Cost-effectiveness of lenalidomide plus dexamethasone vs bortezomib plus melphalan and prednisone in transplant-ineligible US patients with newly-diagnosed multiple myeloma

Abstract

Objective:

To conduct a cost-effectiveness assessment of lenalidomide plus dexamethasone (Rd) vs bortezomib plus melphalan and prednisone (VMP) as initial treatment for transplant-ineligible patients with newly-diagnosed multiple myeloma (MM), from a US payer perspective.

Methods:

A partitioned survival model was developed to estimate expected life-years (LYs), quality-adjusted LYs (QALYs), direct costs and incremental costs per QALY and LY gained associated with use of Rd vs VMP over a patient’s lifetime. Information on the efficacy and safety of Rd and VMP was based on data from multinational phase III clinical trials and a network meta-analysis. Pre-progression direct costs included the costs of Rd and VMP, treatment of adverse events (including prophylaxis) and routine care and monitoring associated with MM. Post-progression direct costs included costs of subsequent treatment(s) and routine care and monitoring for progressive disease, all obtained from published literature and estimated from a US payer perspective. Utilities were obtained from the aforementioned trials. Costs and outcomes were discounted at 3% annually.

Results:

Relative to VMP, use of Rd was expected to result in an additional 2.22 LYs and 1.47 QALYs (discounted). Patients initiated with Rd were expected to incur an additional $78,977 in mean lifetime direct costs (discounted) vs those initiated with VMP. The incremental costs per QALY and per LY gained with Rd vs VMP were $53,826 and $35,552, respectively. In sensitivity analyses, results were found to be most sensitive to differences in survival associated with Rd vs VMP, the cost of lenalidomide and the discount rate applied to effectiveness outcomes.

Conclusions:

Rd was expected to result in greater LYs and QALYs compared with VMP, with similar overall costs per LY for each regimen. Results of this analysis indicated that Rd may be a cost-effective alternative to VMP as initial treatment for transplant-ineligible patients with MM, with an incremental cost-effectiveness ratio well within the levels for recent advancements in oncology.

Keywords::

• Multiple myeloma
• Lenalidomide
• Bortezomib
• Cost-effectiveness
• Cost-benefit analysis
• Drug therapy

Introduction

Multiple myeloma (MM) is a malignant neoplasm that causes plasma cell accumulation in the bone marrow, eventually leading to bone destruction and marrow failure1. According to the American Cancer Society, ∼26,850 persons were newly diagnosed with MM in the US in 20142, representing ∼1.4% of incident cases of all cancer3. MM is most frequently diagnosed among patients aged 65–74 years, with a median age at diagnosis of 69 years3. Over the last decade, substantial improvements in survival outcomes have been achieved4^,5. Today, overall 5-year survival is over 50%3^,6. This increase in survival has been largely attributed to the use of newer treatment options, including lenalidomide, bortezomib, and thalidomide6.

Initial treatment recommendations for newly-diagnosed MM often include induction therapy with novel agents with or without chemotherapy, followed by autologous stem cell transplant (ASCT). However, a substantial number of patients from the US do not receive ASCT as part of their initial treatment due to patient/physician choice or because they are deemed ineligible for ASCT because of age or other reasons (e.g., considerable cardiac, pulmonary, renal, and/or hepatic dysfunction)7. The 2016 guidelines from the National Comprehensive Cancer Network (NCCN) for previously untreated, ASCT-ineligible patients include lenalidomide plus low-dose dexamethasone until disease progression (Rd) and bortezomib, melphalan, and prednisone (VMP) as ‘category 1’ therapy options, which indicate NCCN consensus, based on high-level evidence, that either intervention is appropriate1. (Melphalan, prednisone, and lenalidomide [MPL or MPR] and melphalan, prednisone, and thalidomide [MPT] also are listed as category 1 therapy options1.)

Lenalidomide and bortezomib are two of the most commonly used agents for initial treatment of MM in the US6. The efficacy and safety of Rd as a first-line therapy for ASCT-ineligible patients were demonstrated in the Frontline Investigation of Revlimid and Dexamethasone Versus Standard Thalidomide (FIRST) study. Results of this study showed that continuous treatment with Rd significantly improved progression-free survival (PFS; hazard ratio [HR] = 0.72; p = 0.00006) and provided an overall survival (OS) benefit at interim analysis (HR = 0.70; p = 0.01685) vs fixed duration of MPT in transplant-ineligible patients8. The efficacy and safety of VMP were established in the Velcade as Initial Standard Therapy in Multiple Myeloma (VISTA) study. Median time to progression and PFS were significantly longer for patients treated with VMP than patients treated with melphalan plus prednisone (MP). At the time final survival analyses were conducted (median follow-up = 60.1 months), a total of 49% of patients treated with VMP were alive vs 38% of patients treated with MP (HR = 0.695; p < 0.001); median OS was 56.4 months for patients treated with VMP vs 43.1 months for patients treated with MP9.

While the FIRST and VISTA studies have demonstrated the efficacy and safety of Rd and VMP and the potential for these two regimens to prolong survival, no randomized trials have been undertaken to directly compare Rd with VMP. A recent meta-analysis suggested that OS associated with VMP was equivalent to that of MPT10. This equivalence was also assumed in a previous cost-effectiveness analysis of bortezomib and thalidomide11. Given that FIRST demonstrated a favorable HR for survival for Rd relative to MPT8, results of the meta-analysis may suggest a possible advantage for Rd relative to VMP in terms of OS. This presumed survival benefit is likely to be associated with higher lifetime cost due to the longer survival of patients, as well as longer initial treatment duration, given that Rd is intended to be administered continuously until death or disease progression. Accordingly, the degree to which use of continuous Rd in lieu of fixed duration VMP is cost-effective is unknown. A decision-analytic model was developed to undertake a comparative cost-effectiveness assessment of continuous treatment with Rd vs fixed-duration VMP as first-line treatment for transplant-ineligible patients with MM, from a US payer perspective.

Patients and methods

Model description

A partitioned survival model was developed to examine the cost-effectiveness of Rd vs VMP. As with Markov state transition models, partitioned survival models characterize the condition of interest in terms of mutually exclusive health states. However, unlike Markov models, partitioned survival models do not require explicit calculation or use of transition probabilities, but rather estimate time spent in various health states based on area under the curve methodology. Our model consisted of three health states: (1) PFS; (2) post-progression survival (PPS); and (3) dead. The proportion of patients in each of these health states over time was estimated using survival functions for PFS and OS; PPS was estimated by subtracting PFS from OS. Expected life-years (LYs) were estimated as the area under the OS curve. Outcomes of interest generated by the model included expected lifetime total direct costs and total LYs and quality-adjusted life-years (QALYs). The cost-effectiveness of Rd vs VMP was estimated by dividing the difference in expected total direct costs by the corresponding difference in effectiveness (i.e., LYs or QALYs). The model was programmed in Microsoft Excel 2007.

Model parameters and data sources

Regimens

Consistent with FIRST, Rd consisted of lenalidomide (25 mg or adjusted according renal function) on days 1–21 of every 28-day cycle plus dexamethasone (40 mg or 20 mg/day according to age ≤75 years or >75 years, respectively) on days 1, 8, 15, and 22 of every 28-day cycle8. Use of Rd was modeled based on actual utilization observed in FIRST, which allowed the model to account for early discontinuation of therapy (i.e., discontinuation prior to disease progression or death) as well as missed doses treatment holidays taken by the patient. Additional treatment cut-offs included death, evidence of disease progression, or a maximum of 6 years of treatment, whichever occurred first. Consistent with the VISTA study, VMP consisted of single-use vials of bortezomib (1.3 mg/m²) for 8 days of each of the first four 6-week cycles and on days 1, 8, 22, and 29 of each subsequent 6-week cycle up to a maximum of 10 total cycles (while the original VISTA publication described intended treatment for a maximum of nine 6-week cycles12, a subsequent report indicated that patients were allowed to continue VMP beyond that time13) plus melphalan (9 mg/m²) and prednisone (60 mg/m²) on days 1–4 of each 6-week cycle. The post-progression choice of therapies was assumed to be conditional on the initial treatment received, consistent with information reported in the FIRST and VISTA studies8^,12^,13, as well as previous cost-effectiveness assessments in this therapeutic area11^,14.

PFS and OS

In the absence of randomized studies that compare Rd to VMP as initial therapy for transplant-ineligible patients with MM, an evidence network between the two studies (i.e., FIRST for Rd, VISTA for VMP) was developed using MPT (a comparator in FIRST) and MP (a comparator in VISTA; Figure 1)15. Linkage between MPT and MP was based on three studies—Intergroupe Francophone du Myélome (IFM) 99-06, IFM 01-01, and Sacchi et al.16—that included dose and duration of MPT similar to those used in FIRST16–18.

Figure 1. Evidence network of selected regimens for newly-diagnosed multiple myeloma.

Parametric proportional hazards models were used to estimate the necessary survival functions. Specifically, patient-level data from FIRST were used to derive a number of different parameterized PFS and OS functions for Rd. The best fit to the empirical data reported in FIRST was obtained using a Weibull distribution (Figure 2). Using the evidence network, a network meta-analysis (NMA) was used to estimate the HR for PFS and OS associated with Rd vs VMP. The PFS and OS functions for VMP were then estimated by applying the HRs to the corresponding Rd curves. The HRs for PFS and OS for Rd vs VMP obtained through the NMA were 0.69 (95% CI = 0.49–0.98) and 0.63 (95% CI = 0.44–0.92), respectively15. The fitted Weibull distributions for Rd PFS and OS and those for VMP obtained by applying HRs over a lifetime are presented in Figure 2.

Figure 2. Progression-free survival and overall survival parametric distributions.

Adverse events (AEs)

AEs included in the model were based on information reported in FIRST and VISTA and are listed in Table 1. In all instances, we limited our attention to AEs of grade 3/4 severity reported in ≥5% of patients.

Table 1. Percentages of patients receiving Rd or VMP reporting grade 3/4 adverse events.

Adverse events	Rd*	VMP†
Neutropenia	28	40
Anemia	18	19
Thrombocytopenia	8	38
Lymphopenia	6	20
Leukocytopenia	5	24
Pneumonia	8	7
Asthenia	8	6
Fatigue	7	8
Hypokalemia	7	7
Hyperglycemia	5	Not reported
Back pain	7	Not reported
Dyspnea	6	Not reported
Peripheral sensory neuropathy	1	13
Constipation	2	Not reported
Deep vein thrombosis	8	Not reported
Rash	6	Not reported
Cataract	6	Not reported
Neuralgia	Not reported	9
Diarrhea	4	8

Rd, lenalidomide and dexamethasone; VMP, bortezomib, melphalan, and prednisone.

*From analyses of data from the FIRST study8.

†From the VISTA study19.

For Rd, rates of each AE were based on analyses of patient-level data from FIRST; corresponding values for VMP were based on information from VISTA19. Rates of each AE were examined over discrete intervals of follow-up during FIRST, as follows: (1) cycles 0–6, (2) cycles 7–12, (3) cycles 13–18, and (4) all subsequent cycles. The rates were then converted to probabilities (Table 1)20. For VMP, differential AE rates by cycle were unavailable. Thus, the percentage of patients experiencing each AE as reported in VISTA was converted to an annual rate. This annual rate was expressed as a per-cycle rate by dividing by 13 (i.e., the number of 28-day cycles per year) and subsequently converted back to a probability. Patients were assumed to be at risk for AEs only during time spent on therapy.

Direct costs

The model considered the total direct costs of care for patients treated with Rd or VMP, including medical and pharmacy costs related to treatment of AEs (including prophylaxis), routine care and monitoring associated with pre- and post-progression, and post-progression treatment costs. All costs were estimated from the perspective of a US third-party payer using 2014 US dollars.

The per-cycle cost of lenalidomide was based on the wholesale acquisition cost (WAC) per capsule21, and the expected number of capsules used. Patient-level data from FIRST informed estimates of compliance with Rd, both in terms of monthly utilization as well as the percentage of progression-free patients continuing on therapy over time. The per-cycle cost of dexamethasone was estimated in the same fashion22. The assumed cost per cycle of Rd is shown in Table 212^,21–24.

Table 2. First-line drug costs.

Treatment name	Drug	Planned regimen	Cost per dose (or vial)	Percentage starting	Cost per 28-day cycle		Sources
					For regimen	Weighted average*
Rd	Lenalidomide	25 mg/day for 21 of 28 days 10 mg/day for 21 of 28 days 15 mg/day for 11 of 28 days	$455.61 (25 mg capsule) $455.61 (10 mg capsule) $455.61 (15 mg capsule)	68 23 9	$8744.98 $8744.98 $4372.49	$8355.83	Dose: Benboubker et al.8, 2014 Supplementary Appendix  Drug cost: WAC Wolters Kluwer Health Medi-Span Drug Information Database. Accessed: September 201421  Percentage starting: FIRST Trial Data on File
	Dexamethasone	40 mg/day for 4 of 28 days 20 mg/day for 4 of 28 days	$0.10 (4 mg tablet) $0.10 (4 mg tablet)	65 35	$3.89 $1.94	$3.21	Dose: Benboubker et al.8, 2014 Supplementary Appendix8  Drug cost: WAC Red Book Online Truven Health Analytics Inc. Accessed: 16 July 201422  Percentage starting: FIRST Trial Data on File
VMP	Bortezomib (first 4 cycles)	1.3 mg/m^2 for 8 of 42 days	$1612.84 (3.5 mg vial)	100	$9088.99	$9088.99	Dose: San Miguel et al.12, 2008  Drug cost: ASP + 6% Medicare Part B. Accessed: September 20142 3  Admin cost: CMS. Hospital Outpatient PPS 2014 File. Payment Rate24
	Bortezomib (cycles 5–9)	1.3 mg/m^2 for 4 of 42 days	$1612.84 (3.5 mg vial)	100	$4544.49	$4544.49	Dose: San Miguel et al.12, 2008  Drug cost: ASP + 6% Medicare Part B. Accessed: September 20142 3  Admin cost: CMS. Hospital Outpatient PPS 2014 File. Payment Rate24
	Melphalan	9 mg/m^2 for 4 of 42 days	$9.51 (2 mg tablet)	100	$212.20	$212.20	Dose: San Miguel et al.12, 2008  Drug cost: WAC Red Book Online Truven Health Analytics Inc. Accessed: 16 July 201422
	Prednisone	60 mg/m^2 for 4 of 42 days	$0.18 (50 mg tablet)	100	$1.09	$1.09	Dose: San Miguel et al.12, 2008  Drug cost: WAC Red Book Online Truven Health Analytics Inc. Accessed: 16 July 201422

ASP, average sales price; CMS, Centers for Medicare & Medicaid Services; d, day; mg, milligram; MM, multiple myeloma; PPS, prospective payment system; Rd, lenalidomide and dexamethasone; VMP, bortezomib, melphalan, and prednisone; WAC, wholesale acquisition cost.

*A relative dose intensity (RDI) was applied to lenalidomide and dexamethasone. RDI was defined as the number of pills or vials used divided by the number of pills or vials assigned as the treatment regimen. It was found to be 91.4% for lenalidomide and 94.0% for dexamethasone.

The per-cycle cost of bortezomib was calculated based on its 2014 US Centers for Medicare and Medicaid Services (CMS) payment limit23 per vial (i.e., average sale price [ASP] + 6%) and the assumed number of vials required. While the calculated dose per administration was estimated to be 2.4 mg for a typical patient with a body surface area of 1.86 m², it was assumed that a full vial would be billed, as is common in the US. Patients can receive bortezomib through subcutaneous or intravenous administration; in the absence of published data on relative frequency we arbitrarily assumed that VMP patients would receive an equal distribution of each at relevant CMS reimbursement rates ($108.90 intravenous administration [current procedural terminology (CPT) code 96409] and $73.79 subcutaneous administration [CPT code 96401])24. Acquisition costs of melphalan and prednisone, respectively, were based on the WAC from the Red Book Online22. The assumed cost per cycle of VMP (including acquisition and administration) is shown in Table 2.

Costs of each AE were based on published literature25–36. As necessary, costs were updated to 2014 US dollars by means of the medical care component of the Consumer Price Index (Table 3)25–36.

Table 3. Costs of adverse events.

Adverse event	Cost	Source
Neutropenia	$159	Goulart and Ramsey32, 2011
Anemia	$928	Goulart and Ramsey32, 2011
Thrombocytopenia	$159	Goulart and Ramsey32, 2011
Lymphopenia	$159	Goulart and Ramsey32, 2011
Leukopenia	$159	Goulart and Ramsey32, 2011
Pneumonia	$14,112	Eber et al.29, 2010
Asthenia	$278	Ramsey et al.34, 2006
Fatigue	$278	Ramsey et al.34, 2006
Hypokalemia	$1475	Sorensen et al.36, 2013
Hyperglycemia	$1378	Nichols et al.33, 2008
Back pain	$2499	Gatchel et al.31, 2003
Dyspnea	$275	Ramsey et al.34, 2006
Peripheral sensory neuropathy	$1078	Calhoun et al.26^, 2001
Constipation	$354	Singh et al.35, 2007
Deep vein thrombosis	$20,186	Elting et al.30, 2004
Rash	$140	Ramsey et al.34, 2006
Cataract	$2733	Brown et al.25, 2013
Neuralgia	$6741	Dworkin et al.28, 2008
Diarrhea	$2931	Dranitsaris et al.27, 2005

All patients were assumed to require periodic routine monitoring consisting of various laboratory tests (e.g., metabolic panels, liver function tests) and other assessments, with frequencies in part assumed to be dependent on treatment received, sourced from interviews with physicians; costs were based on relevant CPT-4 codes and CMS reimbursement rates (Table 4)24^,37.

Table 4. Use and cost of routine care and monitoring for multiple myeloma, by treatment.

Test or monitoring name	Cost per test	Number of assessments per year			Source
		Rd	VMP	Post-progression
Routine blood counts	$8.83	12.0	6.0	20.1	CMS. 2014 Clinical Diagnostic Laboratory Fee Schedule. National Limit Payment37
Clotting	$17.72	1.0	1.0	3.9
Internationalized normalized ratio	$5.37	0.5	0.5	2.6
Biochemistry (U&Es)	$14.96	10.0	5.0	17.3
Liver function tests	$11.14	5.0	3.0	14.6
Erythrocyte sedimentation rate	$3.69	1.0	1.0	2.6
Plasma viscosity	$15.92	0.5	0.5	1.6
Uric acid (urate)	$6.16	1.0	1.0	2.7
Immunoglobulin	$10.94	6.0	6.0	9.7
Paraprotein measurements	$30.58	12.0	12.0	11.1
Protein electrophoresis	$14.65	4.0	4.0	9.6
Serum β2 microglobulin	$22.07	2.0	2.0	5.0
C-reactive protein	$7.06	1.0	1.0	3.3
Serum erythropoietin level	$25.65	0.1	0.1	0.5
Immunofixation	$30.48	4.0	3.0	4.8
Creatinine clearance	$12.89	0.5	0.5	2.3
Routine urinalysis	$4.32	2.0	1.0	4.4
24-h urine measurement	$4.09	3.0	2.0	3.0
24-h urine for creatinine	$7.06	0.5	0.5	1.4
24-h total urine protein	$5.00	0.5	0.5	3.2
Urine protein electrophoresis/light chains	$24.33	2.0	2.0	4.9
Urine immunofixation	$40.04	1.0	1.0	2.1
Skeletal survey by X-ray	$90.89	0.0	0.0	1.6	CMS. Hospital Outpatient PPS File. 2014 Payment Rate24
Skeletal survey by X-ray individual sites	$90.89	0.3	0.3	1.6
Magnetic resonance imaging	$492.92	0.0	0.0	0.9
Bone densitometry	$90.16	0.0	0.0	0.1
Bone marrow aspirate	$640.91	0.3	0.3	2.1
Bone marrow trephine biopsy	$640.91	0.0	0.0	2.0
Bacterial investigation	$14.09	0.5	0.5	1.6	CMS. 2014 Clinical Diagnostic Laboratory Fee Schedule. National Limit Payment37
Calcium	$7.04	0.3	0.3	3.0
Albumin	$6.75	0.5	0.5	3.0
Lactase dehydrogenase	$8.23	0.0	0.0	0.3
Cost per cycle per patient	N/A	$105	$88	$411

CMS, Center for Medicare and Medicaid Services; N/A, not applicable; PPS, prospective payment system; Rd, lenalidomide and dexamethasone; U&Es, urea and electrolytes; VMP, bortezomib, melphalan, and prednisone.

Agents used in second- and third-line therapy were assumed to be conditional on prior line(s) of treatment (e.g., patients who received VMP as first-line therapy were assumed to be more likely to receive Rd as second-line therapy). Regimens included as second- and third-line therapy (collectively, ‘subsequent therapy’) included Rd, bortezomib and thalidomide. Duration of subsequent therapy was assumed to be independent of initial treatment received and was estimated based on patterns of utilization of these agents, as reported in clinical trials among patients with relapsed/refractory disease. A decision tree was used to track treatment pathways over time and to calculate weighted costs of second- and third-line therapy (Table 5)21–24^,38–40. The tests and assessments were assumed to be the same as those used during pre-progression. However, the frequency with which they were administered was assumed to be higher than that in the pre-progression phase (Table 4).

Table 5. Cost of second- and third-line therapy.

Regimen name	Cost per dose (or vial)	Cost per 28-day cycle	Duration (months)		Percentage of Patients Post Rd		Percentage of Patients Post VMP		Sources
			2nd line	3rd line	2nd line	3rd line	2nd line	3rd line
Rd	$455.61 (Lenalidomide 25 mg capsule) $0.10 (Dexamethasone 4 mg tablet)	$8853.06	18.0	13.2	25	42	75	42	Durations: Derived from Stadtmauer et al.40, 2009 Len Dose: Stadtmauer et al.40, 2009 Drug cost: WAC Wolters Kluwer Health Medi-Span Drug Information Database. Accessed: September 201421 Dex Dose: Stadtmauer et al.40, 2009 Drug cost: WAC Red Book Online Truven Health Analytics Inc. Accessed: 16 July 201422
V	$1612.84 (3.5 mg vial)	$7687.15	4.1	4.1	75	41	25	41	Durations: Derived from Richardson et al.39, 2005 Dose: Richardson et al.39, 2005 Drug cost: ASP + 6% Medicare Part B. Accessed: September 20142 3 Admin cost: CMS. Hospital Outpatient PPS 2014 File. Payment Rate24
T	$277.41 (200 mg capsule)	$7767.41	4.5	5.8	0	17	0	17	Duration: Derived from Kropff et al.38, 2012 Dose: Kropff et al.38, 2012 Drug Cost: WAC Red Book Online Truven Health Analytics Inc. Accessed: 16 July 201422

ASP, average sales price; CMS, Center for Medicare and Medicaid Services; Dex, dexamethasone; Len, lenalidomide; Rd, lenalidomide and dexamethasone; T, thalidomide; V, bortezomib; VMP, bortezomib, melphalan, and prednisone; WAC, wholesale acquisition cost.

Health-state utilities

Estimates of pre-progression health-state utility for Rd and VMP were based on information collected during FIRST and VISTA, respectively. For Rd, health-state utility values associated with PFS were calculated based on responses to the EuroQol 5-dimension questionnaire (EQ-5D), which was administered at baseline, as well as at various points during follow-up (i.e., 1, 3, 6, 12, and 18 months after treatment initiation, and at study discontinuation) and mapped to utility values using the time trade-off method for the UK general population41. These utility values were used to derive a predictive equation that estimated Rd-specific health-state utility values over the duration of PFS (Figure 3). Because these values were based on the pre-progression experience of the intent-to-treat population enrolled in FIRST, including the impact of AEs in patients who experienced such events, no additional decrement associated with treatment-related AEs was needed.

Figure 3. Health state utilities over time.

A similar approach was taken for VMP. However, because patient-level data from VISTA were unavailable, estimates were based on a previous publication in which European Organization for Research and Treatment of Cancer Quality of Life Questionnaire C30 (EORTC QLQ-C30) values were provided for all randomized individuals during each of the initial nine cycles of treatment as well as at the end-of-treatment visit42. These scores were then mapped to the EQ-5D using methods described by Proskorovsky et al.43 and subsequently converted to health-state utility scores using the time trade-off method for the UK general population44. Similar to the case with Rd, the impact of AEs in patients who experienced such events were included in the QLQ-C30 measurements taken in the VISTA trial; thus, no additional decrement associated with treatment-related AEs was needed.

For both Rd and VMP, the pre-progression health-state utility at baseline was assumed to be 0.53, consistent with that reported among patients enrolled in FIRST41; subsequent pre-progression health-state utility values varied over time, with a maximum value of 0.67 for Rd and 0.65 for VMP (Figure 3). The post-progression health-state utility value of 0.59 was also calculated using information from FIRST. Unlike pre-progression values, the health-state utility accorded to post-progression was assumed to be invariant with respect to first-line treatment received and time since disease progression.

Analyses

For each initial regimen, the model calculated total expected LYs, QALYs, and lifetime costs, respectively. Incremental cost-effectiveness ratios (ICERs) of Rd vs VMP were reported in terms of the incremental cost per LY gained and per QALY gained.

Deterministic sensitivity analyses varied several key input parameters, as shown in Table 6. Except for the discount rates, OS and PFS HRs, baseline and post-progression utilities, and the cost of bortezomib administration, all parameters were varied ±20% of their base case value, an assumption taken to vary the inputs over a wide range of possibilities in lieu of confidence intervals, which were lacking for many of the parameter estimates. The discount rates were varied between 1.0% and 5.0%; OS and PFS, between bounds of their 95% confidence intervals; utilities, also between bounds of their 95% confidence intervals; and cost of bortezomib administration, between the costs of assuming all patients using the subcutaneous formulation and all patients using the intravenous formulation. A probabilistic sensitivity analysis was used to account for uncertainty in model parameter estimates, as shown in Table 6, via a Monte Carlo analysis run with 1000 simulations.

Table 6. Deterministic and probabilistic sensitivity analyses parameters.

	Base case	Lower value for DSA	Upper value for DSA	Source for DSA values	Distribution for PSA	Standard error for PSA	Source for standard error
Annual discount rates
Effectiveness	3.0%	1.0%	5.0%	Assumption	–	–	–
Costs	3.0%	1.0%	5.0%	Assumption	–	–	–
Effectiveness
OS Rd Weibull Intercept	4.7102	–	–	–	Normal*	0.1650	FIRST Trial Data on File
OS Rd Weibull Scale	1.0565	–	–	–	Normal*	0.0847	FIRST Trial Data on File
OS HR of VMP vs Rd	1.587	1.087	2.273	Weisel et al.15, 2015	Log-Normal	0.193	Weisel et al.15, 2015
PFS Rd Weibull Intercept	3.8031	–	–	–	Normal*	0.0732	FIRST Trial Data on File
PFS Rd Weibull Scale	1.1584	–	–	–	Normal*	0.0603	FIRST Trial Data on File
PFS HR of VMP vs Rd	1.449	1.020	2.041	Weisel et al.15, 2015	Log-Normal	0.179	Weisel et al.15, 2015
Mortality risk
First-Line Rd	20.1%	16.1%	24.1%	±20%	Beta	2.1%	Assumption†
First-Line VMP	16.8%	13.4%	20.2%	±20%	Beta	1.7%	Assumption†
Second-Line Rd	17.5%	14.0%	21.0%	±20%	Beta	1.8%	Assumption†
Second-Line Bortezomib	17.5%	14.0%	21.0%	±20%	Beta	1.8%	Assumption†
Adverse event rates
Rd	Multiple‡	−20%	+20%	±20%	Normal	Multiple‡	Assumption†
VMP	Multiple‡	−20%	+20%	±20%	Normal	Multiple‡	Assumption†
Utility
Baseline	0.53	0.51	0.55	FIRST Trial Data on File	Beta	0.01	FIRST Trial Data on File
Post-Progression	0.59	0.56	0.62	FIRST Trial Data on File	Beta	0.01	FIRST Trial Data on File
Costs per dose or vial
Lenalidomide (all strengths)	$455.61	$364.49	$546.73	±20%	–	–	–
Dexamethasone	$0.10	$0.08	$0.12	±20%	–	–	–
Bortezomib	$1612.84	$1290.27	$1935.41	±20%	–	–	–
Melphalan	$9.51	$7.61	$11.41	±20%	–	–	–
Prednisone	$0.18	$0.14	$0.22	±20%	–	–	–
Thalidomide	$277.41	$221.93	$332.89	±20%	–	–	–
Other costs
Bortezomib cost per administration	$91.35	$73.79	$108.90	Assumption	Gamma	$9.32	Assumption†
AE management costs	Multiple‡	−20%	+20%	±20%	Gamma	Multiple‡	Assumption†
Resource use and monitoring costs per cycle per patient
First-Line Rd	$105.42	$84.34	$126.51	±20%	Gamma	$10.76	Assumption†
First-Line VMP	$88.05	$70.44	$105.66	±20%	Gamma	$8.98	Assumption†
Post-Progression	$410.80	$328.64	$492.96	±20%	Gamma	$41.92	Assumption†
Treatment duration
First-Line Rd - Exponential Intercept	3.2983	–	–	–	Normal	0.0811	FIRST Trial Data on File
Second-Line Rd (months)	17.97	14.38	21.57	±20%	Normal	1.83	Assumption†
Second-Line Bortezomib (months)	4.08	3.27	4.90	±20%	Normal	0.42	Assumption†
Third-Line Rd (months)	13.23	10.58	15.87	±20%	Normal	1.35	Assumption†
Third-Line Bortezomib (months)	4.08	3.27	4.90	±20%	Normal	0.42	Assumption†
Third-Line Thalidomide (months)	5.80	4.64	6.95	±20%	Normal	0.59	Assumption†

*Cholesky decomposition used with covariance 0.011439 for OS parameters and 0.001388 for PFS parameters.

†Standard error assumes that bounds of 95% confidence interval are 20% above and below the mean.

‡Multiple indicates a group of parameters were varied.

In addition, three scenario analyses were undertaken to see the effect of several assumptions on the cost-effectiveness estimates. Specifically, to test the impact of allowing some vial sharing in bortezomib-treated patients, we conducted a scenario in which 50% of patients were assumed to share vials (i.e., no drug wastage) and 50% were assumed to use part of a vial, with the remaining medication discarded. To test the impact of using utilities from different sources with different methodologies, we conducted a scenario in which VMP-treated patients were assumed to have the same utility over time as Rd-treated patients (i.e., those derived from the FIRST trial). Lastly, in order to see how much the treatment duration for Rd-treated patients impacted the results, we allowed patients to be treated until death, disease progression, or treatment discontinuation (without any maximum treatment duration imposed).

All analyses were undertaken from the perspective of a US third-party payer. A threshold of $125,000 per QALY was used in evaluations of cost-effectiveness, which was consistent with a prior examination of the cost-effectiveness of trastuzumab in US patients with metastatic breast cancer45 and with an updated assessment of the benchmark value of a life year gained from dialysis46. All future benefits and costs were discounted at 3% annually.

Results

Base case

On an undiscounted basis, relative to VMP, use of Rd as first-line therapy for ASCT-ineligible patients in the US with MM was expected to result in an additional 3.23 LYs (8.60 vs 5.37 for VMP), 1.30 progression-free LYs (PFLYs; 3.72 vs 2.42), 1.94 post-progression LYs (PPLYs; 4.89 vs 2.95) and 2.09 QALYs (5.35 vs 3.27; Table 7). After discounting at an annual rate of 3%, the expected gains were 2.22 LYs, 1.06 PFLYs, 1.16 PPLYs, and 1.47 QALYs.

Table 7. Base case results: cost-effectiveness outcome (average per patient).

	Undiscounted			Discounted
	Rd	VMP	Difference/ICER	Rd	VMP	Difference/ICER
Outcomes (years)
PFLYs	3.72	2.42	1.30	3.28	2.23	1.06
PPLYs	4.89	2.95	1.94	3.52	2.36	1.16
LYs	8.60	5.37	3.23	6.81	4.59	2.22
QALYs	5.35	3.27	2.09	4.26	2.79	1.47
Costs (USD)
Pre-progression treatment and administration	–	–	–	$202,868	$76,660	$126,209
Prophylaxis				$2810	$55	$2755
AE management				$3623	$1787	$1836
Routine care and monitoring				$4502	$2550	$1952
Post-progression	–	–	–	$110,992	$164,767	−$53,775
Total	–	–	–	$324,795	$245,819	$78,977
Total cost per LY	–	–	–	$47,717	$53,611	−$5894
Incr cost per QALY (Rd vs VMP)	–	–	–	–	–	$53,826
Incr cost per LY (Rd vs VMP)	–	–	–	–	–	$35,552

ICER, incremental cost-effectiveness ratio; Incr, increase incremental; LYs, life-years; PFLYs, progression-free life-years; PPLYs, post-progression life-years; QALYs, quality-adjusted life-years; Rd, lenalidomide and dexamethasone; VMP, bortezomib, melphalan, and prednisone.

Discounted lifetime total costs of care per patient were $324,795 for Rd and $245,819 for VMP, a difference of $78,977. The incremental cost was primarily attributable to higher total expected costs of Rd relative to VMP ($213,803 vs $81,052), which were partially offset by higher post-progression costs for VMP relative to Rd ($110,922 vs $164,767). The incremental cost per QALY gained with Rd vs VMP was $53,826; the incremental cost per LY gained was $35,552.

Sensitivity analyses

Results of deterministic sensitivity analyses are displayed in Figure 4. The model was most sensitive to changes in the HR for OS associated with Rd vs VMP, the cost per capsule of lenalidomide (all strengths and lines), and the discount rate applied to effectiveness outcomes. It was least sensitive to the costs per day of prednisone, dexamethasone, and thalidomide, the treatment durations for bortezomib and thalidomide in the third line, and the AE rates for VMP.

Figure 4. Tornado diagram illustrating results of deterministic sensitivity analyses.

Results of probabilistic sensitivity analyses are shown in Figures 5 and6. In probabilistic sensitivity analyses, 99.6% of simulations resulted in a positive incremental cost-effectiveness ratio (i.e., quadrant I), indicating that Rd was associated with greater life expectancy, albeit at a higher cost. At a willingness-to-pay (WTP) threshold of $125,000 per QALY gained, Rd was found to be cost-effective in 90.6% of simulations.

Figure 5. Incremental cost-effectiveness plane for probabilistic sensitivity analyses.

Figure 6. Cost-effectiveness acceptability curve.

Scenario analyses

In a scenario analysis that assumed 50% of bortezomib vials would be re-used by other patients, expected total costs of patients treated with VMP decreased by $10,867 ($234,952 vs the base-case estimate of $245,819). The resulting estimated incremental cost per QALY gained with Rd vs VMP was $61,233; the corresponding value for the incremental cost per LY gained was $40,444.

In a second scenario analysis in which the utilities over time assumed for VMP-treated patients were set equal to those assumed for Rd-treated patients, expected total QALYs of VMP-treated patients increased by 0.07 years (2.86 vs the base-case estimate of 2.79). This resulted in an incremental cost per QALY gained with Rd vs VMP of $56,503, an increase of $2676 from the base-case. As expected, the incremental cost per LY gained remained unchanged from the base case.

In a third scenario analysis in which Rd use was assumed to continue until death, disease progression, or discontinuation (i.e., without a duration cap imposed), expected total healthcare costs in patients treated with this regimen increased by $10,142 ($334,937 vs the base-case estimate of $324,795; no other values changed relative to base-case estimates). The resulting estimated incremental cost per QALY gained with Rd vs VMP was $60,738; the corresponding value for the incremental cost per LY gained was $40,118.

Discussion

While the use of lenalidomide and bortezomib as first-line therapy in MM has improved outcomes, questions remain regarding increased costs. In the absence of direct random-ized clinical comparisons between these two agents, this analysis modeled the long-term implications based on indirect comparison methods. In the base case analyses, use of Rd was estimated to result in a gain of 3.23 LYs relative to VMP; the corresponding gain in QALYs was 2.09 (all undiscounted). The resulting estimates of cost-effectiveness—$53,826 per QALY gained and $35,552 per LY gained—were well within the range of other novel oncology treatments45^,47–49. Given that Rd is intended for continuous use until disease progression or death while VMP has a finite duration of therapy, it is not surprising that our findings were most sensitive to assumptions concerning OS of VMP relative to Rd and the cost of lenalidomide.

As evidenced by the availability of other agents with relatively high ICERs, treatment decisions in oncology in the US are not based on a formal dollar threshold for value. In a recent survey of oncologists regarding willingness to use various hypothetical agents based on price and expected gains in survival, the median response was associated with a treatment that was $125,000 more than standard care and that yielded an additional 6 months of life (this translates to a cost per LY gained of $250,000)50. One benchmark commonly cited in the US (i.e., $50,000 per QALY) was based on the cost of dialysis for 1 year51, this benchmark was re-evaluated and updated in 2009 to $129,000 per QALY46.

Although no other economic models have been conducted to compare these two regimens specifically, other models have been developed to compare different first-line regimens in MM. Garrison et al.14 used a Markov model to evaluate MP, VMP, MPT, and MPR (MPR is identical to MPL; referred to as MPR and MPR-R without and with R maintenance, respectively). Results from their analysis indicated VMP was associated with lower costs and greater benefit than the MPR-R regimen14. However, several aspects of that model warrant further inspection. For example, based on the reported mean total expected cost of VMP ($39,754) and the mean duration of VMP therapy (39 weeks, which would provide a total of 40 doses as opposed to 52 doses associated with the full 54-week course of therapy), it appears that the authors assumed a cost per dose of bortezomib of $994 (i.e., $39,754/40). Since the actual price of bortezomib is nearly twice as high (i.e., ∼$1613 per 3.5 mg vial)23, we surmise that their cost calculations only considered the quantity of bortezomib administered, and not the full cost of the single-use vial. Had full-vial costs been used, VMP costs would have been ∼50% higher. There is also a question concerning the granulocyte colony-stimulating factor (G-CSF) costs ascribed to every patient on the MPR-R regimen in the model, at a cost of $3000 every month for neutropenia treatment and an additional $922 every month for G-CSF prophylaxis, despite the lack of evidence of such in the clinical trial and the fact that routine G-CSF prophylaxis is not recommended for patients receiving lenalidomide. The assumptions of clinical benefit from VMP also warrant closer inspection. In an earlier publication of their analysis52, the authors noted that, of the total survival benefit attributed to VMP, 94% of it was due to LY gains from post-progression treatment and 6% from extension of progression-free survival attributable to VMP; however, this breakout of pre- vs post-progression benefit was not included in the final publication.

Kim et al.53 evaluated the cost-effectiveness of lenalidomide maintenance therapy for patients with MM. Based on use of an unconventional measure evaluating the Average Cumulative Cost per Progression-Free Survivor, it was deemed that the addition of lenalidomide to MP in the induction and also in the maintenance setting was ‘prohibitively costly’. Further, it was assumed that all patients remained on therapy until progression of the disease, which would result in a duration of therapy that would be double that noted in FIRST with Rd, with corresponding impact on drug and AE costs. The present model used conventional economic methods, reported duration and utilization from clinical trial data, and provided transparency regarding LYs gained before and after progression. Using these techniques, the present model showed that Rd provided additional LYs and QALYs at a roughly commensurate higher cost compared with VMP, with a relatively similar cost per LY ($5894 in favor of Rd) and an ICER well within the ratios seen for other advances in oncology, and represented a cost-effective option as first-line treatment for ASCT-ineligible MM relative to VMP.

There are certain limitations associated with this model that merit mention. Effectiveness comparisons were based on an NMA that used MP and MPT to ‘bridge’ between the two relevant studies (i.e., FIRST for Rd; VISTA for VMP). A key assumption of an NMA is the homogeneity of individuals across all included studies. While patients across the five studies were comparable in most respects, patients enrolled in IFM 99-06, IFM 01-01, and the study conducted by Sacchi et al.16 showed trends toward being less likely to have more advanced disease and more likely to have bone lesions (not reported in Sacchi et al.16). These differences should have had little impact on the estimate of relative efficacy of Rd vs VMP, because only relative efficacy (i.e., HRs of OS or PFS) from each study was considered in the NMA analysis to preserve randomization. In addition, the relative treatment efficacy of MPT vs MP had no heterogeneity (I²= 0) across both IFM trials, demonstrating that differences in baseline factors, particularly age, do not have much impact on the relative treatment efficacy15.

Another limitation is that use of a partitioned survival model required complete estimates of PFS and OS for each of the two regimens. The duration of follow-up in FIRST did not extend to the lifetime of all patients enrolled in the study, so data were extrapolated based on parametric models. Information on PFS was relatively complete (∼70% of Rd patients had progressed or died by the end of follow-up), but corresponding information for OS was not (4-year survival was 59% among Rd patients). The degree to which parameterized functions calculated to represent OS curves for Rd and VMP approximate actual long-term survival is not known.

Utility values used in this analysis were calculated from the full patient populations of the FIRST and VISTA trials and are not limited to US patients. In addition, UK population tables were used to generate utility score from collected PRO data. Deterministic sensitivity analysis suggested that utility values had minimal impact on the calculated ICER.

On a related matter, the methods of utility estimation for Rd and VMP differed due to the nature of the available data. Unlike the FIRST trial, where EQ-5D responses were converted directly to health-state utilities, a mapping algorithm was used to convert EORTC QLQ-C30 values obtained during the conduct of VISTA to EQ-5D values, which were subsequently converted to utilities. Furthermore, the derived utilities for VMP were adjusted to ensure both modeled Rd- and VMP-treated patients started with the same baseline utility (i.e., that observed in the Rd arm of the FIRST trial). Although the mapping study was conducted in the target population, we cannot be sure that the mapped EQ-5D values would have been the same had they been collected in the VISTA trial. In order to test the impact of utility values used for VMP, a scenario was conducted in which the utilities of VMP-treated patients were assumed equal to those of Rd-treated patients. This resulted in a relatively modest increase in the ICER of Rd relative to VMP of $2677 per QALY.

Our model assumed constant costs for each time period in a disease state with a particular regimen. A recent analysis by Arikian et al.54 showed that, for patients with MM who received first-line therapy with either lenalidomide or bortezomib, monthly total costs (drug and medical) were roughly $16,000 in the initial quarter and declined steadily each quarter until plateauing at a 68% reduction 18 months later, and then rose sharply upon beginning second-line therapy. This finding, and similar work by Gaultney and Uyl-de Groot55 showing that patients in later lines of therapy had higher monthly costs, suggest an economic benefit for Rd due to the longer extension of time until the next therapy and consequent longer period of time in a below-average cost state.

Conclusions

Continuous first-line therapy has been associated with improved disease management and outcomes. Results of our model showed that use of Rd as first-line therapy in the US was estimated to provide more LYs and QALYs relative to VMP in transplant-ineligible patients with MM at a commensurately higher cost. Despite the additional costs incurred with continuous treatment, the extension of clinical benefit resulted in cost-effectiveness ratios within accepted levels published in other front-line oncology settings. As such, Rd may be a cost-effective alternative to VMP as initial treatment for transplant-ineligible patients with MM.

Transparency

Declaration of funding

This work was supported by research funding provided by Celgene Corporation, Summit, NJ, USA.

Declaration of financial/other relationships

SZU has served as a consultant to Celgene Corporation. AB, SG, and CGP are employed by Evidera Inc., a contract research organization that received funding from Celgene Corporation for this study. GB, WY, AEH, and YN are employed by and have an ownership interest in Celgene Corporation. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgments

The authors thank Meg Franklin of Franklin Pharmaceutical Consulting for medical writing assistance, and MediTech Media, for medical editing assistance, sponsored by Celgene Corporation.