Bevacizumab in combination with oxaliplatin and either fluorouracil plus folinic acid or capecitabine for the treatment of metastatic colorectal cancer

The manufacturer's approach to the decision problem compared bevacizumab plus FOLFOX (B-FOLFOX) or bevacizumab plus XELOX (B-XELOX) with FOLFOX and XELOX without bevacizumab as a first-line treatment. The manufacturer stated that the use of irinotecan in combination with folinic acid and fluorouracil (FOLFIRI) is decreasing and that it is mainly used in the small minority of patients in whom oxaliplatin is contraindicated or who cannot tolerate oxaliplatin. This was based on market research analysis, which indicated that combination chemotherapy regimens including oxaliplatin are the most commonly used in UK clinical practice. However, for completeness, the manufacturer performed an economic evaluation comparing B-FOLFOX or B-XELOX with FOLFIRI. The manufacturer stated that the cost effectiveness of bevacizumab as a second-line treatment could not be demonstrated.

The manufacturer undertook a systematic review of the literature and identified 2 randomised controlled trials: 1 assessed bevacizumab as a first-line therapy (NO16966) and 1 assessed bevacizumab as a second-line therapy (E3200). No evidence of bevacizumab used in lines of treatment beyond second-line therapy was provided by the manufacturer.

The NO16966 study started as a phase 3, multinational, 2-arm, randomised, open-label study. This study was originally designed to demonstrate the non-inferiority of XELOX compared with FOLFOX-4 (that is, the FOLFOX regimen given every 2 weeks, with 2 long infusions in the first 48 hours) in adult patients with histologically confirmed metastatic colorectal cancer who had not been treated before with chemotherapy. After randomisation of 634 patients to XELOX or FOLFOX-4, the original protocol design was amended to include a 2×2 factorial randomised study in which patients were subsequently randomised to either XELOX or FOLFOX plus either bevacizumab or placebo. A further 1401 patients were then recruited (blinded to the allocation of bevacizumab or placebo), and a final total of 2035 patients were randomised in the NO16966 study. The study amendment included an additional objective of demonstrating superiority of bevacizumab in combination with chemotherapy (B-XELOX or B-FOLFOX-4) over placebo in combination with chemotherapy (P-XELOX or P FOLFOX-4). The dosage of bevacizumab was 5 mg/kg every 2 weeks (B-FOLFOX-4) or 7.5 mg/kg every 3 weeks (B-XELOX). Treatment was planned to continue until disease progression, unacceptable toxicity, resection of metastatic disease or for 48 weeks, whichever came first (this being at the discretion of the investigator). Patients who completed the 48-week study treatment phase without progressive disease were eligible to enter the post-study treatment phase and continue their allocated treatment until their disease progressed. Patients whose disease became operable, and underwent resection, were eligible to enter the post-study treatment phase. The NO16966 protocol also allowed continuation of allocated treatment (bevacizumab or placebo) until disease progression (in line with the bevacizumab SPC, which was produced after the NO16966 study) if oxaliplatin treatment was terminated because of adverse events.

All patients in the study had an Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1. The study was stratified to ensure that study arms were balanced with regards to ECOG performance status (0 versus 1), number of organs with metastases at baseline (1 versus more than 1), alkaline phosphatase level at baseline (within normal range versus above normal range), liver as a site of metastasis (yes versus no) and geographic region. The median follow-up was 28 months. The manufacturer acknowledged that patients in the study were slightly younger and fitter than patients with metastatic colorectal cancer in the UK who are, on average, over 60 years of age.

The primary pooled analysis of non-inferiority of the NO16966 study (that is, pooling of all XELOX arms compared with pooling of all FOLFOX-4 arms) showed that the XELOX and FOLFOX-4 regimens were equivalent for overall and progression-free survival. The primary pooled analysis of superiority for the NO16966 study (that is, pooling of the initial 2-arm study and the 2×2 factorial part of the study) showed that the addition of bevacizumab to chemotherapy (B-XELOX and B-FOLFOX-4) significantly improved progression-free survival compared with chemotherapy alone (P XELOX and P-FOLFOX-4 and XELOX and FOLFOX-4 combined). The median progression-free survival was 7.7 months in the placebo plus chemotherapy group and 9.4 months in the bevacizumab plus chemotherapy group (intention-to-treat analysis, hazard ratio [HR] 0.79; 95% confidence interval [CI] 0.72 to 0.87, p=0.0001). The median overall survival was 18.9 months in the placebo plus chemotherapy group compared with 21.2 months in the bevacizumab plus chemotherapy group (HR 0.83; 95% CI 0.74 to 0.93, p=0.0019).

The manufacturer provided a secondary pooled analysis of superiority based only on the 2×2 factorial design (according to the original statistical plan, that is, B-XELOX and B-FOLFOX-4 combined compared with P-XELOX and P-FOLFOX-4 combined). The median progression-free survival was 8.0 months in the placebo plus chemotherapy group and 9.4 months in the bevacizumab plus chemotherapy group (HR 0.83; 95% CI 0.72 to 0.95, p=0.0023). The median overall survival was 19.9 months in the placebo plus chemotherapy group and 21.3 months in the bevacizumab plus chemotherapy group (HR 0.89; 95% CI 0.76 to 1.03, p=0.0769).

The manufacturer reported that the difference in progression-free survival was statistically significant for bevacizumab versus placebo in the XELOX subgroup (HR 0.80; 97.5% CI 0.66 to 0.96, p=0.0059) but not in the FOLFOX-4 subgroup (HR 0.89; 97.5% CI 0.74 to 1.06, p=0.1312). An exploratory analysis submitted by the manufacturer suggested that bevacizumab did not deliver a benefit to patients in the FOLFOX-4 group who had received prior adjuvant treatment (HR 1.75; 97.5% CI 1.15 to 2.65, p value not reported), whereas it did give a benefit to patients in the FOLFOX-4 group who had not had adjuvant therapy (HR 0.72; 97.5% CI 0.58 to 0.90, p value not reported). The manufacturer stated that this difference could be because the patients in the P-FOLFOX-4 group had a greater time between adjuvant treatment and development of metastatic disease than all of the other groups. This suggested that there was an imbalance in the data due to the better prognosis in the P-FOLFOX-4 group because the cancers in this group were growing more slowly than in the other groups. Additional exploratory analyses showed that when the data for all patients who had received adjuvant treatment were removed from all treatment groups (and thereby including the removal of the subgroup of patients that may have had slower tumour progression), the hazard ratios for overall survival and for progression-free survival ranged from 0.83 to 0.85 (all p values <0.03) and from 0.74 to 0.77 (all p values <0.0001).

The manufacturer also conducted a post-hoc subgroup analysis of the impact of bevacizumab treatment on liver resection rates. This analysis suggested that bevacizumab improved liver resection rates, although this was not statistically significant. No analyses of KRAS or other mutations were submitted.

The rates of discontinuation in the NO16966 study were higher in the bevacizumab plus chemotherapy groups than in the placebo plus chemotherapy groups. In the 2×2 factorial part of the study, 29% (203 out of 699) of patients receiving bevacizumab plus chemotherapy and 47% (329 out of 701) of patients receiving chemotherapy alone were treated until progression (despite the protocol allowing treatment to be continued until disease progression, in line with the SPC of bevacizumab).

In the NO16966 study, the most common adverse events of bevacizumab treatment were thromboembolic (7.8% venous thromboembolic events compared with 5.1% in the placebo plus chemotherapy groups and 1.7% arterial thromboembolic events compared with 0.9% in the placebo plus chemotherapy groups). Grade 3 or 4 hypertension, proteinuria and bleeding were more common in the bevacizumab groups than in the placebo plus chemotherapy groups (4% versus 0.8%, 3.5% versus 0.9% and 1.9% versus 1.5% respectively). The incidence of serious and life-threatening (grade 3 and 4) adverse events was higher in the bevacizumab plus chemotherapy groups (79.9%) than the placebo plus chemotherapy groups (74.8%) and higher with FOLFOX-4 regimens than XELOX regimens (although FOLFOX-4 was associated with different adverse events than XELOX). The manufacturer stated that most adverse events were associated with cytotoxic chemotherapy, and the higher incidence in the bevacizumab groups was likely to be a consequence of a longer duration of chemotherapy for patients receiving bevacizumab. No health-related quality-of-life data were collected in the NO16966 study.

The E3200 study was a phase 3, multicentre, 3-arm, randomised, open-label study. It compared B-FOLFOX-4 (n=293), FOLFOX-4 (n=292), and bevacizumab alone (n=244) in adult patients with advanced or metastatic colorectal cancer previously treated with a fluoropyrimidine and irinotecan, either separately or in combination. Following a data monitoring review 18 months after the start of the trial, the bevacizumab-alone arm was terminated because of poorer efficacy. Patients were stratified by ECOG performance status (0 versus 1 or more) and prior radiation therapy (yes versus no). The dosage of bevacizumab was 10 mg/kg every 2 weeks. Median overall survival was 10.8 months in the FOLFOX 4 group and 13 months in the B-FOLFOX-4 group (intention-to-treat analysis, HR 0.751; 95% CI 0.332 to 0.893, p=0.0012). Median progression-free survival increased from 4.5 months with FOLFOX-4 to 7.5 months with B-FOLFOX-4 (HR 0.518; 97.5% CI 0.416 to 0.646, p<0.0001).

The manufacturer produced a Markov model to estimate the cost effectiveness of bevacizumab plus FOLFOX and XELOX compared with FOLFOX and XELOX alone. The model had 4 distinct health states: first-line treatment, after first-line treatment without progression, progressed disease and death. It was assumed that all patients start in the first health state (first-line treatment) in accordance with the NO16966 study. The model had a cycle length of 1 month and the time horizon was 8 years (equivalent to life expectancy in the population of interest). A half-cycle correction was applied to the model. An NHS and personal social services (PSS) perspective was taken.

Costs were mainly derived from 2007 to 2008 national reference costs, BNF 57 and PSSRU (Personal Social Services Research Unit) 2008. FOLFOX-4 was used in the pivotal trial of bevacizumab but the manufacturer stated that FOLFOX-6 (that is, the FOLFOX regimen delivered over 2 weeks but with only 1 long infusion in the first 48 hours) is more commonly used in UK clinical practice. Therefore, the economic model was adjusted to include FOLFOX-6 by assuming similar efficacy to FOLFOX-4 but with reduced costs.

Treatment duration and dose intensity were based on the NO16966 study. Mean and median treatment durations (6 and 7 months respectively) were shorter in the NO16966 study than progression-free survival. Duration of treatment varied between treatment arms and was longer with the addition of bevacizumab and longer in the FOLFOX than in the XELOX arms. Treatment duration was estimated and applied in the model for each arm of the NO16966 study. However, for simplicity it was assumed that oxaliplatin treatment duration was the same as bevacizumab treatment duration in the B-FOLFOX and B-XELOX arms. It was also assumed that treatment was given continuously, as in the NO16966 study, rather than intermittently. Kaplan–Meier estimates from the NO16966 study were used for progression-free and overall survival up to a median survival of 28 months. After this point, a Weibull probability distribution for overall survival and an exponential probability distribution of progression-free survival (based on average hazard for months 13 to 28) were used to model the tails. For overall survival, treatment effect was assumed to continue after the median follow-up period and this was further explored in the sensitivity analysis.

The manufacturer used utility values from NICE's previous technology appraisal guidance on cetuximab for the first-line treatment of metastatic colorectal cancer (TA176; now replaced by NICE's technology appraisal guidance on cetuximab and panitumumab for previously untreated metastatic colorectal cancer). These utility values were taken from a randomised controlled trial comparing cetuximab plus FOLFIRI with FOLFIRI alone in first-line treatment of metastatic colorectal cancer and represented mean utility values from 42 patients using the EQ 5D questionnaire; however, only 37 patients fully completed the questionnaire. A utility value of 0.77 was assigned to the first-line treatment health state, the average of all the EQ-5D completed responses over the study period (this assumption was used in TA176). A utility value of 0.79 was assigned to the health state after first-line treatment (that is, without disease progression). This was based on expert opinion that patients in this state will experience a higher quality of life than patients receiving first-line treatment, because of fewer adverse events, and their utility value will be similar to a person aged 55 to 64 years in the UK general population. A utility value of 0.68 was assigned to the progressed disease state, taken from a trial of cetuximab for the third-line treatment of metastatic colorectal cancer and using the Health Utility Index questionnaire (this assumption was used in TA176).

The manufacturer's original submission included details of a proposed patient access scheme for the first-line use of bevacizumab. The scheme involved supplying bevacizumab at a fixed price per cycle of treatment (£800 for 2-weekly cycles and £1,200 for 3-weekly cycles), with bevacizumab being provided free after 12 months of treatment and with oxaliplatin being provided free of charge throughout. The manufacturer stated that it would take approximately 5 minutes per cycle for the pharmacist to update the scheme's registry system. This equated to £4 per cycle. The Department of Health stated that it has concerns around the complexity of the patient access scheme and believes that the administration costs of the scheme would probably be greater than those set out by the manufacturer.

The manufacturer stated that the incremental cost-effectiveness ratios (ICERs) most relevant to the decision problem were B XELOX compared with XELOX and B-FOLFOX-6 compared with FOLFOX-6. In response to a request for clarification from the ERG, the manufacturer provided a revised base-case analysis. The ERG noted that 19.7% (n=256) and 13.7% (n=96) of patients were alive after median follow-up in the XELOX/FOLFOX-6 and B XELOX/B-FOLFOX-6 arms respectively. The ERG asked the manufacturer to use untruncated data to calculate the estimates of the parameters of the Weibull distribution. The manufacturer used Kaplan–Meier estimates of survival up to month 6 and then a Weibull distribution fitted to untruncated data after month 6 for progression-free survival and a Weibull distribution fitted to untruncated data for overall survival. The model also accounted for oxaliplatin wastage (that is, no vial sharing was assumed). In the manufacturer's base-case analysis (that is, pooling of the initial 2-arm part of the study and the 2×2 factorial part of the study), B-XELOX produced an ICER of £35,912 per QALY gained when compared with XELOX (£84,553 per QALY gained without the patient access scheme), and B-FOLFOX-6 produced an ICER of £36,569 per QALY gained when compared with FOLFOX-6 (£92,634 per QALY gained without the patient access scheme). In the 1-way sensitivity analyses, which were only provided with the patient access scheme, the ICERs were not greatly influenced by variations in any of the parameters. In the probabilistic sensitivity analysis, the mean ICER for the comparison of B-FOLFOX-6 with FOLFOX-6 was £36,907 per QALY gained and for the comparison of B-XELOX with XELOX the mean ICER was £36,205 per QALY gained (95% intervals not provided).

The manufacturer stated that currently only a minority (12%) of patients in the UK receive FOLFIRI as a first-line treatment for metastatic colorectal cancer and that in most of these patients, treatment with oxaliplatin is contraindicated. For completeness, the manufacturer provided a cost-effectiveness analysis comparing bevacizumab in combination regimens containing oxaliplatin with FOLFIRI. The efficacy of FOLFIRI was derived from a mixed-treatment comparison. A constant hazard ratio was applied to the extrapolated progression-free survival and overall survival curves of FOLFOX-4 to derive the survival curves for FOLFIRI. The treatment duration and the drug administration and adverse event costs for FOLFIRI were assumed to be equivalent to those for FOLFOX-4. B-XELOX compared with FOLFIRI was associated with an ICER of £9,192 per QALY gained, B-FOLFOX-6 compared with FOLFIRI was associated with an ICER of £38,835 per QALY gained, and B FOLFOX-4 compared with FOLFIRI was associated with an ICER of £58,575 per QALY gained.

The manufacturer provided a cost-effectiveness analysis of bevacizumab plus regimens including oxaliplatin in the second-line setting. Drug acquisition, administration and pharmacy costs per cycle were taken from the first-line analysis and multiplied by the mean number of cycles reported in the E3200 study. Costs associated with adverse events, third-line treatment and central venous access devices were not included in the second-line analysis and no discounting was applied. As a second-line treatment, B-FOLFOX-4 compared with FOLFOX-4 resulted in an ICER of £102,644 per QALY gained. The manufacturer stated that the larger ICERs reported in the second-line setting were mainly because of the higher doses of bevacizumab used and that bevacizumab could not be considered cost effective for second-line treatment.

The ERG stated that the manufacturer's submission generally followed the NICE reference case. It also highlighted that adequate methods of randomisation and allocation concealment were reported in the NO16966 and E3200 studies. However, the ERG noted that the manufacturer focused on a comparison of oxaliplatin chemotherapy regimens with or without bevacizumab as first-line treatment. This differed from the scope, which included irinotecan chemotherapy regimens without bevacizumab as comparators and bevacizumab plus oxaliplatin regimens as second-line treatment.

The ERG expressed concerns about pooling data from the initial 2-arm part and the 2×2 factorial part of the NO16966 study. It stated that this was inappropriate because of the different designs of the 2 parts of the study. The European Medicines Agency (EMA) in their assessment report for bevacizumab also expressed concerns about the appropriateness of this method of pooling and it noted that this pooled analysis was specified in the protocol in case of borderline significance in progression-free survival in the 2×2 study. The EMA therefore questioned the validity of the results derived from the pooled analysis. The ERG also noted that the number of patients reported as Caucasian and the number of patients with ECOG performance status of 0 were 10% greater in the 2×2 factorial part of the study than the initial 2-arm part and that both are associated with better prognosis. The ERG further highlighted that there was a difference in terms of the overall survival benefit associated with bevacizumab in the primary pooled analysis (overall survival significantly improved) and in the pooled analysis based on the 2×2 factorial design only (overall survival not significantly improved). The ERG suggested that this difference might be because of the imbalance of patients in the 2×2 part of the study who had a slower rate of disease progression (that is, patients whose disease took longer to relapse after prior adjuvant therapy) and the lack of statistical power to assess overall survival.

Additional analyses were provided by the manufacturer using only the 2×2 factorial design. For the comparison of B-XELOX with XELOX, using the 2×2 factorial part of the NO16966 study only (that is, the efficacies of B-XELOX and B-FOLFOX-4 combined compared with P-XELOX and P-FOLFOX-4 combined), B-XELOX produced an ICER of £48,111 per QALY gained (£129,911 per QALY gained without the patient access scheme). Removing the patients who had received prior adjuvant therapy reduced the ICER to £36,006 per QALY gained (£92,698 per QALY gained without the patient access scheme). If the XELOX and FOLFOX arms were not pooled, then the ICER was £35,662 per QALY gained (£90,779 per QALY gained without the patient access scheme). For the comparison of B-FOLFOX-6 with FOLFOX-6, using the 2×2 factorial part of the NO16966 study (that is, the efficacies of B XELOX and B-FOLFOX-4 combined compared with P-XELOX and P-FOLFOX-4 combined), B-FOLFOX-6 produced an ICER of £39,771 per QALY gained (£134,309 per QALY gained without the patient access scheme). Removing patients who had received prior adjuvant therapy was associated with an ICER of £31,174 per QALY gained (£96,687 per QALY gained without the patient access scheme). When the XELOX and the FOLFOX arms were not pooled, then the ICER was £62,714 per QALY gained (£240,324 per QALY gained without the patient access scheme). The manufacturer did not provide any analysis using the 2×2 factorial part of the study with patients who had received prior adjuvant therapy excluded and the XELOX and FOLFOX arms not pooled.

The ERG reviewed the additional analyses submitted by the manufacturer (as outlined in section 3.22) and suggested that the most appropriate analysis was one using the 2×2 factorial design of the NO16966 study with XELOX and FOLFOX not pooled and patients who had received prior adjuvant therapy excluded. However, this analysis was not provided by the manufacturer despite requests by the ERG to do so. The ERG therefore suggested that the next most appropriate analysis was the one using data from the 2×2 factorial design of the NO16966 study with the XELOX and FOLFOX arms pooled and patients who had received prior adjuvant therapy excluded. This analysis gave Kaplan–Meier estimates up to month 6 and then the Weibull distribution was used for extrapolating progression-free survival, and Kaplan–Meier estimates up to month 28 with the Weibull distribution then used for extrapolating overall survival. With this analysis the ERG produced an ICER for B-XELOX of £36,354 per QALY gained compared with XELOX and £31,452 per QALY gained for B-FOLFOX-6 compared with FOLFOX-6 (ICERs without the patient access scheme were not provided). The results of the 1-way sensitivity analyses showed that the ICERs were not greatly influenced by any of the parameter changes.

The ERG noted that the duration of chemotherapy treatment was relatively short (median approximately 6 months) despite the protocol allowing treatment until disease progression or unacceptable adverse events. The protocol also allowed bevacizumab/placebo treatment to continue until disease progression or unacceptable toxicity (as in the bevacizumab SPC) but the duration of therapy with bevacizumab was also relatively short (median 6.5 months). Although the ERG agreed that the manufacturer's economic model was an accurate replication of the NO16966 study, the ERG suggested that in clinical practice treatment with drugs other than oxaliplatin (fluorouracil, capecitabine, bevacizumab) might continue after oxaliplatin regimens stopped. The ERG conducted an exploratory analysis (using the 2×2 factorial design of the NO16966 study with the XELOX and FOLFOX arms pooled and patients who had received prior adjuvant therapy excluded) that examined the impact on ICERs of stopping oxaliplatin 1 month before the other treatment components. Under this scenario, costs in the XELOX and FOLFOX-6 arms were reduced. In the B-XELOX and B-FOLFOX arms, the cost of oxaliplatin remained the same because oxaliplatin is free for these groups. It was also assumed that no change in incremental survival occurred. In this analysis, the ICERs were increased from £36,354 to £43,511 per QALY gained when B XELOX was compared with XELOX and from £31,452 to £39,478 per QALY gained when B-FOLFOX-6 was compared with FOLFOX-6 (ICERs without the patient access scheme were not provided). The ERG also examined the impact of increasing the duration of bevacizumab treatment by 1 month on the ICERs. It was assumed that no change in survival occurred and the treatment duration of the other components remained the same. In this analysis, the ICERs increased from £36,354 to £47,312 per QALY gained when B-XELOX was compared with XELOX and from £31,452 to £41,692 per QALY gained when B-FOLFOX-6 was compared with FOLFOX-6 (ICERs without the patient access scheme were not provided).

The ERG stated that it was not possible to adequately check the sources considered for determining the utility values because the references were incomplete. The ERG suggested that the utility values from NICE's previous technology appraisal guidance on cetuximab (TA176) could be relevant to patients receiving bevacizumab. However, the ERG also commented that the assumption that the utility value of the health state after first-line treatment (that is, off treatment but not yet progressed) is similar to that of people aged 55 to 64 years in the UK general population is unrealistic. This is because after 6 months of chemotherapy, people are often less mentally and physically fit than those of the same age in the general population. In addition, the ERG noted that the utility value for the health state first-line treatment (that is, 0.77) might be an overestimate. This is because the utility value in the UK general population of the same age group is 0.79. The ERG further noted that the model did not take into account the fact that XELOX regimens might be associated with higher health-related quality of life than FOLFOX regimens because the former are considered more convenient. The ERG performed an exploratory analysis that investigated the impact of decreasing the utility values by 20%. This decrease in utility values had a large impact on the ICERs. Using the 2×2 factorial part of the study with patients with prior adjuvant therapy excluded, the ICER for B-XELOX increased from £36,354 to £45,433 per QALY gained when compared with XELOX and the ICER for B-FOLFOX-6 increased from £31,452 to £39,315 per QALY gained when compared with FOLFOX-6 (ICERs without the patient access scheme were not provided).

In response to consultation, the manufacturer provided revised cost-effectiveness estimates. These were based on the ICERs calculated by the ERG of £36,354 per QALY gained for B-XELOX compared with XELOX and £31,452 per QALY gained for B FOLFOX-6 compared with FOLFOX-6 as detailed in section 3.23 (ICERs without the patient access scheme were not provided). The manufacturer revised the time of operating the patient access scheme, based on the number of patients expected to enrol within the first 3 years of the scheme, to 131 minutes and 152 minutes per patient for the XELOX and FOLFOX regimens respectively, based on research within the NHS. This equated to an average cost per patient over years 1 to 3 of £57 and £67 for B-XELOX and B FOLFOX respectively. This increased the ICER for B-XELOX compared with XELOX by £164 per QALY gained and the ICER for B-FOLFOX-6 compared with FOLFOX-6 by £113 per QALY gained. The manufacturer also used a utility value of 0.77 in the health state after first-line treatment (that is, without disease progression) as opposed to a value of 0.79. This increased the ICER for B XELOX compared with XELOX by £647 per QALY gained and the ICER for B-FOLFOX-6 compared with FOLFOX-6 by £560 per QALY gained. The manufacturer also conducted a time-and-motion study of the preparation and administration of bevacizumab infusions at a private hospital. The manufacturer stated that preparation time was divided between the pharmacist and the pharmacy technician. This resulted in a reduction in the bevacizumab administration costs of £42 (as used in the original submission) to £31 per infusion. The ICER for B-XELOX compared with XELOX was reduced by £677 per QALY gained and the ICER for B-FOLFOX-6 compared with FOLFOX-6 was reduced by £1,012 per QALY gained. The cumulative effect of these changes increased the ICER for B-XELOX compared with XELOX from £36,354 to £36,494 per QALY gained and decreased the ICER for B-FOLFOX-6 compared with FOLFOX-6 from £31,452 to £31,122 per QALY gained.

The manufacturer also confirmed that the patient access scheme would apply to bevacizumab given intermittently; that is, bevacizumab would be provided free of charge after 12 months of cumulative treatment had been given. The manufacturer did not state whether this would affect the ICERs. The manufacturer did not provide ICERs from the 2×2 factorial part of the study with patients who had received prior adjuvant therapy excluded and with the XELOX and FOLFOX arms not pooled.

The manufacturer submitted an amended patient access scheme that included all the elements of the original scheme and an additional upfront payment (designated by the manufacturer to be commercial in confidence) to the NHS for each person starting first-line treatment with bevacizumab. When the revised patient access scheme was included, the ICER for B-XELOX compared with XELOX was reduced from £36,494 to £29,975 per QALY gained and the ICER for B-FOLFOX-6 compared with FOLFOX-6 was reduced from £31,122 to £24,604 per QALY gained. Without the patient access scheme, the ICER for B-XELOX compared with XELOX was £104,870 per QALY gained and the ICER for B FOLFOX-6 compared with FOLFOX-6 was £108,267 per QALY gained.

The manufacturer provided an exploration of the effect of the individual components of the revised patient access scheme on the ICERs for B-XELOX compared with XELOX and for B-FOLFOX-6 compared with FOLFOX-6. Providing bevacizumab free after 12 months slightly reduced the ICERs. Providing oxaliplatin free of charge when given with bevacizumab led to a substantial reduction of the ICERs and was the major driver of the impact of the patient access scheme on cost effectiveness. Fixing the price of bevacizumab to £800 per 2-weekly cycle and £1,200 per 3-weekly cycle and the upfront payment resulted in further but less marked reductions in the ICERs. The precise details of the effect of the individual components of the revised patient access scheme were designated by the manufacturer to be commercial in confidence.

The ERG commented on the revised ICERs and patient access scheme submitted by the manufacturer. The ERG noted that the time-and-motion study conducted by the manufacturer to ascertain the administration costs of bevacizumab was based on information from 1 small private hospital and may not fully reflect the true costs to the NHS. The ERG conducted the same analyses as the manufacturer and noted slight differences in the resulting ICERs. The ICER for B-XELOX compared with XELOX was £29,956 per QALY gained and the ICER for B-FOLFOX-6 compared with FOLFOX-6 was £24,577 per QALY gained when the revised patient access scheme was incorporated. The ERG stated that the reasons for the differences between their calculations and those of the manufacturer were unclear but recognised that the differences were small. The ERG further noted that when the manufacturer used higher operating costs of the patient access scheme then the ICERs were slightly increased. The ERG conducted the same analysis as the manufacturer, but it set the cost per patient of operating the patient access scheme to £100. Under this scenario analysis the ICERs were slightly increased; £30,684 per QALY gained when B-XELOX was compared with XELOX and £25,312 per QALY gained when B FOLFOX-6 was compared with FOLFOX-6.

The ERG performed exploratory analyses incorporating discounts on the list price of oxaliplatin. The ICER with the patient access scheme for B XELOX compared with XELOX increased from £29,975 per QALY gained to £68,140 per QALY gained assuming a 90% discount and £70,260 per QALY gained assuming a 95% discount. The ICER with the patient access scheme for B-FOLFOX-6 compared with FOLFOX-6 increased from £24,604 per QALY gained to £70,470 per QALY gained assuming a 90% discount and £73,018 per QALY gained assuming a 95% discount. The ICERs without the patient access scheme applied were reduced slightly (by approximately 5%) when 90% and 95% discounts on the price of oxaliplatin were incorporated to all treatment arms in the model.

Full details of all the evidence are in the manufacturer's submission and the ERG report.