression, and therefore it is essential to consider the distinction between disease progression and treatment discontinuation when evaluating the real incremental cost. In order to account for this difference, patient data on treatment duration were used to produce 'time to off treatment' Kaplan-Meier curves that could be used to determine the proportion of patients still receiving bevacizumab and/or capecitabine each month.

3.22 Progression-free survival health state costs were based on Advanced breast cancer: diagnosis and treatment (NICE clinical guideline 81) 'package 1' with the addition of an outpatient consultation with an oncologist and a computed tomography (CT) scan assumed to occur every 3 months, and were estimated to be £263.55 per month. Progressed disease health state costs were based on NICE clinical guideline 81 'package 2' and estimated to be £804.00 per month. The same costs and utilities were assumed regardless of first-line treatment. Adverse events of grade 3 or 4 severity occurring in greater than 2% of patients were incorporated into the analysis. When clinical advice indicated that the usual response to the adverse event was discontinuation of treatment (for peripheral sensory neuropathy, hand-foot syndrome and proteinuria), it was assumed this had been accounted for elsewhere in the model and no additional costs were accrued. In addition, treatment of diarrhoea was considered to have negligible contribution to costs. Therefore only costs associated with deep vein thrombosis and hypertension were included in the model. All adverse events were assumed to occur in month 1 for both treatment arms and were therefore not discounted.

3.23 The manufacturer did not include terminal care costs in the model, stating that these would refer to costs in the last 2 weeks of life and would therefore have a minimal impact on the ICER irrespective of the regimen received. In addition, no second-line treatment costs were included in the model because it was assumed that the duration of second-line treatment would be the same for a patient receiving first-line bevacizumab plus capecitabine as for a patient receiving first-line capecitabine alone, and the second-line costs in each arm would cancel each other out.

3.24 The base-case results indicated incremental costs of £38,924 and incremental QALYs of 0.5034 for bevacizumab plus capecitabine compared with capecitabine alone. The cost per QALY gained was £77,318 for bevacizumab plus capecitabine compared with capecitabine alone. The manufacturer conducted deterministic sensitivity analyses for a range of parameters. The manufacturer stated that the cost-effectiveness results were most sensitive to the costs and utilities associated with progressed disease.

3.25 The manufacturer conducted a scenario analysis using utility values from Peasgood et al. but this had little impact on the ICER and did not result in it increasing above £79,991 per QALY gained. A second scenario analysis was conducted including different formulations of vinorelbine as the comparator. It was assumed that vinorelbine had an equivalent efficacy and safety profile to capecitabine, with different list prices and costs of administration. The ICER was £58,972 per QALY gained for bevacizumab plus capecitabine compared with oral vinorelbine, £76,198 per QALY gained compared with branded intravenous vinorelbine, and £80,260 per QALY gained compared with generic intravenous vinorelbine.

3.26 The manufacturer conducted a probabilistic sensitivity analysis and concluded that bevacizumab plus capecitabine compared with capecitabine alone had a 0% probability of being cost effective if the maximum acceptable ICER was £30,000 to £50,000 per QALY gained. In response to consultation, the manufacturer provided the ICER based on the probabilistic sensitivity analysis, which was £80,073 per QALY gained for bevacizumab plus capecitabine compared with capecitabine alone (mean incremental costs were £40,161 [95% CI 36,703 to 45,079], mean incremental QALYs were 0.502 [95% CI 0.33 to 0.66]).

3.27 The manufacturer acknowledged that its economic evaluation was only relevant to patients with similar characteristics to those randomised to the capecitabine cohort of the RIBBON-1 trial who had previously been treated with a taxane. The ERG requested additional cost-effectiveness data for the ITT population of the capecitabine cohort for clarification. However, the manufacturer stated that because the submitted analysis calculated an ICER of £77,318 per QALY gained for the subgroup previously treated with a taxane, analysis of the ITT population would result in a larger ICER and therefore would not be considered a cost-effective use of NHS resources.

3.28 The ERG had concerns about the population used in the manufacturer's economic model. The ERG highlighted that the manufacturer had based its economic modelling on the subgroup of patients who had previously been treated with a taxane, because the manufacturer considered this population to represent the population for whom capecitabine is licensed: patients with metastatic breast cancer after failure of taxanes and an anthracycline-containing chemotherapy regimen or for whom further anthracycline therapy is not indicated. The ERG agreed that most patients in this subgroup would probably have previously received an anthracycline in addition to a taxane. However, the ERG questioned whether their treatment would be considered to have failed because the RIBBON-1 trial excluded patients who had received an adjuvant taxane or anthracycline in the last 12 months. The ERG did not consider the subgroup of patients who had previously received a taxane to be the appropriate group of patients. The ERG considered the ITT population in the capecitabine cohort to be the appropriate population because it represents the population in the final scope issued by NICE and the population specified in the marketing authorisation for bevacizumab. In addition, the ERG identified that there appeared to be baseline differences between the subgroup of patients who had previously received a taxane and the ITT population. In particular, the ERG noted from differences in the mean and median age and Eastern Cooperative Oncology Group (ECOG) performance status that the population of patients who had previously received a taxane appeared to be younger and healthier. The ERG also noted that the differences in progression-free and overall survival between the bevacizumab plus capecitabine and capecitabine plus placebo arms appeared to be greater in the subgroup of patients previously treated with a taxane than in the ITT population, as well as being statistically significant. However, the ERG reiterated that because no statistical adjustments were made to control for multiple testing in all subgroups and of all outcomes, these findings may have occurred by chance, and must be interpreted with caution.

3.29 The ERG raised some concerns about the structure and design of the manufacturer's economic model. The ERG noted that the manufacturer adapted a model structure previously used in NICE appraisals of cancer drugs. However, the ERG raised concerns that although the model covered a period of 15 years, no further chemotherapy was considered within the model following disease progression after treatment with bevacizumab plus capecitabine or capecitabine alone. This could have led to substantial bias, because if progression-free survival differed between the arms, the discounted costs and benefits of subsequent treatments would also have differed. Further, if the proportion of patients able to receive subsequent lines of therapy differed between the arms then the costs and outcomes would also have been different.

3.30 The ERG was satisfied that the modelling approach used by the manufacturer to estimate progression-free survival from the RIBBON-1 trial using Kaplan-Meier methods for the first 12 months and assuming an exponential distribution thereafter was credible. The ERG noted that the approach was similar for progressed disease, however the manufacturer had 'uncrossed' the data using the rank preserving structural failure time model to minimise bias. The ERG stated that this approach was unsuitable when a large proportion of patients from both arms cross over. The ERG noted that 44.7% of patients in the bevacizumab plus capecitabine arm and 52.4% of patients in the capecitabine plus placebo arm received bevacizumab after disease progression. Further, patients in the modelled subgroup who previously received a taxane also received other therapies after progression. The ERG stated that given the limitations of the rank preserving structural failure time model and without any other estimate to adjust for crossover, they were unable to confirm the likely effect of the crossover and post-progression therapies on overall survival in this subgroup and caution should be exercised when interpreting the manufacturer's overall survival results.

3.31 The ERG undertook an analysis of the original progressed disease trial data (rather than the 'uncrossed' data) to explore survival during this phase. This analysis separated the bevacizumab plus capecitabine and capecitabine plus placebo arms according to whether patients had crossed over to a different treatment or not. A comparison of survival times during the progressed disease phase indicated that survival is similar in each group and overall, the 4 groups in the RIBBON-1 trial did not show strong evidence of heterogeneity. However, the capecitabine plus placebo group with no crossover appeared to differ when tested pairwise against the other 3 groups. Therefore, the ERG explored 2 different scenarios. The first scenario grouped all patients together and modelled a scenario in which survival after progression was equivalent irrespective of first-line therapy or crossover (common projection scenario). This method resulted in an ICER of £171,411 per QALY gained for bevacizumab plus capecitabine compared with capecitabine alone. The second scenario grouped together all the bevacizumab plus capecitabine patients and the capecitabine plus placebo patients who crossed over, and considered the capecitabine plus placebo patients who did not cross over separately (different projections scenario). This method resulted in an ICER of £92,060 per QALY gained for bevacizumab plus capecitabine compared with capecitabine alone. The ERG stated that the second scenario allowed a clear comparison between patients who did and did not receive bevacizumab during the trial and gives a representation of the effect of crossover. The ERG highlighted that each analysis portrayed an extreme, allowing consideration of a best and worst case scenario for the effect of crossover on post-progression survival.

3.32 The ERG conducted a sensitivity analysis to study the impact of including the licensed dose of capecitabine (1250 mg/m²) rather than the dose widely used in clinical practice (1000 mg/m²). It found that changing the dose of capecitabine to 1250 mg/m² results in an overall incremental increase in drug costs of £3782 and an accompanying increase of £7512 per QALY gained in the ICER estimate. The ERG re-estimated the costs of therapy based on the distribution of patient body weight and body surface area in a UK-specific cohort of patients rather than using a simple average based on trial data. The ERG found that this resulted in an increase in drug costs of £2966 per patient in the bevacizumab plus capecitabine arm and an increase of £50 per patient in the capecitabine alone arm. The adjustment resulted in a revised ICER that was £5793 higher per QALY gained than the manufacturer's base-case ICER. The ERG also added in the costs of terminal care during the last 2 weeks of life, as specified in the guideline on Advanced breast cancer: diagnosis and treatment (NICE clinical guideline 81), and these adjustments resulted in a revised ICER that was £105 lower per QALY gained than the manufacturer's base-case ICER.

3.33 The ERG noted that the utility values used in the manufacturer's model were estimated using the statistical model detailed in a study by Lloyd et al. The ERG noted that there is a lack of consensus among economists in relation to the most appropriate value for age in the Lloyd et al. model, that is, whether it should be the age of the population surveyed in the study or the age of the population taking part in the original health state valuation exercise carried out by Kind et al. (1999). The ERG noted that the manufacturer used 47 years, the mean age of the population taking part in the original Kind et al. study, with the advantage that it was consistent with standard UK EQ-5D tariff scores. However, the ERG stated that the lack of consensus relating to the most appropriate age to use introduces a degree of uncertainty to the utility values used in the model. The ERG also corrected for a typing mistake in the formula used in the manufacturer's model in the capecitabine alone arm and this resulted in a revised ICER that was £786 lower per QALY gained than the manufacturer's base-case ICER.

3.34 The combined impact of the ERG's revisions to the drug costs, terminal care costs and utility estimates in the manufacturer's base case resulted in an ICER of £82,162 per QALY gained for bevacizumab plus capecitabine compared with capecitabine alone. In addition, combining these revisions with the revised progressed disease estimates resulted in an ICER of £181,648 per QALY gained when using the common projection scenario and £97,963 per QALY gained when using the different projections scenario. The ERG also agreed with the manufacturer that the base-case ICER cannot be considered to be generalisable to the whole population covered by the marketing authorisation and that it was likely to be higher than the ICER for the modelled subgroup.

3.35 Full details of all the evidence are in the manufacturer's submission and the ERG report, which are available from http://guidance.nice.org.uk/TA263