Bortezomib for induction therapy in multiple myeloma before high-dose chemotherapy and autologous stem cell transplantation

Clinical effectiveness

3.1 The main clinical evidence submitted by the manufacturer for the bortezomib, thalidomide and dexamethasone regimen came from the PETHEMA trial in which in people with newly diagnosed multiple myeloma who were eligible for autologous stem cell transplantation received up to 6 cycles of bortezomib, thalidomide and dexamethasone, or thalidomide and dexamethasone. The evidence for the bortezomib and dexamethasone regimen came from the IFM trial, which compared 4 cycles of bortezomib and dexamethasone with 4 cycles of vincristine, doxorubicin and dexamethasone. The manufacturer also submitted data from the GIMEMA trial which compared the efficacy and safety of 3 cycles of bortezomib, thalidomide and dexamethasone with 3 cycles of thalidomide and dexamethasone as induction treatment before autologous stem cell transplantation followed by consolidation treatment with 2 cycles of either bortezomib, thalidomide and dexamethasone or thalidomide and dexamethasone. However, the manufacturer highlighted that the PETHEMA trial study design better reflected how the bortezomib, thalidomide and dexamethasone regimen is expected to be used in the UK and therefore was the focus of the manufacturer's submission. Data from the HOVON trial were provided by the manufacturer, but the bortezomib-containing regimen included in this study, bortezomib, doxorubicin and dexamethasone, was not approved by the European Medicines Agency and is therefore not a licensed regimen.

3.2 The PETHEMA trial was a randomised, open-label phase III study that compared the efficacy and safety of bortezomib in combination with thalidomide and dexamethasone against thalidomide and dexamethasone in people with newly diagnosed symptomatic multiple myeloma and measurable disease (serum and/or urine M protein), who were eligible for autologous stem cell transplantation. Patients were randomised to either bortezomib, thalidomide and dexamethasone (n=130) or thalidomide and dexamethasone (n=127), both of which consisted of 6 cycles of 28 days, with each cycle including 4 infusions of bortezomib, oral dexamethasone (40 mg on days 1–4 and 8–11 of each cycle) and oral thalidomide (50 mg daily). After transplantation, patients who continued in the trial were re-randomised to receive 1 of 3 maintenance treatments (interferon alfa‑2b, thalidomide, or bortezomib plus thalidomide). Maintenance therapy was continued for up to 3 years, or until disease progression. Although the PETHEMA trial did not incorporate the discontinuation rule as per the summary of product characteristics, because patients in the GIMEMA trial received only 3 cycles, the manufacturer stated that the PETHEMA trial design better reflected how the bortezomib, thalidomide and dexamethasone regimen is expected to be used in the UK.

3.3 The GIMEMA trial was a randomised, open-label, phase III study in 480 patients with newly diagnosed, previously untreated symptomatic multiple myeloma with measurable disease. The study was designed to compare the efficacy and safety of 3 cycles of bortezomib, thalidomide and dexamethasone with 3 cycles of thalidomide and dexamethasone as induction treatment before autologous stem cell transplantation. It also evaluated subsequent consolidation treatment consisting of 2 cycles of either bortezomib, thalidomide and dexamethasone, or thalidomide and dexamethasone. Maintenance treatment with dexamethasone was continued until disease progression or relapse. Each cycle of induction therapy consisted of 1.3 mg/m² of bortezomib on days 1, 4, 8 and 11, with 100 mg of thalidomide daily for the first 14 days and 200 mg thereafter. Dexamethasone 40 mg was administered on days 1, 2, 4, 5, 8, 9, 11 and 12. Instead of the dosage recommended in the summary of product characteristics, patients randomised to the bortezomib, thalidomide and dexamethasone induction group received only 3 cycles of bortezomib-based treatment.

3.4 The IFM trial was a randomised, open-label study designed to compare the efficacy and safety of bortezomib and dexamethasone (with or without consolidation treatment with dexamethasone, cyclophosphamide, etoposide and cis-platinum) against vincristine, doxorubicin and dexamethasone (with or without intensification). Treatment with bortezomib and dexamethasone consisted of 4 21‑day cycles of 1.3 mg/m² of bortezomib and 40 mg of dexamethasone. However, the manufacturer stated that only the results without the intensification step were relevant to the decision problem. Moreover, as discussed previously, the manufacturer stated that this comparison was not in line with the decision problem because the vincristine, doxorubicin and dexamethasone regimen is not a thalidomide-containing regimen and therefore not an appropriate comparator. The manufacturer also stated that the vincristine, doxorubicin and dexamethasone regimen is not routinely used in UK clinical practice and excluded this from its base-case analysis.

3.5 The primary outcome measures in the PETHEMA, GIMEMA and IFM trials were response rates reported after induction and after transplant. The manufacturer's submission reported 'response' in terms of:

3.6 Patients who received bortezomib (bortezomib, thalidomide and dexamethasone) had a statistically significant difference in overall response rate after induction compared with the thalidomide regimen (thalidomide and dexamethasone) in both the PETHEMA (84.6% compared with 61.4%, p<0.001) and GIMEMA (93.2% compared with 78.6%, p<0.0001) trials. This difference in treatment effect on overall response rate was maintained after transplant (77.7% compared with 56.7%, p<0.001 in the PETHEMA trial and 93.2% compared with 84.5%, p<0.0025 in the GIMEMA trial). Patients receiving bortezomib in PETHEMA and GIMEMA also showed statistically higher post-induction and post-transplant complete response rates than those on the thalidomide-containing regimen. In the PETHEMA trial, 35.4% in the bortezomib, thalidomide and dexamethasone treatment group had a post-induction complete response compared with 13.4% in the thalidomide and dexamethasone group (p<0.001). In the GIMEMA trial, 18.6% had a post-induction complete response in the bortezomib, thalidomide and dexamethasone treatment group compared with 4.6% in the thalidomide and dexamethasone group (p<0.0001). In the post-transplant period, statistically significant differences were maintained for the bortezomib, thalidomide and dexamethasone treatment groups in both the PETHEMA and GIMEMA trials (p<0.001 and p=0.0004 respectively). Both the PETHEMA and GIMEMA trials reported a statistically significant lower proportion of patients experiencing disease progression when treated with bortezomib, thalidomide and dexamethasone compared with patients treated with thalidomide and dexamethasone induction therapy in the post-induction period (6.2% and 23.6%, p=0.0004 in the PETHEMA trial; and 0% and 5.0%, p<0.0005 in the GIMEMA trial).

3.7 In the IFM trial, people who received bortezomib in combination with dexamethasone showed a statistically significant difference in overall response rate after induction compared with vincristine, doxorubicin and dexamethasone (77.1% compared with 60.7%, p<0.001) but this difference was not maintained after stem cell transplantation (79.6% compared with 74.4%, p=0.179).

3.8 Secondary outcomes reported in the PETHEMA, GIMEMA, IFM and MRC Myeloma IX (described in section 3.12) trials included:

3.9 The manufacturer's submission reported the median time to progression and time to progression hazard ratios from the PETHEMA and IFM trials. In the PETHEMA study there was a statistically significant lower hazard of progression in patients treated with bortezomib, thalidomide and dexamethasone compared with thalidomide and dexamethasone (HR 0.64, 95% CI 0.44 to 0.93, p=0.017). No statistically significant difference in median time to progression was reported. In the IFM study there was a numerically lower hazard of progression in patients treated with bortezomib and dexamethasone compared with vincristine, doxorubicin and dexamethasone but this was not statistically significant (HR 0.82, 95% CI 0.63 to 1.06, p value no reported).

3.10 The manufacturer's submission reported unadjusted overall survival hazard ratios for the PETHEMA and IFM trials. Median overall survival was not reached in either the PETHEMA trial (bortezomib, thalidomide and dexamethasone compared against thalidomide and dexamethasone, hazard ratio 0.80, 95% CI 0.48 to 1.34, p=0.393) or IFM trial (bortezomib and dexamethasone compared against vincristine, doxorubicin and dexamethasone, HR 0.8, 95% CI 0.54 to 1.19; p value not reported) and there was no statistically significant difference in overall survival between the treatment arms in each study. The manufacturer's submission highlighted clinical specialist opinion that the durations of the trials were too short to allow differences in overall survival and progression-free survival between treatment groups to be sufficiently captured, given the relatively long survival in this patient population after an autologous stem cell transplant.

3.11 The bortezomib-containing arms of the PETHEMA and GIMEMA trials (bortezomib, thalidomide and dexamethasone) reported higher proportions of patients having stem cell transplantation compared with the thalidomide and dexamethasone arms (80.8% compared with 61.4% in PETHEMA, and 88.0% compared with 82% in GIMEMA). In addition, the bortezomib and dexamethasone arm of the IFM trial reported higher proportions of patients having stem cell transplantation compared with the vincristine, doxorubicin and dexamethasone arm (89.1% compared with 81.8%). However, no statistical tests were reported.

3.12 In the absence of head-to-head trials comparing bortezomib-based regimens against cyclophosphamide in combination with thalidomide and dexamethasone, the manufacturer originally presented an indirect comparison based on the PETHEMA, GIMEMA, HOVON, IFM and MRC Myeloma IX randomised controlled trials. The MRC Myeloma IX trial is the only trial that has compared the efficacy of cyclophosphamide, vincristine, doxorubicin and dexamethasone against cyclophosphamide, thalidomide and dexamethasone in 1111 patients with newly diagnosed symptomatic myeloma. In this trial, patients were randomised to receive induction chemotherapy following either an intensive or non-intensive (attenuated treatment) pathway. The manufacturer considered that the MRC Myeloma IX trial provided the only potential evidence that allowed any form of comparison to be made between bortezomib-based regimens and cyclophosphamide, thalidomide and dexamethasone. The bortezomib plus doxorubicin and dexamethasone regimen (used in the HOVON trial) was included as part of the evidence submission but the manufacturer considered its inclusion in the indirect comparison to be inappropriate because this regimen was not included in the marketing authorisation. The manufacturer stated that a network could not be formed between the available trials, and an indirect comparison with cyclophosphamide, thalidomide and dexamethasone was not possible. In addition, bortezomib and dexamethasone could not be linked to a thalidomide-containing regimen. The manufacturer highlighted that assumptions to overcome the network limitations would generate considerable uncertainties and unreliable results. The manufacturer stated that an incremental analysis of the 2 licensed bortezomib regimens was therefore also not possible and stated that the base case should focus on the comparison of bortezomib, thalidomide and dexamethasone with thalidomide and dexamethasone.

3.13 The manufacturer presented a summary of results for adverse events for the PETHEMA, GIMEMA, IFM and MRC Myeloma IX trials. Only adverse events for the post-induction phase were reported because the manufacturer considered the adverse events for the post-transplant phase not to be relevant. Across all trials a similar proportion of patients reported any adverse event, grade 3/4 adverse events, and serious adverse events in both the bortezomib and comparator treatment arms. However, in the PETHEMA trial a statistically significantly greater number of total treatment-related adverse events was reported during induction with bortezomib, thalidomide and dexamethasone compared with thalidomide and dexamethasone (relative risk [RR]=1.42; 95% CI 1.17 to 1.73). In the GIMEMA trial a statistically significantly greater proportion of patients receiving bortezomib, thalidomide and dexamethasone experienced any grade 3/4 adverse event than those receiving thalidomide and dexamethasone (RR=1.69; 95% CI 1.36 to 2.08). The 2 most common treatment-related adverse events in the PETHEMA trial (pneumonia and peripheral neuropathy) occurred more frequently in the bortezomib, thalidomide and dexamethasone arm than in the thalidomide and dexamethasone arm. The manufacturer's submission highlights that in the 4 bortezomib-based studies, bortezomib was given intravenously. In terms of tolerability, total withdrawals and withdrawals due to disease progression were statistically significantly less in the bortezomib, thalidomide and dexamethasone arm than the thalidomide and dexamethasone arm in the PETHEMA trial (HR 0.51 [95% CI 0.34 to 0.77] and HR 0.45 [95% CI 0.25 to 0.84] respectively).

3.14 No health-related quality of life data were collected in the trials of bortezomib-containing regimens. To inform the cost-effectiveness evidence, the manufacturer conducted a systematic literature search to identify publications relevant to the decision problem in relation to health-related quality of life.

Cost effectiveness

3.18 The manufacturer conducted a systematic search of the literature and identified 3 cost-effectiveness studies relevant to the decision problem. The manufacturer conducted a quality assessment of these studies but did not discuss them further in the submission.

3.19 The manufacturer developed an Excel-based economic model to assess the cost effectiveness of bortezomib-based induction regimens compared with thalidomide-based induction regimens. As discussed previously, the manufacturer's base-case analysis focused on the cost-effectiveness analysis of bortezomib, thalidomide and dexamethasone compared with thalidomide and dexamethasone. The manufacturer presented cost-effectiveness analyses of bortezomib, thalidomide and dexamethasone (including the discontinuation rule stipulated in the marketing authorisation submitted as an addendum submission) compared with thalidomide and dexamethasone, and of bortezomib and dexamethasone compared with vincristine, doxorubicin and dexamethasone. The manufacturer acknowledged that the comparison with vincristine, doxorubicin and dexamethasone did not reflect current best clinical practice in the UK. The manufacturer stated that no comparisons of the bortezomib and dexamethasone regimen with relevant thalidomide-containing regimens (such as cyclophosphamide, thalidomide and dexamethasone) were possible using indirect mixed treatment comparisons.

3.20 The manufacturer chose a state-transition Markov model, with a cycle length of 1 month, to reflect the length of a course of treatment with bortezomib, thalidomide and dexamethasone (28 days) and because clinical outcomes are reported in months. The model did not include a half-cycle correction because the cycle length was short relative to the time horizon used in the model. Costs and quality-adjusted life years (QALYs) were discounted over a lifetime (30 years) time horizon at 3.5% per annum. The manufacturer stated that the model captured the 2 most important outcomes: post-induction response rate and overall survival. However, the manufacturer clarified that because the pivotal trials were not powered to detect a statistically significant difference in overall survival, the model was based on response rate, and the relationship between response rate and overall survival was quantified using long-term survival data from older trials in the same patient population. The model assumed that patients entered at the start of their induction therapy. After induction, patients in the model entered one of 3 health states: complete response, partial response, or non-responders (defined as minimal response, stable disease and progressive disease respectively). Depending on their post-induction response rate, patients subsequently proceeded to high-dose chemotherapy and stem cell transplantation or to the post-induction progression-free survival health state (non-stem cell transplant group). After induction, all patients were assumed to incur the same survival benefit, which was dependent only on their response rate after the induction phase and was independent of the actual induction regimen that they received. On disease progression, patients would then receive a second treatment, followed by third-line and subsequent lines of treatment after further progression.

3.21 Post-induction response rates were used as the main measure of efficacy in the model. Stem cell transplant rates for each response category in each treatment arm were used in the model evaluating bortezomib, thalidomide and dexamethasone compared with thalidomide and dexamethasone. The economic model for bortezomib and dexamethasone compared with vincristine, doxorubicin and dexamethasone used total stem cell transplant rates rather than transplant rates for each response category. The model also included mortality during the induction and transplant periods.

3.22 In order to model long-term survival based on the post-induction response rates, the manufacturer extracted overall survival data from the MRC Myeloma VII trial because overall survival data from the PETHEMA trial were considered immature. The MRC Myeloma VII trial randomised a total of 407 previously untreated multiple myeloma patients to conventional chemotherapy (n=200) or high-dose chemotherapy followed by autologous stem cell transplant (n=201). The 5-year survival in the high-dose chemotherapy followed by autologous stem cell transplant group was 88.6 months (95% CI 61.4, upper CI not reported) for patients who had a complete response, 39.8 months (95% CI 33.8 to 61.4) for patients who had a partial response, and 25.6 months (95% CI 7.0 to 31.3) for patients who had no response. For a scenario analysis, the manufacturer also used long-term survival data from the IFM90 trial from 1996.

3.23 The manufacturer used post-transplant time to progression from the PETHEMA trial to determine the probabilities of transition from the post-transplant progression-free health state to the start of second-line therapies. The manufacturer assumed that time to progression is affected by the interventions because it was modelled using separate parametric curves by treatment and response category. In the base case, time-to-progression transition probabilities were derived from exponential curves fitted to the PETHEMA data. Constant transition probabilities were used for transition from the second-line to the third-line health state across the 2 interventions, the estimates for which were derived from the subgroups of patients who had 1 or 2 lines of treatment respectively in the APEX trial (which compared bortezomib monotherapy with high-dose dexamethasone in patients with relapsed multiple myeloma). Probabilities of transition from third to further lines of treatment were derived by applying an exponential distribution to the time-to-progression data from the APEX trial. The overall survival data from the MRC Myeloma VII trial were used to determine the length of time that patients remained in the further lines of treatment health state before moving to the death state.

3.24 The manufacturer conducted a systematic search of the literature to identify publications that identified health-related quality of life data relevant to the decision problem. Five relevant studies were identified, of which 3 reflected the current UK patient population and clinical practice. The manufacturer selected the van Agthoven study as the base-case source of utility values because the utility values were obtained using EQ‑5D. The study by van Agthoven et al. compared chemotherapy (n=129) with intensive chemotherapy followed by myeloablative chemotherapy with stem cell transplantation (n=132) and total body irradiation treatment regimens. Patients were from the Netherlands and Belgium, under the age of 65 years, and had newly diagnosed and untreated multiple myeloma. They received 3 or 4 cycles of vincristine, doxorubicin and dexamethasone and 2 cycles of intermediate-dose melphalan, after which they were randomised to have either stem cell transplantation and interferon maintenance, or interferon maintenance only. The manufacturer applied a disutility of 0.02 to each patient experiencing an adverse event associated with induction therapy.

3.25 The costs applied in the model were taken from the BNF edition 64 (2012) and the 2012–13 Chemotherapy Regimens List. Administration of chemotherapy drugs, outpatient visits and tests as part of disease and treatment monitoring and the costs relating to stem cell transplantation were taken from the 2011–12 National Schedule Reference costs. The costs associated with treating adverse events were based on inpatient, outpatient or day-case visit National Schedule Reference costs. The manufacturer presented unit costs associated with each of the first-line induction therapies as well as drugs for prophylaxis, administration and monitoring. The total cost, including prophylaxis, administration and monitoring, of the bortezomib, thalidomide and dexamethasone regimen was £28,034, which compared with a total cost of £8,865 for thalidomide and dexamethasone. The total cost of the bortezomib and dexamethasone regimen was £14,104, whereas the total cost of vincristine, doxorubicin and dexamethasone was £2732.

3.26 The manufacturer's economic model estimated a difference in total costs between the bortezomib, thalidomide and dexamethasone and the thalidomide and dexamethasone regimens of £20,682. The bortezomib, thalidomide and dexamethasone regimen was associated with a 1.01 QALY gain compared with thalidomide and dexamethasone. The manufacturer's estimated base-case incremental cost-effectiveness ratio (ICER) for bortezomib, thalidomide and dexamethasone compared with thalidomide and dexamethasone was £20,468 per QALY gained. The incremental cost difference between the bortezomib and dexamethasone and the vincristine, doxorubicin and dexamethasone regimens was £12,710, and bortezomib and dexamethasone was associated with an incremental QALY gain of 0.88 resulting in an estimated ICER of £14,446 per QALY gained for the bortezomib and dexamethasone regimen compared with vincristine, doxorubicin and dexamethasone.

3.27 The manufacturer's deterministic sensitivity analyses highlighted that the results for bortezomib, thalidomide and dexamethasone compared with thalidomide and dexamethasone were most sensitive to the mortality for patients who had a complete response after induction therapy, and to drug costs. If the complete response mortality rate was varied within its 95% confidence interval, other things being equal, the ICER ranged from £17,018 to £28,867 per QALY gained. For the bortezomib, thalidomide and dexamethasone drug costs, sensitivity analyses were conducted using 4, 5 and 6 cycles of induction therapies. This was based on clinical opinion that the number of cycles will vary from one patient to another. The ICER range for the sensitivity analysis varying bortezomib, thalidomide and dexamethasone drug costs was £15,761 to £25,662 per QALY gained. For all other parameters varied in the sensitivity analyses, the ICER remained between £16,000 and £25,000 per QALY gained. Deterministic sensitivity analyses were also presented by the manufacturer for the comparison between the bortezomib and dexamethasone, and the vincristine, doxorubicin and dexamethasone regimens. Here, the results were most sensitive to the mortality for patients with complete response after induction therapy, with ICERs ranging from £10,961 to £18,354 per QALY gained.

3.28 The results of the manufacturer's probabilistic sensitivity analysis showed that, at maximum acceptable ICERs of £20,000 and £30,000 per QALY gained, there was a 35.4% and 71.3% probability respectively of the bortezomib, thalidomide and dexamethasone regimen being cost effective when compared with thalidomide and dexamethasone. The manufacturer estimated that at maximum acceptable ICERs of £20,000 and £30,000 per QALY gained, there was a 68.9% and 83.2% probability respectively of the bortezomib and dexamethasone regimen being cost effective compared with vincristine, doxorubicin and dexamethasone.

3.29 The ERG stated that the structure of the model was consistent with the clinical pathway of care for multiple myeloma and was clearly presented. However, the ERG highlighted that the manufacturer's analysis of bortezomib and dexamethasone compared with vincristine, doxorubicin and dexamethasone was outside the scope. It also highlighted that, given that the comparator in routine use in UK clinical practice is the cyclophosphamide, thalidomide and dexamethasone regimen, the comparison of bortezomib, thalidomide and dexamethasone with thalidomide and dexamethasone was not entirely relevant to clinical practice.

3.30 The ERG also expressed some concerns that the model extrapolated level of response after induction therapy to long-term survival and time to progression based on data from the MRC Myeloma VII trial. The ERG cautioned that the MRC Myeloma VII trial was old and its outcomes may not reflect the more advanced treatments available today. Moreover, the ERG stated that data from the MRC Myeloma VII trial related to maximal response to treatment rather than post-induction response rate, and the resulting survival curves might be confounded to some extent with post-stem cell transplant response. The ERG clinical expert agreed that response rate at induction predicts progression-free survival and overall survival. However, the ERG stated that other surrogate outcomes, such as post-stem cell transplant response rate, may offer a better prediction of progression-free survival and overall survival. The ERG observed that although the model had separate states for those who received a stem cell transplant and those who did not, the model attached no explicit survival benefit to a stem cell transplant other than that achieved by delaying the transition to the post-induction/post-transplant progression-free survival state for the duration of the stem cell transplant period. The ERG clinical expert stated that stem cell transplantation offers a survival benefit of 12–18 months compared with no transplant, and the ERG stated that it would have been more transparent to distinguish the separate effects on survival of post-induction response and stem cell transplantation. Alternatively, post-stem cell transplant response rate could have been considered because it has been shown to be statistically significantly associated with improved overall survival. Overall, the ERG stated that external validity would have been strengthened if the model had been based on overall survival and time to progression Kaplan–Meier curves or post-stem cell transplant response, rather than post-induction response. The ERG was concerned that in the absence of this the results were systematically biased in favour of bortezomib, thalidomide and dexamethasone.

3.31 The ERG stated that, in contrast to the manufacturer's description in the submission, the model implicitly assumed a continuing effect of induction treatment after induction is complete, because separate time to progression curves were used for each induction treatment arm and stem cell transplant mortality was also applied separately by treatment arm. The ERG also highlighted that, contrary to statements in the manufacturer's submission, the probability of receiving a stem cell transplant was not dependent on post-induction response, but only on treatment received.

3.32 The ERG noted the manufacturer's approach to calculating transition probabilities (section 3.23), and stated that the exponential distribution fitted to the PETHEMA complete response time to progression data for bortezomib, thalidomide and dexamethasone resulted in a shorter median survival time (approximately 61 months) than the exponential distribution fitted to complete response time to progression data for patients receiving thalidomide and dexamethasone (median survival approximately 98 months). The ERG stated that this contrasted with overall findings for progression-free survival in the trial publication in which median progression-free survival was statistically significantly higher with bortezomib, thalidomide and dexamethasone than with thalidomide and dexamethasone. The ERG noted that the manufacturer derived transition probabilities for third and further treatment lines using data from the APEX trial, which compared bortezomib monotherapy with high-dose dexamethasone in patients with relapsed multiple myeloma. The ERG commented that because the APEX trial used bortezomib as a monotherapy treatment, it may have had different survival outcomes to those seen with bortezomib combination therapy.

3.33 The ERG considered that the costs included in the model were reasonable. However, the ERG identified that a number of changes to the manufacturer's addendum economic model, submitted to take account of the discontinuation rule stipulated for bortezomib, thalidomide and dexamethasone (see section 3.19), were not documented by the manufacturer. Although the manufacturer's addendum referred to the original submission for discussion of resource identification, measurement and valuation, the ERG noted that many of the costs in the revised model were different from those in the original model. These included costs for drugs, induction, stem cell transplant, second-line treatment, third-line treatment, and some monitoring costs. The ERG noted that although these changes were generally minor, some were substantial. For example, the administration costs for high-dose dexamethasone increased from £168 to £1242 in second-line therapy, and from £168 to £1288 in third-line therapy. The ERG stated that when considering only model changes and assumptions that were documented in the manufacturer's addendum, the ICER was £23,958 per QALY gained for bortezomib, thalidomide and dexamethasone compared with thalidomide and dexamethasone, compared with the £20,468 per QALY gained reported in the manufacturer's addendum.

3.34 The ERG conducted a series of additional exploratory analyses. It considered that the MRC Myeloma VII trial was old and its outcomes may not reflect the more advanced treatments available today (see section 3.30) and noted that the manufacturer's sensitivity analysis used an even older study (the IFM90). Therefore, the ERG obtained data from a study by Alvares and from the NMSG 5/94 study to conduct scenario analyses. The NMSG 5/94 study was a prospective study with 247 patients recruited between 1994 and 1997, and Alvares was a retrospective study with 383 patients in England diagnosed with multiple myeloma between 1985 and 2004. The ERG considered that the Alvares data provided the better fit to the PETHEMA overall survival data. The ERG commented that because median overall survival for partial and non-responders in the Alvares study was much better than in the MRC Myeloma VII trial, this resulted in an increase in the base-case ICER from £20,468 per QALY gained to £30,368 per QALY gained for bortezomib, thalidomide and dexamethasone compared with thalidomide and dexamethasone. The ERG also provided the ICER using the data from the NMSG 5/94 study, but this did not include the discontinuation rule for the bortezomib, thalidomide and dexamethasone regimen and is therefore not presented. However, it resulted in a higher ICER than using the Alvares study data. The ERG also highlighted that data from the MRC Myeloma VII trial related to maximal response to treatment rather than post-induction response rate and that this was arguably more similar to post-stem cell transplant response. The ERG commented that post-stem cell transplant response rates provided a more consistent fit to the MRC Myeloma VII data and would better predict overall survival. Applying post-stem cell transplant response rates alone increased the manufacturer's base-case ICER from £20,468 to £26,292 per QALY gained for bortezomib, thalidomide and dexamethasone compared with thalidomide and dexamethasone. The ERG combined its preferred scenario analyses, that is, using data from Alvares to inform long-term survival and using post-stem cell transplant response rates. This resulted in an ICER of £38,985 per QALY gained for bortezomib, thalidomide and dexamethasone compared with thalidomide and dexamethasone.

3.35 The ERG conducted further exploratory analyses using the relevant economic model outputs from the manufacturer's base-case cost-effectiveness results to calculate ICERs for all treatments compared with thalidomide and dexamethasone and with cyclophosphamide, thalidomide and dexamethasone, the latter of which is a more relevant comparator regimen in a UK population. The ERG commented that all results should be treated with extreme caution as they compare individual arms of separate trials, without adjusting for trial populations. Furthermore there were differences in the trial designs. For these reasons, the results should not be directly compared. Using the manufacturer's base-case model to compare bortezomib, thalidomide and dexamethasone against cyclophosphamide, thalidomide and dexamethasone resulted in an exploratory ICER of £228,159 per QALY gained. The ERG then applied its preferred assumptions, that is, using data from Alvares to inform long-term survival and using post-stem cell transplant response rates. This resulted in an exploratory ICER for bortezomib, thalidomide and dexamethasone compared with cyclophosphamide, thalidomide and dexamethasone of £81,983 per QALY gained. The ERG conducted the same exploratory analyses in order to calculate ICERs for the bortezomib and dexamethasone regimen compared with thalidomide and dexamethasone and with cyclophosphamide, thalidomide and dexamethasone. The ERG applied data from Alvares and used post-stem cell transplant response rates, and this resulted in an ICER of £26,701 per QALY gained for bortezomib and dexamethasone compared with thalidomide and dexamethasone. However, bortezomib and dexamethasone was dominated by (that is, was less effective and more expensive than) cyclophosphamide, thalidomide and dexamethasone.

Manufacturer's response to the appraisal consultation document

3.36 Additional analyses were provided by the manufacturer in response to NICE's request for further work on the comparison between the regimen containing bortezomib and dexamethasone compared with the most relevant comparator (cyclophosphamide, thalidomide and dexamethasone) or an alternative comparator in circumstances when cyclophosphamide, thalidomide and dexamethasone is not suitable. Although the Committee did not request further analyses relating to the bortezomib, thalidomide and dexamethasone regimen, the manufacturer provided an amended model containing some revised assumptions to reflect some of the concerns raised by the ERG and the Committee's considerations in the appraisal consultation document.

3.37 The manufacturer acknowledged that the cyclophosphamide, thalidomide and dexamethasone regimen was the most relevant comparator in UK clinical practice. However, it did not provide a comparison for the bortezomib, thalidomide and dexamethasone regimen with the cyclophosphamide, thalidomide and dexamethasone regimen. It highlighted that for the comparison of the bortezomib, thalidomide and dexamethasone regimen with the thalidomide and dexamethasone regimen presented in the manufacturer's base-case analysis, it was important to evaluate how much additional benefit would be gained in terms of response rates if cyclophosphamide was to be added to thalidomide and dexamethasone. The manufacturer stated that threshold analyses performed in its original submission indicated that the complete response rate for cyclophosphamide, thalidomide and dexamethasone would have to be nearly double that observed in the PETHEMA trial for thalidomide and dexamethasone for the ICER for bortezomib, thalidomide and dexamethasone compared with thalidomide and dexamethasone to reach £30,000 per QALY gained. Therefore the manufacturer stated that the incremental clinical efficacy of cyclophosphamide, thalidomide and dexamethasone would not be substantially greater than thalidomide and dexamethasone.

3.38 The manufacturer updated all the economic models so that:

3.39 For the model comparing the bortezomib, thalidomide and dexamethasone regimen against thalidomide and dexamethasone, the manufacturer provided a revised base-case ICER of £17,841 per QALY gained. Sensitivity analyses using data from alternative sources to inform overall survival were presented; using long-term overall survival data from the Alvares and NMSG 5/94 studies resulted in ICERs of £22,696 and £39,618 per QALY gained respectively. The probabilistic ICER for the manufacturer's revised base-case was £22,289 per QALY gained. The probabilistic ICERs for the sensitivity analyses using Alvares and NSMG5/94 were £22,952 and £39,881 per QALY gained respectively. The manufacturer also carried out sensitivity analyses by fitting parametric curves (exponential, Weibull and log-logistics) to the PETHEMA Kaplan-Meier curves. It selected the parametric functions it thought were most appropriate, providing justification for their suitability based on face validity with the trial data, resulting in a deterministic ICER of £19,359 per QALY gained and a probabilistic ICER of £19,668 per QALY gained. The manufacturer maintained that the most appropriate source to inform overall survival in model was from the MRC Myeloma VII trial.

3.40 For the indirect comparison of bortezomib and dexamethasone compared with cyclophosphamide, thalidomide and dexamethasone, the manufacturer used a 'matching-adjusted indirect comparison' method to account for differences in patients' baseline characteristics between the available trials. This created a new set of post-induction response rates, stem cell transplant rates and the post-transplant response rates for the bortezomib and dexamethasone arm. Using the MRC Myeloma VII trial as the source for long-term survival, the manufacturer's base-case deterministic ICER for the comparison of bortezomib and dexamethasone against cyclophosphamide, thalidomide and dexamethasone was £20,588 per QALY gained. The probabilistic ICER was £22,305 per QALY gained. Using long-term survival data from Alvares and NMSG 5/94 resulted in deterministic ICERs of £24,267 and £33,435 per QALY gained respectively, and the corresponding probabilistic ICERs were £23,816 and £33,107 per QALY gained. The manufacturer presented sensitivity by fitting parametric curves (exponential, Weibull and log-logistics) to the PETHEMA Kaplan-Meier curves (see section 3.39) which resulted in a deterministic ICER of £18,864 per QALY gained and a probabilistic ICER of £19,057 per QALY gained.

3.41 In response to the Committee's request for a comparison of bortezomib and dexamethasone with a relevant comparator when cyclophosphamide, thalidomide and dexamethasone is not suitable, the manufacturer highlighted that the only relevant comparator for which there was direct available evidence was vincristine, doxorubicin and dexamethasone, which might be assumed to be approximately equivalent to cyclophosphamide and dexamethasone or other options lacking thalidomide. Therefore, the manufacturer presented a deterministic base-case ICER for bortezomib and dexamethasone compared with vincristine, thalidomide and dexamethasone (including the model amendments highlighted in section 3.38) of £18,914 per QALY gained, and a probabilistic ICER of £20,096 per QALY gained. Sensitivity analyses were presented for the alternative sources of overall survival using data from the Alvares and NMSG 5/94 studies and fitting parametric curves to the PETHEMA data, which resulted in deterministic ICERs of £25,575, £42,811 and £18,489 per QALY gained respectively. The corresponding probabilistic ICERs were £25,494, £42,528 and £18,761 per QALY gained.

3.42 The ERG noted that the model structure had changed substantially from the original models and that the new approach appeared to be more intuitive. However, it highlighted that the manufacturer had not checked the external validity by validating overall survival against the PETHEMA trial.

3.43 The ERG noted that the manufacturer's analysis of bortezomib, thalidomide and dexamethasone compared with thalidomide and dexamethasone, using the MRC Myeloma VII as the source of overall survival, provided a poor fit for overall survival compared with the observed data in the PETHEMA trial. It stated that the model consistently underestimated overall survival and was systematically biased in favour of the bortezomib, thalidomide and dexamethasone regimen. The ERG noted that the bias appeared even more pronounced in the new analyses. The ERG maintained that the sensitivity analyses using data from the Alvares or NMSG 5/94 trials was a better fit for overall survival than the base-case analysis (using long-term overall survival data from the MRC Myeloma VII trial).

3.44 For the comparison of the bortezomib and dexamethasone regimen with the cyclophosphamide, thalidomide and dexamethasone regimen, the ERG noted that stem cell transplant rates used in the manufacturer's model were 89.1% for the bortezomib arm (taken from the IFM trial) and 66.7% for the cyclophosphamide arm (taken from the MRC Myeloma IX trial). The ERG commented that the stem cell transplant rate for the cyclophosphamide-containing arm was inconsistent with the response data and may have substantially biased the bortezomib and dexamethasone cohort. The ERG explored the impact of assuming that the stem cell transplant rates for the cyclophosphamide-containing arm were similar to the IFM comparator arm (that is, 81.8% for vincristine, doxorubicin and dexamethasone) to better reflect the smaller differences observed between treatment arms in the other bortezomib trials (GIMEMA, HOVON and IFM). This increased the manufacturer's base-case ICER from £20,588 to £36,712 per QALY gained.

3.45 The ERG stated that the manufacturer's analysis of bortezomib and dexamethasone compared with vincristine, doxorubicin and dexamethasone provided a poor fit for overall survival compared with the observed data in the IFM trial. The ERG noted that the manufacturer provided sensitivity analyses using alternative data sources and parametric curves to model overall survival and that it considered that these sensitivity analyses, using the Alvares or NMSG 5/94 trials, were a better fit for overall survival than the base-case analysis (using long term overall survival data from the MRC Myeloma VII trial).

3.46 Full details of all the evidence are in the manufacturer's submission and the ERG report. Further evidence is available in the manufacturer's response to the appraisal consultation document and the ERG critique.