Topotecan for the treatment of recurrent and stage IVB cervical cancer

3 The manufacturer's submission

The Appraisal Committee (appendix A) considered evidence submitted by the manufacturer of topotecan and a review of this submission by the Evidence Review Group (ERG; appendix B).

3.1 In the submission, the manufacturer compared topotecan plus cisplatin with cisplatin alone. The manufacturer also compared topotecan plus cisplatin with paclitaxel plus cisplatin. The manufacturer justified their choice of comparator with data from the IMS Oncology Analyzer database from 2004 to 2008 to show that cisplatin alone is the most frequently used therapy in the group of women for whom topotecan plus cisplatin is licensed (39%). A more recent breakdown of the IMS Oncology Analyzer database from 2006 to 2008 indicates that 27% of patients receive cisplatin alone; 23% receive carboplatin plus paclitaxel. There are a range of other combination therapies, each of which is given to fewer than 10% of patients. In total 57% of patients receive some form of combination therapy.

3.2 The manufacturer identified one phase III, open-label randomised controlled trial (GOG-0179; n = 293) that included women with persistent, recurrent or stage IVB cervical cancer and compared topotecan plus cisplatin with cisplatin alone. These women were followed up for a maximum of 36 months. The trial reported increased median overall survival for topotecan plus cisplatin compared with cisplatin alone: 9.4 versus 6.5 months, respectively (hazard ratio [HR] 0.76; 95% confidence interval [CI] 0.59 to 0.98; p = 0.033), and increased median progression-free survival for topotecan plus cisplatin compared with cisplatin alone: 4.6 versus 2.9 months, respectively (HR 0.76; 95% CI 0.60 to 0.97; p = 0.027).

3.3 The manufacturer reported that the safety profile for topotecan plus cisplatin was predictable and manageable. However, there were four treatment-related deaths in the topotecan plus cisplatin group compared with none in the cisplatin group. Febrile neutropenia occurred in 17.7% of women treated with topotecan plus cisplatin and in 7.5% of women treated with cisplatin alone. Serious adverse events occurred in 10% of women treated with cisplatin alone compared with 14% of women treated with topotecan plus cisplatin.

3.4 The manufacturer presented data on subpopulations of the GOG-0179 trial. The 'licensed population', which consisted of 222 women, was defined as the population for whom topotecan is indicated in the marketing authorisation. Data from the other 71 women in the trial were excluded because they had cervical cancer that was not covered by the marketing authorisation (32 women had persistent disease and in 39 women the sustained cisplatin-free interval was less than 180 days). The median overall survival estimates for the licensed population were 11.9 months for topotecan plus cisplatin (n = 107) and 7.3 months for cisplatin alone (n = 115) (HR 0.65; 95% CI 0.49 to 0.88; p = 0.0041).

3.5 The manufacturer completed further subgroup analyses of the licensed population to consider the benefits of topotecan in women who had never had cisplatin (cisplatin naive; n = 120) and those with a sustained cisplatin-free interval longer than 180 days (n = 102). The median overall survival in the cisplatin-naive group was 14.5 months for topotecan plus cisplatin and 8.5 months for cisplatin alone (HR 0.59; 95% CI 0.39 to 0.88; p = 0.0098). The median overall survival in the sustained cisplatin-free interval group was 9.9 months for topotecan plus cisplatin and 6.3 months for cisplatin alone (HR 0.75; 95% CI 0.49 to 1.16; p = 0.1912).

3.6 The manufacturer identified a trial (GOG-0204) that was not formally included in the clinical-effectiveness review. An abstract reported on this trial, which included a head-to-head comparison of four cisplatin-containing combinations: paclitaxel (n = 103), vinorelbine (n = 108), gemcitabine (n = 112) and topotecan (n = 111). A planned interim analysis recommended early closure of GOG-0204 because the comparator groups were unlikely to demonstrate a statistically significant benefit compared with paclitaxel plus cisplatin. For the comparison of cisplatin plus topotecan with cisplatin plus paclitaxel, the trial reported a hazard ratio for progression-free survival of 1.268 and for overall survival of 1.255. The differences favoured the paclitaxel combination but were not statistically significant.

3.7 The manufacturer identified another trial (GOG-0169) which was used in an indirect comparison of topotecan plus cisplatin and paclitaxel plus cisplatin. This phase III study compared paclitaxel plus cisplatin (n = 130) with cisplatin alone (n = 134) in women with stage IVB, recurrent, or persistent squamous cell cervical cancer. The trial duration was 24 months. The median overall survival was 9.7 months for paclitaxel plus cisplatin and 8.8 months for cisplatin alone. The median progression-free survival was 4.8 months for paclitaxel plus cisplatin and 2.8 months for cisplatin alone.

3.8 The manufacturer submitted two separate cost-effectiveness analyses:

3.9 For the within-trial comparison the manufacturer performed separate analyses for the licensed population and subgroups of this population. The subgroups were women who were cisplatin naive and women who had had a sustained cisplatin-free period. The manufacturer stated that the least potentially biased analysis in the model-based comparison would be between the cisplatin-naive population of GOG-0179, including women with persistent disease, and the overall intention-to-treat population of GOG-0169. The manufacturer considered the within-trial comparison to be the primary analysis within their submission. The model-based comparison was presented as a secondary analysis to include alternative comparators used in England and Wales.

3.10 In the within-trial comparison, the manufacturer included patient-level data for clinical efficacy, safety and quality of life from the GOG-0179 trial. Data on resource use were based on clinical events occurring in the trial supplemented by data from external sources, including expert opinion. Costs were obtained from published sources, including NHS Reference Costs 2006/07. The manufacturer did not give a breakdown of the costs for the within-trial comparison. It was assumed that the cost of topotecan was £488.25 per cycle and the cost of cisplatin was £50.74 per cycle. The cost of topotecan was varied in a sensitivity analysis from £390.60 to £585.90 to reflect minimum wastage of unused topotecan (when vials were reused over the 3-day dosing schedule) and maximum wastage (when vials were discarded immediately after use). The cost of administering topotecan was assumed to be £277 for the first dose of each cycle and £51 for each subsequent dose in each cycle.

3.11 The manufacturer incorporated quality-of-life benefits into the within-trial comparison using an algorithm linking a disease-specific measure of quality of life (Functional Assessment of Cancer Therapy – General [FACT-G]) to utility. Utility values differed depending on whether a woman was treated with cisplatin alone or topotecan plus cisplatin. Values also differed according to the treatment phase: prior to randomisation, prior to cycle 2, prior to cycle 5 and 9 months after randomisation. The values for the cisplatin-alone group were 0.79, 0.73, 0.58 and 0.33, for these four treatment phases respectively. The corresponding values for the topotecan plus cisplatin group were 0.79, 0.72, 0.66 and 0.45. The manufacturer also included a review of the literature of alternative utility data associated with cervical cancer and other gynaecological cancers (including breast cancer). The utility values used in the sensitivity analysis were identified from a study of breast cancer (Brown and Hutton 1998) and were 0.64 at the start of treatment, 0.81 to reflect response to treatment, 0.39 following progression of disease and 0.16 during the last week of life.

3.12 In the model-based comparison the manufacturer based the key analysis on aggregate data from indirectly comparing the GOG-0179 and GOG-0169 trials. GOG-0169 did not report the hazard ratio for overall survival, therefore the manufacturer estimated the hazard ratio from the survival curves (HR = 0.87; 95% CI 0.68 to 1.11). The estimated hazard ratio was then applied to the observed overall survival for the cisplatin group of GOG-0179 to estimate the overall survival for paclitaxel plus cisplatin in the model-based comparison. The hazard ratio for the compared trials was 0.72 (95% CI 0.46 to 1.15). An additional sensitivity analysis included direct data on this comparison from the GOG-0204 trial. Resource use in the model-based comparison was based on the costing algorithms developed for the within-trial comparison. The utility values from the literature review were included in the cost per QALY analyses.

3.13 In the within-trial comparison, the base-case results for the licensed population were an incremental QALY gain of 0.23 at an incremental cost of £4122, giving an incremental cost-effectiveness ratio (ICER) of £17,974 per QALY gained. Probabilistic sensitivity analysis suggested that the probability of topotecan being cost effective at £20,000 and £30,000 per QALY gained was 50% and 88% respectively. For the cisplatin-naive population (including women with stage IVB cervical cancer) the incremental QALY gain was 0.32 at an incremental cost of £3521, giving an ICER of £10,928 per QALY gained. Probabilistic sensitivity analysis suggested that the probability of topotecan being cost effective at £20,000 and £30,000 per QALY gained was 89% and 98% respectively. For the sustained cisplatin-free interval population the incremental QALY gain was 0.13 at an incremental cost of £4145, giving an ICER of £32,463 per QALY gained. Probabilistic sensitivity analysis suggested that the probability of topotecan being cost effective at £20,000 and £30,000 per QALY gained was 31% and 55% respectively.

3.14 In the model-based comparison, the manufacturer only presented results for the cisplatin-naive population (including women with persistent disease). In the base-case results topotecan plus cisplatin dominated paclitaxel plus cisplatin (that is, paclitaxel plus cisplatin was less effective and more expensive), and had an ICER of £19,964 per life year gained compared with cisplatin alone. Using the hazard ratio from GOG-0204 (rather than from GOG-0169), paclitaxel plus cisplatin had an ICER of £982 per life year gained compared with topotecan plus cisplatin. In response to a request for clarification from the ERG, the manufacturer submitted a revised model-based comparison incorporating health-related quality of life and a time horizon of 36 months. When the hazard ratio from GOG-0169 was used, topotecan plus cisplatin dominated paclitaxel plus cisplatin; when the hazard ratio from GOG-0204 was used, paclitaxel plus cisplatin had an ICER of £13,260 per QALY gained compared with topotecan plus cisplatin.

3.15 The ERG identified a number of differences between the inclusion criteria of the clinical trials. GOG-0179 included women who were previously untreated, or had received prior chemotherapy or radiotherapy with or without a radiosensitiser. Approximately 60% of women had received prior cisplatin either as chemotherapy or as a radiosensitiser. GOG-0169 excluded women who had received prior chemotherapy, but included women who had been given chemotherapy as part of radiosensitisation (approximately 30%). However, it was unclear how many women received cisplatin as a radiosensitiser. GOG-0204 also excluded women who had previously received chemotherapy, unless this was given as a radiosensitiser, and the proportion of women who had previously received cisplatin as a radiosensitiser was approximately 70%. The ERG considered that GOG-0204 may be more representative of the UK population than GOG-0169, because of the increasing number of women in the UK who receive cisplatin as first-line treatment or as a radiosensitiser. The ERG stated that the manufacturer had included treatments currently used in the UK, but had not explained why other potentially relevant comparators were not included such as cisplatin plus 5-fluorouracil and cisplatin plus mitoxantrone.

3.16 The ERG stated that it was unclear from the manufacturer's submission whether a complete network of evidence had been identified and investigated. GOG-0179 was a well-conducted randomised controlled trial and it was reasonable for the manufacturer to use this as the direct comparison. However, head-to-head comparisons were also available from GOG-0204. The ERG considered such a direct comparison of topotecan plus cisplatin and paclitaxel plus cisplatin would have been preferable to the indirect comparison used, particularly given the differences in populations between GOG-0169 and GOG-0179. The inclusion of GOG-0204 would also have increased the number of potential comparators and expanded the network of indirect evidence.

3.17 The ERG stated that a complete validation of the within-trial comparison was not possible because complete data sets and coding had not been provided within the timelines of the ERG critique. In addition, the ERG raised concerns about the external validity of this comparison. When comparing the two economic analyses, the ERG noted a difference in the mean costs obtained from the within-trial comparison and the model-based comparison. The ERG was unable to fully investigate the difference because a breakdown of the costs was not provided for the within-trial comparison. The ERG noted that the utility estimates did not appear to have been derived accurately from the trial because of incorrect mapping of FACT-G data to utility values. In addition, there were concerns about the imputation methods and that the impact of mortality may have been double counted. Furthermore, the ERG questioned the appropriateness of the utility values used in the model-based comparison and sensitivity analysis because they were from a study on metastatic breast cancer and not cervical cancer. The ERG raised concerns about the costing in both analyses, particularly costs relating to administration and adverse events.

3.18 The ERG undertook a number of exploratory analyses for both the cisplatin-naive and the licensed populations using the model-based comparison. The ERG amended the utility values, the costs of administering topotecan and the assumed number of vials of topotecan used per treatment cycle. The ERG also performed exploratory analyses that considered dose reduction.

3.19 To address the limitations in the utility values available, the ERG considered three scenarios. The first used the manufacturer's starting utility value (Brown and Hutton 1998; 0.64) adopted for the model-based comparison. The second used a slightly higher starting utility value of 0.67 taken from literature estimates of mean utility values associated with cervical cancer, weighted according to the proportion of patients with each stage of disease in GOG-0179. Values for subsequent health states were calculated using the Brown and Hutton 1998 utility values. This second scenario assumed that utility remained constant from starting treatment to disease progression. The third scenario was the same as the second but used a starting utility value of 0.72 derived from the FACT-G data collected in GOG-0179. The ICERs for topotecan plus cisplatin compared with cisplatin alone for the cisplatin-naive population for the three utility scenarios were £25,309, £26,156 and £24,513 per QALY gained respectively. The ICERs for the licensed population for the three utility scenarios were £55,926, £59,406 and £54,352 per QALY gained respectively. The ERG used the third scenario in all subsequent exploratory analyses because they considered it the most appropriate.

3.20 The ERG considered that the costs of administering topotecan may have been underestimated. The ERG stated that more appropriate estimates of the administration costs for each treatment could be taken from the health resource group code SB14Z for the delivery of complex chemotherapy, including prolonged infusion treatment at first attendance, and code SB15Z for the delivery of subsequent elements of a chemotherapy cycle, given in NHS Reference Costs 2006/07. The cost code SB14Z (£289, inflated to £299 at 2007/08 prices) was assumed to reflect the administration of cisplatin, paclitaxel plus cisplatin, or the first administration of topotecan plus cisplatin. The cost code SB15Z (£189, inflated to £195 at 2007/08 prices) was assumed to reflect the second and third administration of topotecan for each cycle. The total cost of administering topotecan plus cisplatin was £689 per cycle, while the cost of administering cisplatin alone or paclitaxel plus cisplatin was £299 per cycle. The ICER for topotecan plus cisplatin compared with cisplatin alone (including the amended utilities) was £31,831 per QALY gained in the cisplatin-naive population and £68,885 per QALY gained in the licensed population. The revised administration costs were used in subsequent exploratory analyses.

3.21 The ERG had concerns about the number of topotecan vials used and the amount of wastage in the manufacturer's analysis. In the cisplatin-naive population the ICERs for topotecan plus cisplatin compared with cisplatin alone (including the amended utilities and administration costs) were £26,778 and £34,327 per QALY gained, for minimum and maximum wastage respectively. For the licensed population the ICERs were £58,872 and £73,833 per QALY gained respectively.

3.22 The ERG considered that the differences in costs between the within-trial comparison and the model-based comparison may have been because of dose reduction. The ERG therefore calculated the difference in costs between the manufacturer's model-based comparison and the ERG's revised cost estimates. The differences were then applied to the absolute estimates of costs in the within-trial analysis. Both minimum and maximum wastage of vials were considered. When wastage was minimised, the ICER for topotecan plus cisplatin compared with cisplatin alone was £19,815 per QALY gained in the cisplatin-naive population and £53,868 per QALY gained in the licensed population. When maximum wastage of topotecan was assumed, the ICERs were £27,362 and £68,826 per QALY gained respectively.

3.23 The manufacturer's model-based comparison did not report an ICER for any treatment in comparison with cisplatin alone. The ERG integrated the cost and QALY values for cisplatin into the manufacturer's model-based comparison so that cisplatin could be considered as a comparator alongside topotecan plus cisplatin and paclitaxel plus cisplatin. This allowed for a simultaneous incremental analysis to be carried out between the three treatments. The ERG presented two separate scenarios, one using the hazard ratio from the indirect comparison of GOG-0169 and GOG-0179 and another using the hazard ratio from GOG-0204. Both included the amended utility values and administration costs, but neither included dose reduction. When the hazard ratio from the indirect comparison of GOG-0169 and GOG-0179 was used and minimum wastage assumed, the ICER was £26,778 per QALY gained for topotecan plus cisplatin compared with cisplatin alone in the cisplatin-naive population and £58,872 per QALY gained in the licensed population. When maximum wastage was assumed, the ICER was £34,327 per QALY gained for topotecan plus cisplatin compared with cisplatin alone for the cisplatin-naive population. In this scenario paclitaxel plus cisplatin was extendedly dominated (that is, the ICER was higher than that of the next, more effective, alternative). For the licensed population the ICER for topotecan plus cisplatin compared with paclitaxel plus cisplatin was £116,788 per QALY gained, and the ICER for paclitaxel plus cisplatin in comparison with cisplatin alone was £64,865 per QALY gained. When the GOG-0204 hazard ratio was used, topotecan plus cisplatin was dominated by paclitaxel plus cisplatin regardless of the assumption about topotecan wastage (that is topotecan plus cisplatin was more expensive and less effective than paclitaxel plus cisplatin). The ICER for paclitaxel plus cisplatin compared with cisplatin alone was £17,021 per QALY gained for the cisplatin-naive population and £21,926 per QALY gained for the licensed population.

3.24 Following consultation on the appraisal consultation document, the manufacturer of topotecan provided a network meta-analysis of clinical-effectiveness data and further economic analyses. The network meta-analysis pooled estimates of effectiveness derived from direct and indirect comparisons of the three relevant clinical trials for cisplatin, paclitaxel plus cisplatin and topotecan plus cisplatin: GOG-0179, GOG-0169 and GOG-0204. The overall survival hazard ratios for the comparison of cisplatin plus paclitaxel with cisplatin alone were 0.83 (95% CI 0.68 to 1.08) for the cisplatin-naive population and 0.81 (95% CI 0.67 to 1.03) for the licensed population, favouring the paclitaxel combination. The corresponding hazard ratios for the comparison of cisplatin plus topotecan with cisplatin alone were 0.75 (95% CI 0.53 to 0.97) and 0.81 (95% CI 0.62 to 0.98), favouring the topotecan combination. For the comparison of cisplatin plus topotecan with cisplatin plus paclitaxel the hazard ratio for the cisplatin-naive population was 0.98 (95% CI 0.73 to 1.23). A hazard ratio was not presented for the licensed population.

3.25 The manufacturer included the hazard ratios obtained from the network meta-analysis in a revised economic analysis. A further economic analysis was also presented that used the hazard ratio for topotecan from GOG-0179, but data from the meta-analysis for the cisplatin survival curves. In addition, the economic model was updated to provide fully incremental analyses (that is, to provide a simultaneous comparison of all three treatment options) and to include a probabilistic function, to capture the uncertainty of the results. Revised parameter assumptions were also incorporated to include the ERG's preferred utility values, preferred administration costs and assumptions of maximum and minimum wastage. Results were presented for both the licensed population and the cisplatin-naive population.

3.26 For the licensed population, using the hazard ratios from the meta-analysis and assuming maximum wastage, the ICER for topotecan plus cisplatin was £81,756 per QALY gained in comparison with cisplatin alone and was dominated by paclitaxel plus cisplatin. If minimum wastage of topotecan was assumed, the ICER for topotecan plus cisplatin was £63,913 per QALY gained in comparison with cisplatin alone and was dominated by paclitaxel plus cisplatin.

3.27 For the licensed population, using the hazard ratios from GOG-0179 for topotecan survival and hazard ratios from the meta-analysis for the cisplatin survival curves, and assuming maximum wastage, the ICER for topotecan plus cisplatin was £60,903 per QALY gained in comparison with cisplatin alone and £65,364 per QALY gained in comparison with paclitaxel plus cisplatin. If minimum wastage of topotecan was assumed, the ICER was £47,616 per QALY gained in comparison with cisplatin alone, and £7142 per QALY gained in comparison with paclitaxel plus cisplatin.

3.28 For the cisplatin-naive subgroup, using the hazard ratios from the meta-analysis and assuming maximum wastage of topotecan, the ICER for topotecan plus cisplatin was £58,911 per QALY gained in comparison with cisplatin alone, and £49,964 in comparison with paclitaxel plus cisplatin. If minimum wastage of topotecan was assumed, the ICER for topotecan plus cisplatin was £46,054 per QALY gained in comparison with cisplatin alone and £5459 per QALY gained in comparison with paclitaxel plus cisplatin.

3.29 For the cisplatin-naive subgroup, using the hazard ratios from GOG-0179 for topotecan survival and hazard ratios from the meta-analysis for the cisplatin survival curves, and assuming maximum wastage, the ICER for topotecan plus cisplatin was £30,171 per QALY gained in comparison with cisplatin alone and £11,627 in comparison with paclitaxel plus cisplatin. If minimum wastage of topotecan was assumed, the ICER was £23,586 per QALY gained in comparison with cisplatin alone and £1270 in comparison with paclitaxel plus cisplatin.

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