The clinical effectiveness and cost-effectiveness of bortezomib and thalidomide in combination regimens with an alkylating agent and a corticosteroid for the first-line treatment of multiple myeloma: a systematic review and economic evaluation

Prognosis

Myeloma is not curable, but can be treated with a combination of supportive measures and chemotherapy to improve survival and quality of life (QoL). A range of factors affects prognosis. These include factors related to burden of disease [e.g. beta₂-microglobulin (β₂-microglobulin)], characteristics of the myeloma cells’ biology (e.g. the type of cytogenetic abnormality present), the microenvironment surrounding the myeloma cells (e.g. bone marrow microvessel density), patient-related factors (e.g. age and performance status) and treatment response factors [e.g. whether complete response (CR) is achieved with initial therapy]. 5 Because of the number of factors that affect prognosis, survival of patients from the point of diagnosis varies from months to over a decade. 4 In the UK and Ireland, median survival increased from around 2 years in the 1980s and early 1990s to around 4 years in the late 1990s. 11 There is evidence from some cohorts of patients that novel therapies can extend median survival time to 8 years. 12

Epidemiology

Multiple myeloma is the second most common haematological cancer after lymphoma in the UK. In 2007 there were 3357 new diagnoses of MM in England,13 with the highest incidence among those aged 75–79 years (Table 1). In Wales, in the 3 years from 2004 to 2006, an average of 252 new MM diagnoses were recorded. 14 MM is rare before the age of 40 years. The average incidence rates were higher in men than in women, and higher for both sexes in Wales compared with England (Table 2). There are ethnic differences in incidence rates that have been observed in data from the USA; in black people (African American and other black people, but not Hispanic people) the incidence of MM is about twice that of white people, whereas in Asian people the incidence is lower than that of white people. 15 The statistical information team at Cancer Research UK has used incidence and mortality data for 2001–5 to estimate the lifetime risk of developing MM, which is 1 in 148 for men and 1 in 186 for women in the UK. 16 There are currently approximately 10,000–15,000 people living with MM in the UK. 17

The risk factors for developing MM are not well defined but there is evidence for involvement of genetic factors because the first-degree relatives of people with MM are at greater risk of developing MM and related conditions than the first-degree relatives of people without MM. 8,18 Epidemiological studies have looked for evidence of a causal link between a range of potential environmental risk factors and MM but, in general, these have not produced consistent results. 4,8

Diagnosis and staging

Multiple myeloma is typically diagnosed in secondary care using a combination of tests such as urine tests, blood tests, bone marrow examination, imaging, plain radiograph and/or magnetic resonance imaging. If necessary, further tests can be conducted to find out the stage of disease. 1 There are two systems for staging MM. The Durie–Salmon19 (DS) staging system, which has been in use since 1975, is one of the systems but this is gradually being replaced by an updated system, the International Staging System (ISS). 20 This new system is based on measurement of two serum proteins, β₂-microglobulin and albumin (Table 3). A patient with stage I disease will not necessarily proceed linearly through disease stages. Stage III disease can be reached without a requirement to pass through stage II first. It is also noteworthy that staging does not have a significant influence on treatment. If MM is symptomatic, treatment is required irrespective of disease stage.

Place of the interventions in the treatment pathway

In this assessment bortezomib and thalidomide are being considered for use in combination therapy with an alkylating agent and a corticosteroid as a first-line treatment for MM in patients who are not eligible for HDT with autologous SCT.

Study design

• For the systematic review of clinical effectiveness, randomised controlled trials (RCTs) were eligible for inclusion. In addition, evidence from good-quality observational studies was also eligible for consideration if the data from available RCTs were incomplete (e.g. absence of data on outcomes of interest).

• For the systematic review of cost-effectiveness economic evaluations (such as cost-effectiveness studies, cost–utility studies, cost–benefit studies) were eligible for inclusion.

• Abstracts or conference presentations of studies were eligible for inclusion only if sufficient details were presented to allow an appraisal of the methodology and the assessment of results to be undertaken.

• Case series, case studies, narrative reviews, editorials and opinions were excluded, as were non-English-language studies. Systematic reviews and clinical guidelines were used only as a source of references.

Intervention(s)

• Bortezomib in combination with an alkylating agent and a corticosteroid for first-line treatment of MM.

• Thalidomide in combination with an alkylating agent and a corticosteroid for first-line treatment of MM.

• Studies of treatment with either bortezomib or thalidomide as a single agent were excluded.

Comparator(s)

• Interventions described above compared with each other.

• Melphalan or cyclophosphamide in combination with prednisolone/prednisone or dexamethasone.

• Other chemotherapy regimens or SCT were excluded.

Population

• People with previously untreated MM who are not candidates for HDT with SCT.

• Studies of MM patients who had received previous treatment(s) were excluded.

Outcomes

Studies were included if they reported on one or more of the following outcomes:

• overall survival

• progression-free survival (deaths counted as events)

• time to progression (deaths are excluded from the calculation of this outcome)

• response rates

• health-related quality of life

• cost-effectiveness [such as incremental cost per quality-adjusted life-year (QALY) gained]

• adverse events of treatment were reported when available within the trials that met the inclusion criteria.

Response definitions

• Response to treatment is usually assessed based on changes in serum levels of M-protein and/or urinary light chain excretion. Two different systems for categorising response are included in this report, the European Group for Blood and Marrow Transplantation (EBMT) criteria39 and the Intergroupe Francophone du Myélome (IFM) criteria. 23 Where there are differences in the two systems, in general the EBMT criteria require a slightly greater improvement. For example, in the definition of partial response (PR) one of the IFM requirements is more than a 75% reduction in 24-hour urinary light chain excretion, whereas one of the EBMT criteria for PR is a 90% decrease in urinary light chain excretion. The EBMT and IFM criteria for judging response are provided in Appendix 3.

Adverse event definitions

• Two slightly different National Cancer Institute (NCI) criteria have been used to grade AEs, the NCI Common Terminology Criteria for Adverse Events (CTCAE) version 4, and the NCI Common Toxicity Criteria (CTC) version 2. The NCI CTCAE version 4 grades AEs on a five-point scale (1–5) and the NCI CTC version 2 grades AEs on a six-point scale, as 0 is included (0 = no AE or within normal limits). Events of a higher grade are more serious than those at a lower grade, with a grade 1 event described as ‘mild’, grade 2 ‘moderate’, a grade 3 event would be considered ‘severe’, while a grade 4 event could be ‘life threatening’. Grade 5 is reserved for deaths related to an AE.

Quantity and quality of research available

Titles and, where available, abstracts of a total of 1436 records were screened and full copies of 40 references were retrieved. Of these, six were excluded after inspection of the full article (see Appendix 4). Two of these articles were excluded because they were not clinical trial reports, two were abstracts that were excluded because they described maintenance therapy with thalidomide, another abstract was excluded because it did not report on any of the outcomes of interest, and a sixth abstract described a systematic review with meta-analysis. Five full papers described four RCTs that met the inclusion criteria of the review (Figure 1 and Table 4). Each RCT was described by at least one full paper, with linked abstracts also being available. As the full papers provided the most complete data these were the primary source of information for the review.

FIGURE 1.

Flow diagram of reference screening processes. a, Additional information was received from the trialists providing details for one RCT only described in conference abstracts. The additional details allowed us to appraise the study methodology and make judgements about study quality. Results from this RCT could therefore be considered for inclusion in the systematic review of clinical effectiveness. b, Outcomes from these studies could not be included because of insufficient details about study methodology and insufficient details about study quality. These studies, which are all ongoing, are briefly summarised below (see Ongoing studies).

One ongoing RCT, the Myeloma IX (MMIX) Trial, which is a UK-based MRC collaborative RCT with two treatment pathways, appeared to meet the inclusion criteria of the review and was described in conference abstracts. The search for studies of clinical effectiveness identified four abstracts for this RCT,42–45 with a further three abstracts identified in the MSs. 46–48 Four of the abstracts,42–44,47 were excluded because they described the intensive pathway or thalidomide maintenance treatment that did not meet the inclusion criteria for this review. Three abstracts45,46,48 described the non-intensive pathway that met the inclusion criteria. The final results from the 3-year median follow-up have not yet been published. However, because of this RCT’s potential relevance to our inclusion criteria, the Clinical Trials Research Unit (CTRU) at the University of Leeds, who are coordinating the RCT, provided the trial protocol,49 additional background information50,51 and trial baseline data,52 and have also made the results from the non-intensive treatment pathway53–58 available to NICE and the authors of this report in academic confidence. As the trial protocol and results provided directly by the CTRU provided the most complete and up-to-date data these were used as the primary source of information for the review.

Four additional ongoing RCTs were described in conference abstracts but it was unclear whether these met the inclusion criteria for this review. These ‘unclear’ studies are briefly described later in this chapter (see Ongoing studies). The total number of records assessed at each stage of the systematic review screening process is shown in the flow chart in Figure 1.

One of the included RCTs evaluated VMP (VISTA trial),26 while three RCTs evaluated MPT [IFM 01/01 trial,59 IFM 99/06 trial,23 and the Gruppo Italiano Malattie Ematologiche dell’Adulto (GIMEMA – Italian Group for Adult Hematologic Diseases) trial24]. The fifth RCT, the MMIX Trial (non-intensive pathway), evaluated CTDa. The comparator in all five of the included RCTs was MP.

Quality assessment of included studies

The outcome of the quality assessment of included RCTs is summarised in Table 5.

Assessment of effectiveness

Overall survival

Overall survival was a secondary outcome in the VISTA RCT of VMP versus MP (Table 6) and was calculated from randomisation. A statistically significant survival benefit for VMP compared with MP is reported in an abstract60 after a median follow-up of 25.9 months (HR = 0.64, p = 0.0032). Three-year survival rates in a more recent abstract61 after a median follow-up of 36.7 months were 68.5% versus 54%, respectively. At the earlier median follow-up of 16.3 months, reported in the published paper,26 median OS had not been reached. However, San Miguel and colleagues26,60 stated that a survival benefit was associated with bortezomib because 45 patients (13%) in the VMP group had died in comparison with 76 patients (22%) in the MP group (HR 0.61, p = 0.008) (despite 43%60 of MP patients receiving subsequent bortezomib therapy after disease progression – see Table 21). The most recent abstract reports that median OS is 43.1 months in the MP group but not estimable in the VMP group. 61

TABLE 6 - Overall survival

At the initial analysis (median follow-up 36.8 months) no difference in OS was recorded as a function of initial thalidomide dose (≤ 200 mg per day vs > 200 mg per day, p = 0.93).

Myeloma progression was considered to be the major cause of the majority of deaths in both study arms (36/58 deaths in the MPT group, 54/76 deaths in the MP + placebo group).

Study	Median follow-up (months)	Treatment arms		HR and p-value
San Miguel et al.26 VISTA
		VMP (n = 344)	MP (n = 338)
OS (abstract61)	36.7	Not estimable	43.1 months	0.653; 0.0008
OS (abstract60)	25.9	NR	NR	0.64; 0.0032
OS26	16.3	Median survival not reached	Median survival not reached	0.61; 0.008
Deaths26		45/344 (13%)	76/338 (22%)
Three-year survival rate (abstract61): %		68.5	54.0	NR
Three-year survival rate (abstract60): %		72	59	NR
Facon et al.23 IFM 99/06
		MPT (n = 125)	MP (n = 196)
OS:a median (SE, IQR)	51.5 (IQR 34.4–63.2)	51.6 (4.5, 26.6 to ‘not reached’) months	33.2 (3.2, 13.8 to 54.8) months	0.59 (95% CI 0.46 to 0.81); 0.0006
Deaths		62/125 (50%)	128/196 (65%)
Hulin et al.59 IFM 01/01
		MPT (n = 113)	MP + placebo (n = 116)
OS: median (95% CI)	47.5	44.0 (33.4 to 58.7) months	29.1 (26.4 to 34.9) months	0.68 (95% CI not reported); 0.028
Deathsb		58/113 (51%)	76/116 (65.5%)	0.03

Two23,59 of the three RCTs investigating MPT versus MP alone reported OS as their primary outcome. Both RCTs calculated OS from randomisation but only one of them, Hulin and colleagues,59 explained that data on patients who were alive at the time of analysis were censored in the survival analysis on the last date they were known to be alive. For the third RCT,24 OS was a secondary outcome and was not eligible for inclusion in this systematic review because participants received maintenance therapy with thalidomide after the six 4-week cycles of MPT were completed.

A statistically significant difference in OS in favour of the MPT group was found by both RCTs (see Table 6). Facon and colleagues23 reported their results after a median follow-up of 51.5 months. In the MPT group there were 62 events (deaths) and median survival was 51.6 months [interquartile range (IQR) 26.6 to ‘not reached’], whereas in the MP group, where there were 128 events, median survival was 33.2 months (IQR 13.8–54.8). The difference in OS was statistically significant, with an estimated HR for median OS in favour of MPT of 0.59 (95% CI 0.46 to 0.81, p = 0.0006). When adjusting for prognostic factors (e.g. WHO performance index; β₂-microglobulin, albumin, etc.) the results showed that MPT remained the superior treatment in terms of the specified outcome OS (HR 0.49, 95% CI 0.33 to 0.73, p = 0.0002) (see Appendix 5). Similarly Hulin and colleagues,59 reporting after a slightly shorter median follow-up of 47.5 months, found that the median survival of 44 months (95% CI 33.4 to 58.7) in the MPT group was statistically significantly longer than in the MP + placebo group where median survival was 29.1 months (95% CI 26.4 to 34.9). In this RCT the reported HR for median OS in favour of MPT was 0.68 (95% CI for the HR not reported, p = 0.028).

As noted above, neither RCT reported on the amount of censored data, or the reasons for this. It is therefore not possible to determine whether censored data had any impact on the outcome of OS.

The MMIX RCT49 OS outcome was not eligible for inclusion in this systematic review because participants were entered into a second randomisation to receive either maintenance therapy with thalidomide or no maintenance therapy after they had completed first-line treatment with either CTDa or MP.

Two MPT versus MP RCTs23,59 reported OS outcome data that met the inclusion criteria of the review. A fixed-effects meta-analysis was conducted and, as can be seen in Figure 2, the I²-test suggests there is little or no heterogeneity between the two RCTs for this outcome. The summary OS HR was 0.62 (95% CI 0.50 to 0.77) in favour of MPT.

FIGURE 2.

Melphalan, prednisolone/prednisone plus thalidomide vs MP OS.

The Facon study23 CIs shown in Figure 2 obtained from Review Manager are slightly different to those reported by the published paper and shown in Table 6. The difference arises from the estimating method41 used to obtain the SEs for the log-HRs needed to undertake the meta-analysis.

Response to treatment

Various response to treatment rates are reported as secondary outcomes in the VISTA RCT of VMP (Table 8),26 although the analysis is not ITT as previously explained. The time at which response was assessed is not reported. Rates of PR or better (according to EBMT criteria, Appendix 3) were 71% in the VMP group and 35% in the MP group (p < 0.001), and the CR rates were 30% and 4%, respectively (p < 0.001). The rate of PR was 40% in the VMP group and 31% in the MP group and minimal response (MR) rates were 9% and 22%, respectively. Stable disease rates were 18% in the VMP group and 40% in the MP group, and progressive disease rates were 1% and 2%, respectively.

TABLE 8 - Response to treatment

Study	Treatment arms		p-value
San Miguel et al.26 VISTA
	VMP (n = 344): n/N (%)	MP (n = 338): n/N (%)
Rate of PR or better	238/337 (71)	115/331 (35)	< 0.001
Rate of CR	102/337 (30)	12/331 (4)	< 0.001
Rate of PR	136/337 (40)	103/331 (31)	NR
MR	32/337 (9)	72/331 (22)	NR
Stable disease	60/337 (18)	113/331 (40)	NR
Progressive disease	3/337 (1)	7/331 (2)	NR
Facon et al.23 IFM 99/06
	MPT (n = 125): n/N (%)	MP (n = 196): n/N (%)
CR at 12 months	10/75 (13)	4/165 (2)	0.0008
At least PR at 12 months	57/75 (76)	57/165 (35)	< 0.0001
At least very good PR at 12 months	35/75 (47)	11/165 (7)	< 0.0001
Hulin et al.59 IFM 01/01
	MPT (n = 115): n/N (%)	MP + placebo (n = 117): n/N (%)
CR at 12 months	7/107 (7)	1/112 (1)	< 0.001
At least PR at 12 months	66/107 (62)	35/112 (31)	< 0.001
At least very good PR at 12 months	23/107 (21)	8/112 (7)	< 0.001
Palumbo et al.24 GIMEMA
	MPT (n = 167, 129 analysed): n/N (%)	MP (n = 164, 126 analysed): n/N (%)	Absolute difference MPT–MP: % (95% CI)
CR at 6 months	20/129 (15.5)	3/126 (2.4)	13.1 (6.3 to 20.5)
Complete or PR at 6 months	98/129 (76.0)	60/126 (47.6)	28.3 (16.5 to 39.1)
PR	78/129 (60.4)	57/126 (45.2)	15.2 (3.0 to 26.9)
Near CR	16/129 (12.4)	6/126 (4.8)	NR
90–99% M-protein reduction	11/129 (8.5)	6/126 (4.8)	NR
50–89% M-protein reduction	51/129 (39.5)	45/126 (35.7)	NR
MR at 6 months	7/129 (5.4)	21/126 (16.7)	–11.2 (–19.2 to –3.6)
No response at 6 months	7/129 (5.4)	19/126 (15.1)	–9.7 (–17.4 to –2.2)
Progressive disease at 6 months	10/129 (7.8)	21/126 (16.7)	–8.9 (–17.2 to –0.8)
Not available	7/129 (5.4)	5/126 (4.0)	NR
MMIX49,53,54
	CTDa (AiC/CiC information has been removed)	MP (AiC/CiC information has been removed)	p-value
(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)
(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)
(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)
(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)
(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)
(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)
(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)
(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)
(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)
(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)

All three RCTs investigating MPT reported on response to treatment (see Table 8). 23,24,59 The two IFM RCTs23,59 reported the response at 12 months as a secondary outcome, and response was judged according to their own criteria. These criteria are very similar, but not identical, to the EBMT criteria that were used in the RCT by Palumbo and colleagues24 to assess response at 6 months, which was the primary outcome of this RCT (see Appendix 5). Facon and colleagues23 stated that all analyses were undertaken on an ITT basis; it is therefore unclear why response to treatment outcomes are reported for only 60% of the MPT group (75 of the 125 participants enrolled to this group) and 84% of the MP group (165 of 196 enrolled). Hulin and colleagues59 did not indicate that all analyses were ITT (only survival analyses were clearly stated to be ITT) but response to treatment is reported for 93% of the MPT group and 96% of the MP + placebo group. Palumbo and colleagues24 reported on all of those who contributed to the 6-month follow-up results. However, as noted earlier, not all of the randomised participants contributed data to this outcome because some participants had not achieved 6 months of follow-up when these data were analysed.

At 12 months, statistically significant differences in CR in favour of the MPT group were observed in both the IFM RCTs. 23,59 Facon and colleagues23 reported 13% of 75 participants in the MPT group had achieved CR at 12 months in comparison with just 2% of 165 participants in the MP group (p = 0.008). Caution must be applied in interpreting these results, however, which appear to be based on a small proportion of the participants. The difference between the groups reported by Hulin and colleagues59 was less marked but still statistically significant (MPT 7% of 107 participants CR versus 1% of 112 participants in MP + placebo group, p < 0.001). Palumbo and colleagues24 reported an absolute difference in CR MPT–MP at 6 months of 13% (95% CI 6.3 to 20.5).

When response categories were combined, the percentage of participants in the IFM RCT MPT groups achieving at least a PR at 12 months was double the percentage achieving this level of response in the MP group (Facon and colleagues,23 MPT 76% vs MP 35%, p < 0.0001; Hulin and colleagues,59 MPT 62% vs MP + placebo 31%, p < 0.001). At 6 months in the RCT by Palumbo and colleagues24 there was a difference in favour of the MPT group of 28.3% (95% CI 16.5 to 39.1) for participants achieving either a CR or PR.

Each MPT RCT reported on a subcategory of participants with a PR. In the two IFM RCTs, only one subcategory of participants was reported on who were described as having a very good PR. 23,59 These participants had more than a 90% decrease in monoclonal protein in serum and urine. Palumbo and colleagues24 reported on three subcategories of participants with PR. 24 Those with a near CR had disappearance of M-protein from serum and urine but still detectable by immunofixation (immunofixation positive); the remaining participants with a PR were divided into those with a 90–99% M-protein reduction and those with a 50–89% M-protein reduction. Facon and colleagues23 reported at least a very good PR at 12 months in 35 out of 75 participants (47%), which was statistically significantly better than in the MP group where only 7% (11/165) of participants met the criteria (p < 0.001). Hulin and colleagues59 also reported a statistically significant difference in favour of the MPT group at 12 months when 21% (23/107) met the criteria for at least a very good PR, in comparison to 7% (8/112) in the MP group (p < 0.001). Palumbo and colleagues24 report greater proportions of participants in the MPT group than in the MP group at each subcategory of PR after 6 months of follow-up. Of the 78 participants (60.4%) in the MPT group with a PR most (n = 51) had experienced a 50–89% M-protein reduction, 11 participants had a 90–99% M-protein reduction, and 16 participants had a near CR. In contrast, only 57 (45.2%) of MP group participants achieved a PR with the majority (n = 45) having a 50–89% M-protein reduction, six participants having a 90–99% M-protein reduction, and six participants achieving a near CR.

Facon and colleagues23 and Hulin and colleagues59 gave no details about participants who achieved less than a PR at 12 months. Palumbo and colleagues,24 however, provided information on each of the remaining three EBMT categories – minimal response at 6 months, no response at 6 months, and progressive disease at 6 months – as well as indicating the proportion of data that was not available (see Table 8). There were greater proportions of participants from the MP group than the MPT group in the final three categories.

The MMIX RCT49,53,54 assessed maximal response after induction chemotherapy with either CTDa or MP. Response was categorised using EBMT definitions (see Appendix 3). Response was one of the three co-primary outcomes of this RCT. (AiC/CiC information has been removed) (Table 8).

Risk ratios for complete response

Three MPT versus MP RCTs23,24,59 reported CR outcome data that could be meta-analysed. A fixed-effects meta-analysis was conducted and, as can be seen from Figure 3, the I²-test suggests that there is little or no heterogeneity between the three RCTs for this outcome. The outcome is reported as a risk ratio (RR) because a summary relative risk was required for the cost-effectiveness model (see Chapter 5, SHTAC data sources, Complete response). The Facon results were entered using the original group sizes to generate a conservative estimate of overall treatment effect for use in the cost-effectiveness model. The overall effect for the outcome of CR favours MPT (RR 5.49, 95% CI 2.55 to 11.83).

FIGURE 3.

Melphalan, prednisolone/prednisone plus thalidomide vs MP CR.

Risk ratios for the outcome of CR were also obtained for the single RCTs for the VMP versus MP comparison and the CTDa versus MP comparison using the data reported in Table 8 and Review Manager software. These RRs were needed for the cost-effectiveness model [CR VMP vs MP RR 8.35, 95% CI 4.68 to 14.89, (AiC/CiC information has been removed)].

Progression-free survival

Progression-free survival in the VISTA RCT of VMP was defined by San Miguel and colleagues26 as the time between randomisation and either disease progression or relapse from CR by EBMT criteria, or death due to any cause, whichever occurred first. Median PFS by investigator assessment based on central laboratory data and applying EBMT criteria was 21.7 months in the VMP group and 15.2 months in the MP group (HR 0.56, p < 0.001) – see Table 10.

TABLE 10 - Progression-free survival

Median follow-up not explicitly stated, assumed to be the same as that for OS.

The Janssen–Cilag MS reported different PFS values and Janssen–Cilag informed SHTAC that the published value is incorrect although the reason why the published value is incorrect was not provided.

At the initial analysis (median follow-up 36.8 months) no difference in PFS was recorded as a function of initial, maximum or average thalidomide doses (p = 0.22, p = 0.75, p = 0.92, respectively).

Study	Median follow-up (months)	Treatment arms		HR and p-value
San Miguel et al.26 VISTA
		VMP (n = 344)	MP (n = 338)
PFS (months): median	16.3a	21.7	15.2	0.56; < 0.001
PFS (months): median (reported in Janssen–Cilag submission to NICE)b	16.3	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)
Facon et al.23 IFM 99/06
		MPT (n = 125)	MP (n = 196)
PFS (months): median (SE)	51.5	27.5 (2.1)	17.8 (1.4)	0.51 (95% CI 0.39 to 0.66); 0.0001c
Hulin et al.59 IFM 01/01
		MPT (n = 113)	MP + placebo (n = 116)
PFS (months): median (95% CI)	47.5	24.1 (19.4 to 29.0)	18.5 (14.6 to 21.3)	0.62; 0.001

The two included IFM RCTs23,59 reported on PFS and both calculated PFS from randomisation to either progression, or death without progression (Table 10). Hulin and colleagues59 censored data on patients who had not experienced progression to the last day that they were known to be alive and progression free. Facon and colleagues23 did not comment on methods for censoring data.

After a median follow-up of 51.5 months, 92 of the 125 participants in the MPT group of the Facon and colleagues23 RCT had either experienced disease progression or they had died. The median PFS of the MPT group was 27.5 months (SE 2.1). In comparison, in the MP group 171 of 196 participants had disease progression or had died and the median PFS was 17.8 months (SE 1.4). The difference in PFS was statistically significant (p = 0.001) with a HR for median PFS in favour of MPT of 0.51 (95% CI 0.39 to 0.66).

Hulin and colleagues59 also found that the difference in PFS between MPT and MP + placebo groups after a median follow-up of 47.5 months was statistically significant with a HR of 0.62 (p = 0.001). In the MPT group median PFS was 24.1 months (95% CI 19.4 to 29.0) in comparison to 18.5 months (95% CI 14.6 to 21.3) in the MP + placebo group.

The event-free24 and progression-free49 survival outcomes reported by Palumbo and colleagues24 and the MMIX RCT49 were not eligible for inclusion in this systematic review because participants received maintenance therapy with thalidomide after first-line treatment had been completed.

Two MPT versus MP RCTs23,59 reported PFS outcome data that was included in a fixed-effects meta-analysis. As can be seen in Figure 4, the I²-test suggests there is little or no heterogeneity between the two RCTs for this outcome. The summary PFS HR was 0.56 (95% CI 0.46 to 0.67) in favour of MPT.

FIGURE 4.

Melphalan, prednisolone/prednisone plus thalidomide vs MP PFS.

The Facon study CIs shown in Figure 4 obtained from Review Manager are slightly different to those reported by the published paper and shown in Table 10. The difference arises from the estimating method used to obtain the SEs for the log-HRs needed to undertake the meta-analysis.

VMP versus MP alone

• The quality of the RCT26 was difficult to determine. Risk of allocation bias and of unbalanced confounding factors could not be judged because details on these aspects were not reported. Most, but not all, analyses had followed the ITT principle but the methods used to account for any missing data were not described. It was not possible to determine whether the amount and pattern of censored data was similar between trial arms.

• Time to disease progression was the primary outcome of the RCT and a statistically significant effect in favour of the VMP group was reported.

• Overall survival was a secondary outcome. A survival advantage for the VMP arm in comparison with the comparator MP was reported.

• Statistically significantly more participants in the VMP group achieved CR, or achieved a PR or better. This outcome was not analysed by ITT principles.

• Median PFS was statistically significantly longer in the VMP group than in the MP group.

• Limited data on HRQoL was available. This indicated that, after the onset of best response, participants treated with VMP had a higher sustained HRQoL improvement rate in 14 of the 15 EORTC QLQ-C30 scores than those participants receiving therapy with MP.

• Adverse events occurred in both trial arms. Although the occurrence of any AE and any grade 4 AE was similar in the two groups, there was a statistically significant increase in grade 3 AEs in the VMP group.

• Subgroup analyses were conducted. The RCT may not have been powered for these analyses so the results, which indicate that the reported benefits of bortezomib for TTP apply to each of the seven subgroups of participants assessed, should be interpreted with caution.

MPT versus MP alone

• The quality of the three RCTs23,24,59 was variable. Risk of allocation bias could not be judged for one RCT, and the risk of allocation bias and of unbalanced confounding factors could not be judged for another RCT because the necessary details were not reported. Although all RCTs stated that ITT analyses had been conducted, the details of these analyses and the methods used to account for missing data were, in general, poorly described. It was not possible to determine whether the amount and pattern of censored data was similar between trial arms for any of the included RCTs.

• Overall survival was a primary outcome in two of the RCTs. 23,59 Both reported a survival advantage for the MPT arm in comparison with the comparator MP alone. As expected, meta-analysis of the OS data from two RCTs of MPT versus MP provided a HR in favour of MPT for the OS outcome. The third RCT of MPT included maintenance therapy with thalidomide and therefore OS, which was a secondary outcome of this RCT, was not eligible for inclusion in the review.

• Response to treatment was the primary outcome of one RCT24 (at a 6-month time point) and a secondary outcome in two RCTs23,59 (both at a 12-month time point). At 6 months more participants in the MPT group achieved CR or achieved a PR or better, but a p-value for the comparison is not reported. At 12 months, two RCTs reported that a statistically significant greater proportion of participants had achieved CR or had achieved at least a PR. However, it was noteworthy that in one of these RCTs the numbers of participants contributing data to this outcome was low. Outcomes for CR from three RCTs were combined by meta-analysis, which confirmed that MPT was superior in comparison with MP in terms of the proportion of patients achieving CR.

• Two RCTs23,59 reported a statistically significant advantage in the MPT group in comparison with the MP group for the outcome of PFS. The PFS data were combined by meta-analysis, which confirmed that MPT was superior in comparison with MP for this outcome.

• Adverse events were reported in different ways so it was difficult to summarise the results across the RCTs. Because one RCT had included maintenance therapy with thalidomide, few AE data could be included, so the majority of the data come from just two RCTs. AEs with a statistically significant greater occurrence in the MPT arm that was reported by two RCTs included neutropenia and peripheral neuropathy. One RCT found that overall non-haematological toxic effects were statistically significantly more likely in the MPT group. For the outcomes of thrombosis or embolism, somnolence, constipation and infections, the results were inconsistent between RCTs (no significant difference in incidence reported by one RCT, but statistically significantly more in the MPT arm reported by the other RCT). This inconsistency may be a consequence of the different methods of reporting AEs. Some outcomes were only reported by one RCT, such as anaemia and thrombocytopenia (no statistically significant differences).

CTDa versus MP

• This RCT49 was judged to be at low risk from allocation bias and bias due to unbalanced confounding factors. Analyses had been conducted by ITT principles and some information was provided on the methods used to handle missing data. It was not reported whether the amount and pattern of censoring was comparable between the groups.

• Response was one of three co-primary outcomes of this RCT. (AiC/CiC information has been removed.) The remaining two co-primary outcomes, OS and PFS, and also the HRQoL outcomes, were not eligible for inclusion because participants were randomised to maintenance therapy with thalidomide after induction chemotherapy and this treatment did not meet the inclusion criteria of the systematic review.

• Adverse events occurred in both RCT arms. (AiC/CiC information has been removed.)

• Subgroup analyses were conducted although numerical data were not presented. (AiC/CiC information has been removed.)

Methods for the systematic review of cost-effectiveness

A systematic literature search was undertaken to identify economic evaluations for first-line treatment with either bortezomib or thalidomide in combination with an alkylating agent and a corticosteroid in people with MM, who are ineligible for HDT with SCT, compared with existing treatments. The details of the search strategy and the methods for the systematic review of cost-effectiveness studies are outlined in Chapter 3 and Appendix 1.

Results of the systematic review of cost-effectiveness

Searches for economic evaluations identified the titles and abstracts of 183 potentially relevant studies. The full text of seven papers was retrieved for further consideration, with none of the studies meeting the a priori inclusion criteria. A summary of the selection process and the reasons for exclusion are presented in Figure 5 and a list of excluded studies is presented in Appendix 8. Two studies were excluded as they assessed a different intervention and/or population group from that specified in the research protocol. 70,71 Although five studies reported as abstracts appeared to meet the a priori inclusion criteria,72–76 they did not contain sufficient information on the methods used and the results to justify formal data extraction or critical appraisal. Given the apparent relevance of these five studies, a brief summary of the abstracts is presented below.

FIGURE 5.

Flow chart of identification of studies for inclusion in the review of cost-effectiveness. a, The five abstracts provided insufficient details of methods and results to allow inclusion in a formal systematic review. However, as the abstracts met other inclusion criteria they are discussed for information.

Deniz and colleagues72 estimated the lifetime health and cost consequences of MPT compared with MP in people in Scotland with previously untreated MM. They developed a Markov model for a cohort of patients receiving a course of MPT or MP, conceptualising the disease by four health states: preprogression without AE, preprogression with AE, progressive disease and death. Progression between health states as well as treatment duration, dose and AE risks were derived from a long-term RCT (see Chapter 4, Results of the systematic review of clinical effectiveness). 23 Patient cohorts received a maximum of 12 six-week cycles of treatment, until progression or treatment-limiting toxicity. The abstract indicates that health-state utilities associated with disease states and AEs were obtained from the literature, but no sources are provided. Thalidomide costs were from UK list prices and routine disease management costs reflected current practice in Scotland. Costs and health outcomes were discounted at 3.5% per annum. The model estimated improvements in health outcomes with MPT with a median TTP of 25 months versus 12 months with MP. Estimated median OS was 4.03 years with MPT versus 2.88 years with MP. These translated to a gain of 0.91 QALYs for MPT (3.24 QALYs) compared with MP (2.32 QALYs). There were increased costs with MPT of £25,199 per patient compared with £8935 per patient for MP, leading to an incremental cost-effectiveness ratio (ICER) of £17,847 per QALY and £14,803 per life-year gained. The authors state that sensitivity analyses showed that these results were consistent through changes in model parameters, although no information is presented. The authors conclude that MPT improves PFS and OS compared with MP and the results are cost-effective. A similar study comparing lifetime health and cost consequences of MPT compared with MP was completed for untreated MM patients in Wales. 74 While this evaluation used the same clinical outcomes for OS and PFS, it used slightly different QALY gains (0.9 QALYs) and lifetime costs specific to managing the disease in Wales (£16,937 per patient for MPT vs £1524 per patient for MP). The study produced a slightly more favourable ICER of £17,002 per QALY and £13,346 per life-year gained. It was reported that sensitivity analyses showed that findings were robust, with 95% of outcomes between £12,750 and £26,500 per QALY gained. Both studies were funded by the manufacturer of thalidomide.

De Abreu Lourenco and colleagues75 assessed whether MPT was cost-effective compared with MP for people in Australia who had been newly diagnosed with MM as part of an application to the Pharmaceutical Benefits Advisory Committee (PBAC). They extrapolate Kaplan–Meier survival curves from an unspecified Phase III study to a lifetime horizon to estimate the mean survival time. Costs included drugs, medical services and treatment for thalidomide-related AEs. These data were incorporated into a cost-effectiveness analysis adopting an Australian health-care system perspective, with costs and benefits discounted at 5% (AUS$2008). The modelled analysis estimated an incremental gain in average survival of 1.47 years and 1.14 QALYs with an associated average incremental cost of AUS$23,953. This results in an ICER of AUS$20,998. The authors concluded that the analysis had resulted in a positive recommendation from PBAC to fund thalidomide for the treatment of patients newly diagnosed with myeloma.

Yoong and colleagues73 estimated the cost-effectiveness of bortezomib in combination with MP (VMP) compared with MP and MPT in previously untreated people with MM in Canada who are unsuitable for SCT. Clinical outcomes originated from the VISTA study26 for VMP compared with MP and from an unspecified indirect comparison of VMP and MPT. The economic model projected OS over a 10-year horizon for VMP, MP and MPT using data from relevant studies and survival HRs. Resource use data included costs of drugs, outpatient cancer clinic, management of AEs, supportive care and subsequent lines of treatment, although sources were not specified. The discounted QALYs were 3.51 for VMP, 2.84 for MP and 3.29 for MPT. The total cost of treatment per patient was CAN$59,117 for VMP, CAN$27,026 for MP and CAN$52,226 for MPT. The ICER for VMP versus MP was CAN$48,294 per QALY gained, and it was CAN$31,975 per QALY gained for VMP versus MPT. The study states that sensitivity analyses showed that survival difference was the most influential factor. The authors concluded that the VMP regimen indicates good value for money, and it is being adopted by public cancer agencies in Canada.

Wang and colleagues76 also compared the cost-effectiveness of VMP, MPT and MP as first-line therapy for people with MM in the USA who were ineligible for autologous SCT. A lifetime (20 years) Markov model from the US payer’s perspective was developed with seven health states respresenting periods of treatment response (stable disease/MR, PR or CR), treatment-free interval, progressive disease, second-line treatment and death. Monthly transition probabilities were estimated from the VISTA trial data for VMP and MP,26 and from the IFM 99/06 trial for MPT. 23 Costs included drug and medical costs, treatment-related AEs, second-line treatment and resource utilisation during treatment-free intervals and progressive disease. All costs were adjusted to 2009 and presented in US dollars. State-specific utility estimates were derived from patient-level European Quality of Life-5 Dimensions (EQ-5D) data from the VISTA RCT. 26 Cost and health outcomes were discounted at 3%. The discounted QALY was 2.99 for VMP, 2.09 for MP and 2.95 for MPT. The total costs were US$110,870 for VMP, US$57,864 for MP and US$129,902 for MPT. The ICER of VMP versus MP was US$56,109 per QALY gained. VMP was dominant compared with MPT (greater benefit and lower cost). One-way sensitivity analyses were reported to show that the ICERs were robust, with the key drivers being the HR for VMP versus MP for the transition between second-line treatment and the HRs for MPT versus MP for treatment discontinuation. The authors concluded that VMP is cost-effective compared with MP in the USA.

Summary and conclusions of the health-related quality-of-life review

The systematic review did not find any generic preference-based HRQoL studies that were directly related to the population of interest. The utility estimates from HRQoL studies in patients with MM who had intensive therapy vary between 0.38 and 0.69, with the lower utility estimates during treatment periods or immediately after discharge from treatment. 96 The longer-term HRQoL estimates after discharge from treatment range from 0.64 to 0.69. This may indicate that HRQoL is lower during the treatment period and improves after treatment has finished. Furthermore, long term HRQoL may be stable over time. It is unclear whether patients with CR following treatment have a higher HRQoL than those with other responses.

Modelling approach

A decision-analytic cost–utility model, developed in Microsoft Excel, was used in this submission. The model uses a cohort of newly diagnosed myeloma patients treated with MP as the baseline treatment. Treatment effects for VMP, MPT and CTDa are then modelled over time by adjusting the baseline patient experience via HRs. A survival model appears to be used, which estimates OS and PFS curves for each of the comparators. The model also includes further lines of treatment (second- and third-line) to estimate the total treatment costs.

The analytic framework was based on a variant of Quality-Adjusted Analysis of Time Without Symptoms or Toxicity (Q-TWiST97) using partitioned survival analysis, and utilises the area under, and the difference between, time-to-event curves to estimate mean durations spent within the disease states of interest.

Survival is partitioned into three different states: (1) prior to response to treatment; (2) response but no progression; and (3) post progression. Death represents the final state. The time to response or death was estimated from life tables constructed directly from the VISTA trial patient-level data. 26 PFS for MP was estimated from a meta-analysis of the MP arms of included RCTs to compute MP PFS values at 6, 12, 18 and 24 months. PFS was extrapolated beyond 24 months, assuming an exponential survival distribution, using the hazard rate for all time periods beyond 24 months equal to the hazard rate calculated between months 18 and 24. OS for MP was estimated in a similar way to PFS, but using 48 months of summary survival data from the MP arms of the included RCTs.

For the comparator treatments, relative HRs were taken from the random effects results of the meta-analysis that used OS and PFS summary data. OS and PFS hazard rates were computed for each time period by multiplying the VMP–MP, MPT–MP and CTDa–MP HRs by the appropriate hazard rate for that time period. The computed hazard rates were then used to generate the VMP, MPT and CTDa OS and PFS life tables that extend out to the end of the 30-year lifetime horizon of the model.

The HRs were estimated using a piece-wise constant hazard model using derived survival data from the Kaplan–Meier curves for each of the included RCTs. HRs were estimated at 48 months for OS for each of the RCTs, except the VISTA Trial, which had only 36 months’ follow-up. For estimation of the OS hazard for thalidomide, data from five RCTs were used, which included RCTs that had included thalidomide maintenance. Data were synthesised using Bayesian meta-analysis with fixed and/or random effects models. Results from the random effects model were used in the cost-effectiveness model. (AiC/CiC information has been removed.)

Following the first-line therapy, and upon disease progression, it was assumed that the second-line treatment would consist of bortezomib plus high-dose dexamethasone (HDD), CTDa or HDD. Most patients received CTDa after first-line VMP and bortezomib and HDD for all other first-line therapies. All patients received lenalidomide plus dexamethasone as third-line treatment.

Adverse events were included in the analysis by estimating the incidence of AEs (grades 3 and 4) across the RCTs for each of the comparators and combining this with the unit costs of treating the AEs. Unit costs were mostly based upon those used in a previous NICE report for lenalidomide (TA171). 31 The most common AEs for MPT were non-haematological toxicity, neutropenia and deep venous thrombosis; for MP they were neutropenia, anaemia and thrombocytopenia; and for VMP they were neutropenia, oedema, leucopenia and thrombocytopenia.

Assumptions

The manufacturer’s model makes the following assumptions:

• Dose of thalidomide of 150 mg per day for MPT and 167 mg per day for CTDa.

• Adverse events are included in the model as the cost of treating them; the incidence of AEs does not influence the treatment duration, efficacy or patient utility.

• Costs included for second- and third-line treatments. Most patients who received VMP as first-line treatment receive CTDa as second-line treatment and most who did not receive VMP as first-line treatment do receive it as second-line treatment.

• Thalidomide RCTs that included maintenance therapy with thalidomide were included in the meta-analysis.

Appraisal of the manufacturer’s cost-effectiveness analysis

The Janssen–Cilag MS was appraised for methodological quality and generalisability to the UK NHS using a checklist adapted from the NICE reference case requirements78 and the Philips and colleagues checklist (see Appendix 12). 98 The submission meets all of the requirements for methodological quality and generalisability, except that it did not provide any evidence that the economic model had been validated.

The evaluation provided a clear statement of the decision problem to be addressed, including the population, which appeared to follow the scope for the appraisal issued by NICE. The comparators included (VMP, CTDa, MP and MPT) were appropriate as these are being routinely used or considered for use within the NHS in England and Wales. The perspective for the model was the NHS and PSS. A survival modelling methodology was used which seemed appropriate given the clinical nature of MM. The lifetime horizon used in the model reflects NICE guidance. The model structure was clearly presented with a description and justification of the key assumptions and data inputs used. Measures of clinical effectiveness are from a systematic review of RCTs with an MTC. Benefits for the model are measured in QALYs using the EQ-5D for measuring utility. All benefits and costs are discounted at 3.5% as outlined in NICE guidance. 78 Data on PPS were extrapolated from observed data using an exponential distribution. Uncertainty was assessed through a one-way deterministic sensitivity analysis and probabilistic sensitivity analysis (PSA). It was unclear if the model had been fully validated, as no details were provided.

Estimation of quality-adjusted life-years

Health-related quality of life utility values are assigned to each of the states: prior to response to treatment, response to treatment without progression, and post progression, based on a study evaluating chemotherapy followed by SCT in people with MM. 93 For the response state, a utility value of 0.81 was used, based on the utility of the general public at an age (median 54 years) corresponding to that of the patients in the study. A utility value of 0.64 was applied to the post-progression disease state. A utility value of 0.77 was applied to patients prior to the response to treatment. The submission considered this the most appropriate source of utility values because it is the only study that reports utility values according to response and progression status and, secondly, utility values were derived using the EQ-5D rather than the less methodologically robust indirect mapping approaches used in the other studies. However, as shown in the systematic review of HRQoL, there are several other more relevant HRQoL studies. In particular, it is unlikely that patients with MM would have the same HRQoL as the general population.

Estimation of costs

Treatment unit costs and doses were based on the British National Formulary (BNF) No. 5735 and MIMS 2009. 99 The duration of treatment was based upon the mean treatment duration in the trials and was assumed to incorporate discontinuation of treatment due to progression, death and AEs. The duration of treatment with MP was seven cycles as per the VISTA trial. 26 For bortezomib, (AiC/CiC information has been removed) vials were used per patient (VISTA trial26); the reason why the number of vials used is far fewer than the full treatment course of 52 vials is not given. The submission used an average dose of 150 mg per day for thalidomide obtained from the five MPT RCTs included in the meta-analysis of the MS. Within the CTDa combination, a daily dose of 167 mg was used for thalidomide. This is the weighted average as per protocol escalating dose from the MMIX RCT prior to the maintenance phase. A mean duration of treatment with thalidomide of 315 days was used, based on the duration reported in the MPT RCTs.

The resource use cost for the management of first-line MM was assumed to be the same for patients receiving VMP, MPT, CTDa and MP. There was an outpatient cost of £102 per visit and a total of nine outpatient visits. In addition, patients receiving VMP had this outpatient cost each time they were administered bortezomib.

Cost-effectiveness results

Table 26 shows the base-case results from the submission. The ICER for VMP versus MP is estimated to be £10,498. Furthermore the ICERs of VMP versus MPT and VMP versus CTDa are estimated to be £11,907 and £10,411, respectively. The submission states that the incremental analysis shows extended dominance of MTP over CTDa. However, the assessment group has found an error in the calculation of third-line costs for CTDa (correct cost £24,978 instead of £16,652). Correction of this error resulted in an ICER of £51,552 per QALY gained for CTDa versus MP.

One-way sensitivity analyses were undertaken for a limited number of parameters, including different survival distributions for OS and PFS, alternative HRs for OS, dose and duration of thalidomide, utilities, time horizon and discounting rate. The results are generally robust to changes in the sensitivity analyses. The model is most sensitive to the following parameters: underlying MP survival hazard, HRs for OS, dose of thalidomide, and duration of treatment with thalidomide in the MPT arm.

A PSA was undertaken using Monte Carlo simulation with 10,000 iterations. All parameters in the model were included except medication costs. For the PSA, at the £20,000 and £30,000 willingness to pay thresholds, VMP has the highest probability of being cost-effective: 64% and 75%, respectively.

Two scenario analyses were conducted. Scenario A did not include the costs of subsequent therapy after first-line treatment. In this scenario, the cost-effectiveness results were less favourable for each of the treatments and the ICERs increase to £48,437, £16,956 and £21,099 per QALY gained for CTDa, MPT and VMP compared with MP, respectively. Scenario B assumed the same second-line therapies as those treated with MP in the VISTA26 RCT. The results were similar for this scenario to the base-case analyses.

Summary of general concerns

• Hazard ratio used for OS for thalidomide was derived from a meta-analysis that included RCTs with thalidomide maintenance.

• The utility estimates were from a study with the wrong population, i.e. younger patients who received high-dose therapy. Furthermore, patients who had responded to treatment were assumed to have the same utility as the general population.

• There was an error in the calculation of third-line costs for CTDa.

• There was no evidence provided of model validation.

Overview

A structured data extraction form was used to guide the review of the Celgene submission to NICE (see Appendix 11). The submission states that its objective is to provide an evaluation comparing the costs and benefits of MPT with those of VMP and MP in patients with MM who are older than 65 years or who are ineligible to receive HDT. The evaluation has two stages. First, a short unsystematic review examines the literature for any relevant cost-effectiveness models in general and specifically in previously untreated MM patients who are not eligible for HDT. The review of cost-effectiveness studies indicates that a literature search was undertaken, although no details of the search strategy or methods for the review are provided. Searches identified five publications, with only one having relevance to the scope of the appraisal. The study by Deniz and colleagues72 compared MPT with MP as first-line treatment for MM in Scotland and provided the basis for the model developed for the submission by Celgene. Second, an economic model has been developed using data on the clinical effectiveness of MPT23,59 and VMP26 through a Bayesian MTC. The perspective of the economic evaluation is stated as being that of the NHS and PSS, including direct costs and benefits only. The analysis takes a lifetime horizon (30 years), presenting costs and outcomes (i.e. years of life gained and QALYs gained) for the three treatment arms of MPT, MP and VMP and an incremental analysis of costs and outcomes for MP and VMP when compared with MPT.

Modelling approach

A Markov model was developed to compare the difference in the progression of MM and in the costs of treatment when managed with the three different treatment options of MPT, VMP or MP through a series of different health states. It was developed from a model produced by Deniz and colleagues,72 which compared MPT and MP as first-line treatment for MM in Scotland.

The model has four different health states that are defined by the stage of disease progression or the occurrence of AEs. The four states are preprogression without AEs, preprogression with AEs, post progression and an absorbing state of death. All patients start in the preprogression without AEs state and move to other states if their condition worsens or they incur an AE. As MM is a progressive condition, people can only move to a worse state or remain in the same state. The submission provides limited discussion of the rationale for the approach or of the basis for the transition probabilities used to determine progression between states and other approaches assessing the phases of treatment may reflect variations in HRQoL more closely.

The model has a cycle length of 6 weeks (42 days) with a maximum of 12 cycles for MPT and MP and nine cycles for VMP. The cycle length and the number of cycles correspond to those used in clinical RCTs. 23,26,59 The time horizon used in the model equates to a lifetime horizon, although characteristics of the cohort used in the model are not clearly stated. The consequences of a shorter time horizon of 5 years were examined in the sensitivity analysis.

Treatment effects were calculated from a random effects Bayesian MTC of data originating from three RCTs. 23,26,59 The MTC was undertaken despite differences in the dosage used in the RCTs comparing MP with MPT. It used measures of survival time before and after progression as the primary outcomes. TTP and PFS were used and assumed to be equivalent. The outcome from the MTC was a measure of the risk of progression, provided through the percentage of patients experiencing PFS at 6-month intervals up to 30 months, with extrapolation beyond this point using an exponential distribution. It was assumed that post-progression survival (PPS) would be the same irrespective of preprogression treatment, with the different arms assumed to receive the same alternative treatment after progression (i.e. second- and third-line treatments). PPS was calculated by combining the MPT, MP and MEL100 (VAD, cyclophosphamide and melphalan 100 mg/m2) arms from the IFM 99/06 trial to create an average survival curve. 23 Average survival at different time points was then extrapolated with an exponential distribution. Treatment interruptions or discontinuations were encompassed in the trial efficacy data for MP and MPT, with no alteration to costs in the base case. Changes in cost were encompassed in sensitivity analyses through a reduction in dose as they are likely to reflect clinical practice. No data were available for VMP on discontinuation.

Adverse events were included for people on active treatment only if they were treatment related and considered to be clinically significant (i.e. grade 3 or above or occurred in 2% or more of patients in either arm). Those associated with disease progression were not incorporated into the model. The treatment-related AEs were included in the model through an estimate of the risk of AEs per cycle based on trial data. 23,26 The effects of AEs on HRQoL were also included in the model. A literature search revealed no HRQoL data specific to MM, and so HRQoL decrements were obtained for different patient populations. Costs of AEs were also included.

Assumptions

The manufacturer’s model makes the following additional assumptions:

• Post-progression survival is modelled to be the same across different treatment strategies.

• Patients assumed to discontinue first-line treatment upon disease progression.

• Deaths can only occur at or after progression and are assumed to be due to disease-related deterioration.

• Adverse events are included in the model as a utility decrement at the time of the event and the cost of treating them. They are assumed to not affect the disease progression rate or OS, or treatment duration, efficacy or dose.

• Assumes venous thromboembolism (VTE) antithrombotic prophylaxis for 5 months for patients receiving MPT with no resultant risk in incidence of VTEs, and antiviral prophylaxis for VMP.

Appraisal of the manufacturer’s cost-effectiveness analysis

The Celgene MS was appraised for methodological quality and generalisability to the UK NHS using a checklist adapted from the NICE reference case requirements78 and the Philips and colleagues checklist (see Appendix 12). 98 Although the economic evaluation lacked detail on some criteria, it adhered to the scope of the appraisal and followed the many aspects of the NICE reference case.

The evaluation provided a clear statement of the decision problem to be addressed, which appeared to follow the scope for the appraisal issued by NICE. Despite stating that the model focused on first-line therapy for people with MM who are ineligible for HDT and/or are aged over 65 years, insufficient details were provided of the population cohort used in the model itself. The comparisons of MP, MPT and VMP were appropriate as these are being routinely used or considered for use within the NHS in England and Wales. The setting for the evaluation was England and Wales and the perspective for the model was that of the NHS and PSS. A Markov modelling methodology was used which was developed from a previous evaluation. 72 The methodology seemed appropriate given the progressive nature of MM through distinct stages. The lifetime horizon (30 years) used in the model reflects NICE guidance.

Although the model structure was presented, limited details are given linking the model structure to the baseline risk of the condition. The submission outlines and justifies the assumptions used in the model and the different benefit, resource and cost inputs and their sources. Measures of clinical effectiveness are from a systematic review of RCTs with an MTC. Benefits for the model are measured in QALYs using the EQ-5D for measuring utility. All benefits and costs are discounted at 3.5%, as outlined in NICE guidance. 78 Data on PPS were extrapolated from observed data using an exponential distribution. While uncertainty has been assessed through a one-way deterministic sensitivity analysis, no probabilistic analysis or model validation processes were undertaken. As a consequence, the analysis provides only a partial assessment of the uncertainty in the model with the possibility of correlation between parameters and difficulty in summarising the implications of uncertainty.

Estimation of quality-adjusted life-years

No systematic review was undertaken to identify HRQoL values associated with the benefits of the treatment, but a literature search was conducted to identify utility decrements for AEs. The HOVON 24 study,93 an RCT of intensive chemotherapy followed by myeloblative therapy with autologous stem cell rescue compared with intensive chemotherapy, provided HRQoL data using the EQ-5D to assess the benefits of treatment for people with MM. Although not directly relevant in terms of the population and treatments included in the scope for the technology appraisal, it does provide an indication of the possible utilities for managing people with MM when specific assumptions are applied. The utility values used were 0.64 for people not responding to treatment and 0.81 for people who did respond (using general public utility for same age group). A utility value of 0.77 at 24 months was used for those who continue to respond to treatment with intensive chemotherapy and had not progressed. An assumption was made that preprogression patients and post-progression patients matched responders and non-responders in the HOVON trial. 93 However, other more relevant HRQoL studies (see Systematic review of health-related quality-of-life studies, above) show that the utility values used in the MS are higher than would be experienced by people with MM, whether newly diagnosed (0.52), undergoing treatment (0.38–0.55) or after treatment at 6 months (0.64) and 12 months (0.69).

The literature search for utility decrements for AEs did not identify specific values for people with MM and so utility values from different population groups were used (e.g. breast, colon and rectal cancer). Average per cent reduction in utility by each AE was calculated from these values and applied to the cohort in the model.

Estimation of costs

Resources and costs were obtained from several sources. NHS resources were from an unpublished survey of UK haematologists by Celgene Ltd. Inpatient, outpatient and day-case hospitalisation costs were derived from NHS Reference Costs,100 including inpatient and day-care costs for disease-related complications and treatment-related AEs, outpatient consultations and disease monitoring tests and treatment care costs in primary care. Costs of medicines were from the BNF (No. 57)35 and costs of blood transfusions from Wilson and colleagues101 with costs inflated to 2008. 102 When on active treatment, patients receive the mean observed treatment dose from the trials. Other resource use and cost data were provided for outpatient consultations, disease monitoring and treatment of AEs/complications. No indirect costs were included in the model. The costs of AEs were calculated by combining resource use data from the survey of haematologists with unit costs to estimate total costs. These costs and trial data on the frequency of AEs23 were then used to calculate a weighted average cycle cost. The methods for deriving resources and costs used and the sources were clearly described.

Cost-effectiveness results

The submission reports the benefits [i.e. TTP, patients progressed, deaths, proportion of patients with AEs, median OS, mean survival in years of life (life-years) and total QALYs] and the total costs (i.e. medication, monitoring and management of AEs) separately for each treatment pathway in the model.

Comparison of the benefits used for the model showed considerable benefit for those receiving MPT or VMP over MP on median TTP, median OS, total life-years and total QALYs. In contrast, more people receiving MPT (43.2%) or VMP (40.9%) suffered AEs compared with those receiving MP (13.4%). The total costs of the different treatment strategies used within the model showed considerable variation between MP (£1365) and VMP (£42,616). The cost of the medications was the main reason for these differences.

The base-case analyses (Table 27) produced two comparisons, MPT versus MP and VMP versus MPT, with differing outcomes. When compared with MP, MPT had an ICER of £18,188 per life-year gained and £23,381 per QALY gained. In contrast, the comparison of VMP with MPT showed that VMP produced a small benefit in additional life-years and QALYs at a large additional cost (£21,483). The resultant ICERs were £200,237 per life-year gained and £303,845 per QALY gained.

The submission assessed uncertainty through one-way deterministic sensitivity analyses. No PSA was conducted as the manufacturer stated that the efficacy of MPT and VMP were essentially the same and that the cost differences would be the key driver for the model. The submission included a number of one-way sensitivity analyses for parameter values and model structure. The parameters with the greatest effect on the model results were for the changes in treatment efficacy with a range of £16,586–33,275 per QALY gained for MPT versus MP and a range of £148,873–1,000,435 per QALY gained for VMP versus MPT.

The submission concludes that MPT represents a cost-effective use of NHS resources compared with MP as a first-line therapy for people with MM who are not eligible for HDT and/or are aged over 65 years. In contrast, when comparing MPT and VMP the manufacturer stated there was negligible clinical benefit from VMP at an additional cost that resulted in the ICERs exceeding £300,000 per QALY. When these findings were assessed through sensitivity analysis, the ICERs were reasonably robust.

Summary of general concerns

• The economic evaluation focuses on the effectiveness of first-line treatment for people with MM who were ineligible for HDT, reflecting the scope for the NICE technology appraisal. Exclusion of second- or third-line treatment options may oversimplify the evaluation with consequences for the incremental benefits and costs that would result from different possible options available. Given that first-line treatment may, in part, determine subsequent treatment options, it would be helpful to include these in the evaluation.

• All deaths are assumed to be caused by disease-related deterioration and occur only at or after progression. In practice, deaths may and do occur prior to progression and, as such, the evaluation may overestimate the benefits that are accrued.

• Post-progression survival was the same irrespective of preprogression treatment, which would affect the incremental benefits.

• No HRQoL studies relevant to the evaluation were identified by the manufacturer and utility values from comparisons of different MM populations using alternative management strategies were used.

Comparison of manufacturers’ results

The manufacturers’ economic models had similar structures but used different methodology: one used a survival model and the other a Markov model. Both models compared first-line treatment with VMP, MPT and MP. Janssen–Cilag also included CTDa as a comparator. The ICERs produced by the Janssen–Cilag and Celgene submissions vary considerably from £11,907 to £303,845 per QALY gained for VMP versus MPT. These differences stem from the number of vials used for treatment with bortezomib, the HRs for thalidomide and the inclusion of second- and third-line treatments.

Overview

We developed a new model to estimate the costs, benefits and cost-effectiveness of MPT, VMP and CTDa compared with MP, in newly diagnosed patients with MM ineligible for HDT–SCT. CTDa was included in order to compare all relevant comparators; however, there are limitations to the effectiveness data, as the effectiveness estimate for OS was not statistically significant and the MMIX RCT included a second randomisation to thalidomide maintenance for some patients. The model was populated with clinical effectiveness data from the included RCTs in our systematic review of effectiveness (Chapter 4), HRQoL data from a systematic review of HRQoL studies (see Systematic review of health-related quality-of-life studies) and cost data derived from published studies (where available) and from national and local NHS unit costs.

The economic evaluation was from the perspective of the NHS and PSS, as only these direct costs were included. The model estimates the lifelong costs and benefits from each of the treatments. The costs and benefits were discounted at 3.5%, as recommended by NICE. 78 The base-price year for the costs was 2009. The intervention effect in terms of improvement in OS and PFS was derived from the systematic review of effectiveness reported in Chapter 4 (see Overall survival and Progression-free survival). The outcome of the economic evaluation is reported as cost per QALY gained.

Description of the SHTAC model

A survival model was used to compare the cost-effectiveness estimates of VMP, CTDa and MPT versus MP. The model uses a survival analysis approach to estimate the mean OS and PFS for each of the interventions for a cohort of patients with newly diagnosed MM. The model consisted of cycles of 6 weeks in length to be consistent with the cycle lengths used for chemotherapy treatment. A lifetime horizon of 30 years was modelled to capture all clinical events using partitioned survival analysis for OS and PFS. Two survival curves were constructed for OS and PFS (Figure 7b), based on the derived probability of death and progression in each model cycle, respectively. The mean time spent in each state was calculated from the survival curves for OS and PFS (see Figure 7a).

FIGURE 7.

The survival model adopted for the cost-effectiveness model. Treatment and post-treatment health states refer to first-line treatment.

Survival was classified into three health states, and the mean time spent in each state is as follows:

• Treatment (T_treat) is the mean duration of first-line treatment.

• Post treatment (T_{Post_treat}) is the mean time from stopping first-line treatment until progression, i.e. T_PFS – T_treat

• Post progression (T_prog) is the time from disease progression until death, i.e. T_OS – T_PFS, where T_OS is mean OS and T_PFS is mean PFS.

Each health state was associated with a HRQoL utility estimate that was multiplied by the length of time spent in that state. The total QALYs over the lifetime of a patient were calculated by aggregating the estimated QALYs from each health state.

Due to lack of data on subsequent therapies, it was unclear how the subsequent therapies affected HRQoL and survival and therefore second-line therapy is only included in the model as a cost.

The methodology used for deriving the parameters for the survival curves for the alternative treatments is as follows:

1. Construct the baseline survival curves for MP using the adjusted event probability for each time interval.

2. Construct the survival curves for other treatments by using the event probability for each time interval; i.e. event probability for MP multiplied by HR for treatment option.

For the baseline MP treatment, OS and PFS at regular time points were derived for each of the included studies from our meta-analysis of the clinical RCTs. The data from the RCTs were combined to form baseline MP OS and PFS curves. These curves provided the probability of an event (death or disease progression), i.e. hazard rate, for MP in each time interval (see Baseline MP curves).

The treatment effects for the other interventions compared with MP (HRs) were taken from our systematic review of clinical effectiveness (see Assessment of effectiveness). As the HR of the treatments versus MP varied over time, a constant HR was not appropriate. We estimated the HR for each 6-monthly period for each of the treatments versus MP.

The hazard rate for death was derived for each of the treatments by multiplying the baseline MP probability of death by the HRs for each time interval. The hazard rate for disease progression was derived in a similar manner. This method provided a closer fit to the trial data than approximations, such as fitting distributions. Parameters used in the model and the data sources used to derive them are described in more detail below (see SHTAC data sources). The methodology used for deriving the survival curves is described in more detail in Appendix 13.

The costs in the model comprise drug treatment, consultation, monitoring costs, and costs for treating AEs. Patients remained on drug treatment unless their disease progressed or they died. All patients who had not died received second-line therapy and this was assumed to start at the mean time of disease progression for the cohort. Third-line therapy was not included as it was assumed that most patients would receive lenalidomide, irrespective of the initial treatment. Costs used in the model are described in more detail below (see Estimation of costs).

A list of the model assumptions is given below. Assumptions are applied to all treatment options unless explicitly stated otherwise. All assumptions were tested in sensitivity analyses.

The model includes the following assumptions:

• For bortezomib, each patient receives one vial per administration.

• Costs included for second-line treatments. Most patients who received VMP as first-line treatment receive CTDa as second-line treatment and most who did not receive bortezomib as first-line treatment receive it as second-line treatment.

• Costs and outcomes of third-line and subsequent treatments are assumed to be the same between arms.

• Patients discontinue first-line treatment upon disease progression.

• Health-related quality of life is better for those with CR than those with less than CR and is assumed to improve when patients stop treatment.

• AEs are not modelled explicitly in the model for patient outcomes, i.e. OS and PFS, but are included as additional cost for treating the AEs in the model.

In each cycle the total costs and QALYs are calculated by multiplying the individual costs and HRQoL by the number of people in the cohort still alive for each of the treatments. The total lifetime costs and QALYs are calculated by aggregating the costs and QALYs for all cycles. The total discounted QALY gain and cost of treatments are calculated. Thus, the cost-effectiveness of each of the treatments is calculated,

SHTAC data sources

Baseline MP curves

The baseline MP OS curve was generated using the MP OS curves from the RCTs included in our systematic review of clinical effectiveness (see Chapter 4, Overall survival). Survival probabilities (at 6-month intervals) were extracted from a scanned copy of the Kaplan–Meier plots for each MP group using the digitising software Engauge104 (Appendix 13). A weighted average of the survival probabilities for each time point was calculated to provide a summary MP OS curve (Table 28 and Figure 8) using the number of participants in the trials as weights.

FIGURE 8.

Summary curve for MP OS obtained from weighted average of individual MP curves.

TABLE 28 - Overall survival probabilities extracted from Kaplan–Meier plots for MP study arms using digitising software

Weighted by size of trial arm. The average could not be weighted by numbers at risk at each time point because Palumbo25 and the Vista trial CSR105 did not report this for 6-month time point intervals, and MMIX53 did not report numbers at risk at all.

Trial MP arm	No. in MP arm	Time (months)
		6	12	18	24	30	36	42	48	54	60	66	72
Facon et al.23	196	0.88	0.78	0.72	0.63	0.55	0.48	0.39	0.31	0.26	0.23	0.19	0.17
Hulin et al.59	116	0.90	0.83	0.71	0.62	0.47	0.39	0.38	0.30	0.28
Palumbo et al.24	164	0.96	0.84	0.75	0.71	0.65	0.59	0.53	0.49	0.49	0.36
VISTA Trial105	338	0.92	0.82	0.76	0.69	0.64	0.54	0.51	0.41
MMIX53	426	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)	(AiC/CiC information has been removed)
MP weighted averagea		0.90	0.79	0.72	0.65	0.56	0.48	0.42	0.36	0.31	0.27	0.17	0.17

A baseline MP PFS curve was generated using the PFS curves from the trial data included in our systematic review of clinical effectiveness (see Chapter 4, Progression-free survival) in a similar way to the baseline OS curves – see Appendix 13 and Table 29. A weighted average of the PFS probabilities for each 6-month time point was calculated to provide a summary MP PFS curve (Figure 9) using of participants in the trials as weights.

TABLE 29 - Progression-free survival probabilities extracted from Kaplan–Meier plots using digitising software

Weighted by size of trial arm. MMIX data could not be included because a PFS curve was not available.

Study MP arm	No. in MP arm	Time (months)
		6	12	18	24	30	36	42	48	54
Facon et al.23	196	0.77	0.63	0.49	0.34	0.22	0.14	0.08	0.06	0.05
Hulin et al.59	116	0.8	0.66	0.51	0.33	0.17	0.12	0.06	0.05	0.05
Palumbo et al.24	164	0.88	0.60	0.41	0.28	0.22	0.19	0.18	0.18
VISTA trial26	338	0.77	0.63	0.43	0.28
MP weighted averagea		0.78	0.63	0.46	0.30	0.21	0.15	0.11	0.10	0.05

FIGURE 9.

Summary curve for PFS obtained from weighted average of individual PFS data in MP arms of trials.

The probability of an event at each time interval for the MP treatment arm (hazard OS and hazard PFS) is calculated from the baseline MP OS and MP PFS curves. These probabilities are shown in Table 30. The hazard rate for an event for MP per cycle is estimated for each time point, t_i:

^{[Equation 2]} h ( t i ) = 1 − ( s ( t i ) s ( t i − 1 ) ) 1 ( t i − t i − 1 )

TABLE 30 - Hazard rate for MP for OS and PFS (event rate per cycle)

Months	Cycles	Hazard rate
		OS	PFS
6	4.4	0.024	0.056
12	8.7	0.030	0.048
18	13.0	0.021	0.070
24	17.4	0.023	0.094
30	21.7	0.034	0.067
36	26.1	0.035	0.067
36+	26+	0.028	0.067

where s(t) is the survival function over time t.

For OS, few individuals were followed up for more than 36 months, and so a constant hazard rate was assumed after 36 months using the hazard rate in the first 36 months. For PFS, few individuals were followed up for more than 24 months, and so a constant hazard rate was assumed after 24 months using the hazard rate in the first 24 months. The methodology used to derive the survival curves is described in more detail in Appendix 13.

Health-related quality of life

Although our systematic review of HRQoL studies (see Systematic review of health-related quality-of-life studies) did not find any generic preference-based HRQoL studies of people with untreated MM who were not eligible for HDT with SCT, it did identify two studies that assessed HRQoL in this group using the EORTC QLQ-C30. A targeted search was therefore conducted for studies that mapped data from the EORTC QLQ-C30 onto the EQ-5D to enable the estimation of health state values based on EORTC QLQ-C30 data. The EORTC QLQ-C30 is the most commonly used instrument to measure the HRQoL of cancer patients. Two studies were identified. 106,107

McKenzie and van der Pol107 used an ordinary least squares (OLS) regression analysis with data from an RCT of palliative therapies for 199 patients with inoperable oesophageal cancer, with an average age of 74.8 years. The regression results for the mapping are shown in Table 34.

TABLE 34 - Regression results for mapping between EORTC QLQ-C30 and EQ-5D from McKenzie and van der Pol107

Dimension	Coefficient
Global QoL	0.0016
Physical functioning	0.0004
Role functioning	0.0022
Emotional functioning	0.0028
Cognitive functioning	0.0009
Social functioning	0.0002
Fatigue	–0.0021
Nausea	–0.0005
Pain	–0.0024
Dyspnoea	0.0004
Insomnia	0.00004
Appetite loss	0.0003
Constipation	0.0001
Diarrhoea	–0.0003
Financial difficulties	–0.0006
Constant	0.2376

Kontodimopoulos and colleagues106 used an OLS regression with data from 48 patients with gastric cancer, split into equal subgroups by age, sex and chemotherapy scheme. Three scales were significant predictors (p < 0.05 or better) of EQ-5D indices: physical functioning, emotional functioning and global health status. The regression results for the mapping are shown in Table 35.

TABLE 35 - Regression results for mapping between EORTC QLQ-C30 and EQ-5D from Kontodimopoulos and colleagues106

Dimension	Coefficient
Physical functioning	0.00508
Emotional functioning	0.00313
Global health status	0.00546
Constant	–0.18143

Our systematic review of HRQoL studies found two studies in the population of interest but these used the EORTC QLQ-C30. 91,92 For both studies, we mapped the EORTC QLQ-C30 HRQoL scores to the EQ-5D using each of the mapping algorithms described above (Tables 36 and 37).

TABLE 36 - European Quality of Life-5 Dimensions utility values derived by mapping from EORTC QLQ-C30 HRQoL scores from Strasser-Weipl and Ludwig92

Mapping algorithm	MM	Reference population
McKenzie and van der Pol107	0.59	0.82
Kontodimopoulos et al.106	0.58	0.88

TABLE 37 - European Quality of Life-5 Dimensions utility values derived by mapping from EORTC QLQ-C30 HRQoL scores from Gulbrandsen et al.91

Mapping algorithm	Reference population	Time (months)
		0	1	6	12	24	36
McKenzie and van der Pol107	0.81	0.55	0.58	0.68	0.68	0.68	0.69
Kontodimopoulos et al.106	0.86	0.52	0.58	0.69	0.69	0.69	0.71

Gulbrandsen and colleagues91 provide HRQoL at different time points. Based on this study, it appears that HRQoL is lower during the treatment period and improves after treatment has finished and this is consistent with HRQoL results from Uyl-de Groot and colleagues. 96 Long-term HRQoL appears to be stable over time. In addition, the utility estimates from the HRQoL studies in populations treated with HDT are similar to those from Gulbrandsen and colleagues. 91

The accuracy of the mapping studies was assessed for the study by Uyl-de Groot and colleagues,96 which reported EORTC QLQ-C30 and EQ-5D results. Figure 10 shows the comparison between the EQ-5D utility estimates using the two mapping methods compared with the EQ-5D data from Uyl-de Groot and colleagues. For these data, the mapping algorithm by McKenzie and van der Pol provides the better fit and for most time points is a good fit to the data.

FIGURE 10.

Comparison of results from mapping studies from EORTC QLQ-C30 to EQ-5D with EQ-5D data from Uyl-de Groot and colleagues. 96

We suggest that the most appropriate source of HRQoL data for the treatment period and post-treatment values is from Gulbrandsen and colleagues91 from the mapping by McKenzie and van der Pol. These utility estimates are shown in Table 37. The utility estimates for the treatment period are for the 1-month time point, i.e. 0.58, and for post treatment (and post progression) is an average of the 6- to 36-month time points, i.e. 0.68.

Complete response

Health-related quality-of-life data from the MMIX RCT57 were analysed to determine whether patients with CR had a better HRQoL after response than those with other levels of response. EORTC QLQ-C30 data were available for 0, 3, 6 and 12 months after initial treatment commenced. We mapped the EORTC QLQ-C30 data to EQ-5D health utilities using the algorithm from McKenzie and van der Pol. 107 For the first three periods, the EQ-5D utility scores were similar for both CTDa and MP groups, and similar to those from Gulbrandsen and colleagues. 91 (AiC/CiC information has been removed.)

In the model we estimate the utility for the post-treatment health state (until disease progression) as a weighted average of those who had a CR (AiC/CiC information has been removed) and those with a lesser response (AiC/CiC information has been removed).

Results of SHTAC independent economic evaluation

This section reports the cost-effectiveness results for a typical person with MM who received treatment with bortezomib in combination with MP or thalidomide in combination with MP compared with those receiving MP. Results for costs and QALYs are presented for each treatment, with costs and benefits discounted at 3.5%. 78 The survival curves for OS from the model are shown in Figure 11. The results show increased survival for MPT, VMP and CTDa versus MP. The cost-effectiveness is presented as incremental cost per QALY compared with existing treatment with MP. The summary results of the non-discounted treatment effects are shown in Table 43. In the base-case analysis, OS varied from 4.20 years for MP to 6.66 years for MPT. Survival for MPT is slightly longer than for VMP. The cost-effectiveness results for CTDa should be treated with caution, (AiC/CiC information has been removed). The summary results of the undiscounted costs are shown in Table 44 for each treatment. First-line treatment costs ranged from £112 for MP to £43,824 for VMP. Second-line treatment costs were the same for MP, MPT and CTDa, and about £10,000 lower for VMP. The total costs of the treatments ranged from £23,248 for MP to £59,644 for VMP.

FIGURE 11.

Overall survival curves for MP, MPT, VMP and CTDa.

The baseline discounted cost-effectiveness results are shown in Table 45. Each of the treatments is more expensive than MP, with the additional cost ranging from £8,600 (CTDa) to more than £35,000 (VMP) over a patient lifetime. The incremental cost-effectiveness versus MP for MPT, VMP and CTDa figures are £9135, £29,820 and £33,031 per QALY gained, respectively.

Each comparator is presented in successive rows ordered by the number of QALYs generated. Each option is then compared to the next best option. In summary the incremental analysis suggests extended dominance of MPT over CTDa, and MPT dominates VMP as it is more effective and cheaper (Figure 12). The comparison of VMP versus MPT suggests that VMP and CTDa are unlikely to be cost-effective treatment options at the conventional willingness-to-pay threshold of £20,000–30,000 per QALY gained. However, there is much uncertainty around the results for CTDa because the OS effectiveness estimates were not statistically significant and the results from the MMIX RCT included those of participants who had received thalidomide maintenance therapy.

FIGURE 12.

Cost-effectiveness plane for treatments MP, CTDa, VMP and MPT.

Sensitivity analysis

Probabilistic sensitivity analysis

In the PSA, all parameters were sampled probabilistically from an appropriate distribution using similar ranges as used in the deterministic sensitivity analyses. The parameters sampled were discount rate, number of treatment cycles, utility values, CR rate, cost of AEs, parameters for the survival curves and the proportions of patients receiving bortezomib as second-line therapy. The distribution assigned to each variable included in the PSA and the parameters of the distributions are reported in Appendix 14.

One thousand simulations were run. The PSA results are presented in Table 50 and show similar results to the deterministic analyses (Tables 46–48). The scatter plots for cost and health outcomes for the treatment options for the PSA are shown in Figure 13. The CEAC is shown in Figure 14, and indicates that at the £20,000 and £30,000 willingness-to-pay thresholds MPT has the highest probability of being cost-effective of 0.95 and 0.95, respectively.

TABLE 50 - Baseline PSA cost-effectiveness results vs MP

Result	MP	MPT	VMP	CTDa
Total cost (£)	21,620	33,050	57,545	30,371
Total QALY	2.44	3.68	3.66	2.70
Incremental cost vs MP (£)	–	11,495	35,991	8816
Incremental QALY vs MP	–	1.26	1.24	0.28
ICER vs MP (£)	–	9124	29,102	31,612

FIGURE 13.

Scatter plots of the costs and health benefits from PSA for MP, MPT, VMP and CTDa.

FIGURE 14.

Cost-effectiveness acceptability curve from the PSA.

Summary of cost-effectiveness

• A systematic search of the literature found five abstracts of economic evaluations of treatment for patients with previously undiagnosed MM, who were ineligible for HDT-SCT. None of the studies contained sufficient information for critical appraisal. Three of the abstracts compared MPT with MP in patients in Scotland, Wales and Australia. Each abstract concluded that MPT was a cost-effective alternative to MP. Two abstracts compared VMP, MPT and MP in Canadian and US patients. Both studies concluded that the VMP regimen was cost-effective compared with MP and MPT. The latter study stated that VMP dominated MPT (i.e. more effective at a lower cost). All studies were industry funded.

• A systematic review of studies of QoL for patients with MM identified six studies: only two of these studies were for the population of interest and both studies did not include generic preference-based utility measures; the other four QoL studies provided utility estimates for patients with MM who had intensive therapy.

• Two manufacturers submitted evidence to be considered for the appraisal of bortezomib and thalidomide treatment. Janssen–Cilag, the manufacturer of bortezomib, constructed a survival model that estimated OS and PFS based on treatment effects from a MTC of the RCTs. They included second- and third-line treatment. The base-case results from the submission found all treatments (VMP, MPT and CTDa) to be cost-effective. The ICER for VMP versus MP is estimated to be £10,498. Furthermore, the ICERs of VMP versus MPT and VMP versus CTDa are estimated to be £11,907 and £10,411, respectively.

• Celgene, the manufacturer of thalidomide, constructed a Markov model with health states for preprogression (with or without AEs), post progression and death. They assumed that survival after disease progression was the same irrespective of first-line treatment. Treatment effects for disease progression were calculated from a random effects MTC. The base-case results from the submission estimated an ICER of £23,381 per QALY gained for MPT versus MP and £303,845 per QALY for VMP versus MPT.

• The authors of this report developed an independent survival model. The survival model consisted of two survival curves which estimated the mean time to death and disease progression. These survival durations were used to derive the time spent in three health states: treatment, post treatment and progression. Utility values were applied to these health states to estimate total QALYs for each treatment option. Costs were included for medications and outpatient costs and AEs. The model base-case results showed increased survival for each of the treatments compared with MP at an increased cost. The OS was marginally longer for MPT than for VMP at a considerably lower cost. The cost-effectiveness estimates for MPT, VMP and CTDa versus MP were £9135, £29,820 and £33,031 per QALY gained, respectively. However, MPT dominated VMP as it was cheaper and more effective.

• The effects of a range of parameter values in the economic model were evaluated in sensitivity analyses. The model results were found to be robust to changes in the parameter values. The model results are most sensitive to changes in the parameter values of the HRs for OS.

• The PSA estimated the probability of each of the treatments to be cost-effective at the £20,000 and £30,000 willingness-to-pay thresholds. MPT has the highest probability of being cost-effective, with probabilities of 0.95 and 0.95, respectively.

Clinical effectiveness

• Five RCTs were included in the systematic review of clinical effectiveness. One examined the effectiveness of VMP, three examined the effectiveness of MPT, and one examined the effectiveness of CTDa. The comparator in all RCTs was MP. Two RCTs had a maintenance phase with thalidomide which followed the initial treatment phase. The maintenance phase did not meet the inclusion criteria. Reporting on the results of these RCTs was therefore limited to outcomes that had been reported at a time point prior to the start of maintenance therapy with thalidomide.

• Judgements about aspects of study quality could not be made for some studies because of a lack of detailed reporting in the published papers. Consequently, there is uncertainty about some aspects of study quality. In particular, it was not possible to determine whether the amount and pattern of censoring in the RCTs was comparable between study groups.

• Overall survival was increased in the intervention group in comparison to the groups receiving MP in both of the MPT versus MP RCTs that provided data for this outcome. Meta-analysis of the OS data from two RCTs of MPT versus MP confirmed the superiority of MPT and was in agreement with a published meta-analysis of three MPT versus MP trials. 109 OS was also increased in the single VMP versus MP RCT. Because OS data for the single RCT of VMP are not as mature as those for the two RCTs of MPT it was not possible to determine whether OS was greater with MPT or VMP.

• More participants in the intervention arms of the included RCTs achieved a CR to treatment than in the MP comparator arms. The difference was reported to be statistically significant in four of the included studies, with a fifth study not reporting a p-value. It should be noted, however, that the proportion of participants achieving a CR to treatment was not assessed according to ITT principles in one RCT, which reported data for only approximately three-quarters of the enrolled participants, and the proportions of data missing from each trial arm appeared to be unequal but no explanation for this was provided. The remaining four RCTs reported results for approximately 95% or more of the participants and the proportion of data missing from each arm seemed comparable. A meta-analysis of the CR outcome data from three MPT versus MP RCTs confirmed that MPT was superior in comparison with MP in terms of the proportion of patients achieving a CR. As there were only single trials for VMP versus MP and for CTDa versus MP no meta-analyses for these comparisons were undertaken.

• Progression-free survival was reported to be statistically significantly longer in the intervention group in comparison with the groups receiving MP in both of the MPT versus MP RCTs that provided data on this outcome, and the single VMP versus MP RCT. Only the RCT of VMP versus MP reported on time to disease progression, which was the primary outcome of this trial. There was a statistically significant difference in median time to disease progression in favour of the VMP group.

• Adverse events occurred with all treatments. Some AEs were statistically significantly increased in trial intervention arms. The combination of bortezomib and MP was associated with a statistically significant increase in grade 3 AEs in comparison with the MP group. AE outcomes for thalidomide in combination with MP varied between the RCTs, but the two trials which reported on peripheral neuropathy reported a statistically significant increase in this AE in the treatment groups receiving thalidomide.

Cost-effectiveness

• A systematic search of the literature found five abstracts of economic evaluations of treatment for patients with previously undiagnosed MM, ineligible for HDT–SCT. None of the studies contained sufficient information for critical appraisal.

• A systematic search for published studies of QoL for patients with MM identified six studies: only two of these studies were for the population of interest and neither study included generic preference-based utility measures; the other four QoL studies provided utility estimates for patients with MM who had intensive therapy.

• Two manufacturers submitted evidence to be considered for this review for bortezomib and thalidomide treatment. Janssen–Cilag, the manufacturer of bortezomib, constructed a survival model that estimated OS and PFS based on treatment effects from a MTC of the trials. The base-case results from the submission found all interventions to be cost-effective, with ICERs of less than £11,000 per QALY gained versus MP for MPT, VMP and CTDa.

• Celgene, the manufacturer of thalidomide, constructed a Markov model with health states for preprogression (with or without AEs), post progression and death. The base-case results from the submission estimated MPT to be cost-effective compared with MP, while the ICER for VMP versus MPT was more than £300,000 per QALY gained.

• The authors of this report developed an independent survival model. From this independent model, the incremental cost-effectiveness versus MP for MPT, VMP and CTDa is estimated as £9135, £29,820 and £33,031 per QALY gained, respectively. However, MPT dominated VMP as it was cheaper and more effective. The model results are most sensitive to changes in the parameter values of the HRs for OS. The PSA showed that MPT has the higher probability to be cost-effective at the £20,000 and £30,000 willingness-to-pay thresholds.

Discussion of cost-effectiveness results

The results for the manufacturers’ and SHTAC’s economic analyses are shown in Table 51. The results of the analyses vary considerably. The costs vary substantially between the analyses; for example the cost of MP varies between £1365 for the Celgene submission to £54,434 for the Janssen–Cilag submission. The costs from the Celgene analysis were lower as they had not included any subsequent treatment costs, whereas the SHTAC analysis included costs for second-line treatment and the Janssen–Cilag analysis included costs for second- and third-line treatment.

The incremental costs for MPT versus MP vary between £4888 (Janssen–Cilag) and £19,768 (Celgene). The Celgene submission uses higher dosages of thalidomide (238 mg/day) for longer periods (11 cycles) than the other two analyses. The incremental costs for VMP versus MP vary between £12,242 (Janssen–Cilag) and £41,251 (Celgene). These differences are largely due to the assumptions around the number of vials of bortezomib used, with Janssen–Cilag assuming a mean of (AiC/CiC information has been removed) vials used per person, whereas the mean number of vials used is over 40 in the SHTAC and Celgene economic evaluations. The incremental costs for CTDa versus MP vary between £2234 (Janssen–Cilag) and £8544 (SHTAC), and these differences are due to an error in the cost calculation for third-line therapy for CTDa in the Janssen–Cilag analysis.

The total QALY estimates between the studies are reasonably similar with estimates for all treatment arms varying between 2.42 and 4.03 QALY. The incremental QALY estimates for MPT versus MP vary widely and these differences are due to the estimates chosen for the HR for OS compared with MP. The incremental QALY estimates for MPT versus MP range from 0.55 (Janssen–Cilag) to 1.22 (SHTAC).

The different assumptions and methodology described above results in a range of estimates for the cost-effectiveness of the treatment options. The ICER for MPT versus MP varies between £9135 (SHTAC) and £23,381 (Celgene) per QALY gained. The ICER for VMP versus MP varies between £10,498 (Janssen–Cilag) and £44,838 (Celgene) per QALY gained. The ICER for CTDa versus MP varies between £10,905 (Janssen–Cilag) and £33,031 (SHTAC) per QALY gained.

The results have also been estimated for VMP versus MPT. These results also vary widely for the reasons given above. The ICER for VMP versus MPT was estimated as £11,907 (Janssen–Cilag) and £303,845 (Celgene) per QALY gained. For the SHTAC economic analysis, MPT dominated VMP, i.e. cheaper and more effective, for the base case and most sensitivity analyses.