Advertisement

The Cost-Effectiveness of Bevacizumab in Advanced Ovarian Cancer Using Evidence from the ICON7 Trial

Open ArchivePublished:March 24, 2016DOI:https://doi.org/10.1016/j.jval.2016.01.013

      Abstract

      Background

      Bevacizumab is used extensively in the treatment of cancer, including advanced ovarian cancer, for which results of the International Collaborative Ovarian Neoplasm (ICON) 7 trial have been recently reported. The National Institute for Health and Care Excellence’s (NICE’s) recent decision not to recommend bevacizumab for advanced ovarian cancer was not based on evidence related to the unlicensed lower dosage (7.5 mg/kg) of the drug despite its use in the English National Health Service (NHS) and the ICON7 trial.

      Objective

      To report on the findings of an analysis that considered whether the lower dose is cost-effective.

      Methods

      Cost-effectiveness analysis is assessed from the perspective of the English NHS and health outcomes expressed in terms of quality-adjusted life-years (QALYs). The analysis focuses on a clinically predefined high-risk subgroup of the ICON7 trial. The price at which the lower dose of bevacizumab could be considered cost-effective for the English NHS is presented for a range of scenarios to inform decisions about price negotiations by international health systems.

      Results

      In the base-case analysis, bevacizumab has an incremental cost-effectiveness ratio of £48,975 per additional QALY, which is above NICE’s standard cost-effectiveness threshold (£20,000–£30,000 per QALY). The official price of bevacizumab in 2013 was between £2.31 and £2.63 per milligram. A price reduction of between 46% and 67%, dependent on the NICE threshold, would be required for the product to be cost-effective in the high-risk subgroup.

      Conclusions

      The lower dose of bevacizumab for advanced ovarian cancer is not cost-effective based on the product’s list price and using NICE’s cost-effectiveness thresholds. Significant price discounts would be needed to make the drug affordable to the NHS.

      Keywords

      Introduction

      In the United Kingdom, ovarian cancer represents the fifth most prevalent gynecological cancer, with 7116 new cases diagnosed in 2011 [

      Cancer Research UK. Ovarian cancer incidence statistics. Available from: 〈http://www.cancerresearchuk.org/cancer-info/cancerstats/types/ovary/incidence/uk-ovarian-cancer-incidence-statistics〉. [Accessed July 8, 2013].

      ]. Until 2012 the standard treatment for ovarian cancer in the National Health Service (NHS) had remained relatively unchanged over the previous decade [
      • Ozols R.
      Chemotherapy for ovarian cancer.
      ], with debulking surgery where possible followed by carboplatin chemotherapy usually given in combination with paclitaxel [
      National Institute for Health and Care Excellence
      CG 122: The Recognition and Initial Management of Ovarian Cancer.
      ,
      National Institute for Health and Care Excellence
      TA 55: Review of the Clinical Effectiveness and Cost Effectiveness of Paclitaxel for Ovarian Cancer.
      ]. Bevacizumab, as an angiogenesis inhibitor, represents an adjunctive therapy to standard chemotherapy.
      Bevacizumab currently holds licenses from the European Medicines Agency in a range of oncological indications, including colorectal, lung, and breast, in addition to the license held for ovarian cancer. Its license in ovarian cancer is limited to advanced disease, at a recommended dose of 15 mg/kg. The National Institute for Health and Care Excellence (NICE) has not recommended bevacizumab in any of the cancers it has assessed [
      National Institute for Health and Care Excellence
      TA 212: Bevacizumab in Combination with Oxaliplatin and Either Fluorouracil Plus Folinic Acid or Capecitabine for the Treatment of Metastatic Colorectal Cancer.
      ,
      National Institute for Health and Care Excellence
      TA 148: Bevacizumab for the Treatment of Non-Small-Cell Lung Cancer.
      ,
      National Institute for Health and Care Excellence
      TA 214: Bevacizumab in Combination with a Taxane for the First-Line Treatment of Metastatic Breast Cancer.
      ], with its use in advanced ovarian cancer bring rejected on the grounds of poor value for money to the NHS [
      National Institute for Health and Care Excellence
      TA 284: Bevacizumab in Combination with Paclitaxel and Carboplatin for First-Line Treatment of Advanced Ovarian Cancer.
      ]. Although NICE’s recommendation was limited to the higher dose, interim trial results were reported during the technology assessment suggesting that it was not cost-effective at the lower dose either. Bevacizumab is, however, currently funded by the NHS at the lower dose (7.5 mg/kg) through the Cancer Drugs Fund. A recent update of the fund’s list of approved treatments confirmed that bevacizumab will continue to be funded in the short-term [
      NHS England
      National Cancer Drugs Fund List Ver3.0.
      ]. However, as the Cancer Drugs Fund reaches the end of its funding period (March 2016), understanding the cost-effectiveness of treatments that it currently funds will become important to guide decisions about resourcing in the future.
      A recent publication has reported the mature clinical results of the International Collaborative Ovarian Neoplasm (ICON) 7 trial [
      • Oza A.
      • Cook A.D.
      • Pfisterer J.
      • et al.
      Standard chemotherapy with or without bevacizumab for women with newly diagnosed ovarian cancer (ICON7): overall survival results of a phase 3 randomised trial.
      ], a two-arm, multicenter randomized controlled trial of bevacizumab in ovarian cancer, which followed 1528 patients, recruited between 2006 and 2009, with high-risk early- or advanced-stage epithelial ovarian carcinoma, for up to 5 years [
      • Perren T.
      • Swart A.M.
      • Pfisterer J.
      • et al.
      ICON7 Investigators A phase 3 trial of bevacizumab in ovarian cancer.
      ]. This analysis found modest gains from bevacizumab in progression-free survival (PFS) in the full trial population (a restricted mean difference of 1.5 months between the arms; P = 0.004), with no clinically or statistically significant difference in overall survival. In contrast, evidence was found for differences in both PFS (a restricted mean difference of 3.5 months; P < 0.001) and overall survival (a restricted mean difference of 4.8 months; P = 0.03) between the two treatment arms in a predefined high-risk subgroup.
      This analysis reports the findings of a cost-effectiveness analysis of the predefined high-risk subgroup in the ICON7 trial, and as such represents the first cost-effectiveness analysis of mature data from a trial of bevacizumab for ovarian cancer at the lower dose. The analysis focuses on the high-risk subgroup, as predefined in the trial analysis, due to the lack of significant differences in PFS or overall survival between the trial arms in the full trial population.

      Methods

      Overview

      Cost-effectiveness analysis compares the benefits (generally gains in health) offered by a given intervention with the benefits other patients must forgo as a result of services that are displaced when additional costs are imposed on a health care system with a constrained budget, such as the NHS. This is achieved by considering the incremental health gains and costs associated with a new intervention, as represented by the incremental cost-effectiveness ratio (ICER), against a cost-effectiveness threshold representing forgone health associated with the additional cost.
      This evaluation is consistent with NICE methods guidelines [
      National Institute for Health and Care Excellence
      Guide to the Methods of Technology Appraisal.
      ], considering costs from the perspective of the NHS and Personal Social Services, expressed in UK pounds sterling at 2013 prices. Health outcomes are expressed in terms of quality-adjusted life-years (QALYs), with both costs and outcomes discounted at 3.5% per annum. A cycle length of a week was used to account for the rapid development of advanced ovarian cancer.

      Patient Population

      The ICON7 trial recruited patients who had recently undergone debulking surgery for previously untreated advanced ovarian cancer, or who had early-stage disease deemed to be at high risk of progression. This analysis focuses on the predefined high-risk subgroup represented by those with Federation of Gynecology and Obstetrics (FIGO) stage IV disease or FIGO stage III disease and more than 1.0 cm of residual disease after debulking surgery, plus those who were deemed ineligible for surgery due to the extensive growth of their cancer [
      • Oza A.
      • Cook A.D.
      • Pfisterer J.
      • et al.
      Standard chemotherapy with or without bevacizumab for women with newly diagnosed ovarian cancer (ICON7): overall survival results of a phase 3 randomised trial.
      ]. Of 1528 women included in the ICON7 trial, 502 were categorized into the high-risk subgroup. Further details of patient characteristics have been published elsewhere [
      • Oza A.
      • Cook A.D.
      • Pfisterer J.
      • et al.
      Standard chemotherapy with or without bevacizumab for women with newly diagnosed ovarian cancer (ICON7): overall survival results of a phase 3 randomised trial.
      ].
      Current first-line chemotherapy for advanced ovarian cancer in the NHS is done with the combination of carboplatin and paclitaxel [

      NHSChoices. Ovarian Cancer. Available from: 〈http://www.nhs.uk/Conditions/Cancer-of-the-ovary〉. [Accessed July 8, 2013].

      ]. This is consistent with the design of the ICON7 trial, which compared carboplatin (Area Under the Curve (AUC) 5 or 6) and paclitaxel (175 mg/m2 of body surface area) 3 weekly for 6 cycles, with the same regimen plus bevacizumab (7.5 mg/kg of body weight) given concurrently every 3 weeks and continued for 12 additional cycles or until disease progression if earlier.

      Analytical Methods

      Clinical analysis of the ICON7 trial has found that in the high-risk subgroup, there remains a small difference in the rate of overall survival at 5 years in the two treatment arms despite the convergence of the PFS curves [
      • Oza A.
      • Cook A.D.
      • Pfisterer J.
      • et al.
      Standard chemotherapy with or without bevacizumab for women with newly diagnosed ovarian cancer (ICON7): overall survival results of a phase 3 randomised trial.
      ]. This persistent difference can be seen in Figure 1, which presents Kaplan-Meier curves for the two arms of the trial for both PFS and overall survival, alongside the extrapolations discussed in the next section. Given the difference in overall survival at 5 years, an estimate of differences in mean quality-adjusted life expectancy requires a longer time horizon, based on extrapolation of survival estimates beyond the trial evidence. Such extrapolation is also needed to estimate mean differences in lifetime costs. A partitioned survival model [
      • Billingham L.J.
      • Abrams K.R.
      • Jones D.R.
      Methods for the analysis of quality-of-life and survival data in health technology assessment.
      ,
      • Glasziou P.P.
      • Simes R.J.
      • Gelber R.D.
      Quality adjusted survival analysis.
      ] is adopted for this purpose, which considers a three-state model of disease (preprogression, postprogression, and death). The areas under the PFS and overall survival curves shown in Figure 1 are used to estimate mean time in the preprogression state and mean overall survival, respectively. The difference between these two mean durations is the mean time in the postprogression state.
      Figure thumbnail gr1
      Fig. 1Kaplan-Meier and extrapolated survival curves used to inform partition survival analysis. OS, overall survival; PFS, progression-free survival. (Color version of figure available online).

      Evidence Sources

      The clinical outcomes of progression and mortality that inform the analysis are drawn directly from the ICON7 trial, and have been reported elsewhere [
      • Oza A.
      • Cook A.D.
      • Pfisterer J.
      • et al.
      Standard chemotherapy with or without bevacizumab for women with newly diagnosed ovarian cancer (ICON7): overall survival results of a phase 3 randomised trial.
      ]. The partition survival model applied to the high-risk subgroup is constructed around the PFS and overall survival Kaplan-Meier curves estimated from the trial, and presented in Figure 1. Given the need for a lifetime time horizon and because of censoring in the trial, there is a need to estimate the rate of mortality and progression in patients beyond trial follow-up. In the base-case analysis, overall survival is extrapolated beyond the trial period by assuming that the rate of mortality in those patients surviving at 5 years is assumed to be the same as that found in a large observational study of epithelial ovarian cancer [
      • Baldwin L.A.
      • Huang B.
      • Miller R.W.
      • et al.
      Ten-year relative survival for epithelial ovarian cancer.
      ]. The rate of progression in those women who were progression free at 5 years is assumed to be the same as in patients subject to long-term follow-up in the ICON3 trial [
      • Ozols R.F.
      • Markman M.
      • Thigpen J.T.
      ICON3 and chemotherapy for ovarian cancer.
      ]. The analysis assumes that patients receive treatment at the age of 60 years, the median age of high-risk patients in the ICON7 trial. All extrapolation assumptions are tested using a range of alternative scenarios. Further details of these extrapolation methods can be found in Appendix 1 in Supplemental Materials found at doi:10.1016/j.jval.2016.01.013.
      Health-related quality of life (HRQOL) is based on the three-level EuroQol five-dimensional questionnaire completed by patients at multiple time points throughout the ICON7 trial. A maximum of 19 responses were available for each patient, up to 18 over time in patients who were progression free plus 1 at the point of progression of disease. Patients’ responses are valued using the UK scoring algorithm to provide a single HRQOL score per-patient response, which is bound by −0.594 and 1, where 0 and 1 represent death and good health, respectively [
      • Kind P.
      • Hardman G.
      • Macran S.
      UK population norms for EQ-5D. Working Papers, Centre for Health Economics.
      ].
      Mean HRQOL before patients have progressed was estimated for each of the 18 time points collected in the trial to allow for a possible changing impact of adverse events over time and potential accommodation. Data on postprogression HRQOL were collected in the trial only at the time of diagnosis of progression and at 3 years after baseline. The mean HRQOL at these time points was calculated to inform the postprogression HRQOL in the model.
      The base-case analysis assumed that the HRQOL of the patient population reflects disease stage and time since randomization only and is independent of initial treatment. This assumption of treatment independence was tested in a scenario analysis. The HRQOL of women in the general population, stratified by age (pooled quality-of-life estimates provided by Dr. Anju Keetharuth, University of Sheffield), was used as an estimate of the longer-term HRQOL of those women whose disease has not progressed at 5 years.
      A complete set of resource use data on all patients was collected in the ICON7 trial, and this was used to estimate the costs associated with the two treatment arms. The following cost categories were considered: trial drugs, other treatments, clinical investigations, and laboratory tests. An average cost of treatment was estimated for the two arms and assumed to be incurred within the first year since treatment initiation. Mean daily costs during the progression-free stage of the high-risk model were estimated in each of these categories and for each treatment arm over three time periods (baseline to 1 year, 1–3 years, and 3–5 years). As with HRQOL, nontreatment cost for patients who have progressed is an average of the observed costs from the trial. Further details of these costs can be found in Appendix 2 in Supplemental Materials found at doi:10.1016/j.jval.2016.01.013. The list price of bevacizumab ranges from £2.31 to £2.63 per milligram depending on the size of the vial. A range of possible prices, reflecting alternative vial sizes and combinations, was used assuming minimal drug wastage [
      Joint Formulary Committee. British National Formulary (online). London: BMJ Group and Pharmaceutical Press, 2015.
      ].

      Statistical Methods

      Missing data in clinical trials can result in bias and inefficient use of the data if handled incorrectly [
      • O’Neill R.T.
      • Temple R.
      The prevention and treatment of missing data in clinical trials: an FDA perspective on the importance of dealing with it.
      ,
      • Faria R.
      • Gomes M.
      • Epstein D.
      • White I.R.
      A guide to handling missing data in cost-effectiveness analysis conducted within randomised controlled trials.
      ], necessitating its quantification and appropriate analysis in estimating costs and HRQOL. The analysis uses multiple imputation with chained equations [
      • Faria R.
      • Gomes M.
      • Epstein D.
      • White I.R.
      A guide to handling missing data in cost-effectiveness analysis conducted within randomised controlled trials.
      ], which assumes data are missing at random. Further details are available in Appendix 3 in Supplemental Materials found at doi:10.1016/j.jval.2016.01.013.
      Uncertainty is considered using three approaches. First, a probabilistic sensitivity analysis is conducted, in which distributions representing uncertainty in each parameter are repeatedly resampled (3000 times in this analysis). An average is taken across these simulations to generate an expected ICER; the simulations also provide a means of expressing the parameter uncertainty cost-effectiveness, which is presented as the probability each treatment is cost-effective based on alternative cost-effectiveness thresholds, here based on NICE range of £20,000 to £30,000 per QALY gained [
      National Institute for Health and Care Excellence
      Guide to the Methods of Technology Appraisal.
      ]. Appendix 4 in Supplemental Materials found at doi:10.1016/j.jval.2016.01.013 provides more details of the uncertainty analysis. Second, an analyses of the price per milligram at which bevacizumab would have to be set for it to be cost-effective is presented. This analysis was conducted to inform decisions about the magnitude of discount that would be necessary to result in a cost-effective outcome. Finally, a range of scenario analyses is conducted. Table 1 details the seven scenarios considered, alongside the assumptions made in the base-case analysis. These scenarios are used to test the impact of key structural uncertainty, where this might not be sufficiently captured by the probabilistic sensitivity analysis approach. The scenarios relate to the extrapolation of survival curves (extrapolation scenarios 1 and 2), the treatment independence of HRQOL (HRQOL scenario 1), assumptions around long-term mortality (long-term mortality scenarios 1 and 2), and the impact of vial-sharing assumptions (vial-sharing scenarios 1 and 2). Probabilistic sensitivity analysis and price threshold analyses are conducted for all the scenarios.
      Table 1Scenarios in the high-risk population model
      ScenarioWithin trial phase (0–5 y)Extrapolation phase (>5 y)
      CostsHRQOLPFSOSCostsHRQOLPFSOS
      Base caseTreatment and state-specific mean costs as observed in the trial. Vial sharing is assumed for chemotherapy treatmentTreatment- independent mean EQ-5D observed in the trial, for states without and with progressionTreatment-specific KM function estimated from the ICON7 trialTreatment-specific KM function estimated from the ICON7 trialNo difference in cost per day for without progression; trial-based treatment- dependent cost per day for progressionTrial-based treatment- independent mean EQ-5D, for progression. HRQOL of general population for without progressionRate of progression (given progression-free at 5 y) from the ICON3 trial
      • Ozols R.F.
      • Markman M.
      • Thigpen J.T.
      ICON3 and chemotherapy for ovarian cancer.
      . Treatment independent
      Relative mortality (given survival at 5 y) from Baldwin et al.
      • Baldwin L.A.
      • Huang B.
      • Miller R.W.
      • et al.
      Ten-year relative survival for epithelial ovarian cancer.
      . Treatment independent
      Extrapolation scenario 1, no extrapolationAs base caseAs base caseAs base caseAs base caseNo extrapolation of model
      Extrapolation scenario 2, long-term treatment independenceAs base caseAs base caseAs base caseAs base caseTreatment independentAs base caseAs base caseAs base case
      HRQOL scenario 1, treatment- dependent adjusted HRQOLAs base caseHRQOL in all states assumed to be treatment dependent, adjusted for baseline scoreAs base caseAs base caseAs base caseAs base caseAs base caseAs base case
      Long-term mortality scenario 1, patients “cured” after 10 yAs base caseAs base caseAs base caseAs base caseAs base caseAs base caseAs base case10 y after randomization progression- free patients are assumed “cured”
      Long-term mortality scenario 2, patients “cured” after 5 yAs base caseAs base caseAs base caseAs base caseAs base caseAs base caseAs base case5 y after randomization progression- free patients are assumed “cured”
      Vial-sharing scenario 1, vial sharing for all treatmentsThe cost of treatment assumes no wasted medicationAs base caseAs base caseAs base caseAs base caseAs base caseAs base caseAs base case
      Vial-sharing scenario 2, no vial sharing for either treatmentThe cost of treatment assumes the use of a minimum number of whole vials of medicationAs base caseAs base caseAs base caseAs base caseAs base caseAs base caseAs base case
      EQ-5D, EuroQol five-dimensional questionnaire; HRQOL, health-related quality of life; ICON, International Collaborative Ovarian Neoplasm; KM, Kaplan-Meier; OS, overall survival; PFS, progression-free survival.

      Results

      Costs

      Table 2 presents the estimated mean costs (and standard errors [SEs]) per day from the ICON7 trial used in the model. The costs per day in the chemotherapy-only arm were found to be higher in the first year preprogression and in the postprogression periods, but much lower in the later preprogression periods. After the first year preprogression, the mean cost per day (excluding trial drugs) was greater in the bevacizumab arm.
      Table 2Mean costs per day estimated for the high-risk subgroup model
      Model stateMean cost (£) (SE)
      Chemotherapy aloneChemotherapy plus bevacizumab
      Preprogression
      0–1 y15.11 (1.67)12.98 (2.06)
      1–2 y3.28 (1.18)7.51 (2.37)
      2–5 y1.25 (1.06)5.81 (3.09)
      Postprogression3.00 (0.56)2.40 (0.65)
      Trial drugs20.19 (0.66)72.67 (1.64)
      SE, standard error.

      Health-Related Quality of Life

      Figure 2 shows the HRQOL scores over time used in the progression-free state of the base-case analysis, in which HRQOL is assumed to be treatment independent. The figure shows a trend toward higher levels of HRQOL over time in those who remain progression free, potentially a result of sicker patients (i.e., those with lower HRQOL) progressing earlier. The impact of treatment-related adverse events is shown through the large variations in HRQOL during the first 6 months after randomization.
      Figure thumbnail gr2
      Fig. 2HRQOL scores over time while patients are in the preprogression state (means and 95% confidence intervals). HRQOL, health-related quality of life.
      As with HRQOL before progression, postprogression HRQOL is assumed to be treatment independent in the base-case analysis. Postprogression HRQOL is also assumed to be independent of time since randomization, and is estimated from the trial data as 0.74 (SE 0.013).
      HRQOL scenario 1 considered the assumption of treatment independence of preprogression HRQOL scores. The analysis found that there was a statistically significant difference in the HRQOL between the two arms (0.066 higher in the chemotherapy-alone arm; SE 0.061) at baseline, such that patients in the chemotherapy-alone arm were typically healthier before the commencement of treatment. This difference was adjusted for by reducing all subsequent HRQOL scores for the chemotherapy-only arm by 0.066. Figure 3 shows the adjusted mean difference in preprogression HRQOL scores and 95% confidence intervals. The figure shows that after adjusting for baseline differences, there was little difference in HRQOL between the treatment groups up to around 6 months. After this, HRQOL deteriorates in the chemotherapy arm relative to the bevacizumab arm. However, the 95% confidence intervals in Figure 3 show that the difference is never statistically significant, strengthening the assumption of treatment-independent HRQOL scores made in the base case.
      Figure thumbnail gr3
      Fig. 3Differences in mean preprogression HRQOL and the associated 95% confidence intervals when a treatment-dependent HRQOL is assumed (as applied in HRQOL scenario 1). HRQOL, health-related quality of life; QOL, quality of life.
      There was a small difference in postprogression HRQOL between the two arms, being slightly higher in the chemotherapy-alone arm (0.75; SE 0.016) when compared with the bevacizumab arm (0.71; SE 0.020). However, as with progression-free HRQOL, there was no evidence that this difference was statistically significant (P = 0.095).

      Cost-Effectiveness

      Table 3 provides the cost-effectiveness results from the base-case and alternative scenarios considered.
      Table 3Results of the high-risk subgroup model: Base-case analysis and scenarios
      ScenarioTotal costs (£)Incremental costs (£)Total QALYsIncremental QALYsIncremental cost per QALY gained of bevacizumab (£)Probability of bevacizumab being cost-effective at thresholds (per QALY) ofPrice per milligram at which bevacizumab is cost-effective
      The fixed price per milligram that bevacizumab would have to be set at to be expected to be cost-effective at cost-effectiveness thresholds of £20,000 and £30,000 per QALY. The calculation assumes a fixed cost per milligram of bevacizumab and uses a deterministic model structure
      (£)
      Chemotherapy aloneChemotherapy plus bevacizumabTrial drug costsOther costsChemotherapy aloneChemotherapy plus bevacizumabPreprogressionPostprogressionLong term£20k£30k£20k£30k
      Base-case probabilistic12,87631,56017,7609242.8203.2010.2210.0660.09548,9750.0060.1250.751.25
      Extrapolation scenario 112,01330,76517,7609932.0212.3070.2210.0660.00065,5190.0000.0000.520.92
      Extrapolation scenario 212,79231,66717,7601,1142.8203.2010.2210.0660.09549,4740.0060.1180.721.22
      HRQOL scenario 112,87631,56017,7609242.8713·2540.2810.0080.06048,7010.0080.1360.751.25
      Long-term mortality scenario 113,60232,22117,7608593.4223.8770.2210.0660.16840,9810.0670.3020.931.52
      Long-term mortality scenario 214,63833,16717,7607704.3334.8960.2210.0660.27732,8980.2090.4461.201.92
      Vial sharing of all trial drugs12,87629,60115,8019242.8203.2010.2210.0660.09543,8400.0200.2050.821.38
      No vial sharing of trial drugs13,26331,98517,7989242.8203.2010.2210.0660.09549,0760.0080.1250.741.25
      HRQOL, health-related quality of life; QALY, quality-adjusted life-year.
      low asterisk The fixed price per milligram that bevacizumab would have to be set at to be expected to be cost-effective at cost-effectiveness thresholds of £20,000 and £30,000 per QALY. The calculation assumes a fixed cost per milligram of bevacizumab and uses a deterministic model structure
      In the base-case analysis, the bevacizumab arm was associated with higher costs than chemotherapy alone (incremental costs £18,684), most of which (£17,760) was the additional drug-related cost associated with bevacizumab. The remaining incremental difference (£924) is the result of the higher mean survival duration associated with bevacizumab, with patients incurring the cost of continued care for longer. This ratio of drug costs to other costs is consistent across the scenarios tested, with total and incremental costs changing little.
      Similarly, in the base-case analysis, bevacizumab is associated with a larger total number of QALYs than chemotherapy only (incremental QALYs 0.381), most of which is the result of greater gains in QALYs in the preprogression period of the analysis (0.221), with smaller gains in the postprogression (0.066) and long-term (0.095) periods. Only three of the scenarios impact the QALY gains: HRQOL scenario 1 and long-term mortality scenarios 1 and 2. In all three scenarios, the bevacizumab arm remains more effective (i.e., greater QALYs). Long-term mortality scenario 2, which assumes that surviving patients return to the mortality risk of the general population 5 years after treatment initiation, results in a large increase in total QALYs in both arms but proportionately larger in the bevacizumab arm, resulting in an increase in the incremental QALY gain to 0.563.
      The results demonstrate that in none of the scenarios considered is bevacizumab cost-effective at NICE’s conventional cost-effectiveness thresholds. The base-case analysis results in an ICER of £48,975 per QALY gained, with a probability of bevacizumab being cost-effective of 0.01 and 0.13 at thresholds of £20,000 and £30,000 per QALY, respectively. The per milligram price at which bevacizumab would be cost-effective was £0.75 and £1.25, respectively, for these two thresholds.
      Most of the scenarios considered did not have a significant impact on the ICER, with four of the seven impacting the ICER by less than £5000 per QALY. Where no extrapolation of the survival curves is assumed (i.e., extrapolation scenario 1), the ICER increases significantly to £65,519 because none of the long-term survival gains associated with bevacizumab is incorporated.
      Of all the scenarios considered, only one results in an ICER below £40,000 per QALY, long-term mortality scenario 2, and is associated with an ICER of £32,898 per QALY and probabilities of being cost-effective of 0.21 and 0.45 at thresholds of £20,000 and £30,000 per QALY, respectively.

      Discussion

      This analysis found that bevacizumab (7.5 mg/kg), in combination with carboplatin, and paclitaxel when used as a first-line treatment for advanced ovarian cancer is not cost-effective at its list price. In the predefined high-risk subgroup considered in the ICON7 trial, bevacizumab is associated with a base-case ICER of £48,975 per QALY. Although the ICON7 trial found that there were gains from bevacizumab in both overall survival and PFS in this subgroup, the short duration of these gains and the significant acquisition cost associated with bevacizumab resulted in only small gains in expected QALYs but a high incremental cost.
      Patients with advanced ovarian cancer have poor prognosis in terms of overall survival, with a mean survival of 34.5 months in the comparator arm of the ICON7 trial [
      • Oza A.
      • Cook A.D.
      • Pfisterer J.
      • et al.
      Standard chemotherapy with or without bevacizumab for women with newly diagnosed ovarian cancer (ICON7): overall survival results of a phase 3 randomised trial.
      ]. The provision of bevacizumab in advanced ovarian cancer does not, however, satisfy NICE’s criteria for end-of-life consideration [
      National Institute for Health and Care Excellence
      TA 284: Bevacizumab in Combination with Paclitaxel and Carboplatin for First-Line Treatment of Advanced Ovarian Cancer.
      ]. As such, NICE would consider there to be no special circumstance to accept an ICER above its usual cost-effectiveness thresholds.
      The analysis considers the price reductions that can be expected to make bevacizumab cost-effective in the high-risk subgroup (Table 3). These show that a price reduction of between 46% and 67% (depending on the cost-effectiveness threshold applied) would be necessary in the base case. However, in the scenario associated with the lowest ICER, long-term scenario 2, a reduction of between 21% and 45% would be sufficient. The Cancer Drugs Fund in England could use these price analyses to inform any future decisions about the product in this indication. Similar decisions are necessary internationally and can be guides by the analysis. It should be noted that the price reductions are the minimum that should be necessary to receive positive guidance, with the possible need for greater reductions to reflect uncertainty in cost-effectiveness [
      • Claxton K.
      • Palmer S.
      • Longworth L.
      • et al.
      Informing a decision framework for when NICE should recommend the use of health technologies only in the context of an appropriately designed programme of evidence development.
      ].
      As with most cost-effectiveness analysis of new cancer medications, the main limitation of the analysis is that the ICON-7 trial included a proportion of patients who remained alive and, in some cases, without progression at the end of study follow-up. To estimate costs and relevant health outcomes over a lifetime time horizon, assumptions are necessary regarding the extrapolation of rates of PFS and mortality seen in the trial. The significant impact of extrapolation scenario 1 and long-term morality scenarios 1 and 2 on the cost-effectiveness result highlights that the use of different assumptions has a large impact on cost-effectiveness, although ICERs do not fall below the upper bound of the NICE cost-effectiveness threshold. In addition, some authors have suggested that partition survival models, as used here, are biased [
      • Coyle D.
      • Coyle K.
      PRM74: the inherent bias from using partitioned survival models in economic evaluation.
      ]; however, the authors suggest that such bias is limited to cases in which only PFS differs between treatment arms, and not overall survival as is evident with bevacizumab. In addition, there has been little research into the robustness of this argument, making the generalizability of its inference impossible to comment on.
      Future research should be aimed at directly comparing the costs and outcomes of using the lower dose of bevacizumab in advanced ovarian cancer with those of the product at its higher dose. The higher dose imposes higher costs but potentially improves outcomes further in the high-risk subgroup. We considered incorporating the higher dose into our analysis via an indirect comparison, using findings from the GOG-0218 trial [
      • Burger R.A.
      • Brady M.F.
      • Bookman M.A.
      • et al.
      Incorporation of bevacizumab in the primary treatment of ovarian cancer.
      ]. However, a different trial population, widespread crossover from control to bevacizumab after progression in GOG-0218, and differences between UK practice and US practice made meaningful synthesis impossible. However, consideration of the published median PFS results [
      • Oza A.
      • Cook A.D.
      • Pfisterer J.
      • et al.
      Standard chemotherapy with or without bevacizumab for women with newly diagnosed ovarian cancer (ICON7): overall survival results of a phase 3 randomised trial.
      ,
      • Burger R.A.
      • Brady M.F.
      • Bookman M.A.
      • et al.
      Incorporation of bevacizumab in the primary treatment of ovarian cancer.
      ] and cost-effectiveness estimates here and elsewhere [
      National Institute for Health and Care Excellence
      TA 284: Bevacizumab in Combination with Paclitaxel and Carboplatin for First-Line Treatment of Advanced Ovarian Cancer.
      ] suggests that the lower dose of bevacizumab represents the more cost-effective option despite not being cost-effective given a NICE threshold. However, a robust trial or meta-analysis of published data is required to produce a robust estimate of the efficacy of the two doses.
      The NICE single technology appraisal of bevacizumab in advanced ovarian cancer reported estimates of the cost-effectiveness of bevacizumab at both the lower (7.5 mg/kg) and higher (15 mg/kg) doses [
      National Institute for Health and Care Excellence
      TA 284: Bevacizumab in Combination with Paclitaxel and Carboplatin for First-Line Treatment of Advanced Ovarian Cancer.
      ]. However, because of NICE’s restrictive remit, it was unable to comment on the results produced for the lower unlicensed dose. The manufacturer’s report to NICE estimated ICERs for both doses in the high-risk subgroup, reporting £144,682 per QALY at 15 mg/kg and £32,683 per QALY at 7.5 mg/kg [
      National Institute for Health and Care Excellence
      TA 284: Bevacizumab in Combination with Paclitaxel and Carboplatin for First-Line Treatment of Advanced Ovarian Cancer.
      ], very close to NICE’s upper threshold. The latter estimate is considerably lower than that in the base case reported here (£48,975). This is likely to be due to Roche’s estimate being based on an early “cut” of the ICON7 trial data, although there were also some differences in methods (e.g., the use of elicited resource use profiles for hospitalization and supportive care rather than from the trial evidence).
      In conclusion, the lower dose of bevacizumab for advanced ovarian cancer is not cost-effective based on the product’s list price and using NICE’s cost-effectiveness thresholds. Significant price discounts would be needed to make the drug affordable to the NHS. The Cancer Drugs Fund in England will be informed by such analysis in any future price negotiations, as will decision-making organizations internationally.
      Source of financial support: This work was funded as part of ICON7, an Medical Research Council (MRC)-sponsored, academic-led, and Roche-supported trial to investigate the use of bevacizumab and to support licensing.

      Supplementary Materials

      References

      1. Cancer Research UK. Ovarian cancer incidence statistics. Available from: 〈http://www.cancerresearchuk.org/cancer-info/cancerstats/types/ovary/incidence/uk-ovarian-cancer-incidence-statistics〉. [Accessed July 8, 2013].

        • Ozols R.
        Chemotherapy for ovarian cancer.
        Semin Oncol. 1999; 26: 34-40
        • National Institute for Health and Care Excellence
        CG 122: The Recognition and Initial Management of Ovarian Cancer.
        National Institute for Health and Care Excellence, London2011
        • National Institute for Health and Care Excellence
        TA 55: Review of the Clinical Effectiveness and Cost Effectiveness of Paclitaxel for Ovarian Cancer.
        National Institute for Health and Care Excellence, London2003
        • National Institute for Health and Care Excellence
        TA 212: Bevacizumab in Combination with Oxaliplatin and Either Fluorouracil Plus Folinic Acid or Capecitabine for the Treatment of Metastatic Colorectal Cancer.
        National Institute for Health and Care Excellence, London2010
        • National Institute for Health and Care Excellence
        TA 148: Bevacizumab for the Treatment of Non-Small-Cell Lung Cancer.
        National Institute for Health and Care Excellence, London2008
        • National Institute for Health and Care Excellence
        TA 214: Bevacizumab in Combination with a Taxane for the First-Line Treatment of Metastatic Breast Cancer.
        National Institute for Health and Care Excellence, London2011
        • National Institute for Health and Care Excellence
        TA 284: Bevacizumab in Combination with Paclitaxel and Carboplatin for First-Line Treatment of Advanced Ovarian Cancer.
        National Institute for Health and Care Excellence, London2013
        • NHS England
        National Cancer Drugs Fund List Ver3.0.
        NHS England, London2015
        • Oza A.
        • Cook A.D.
        • Pfisterer J.
        • et al.
        Standard chemotherapy with or without bevacizumab for women with newly diagnosed ovarian cancer (ICON7): overall survival results of a phase 3 randomised trial.
        Lancet Oncol. 2015; 16: 928-936
        • Perren T.
        • Swart A.M.
        • Pfisterer J.
        • et al.
        ICON7 Investigators A phase 3 trial of bevacizumab in ovarian cancer.
        N Engl J Med. 2011; 365: 2484-2496
        • National Institute for Health and Care Excellence
        Guide to the Methods of Technology Appraisal.
        National Institute for Health and Care Excellence, London2013
      2. NHSChoices. Ovarian Cancer. Available from: 〈http://www.nhs.uk/Conditions/Cancer-of-the-ovary〉. [Accessed July 8, 2013].

        • Billingham L.J.
        • Abrams K.R.
        • Jones D.R.
        Methods for the analysis of quality-of-life and survival data in health technology assessment.
        Health Technol Assess. 1999; 3: 1-152
        • Glasziou P.P.
        • Simes R.J.
        • Gelber R.D.
        Quality adjusted survival analysis.
        Stat Med. 1990; 9: 1259-1276
        • Baldwin L.A.
        • Huang B.
        • Miller R.W.
        • et al.
        Ten-year relative survival for epithelial ovarian cancer.
        Obstet Gynecol. 2012; 120: 612-618
        • Ozols R.F.
        • Markman M.
        • Thigpen J.T.
        ICON3 and chemotherapy for ovarian cancer.
        Lancet. 2002; 360: 2086-2087
        • Kind P.
        • Hardman G.
        • Macran S.
        UK population norms for EQ-5D. Working Papers, Centre for Health Economics.
        University of York. 1999;
      3. Joint Formulary Committee. British National Formulary (online). London: BMJ Group and Pharmaceutical Press, 2015.
        • O’Neill R.T.
        • Temple R.
        The prevention and treatment of missing data in clinical trials: an FDA perspective on the importance of dealing with it.
        Clin Pharmacol Ther. 2012; 91: 550-554
        • Faria R.
        • Gomes M.
        • Epstein D.
        • White I.R.
        A guide to handling missing data in cost-effectiveness analysis conducted within randomised controlled trials.
        Pharmacoeconomics. 2014; 32: 1157-1170
        • Claxton K.
        • Palmer S.
        • Longworth L.
        • et al.
        Informing a decision framework for when NICE should recommend the use of health technologies only in the context of an appropriately designed programme of evidence development.
        Health Technol Assess. 2012; 16: 1-323
        • Coyle D.
        • Coyle K.
        PRM74: the inherent bias from using partitioned survival models in economic evaluation.
        Value Health. 2014; (A194): 17
        • Burger R.A.
        • Brady M.F.
        • Bookman M.A.
        • et al.
        Incorporation of bevacizumab in the primary treatment of ovarian cancer.
        N Engl J Med. 2011; 365: 2473-2483