Cost-effectiveness analysis of bortezomib in combination with rituximab, cyclophosphamide, doxorubicin, vincristine and prednisone (VR-CAP) in patients with previously untreated mantle cell lymphoma

Mantle cell lymphoma (MCL) is a rare, incurable and aggressive sub-type of non-Hodgkin’s lymphoma (NHL), accounting for approximately 6 % of all NHL cases [1]. The incidence of MCL in the UK is 0.9 per 100,000 [1]. The general pattern of disease progression in MCL is one of relapse and remission, with each relapse becoming more difficult to treat, and the depth and durability of any subsequent remissions achieved invariably inferior to those achieved with first-line treatment [2–6].

In patients first presenting with aggressive disease requiring treatment, the initial treatment decision is whether patients are suitable for high-intensity induction therapy, to be followed by haematopoietic stem cell transplantation (HSCT). There are no strict criteria against which patients are assessed; rather, haematologists will assess eligibility on a patient-by-patient basis, taking into account factors such as patient age, performance status and disease prognosis, disease severity, co-morbidities, and clinical risk [2, 5–10].

For patients who are not eligible for high-intensity induction therapy, that is those for whom HSCT is unsuitable, there had been no licensed induction therapy regimens prior to bortezomib. Rituximab, cyclophosphamide, doxorubicin, vincristine and prednisone (R-CHOP) became the preferred first-line induction therapy in UK clinics because the large scale European MCL Elderly trial [11] demonstrated a survival benefit for R-CHOP when compared with rituximab in combination with fludarabine and cyclophosphamide (R-FC). Alternative rituximab-based chemotherapy induction regimens are also administered in the first-line setting, but usually only for the frailest of patients considered unsuitable for R-CHOP therapy; while alternatives are considered to be associated with lower toxicity, the evidence base supporting their use is considerably weaker [12]. Median progression-free survival (PFS) associated with chemotherapy is less than 2 years, and median overall survival (OS) is less than 5 years [10, 13–19].

Bortezomib is the first product to be licensed for the treatment of patients with previously untreated MCL for whom HSCT is unsuitable. Bortezomib is administered in combination with the rituximab, cyclophosphamide, doxorubicin, prednisone backbone familiar to clinicians as part of the R-CHOP regimen. A randomised, open-label, multicentre Phase III study (LYM-3002) comparing bortezomib, rituximab, cyclophosphamide, doxorubicin and prednisolone (VR-CAP) to R-CHOP showed a significant improvement in PFS (24.7 versus 14.4 months; hazard ratio [HR] = 0.63, p < 0.001) based on the primary assessment of PFS by the independent review committee (IRC) [20]. Duration of overall response for VR-CAP was more than double that of R-CHOP (median of 36.5 versus 15.1 months), resulting in an increase in the treatment free interval (TFI) of almost 20 months versus R-CHOP (median of 40.6 versus 20.5 months; HR = 0.50, p < 0.001) [20].

There have been no previous technology appraisals by the National Institute of Health and Care Excellence (NICE) within MCL; other therapies that are frequently used such as bendamustine and temsirolimus did not go through the UK health technology assessment (HTA) process due to lack of marketing authorisation approval and manufacturer non-submission, respectively. To gain NICE recommendation for VR-CAP, the cost effectiveness of VR-CAP had to be assessed over the long term and beyond the duration of clinical trial follow up. As median survival for VR-CAP had not been reached in the LYM-3002 trial, it was challenging to provide realistic and robust estimates of long-term OS. This challenge is common in UK HTAs and will become more pronounced as regulatory and HTA bodies come under pressure to provide earlier access to promising drugs.

The objective of this study was to assess the cost effectiveness of VR-CAP compared to R-CHOP, in a UK setting, which is currently seen as standard first-line treatment for patients with MCL.

As presented in Table 6, VR-CAP is associated with higher costs and greater efficacy compared to R-CHOP. The base case results demonstrate that VR-CAP is a cost effective treatment at the conventional UK willingness-to-pay threshold of £20,000–£30,000 per QALY [21] with an ICER of £20,043. The PSA indicated that there was a probability of 88.9 % that the ICER lies below the threshold of £30,000 per QALY. Figure 6 indicates that most uncertainty in the model comes from uncertainty in efficacy.

	VR-CAP	R-CHOP	Differencea
Deterministic results
QALYs	4.10	3.29	0.81
Progression-free survival from first-line treatment	2.70	1.54	1.16
Progressed from first-line treatment	0.14	0.10	0.03
Progression-free survival from second-line treatment	0.12	0.15	−0.03
Progressed from second-line treatment	1.15	1.50	−0.35
Costs	£45,842	£29,630	£16,212
First line therapy medication costs	£22,606	£8,041	£14,566
Administration of first line therapy	£5,817	£1,564	£4,253
Adverse events, transfusions & concomitant medication	£1,472	£1,105	£367
Disease management costs	£4,191	£4,676	−£486
Second line treatment (medication and administration)	£7,152	£9,423	−£2,271
Terminal care	£4,605	£4,821	−£217
Deterministic ICER (£/QALY gained)			£20,043
Probabilistic results
QALYs	4.09	3.28	0.81
Costs	£45,482	£29,285	£16,196
Probabilistic ICER (£/QALY gained)			£19,889

Abbreviations: ICER incremental cost-effectiveness ratio, QALY quality-adjusted life year, R-CHOP rituximab with cyclophosphamide, doxorubicin, vincristine and prednisolone, VR-CAP bortezomib with rituximab, cyclophosphamide, doxorubicin and prednisolone

^aSome differences due to rounding. First line therapy costs: medication costs of first line treatment; Administration costs: costs of administration of first-line therapies; Adverse events and concomitant medication costs: costs associated with adverse events (treatment of adverse events, concomitant medication and transfusions); Medical resource use; all costs for disease management, such as follow up visits and tests; Second line treatment costs; costs for medication and administration of the subsequent line of treatment; Terminal care costs: costs for end-of-life care

Total costs and QALYs, as well as the ICERs are presented in bold text

Table 6 shows that VR-CAP patients have a longer PFS, whereas R-CHOP patients spend more time in the ‘progressed from second-line treatment’ health state than VR-CAP patients. This is due to the difference in PFS, while PPS is assumed to be equal between arms, generating a smaller difference in OS than PFS. The treatment cost accounts for the majority of the overall costs (Table 6), and therefore uncertainty around resource use and cost sources other than drug costs will have only a minor impact on model outcomes.

One-way sensitivity analysis showed that uncertainty in the parameters used within the model for PFS projections had the biggest impact on model outcomes together with the utility value applied to the ‘progressed from second-line treatment’ health state (Fig. 7).

The base case ICER of £20,043 indicates that VR-CAP is a cost-effective treatment option for patients with previously untreated MCL, using the standard UK threshold of £20,000–30,000 per QALY.

In the analysis, PFS is used as a surrogate for OS. This approach assumes that there is no survival benefit after a patients disease has progressed following treatment. When OS data were used directly to model cost effectiveness, the ICER increased slightly to £21,357. In this scenario it is assumed that there is a continued benefit of VR-CAP over R-CHOP after disease progression. The observation that OS, as modelled in the base case, shows a good reflection of the LYM-3002 data supports the use of PFS as a surrogate in the base case. A targeted literature review of NICE appraisals for cancer drugs from 2010 onwards identified two recent examples where PFS was used as a surrogate for OS either directly or indirectly (by assuming the same post-progression survival [PPS]) [37, 38]. In both cases, this methodology came under substantial scrutiny. Additionally, three submissions were identified where the same PPS was applied for all treatment arms [39–41].

There are some differences between the LYM-3002 trial population and MCL patients in the UK. As is often the case in clinical trials, the mean age of participants in LYM-3002 (64 years) was relatively low, compared with most patients who present at a median age of 73.5 in clinical practice in the UK [42]. Additionally, only 30 % of patients enrolled in LYM-3002 came from the European Union or North America, with no patients included from the UK. However, efficacy results showed consistency between geographic regions both in the size of benefit with VR-CAP and the absolute PFS for R-CHOP. It is therefore unlikely that the geographic spread of countries included in the trial and the lack of UK patients had any relevant impact upon the results.

The status of the OS data is the main uncertainty in assessing the cost effectiveness of treatment. Despite the conclusion that modelled OS was reasonably comparable to long-term datasets, OS data for VR-CAP are immature. Once the final analysis of OS for LYM-3002 is available, the model could be re-assessed to confirm robustness of the current analysis.

The model does not take into account rituximab maintenance (R-maintenance) treatment for patients that respond to induction therapy, which has been adopted in clinical practice in recent years based on the findings of the European MCL Elderly trial [11]. At the time of initiation of LYM-3002, R-maintenance was not commonly adopted and thus was not included in the trial design. There is a believe that R-maintenance therapy results in similar benefit after any CHOP-like induction regimen, and therefore we would expect to be able to give R-maintenance after VR-CAP induction with a similar extension to median survival times as observed with R-maintenance after R-CHOP induction [43]. As the European MCL Elderly trial was not designed to assess the clinical efficacy of induction therapy with versus without maintenance therapy, it could not be used to model R-maintenance.

When submitted to NICE, the evidence review group agreed that immature data may bias the extrapolation of survival data, and had some concerns about the methods used to overcome this. It was argued that if data are too immature to model OS for all patients, it would be questionable whether sufficient data are available to separately estimate long-term survival for patients with and without progression. However, the uncertainty was reduced for patients who had progressed as a smaller proportion of patients at risk were still alive at the time of evaluation. Furthermore, the data for the two treatment arms is pooled and thereby the total sample size is increased. The uncertainty of survival for patients who had not progressed may be increased by using this method, but this was accounted for by including general population mortality for patients that had not yet progressed. In doing so, it was assumed that all deaths in the PrePS curves (prior to adjustment for background mortality) in the trial were deaths from MCL. This was a reasonable assumption as the number of deaths reported in the LYM-3002 trial that were not due to progression or toxicity was very low. Of the 69 deaths in total in the VR-CAP group, there were only eight deaths that were not due to progression or AEs. In the R-CHOP group, there were a total of 87 deaths, of which 14 were not due to progression or AEs [22].

A submission for HTA was also made to the Scottish Medicines Consortium (SMC), who also noted that there are limitations arising from the maturity of the survival data, but found it unlikely that the approach taken would cause substantial bias in favour of VR-CAP. The SMC noted that this was supported by the literature providing evidence of an association between PFS and OS in MCL. In addition, it was noted that the modest impact on the ICER from uncertainty associated with varying survival inputs meant that the ICER for VR-CAP was robust [44].

In 2015 both NICE and the SMC accepted the overall approach taken in the cost-effectiveness model as a basis for their conclusion that VR-CAP represents a cost-effective treatment option for previously untreated MCL for whom HSCT is unsuitable, in the UK [22, 44]. VR-CAP is now recommended for use within the National Health Service.

The current model shows that VR-CAP is a cost effective treatment option for patients with previously untreated MCL, for whom haematopoietic stem cell transplantation is unsuitable, in the UK. Both NICE and SMC have recommended the use of VR-CAP in these patients.

AE, adverse event; ECOG, Eastern Cooperative Oncology Group; HMRN, Haematological Malignancy Research Network; HSCT, haematopoietic stem cell transplantation; HTA, health technology assessment; ICER, incremental cost-effectiveness ratio; IRC, independent review committee; IV, intravenous; MCL, mantle cell lymphoma; NHL, non-hodgkin lymphoma; NHS, National Health Services; NICE, National Institute for Health and Care Excellence; OS, overall survival; PFS, progression-free survival; PPS, post-progression survival; PrePS, pre-progression survival; PSA, probabilistic sensitivity analysis; QALY, quality-adjusted life year; R, rituximab; R-CHOP, rituximab, cyclophosphamide, doxorubicin, vincristine and prednisolone; R-FC, rituximab, fludarabine and cyclophosphamide; SEER, Surveillance, Epidemiology, and End Results Program; SMC, Scottish Medicines Consortium; TFI, treatment-free interval; UK, United Kingdom; VR-CAP, bortezomib, rituximab, cyclophosphamide, doxorubicin and prednisolone