A cost-effectiveness analysis of denosumab for the prevention of skeletal-related events in patients with multiple myeloma in four European countries: Austria, Belgium, Greece, and Italy

Abstract

Aim: The approved indication for denosumab (120 mg) was expanded in 2018 to include skeletal-related event (SRE) prevention in patients with multiple myeloma (MM). Therefore, a cost-effectiveness analysis was conducted comparing denosumab with zoledronic acid (ZA) for SRE prevention in patients with MM from the national healthcare system perspective in a representative sample of European countries: Austria, Belgium, Greece, and Italy.

Methods: The XGEVA global economic model for patients with MM was used to calculate incremental cost-effectiveness ratios (ICERs) for denosumab vs ZA over a lifetime horizon. Clinical inputs were derived from the denosumab vs ZA randomized, phase 3 study (“20090482”) in patients newly-diagnosed with MM, and comprised real-world adjusted SRE rates, serious adverse event (SAE) rates, treatment duration, dose intensity, progression-free survival (PFS), and overall survival (OS). Economic inputs comprised country-specific denosumab and ZA acquisition and administration costs, SRE and SAE management costs, and discount rates. Health utility decrements associated with MM disease progression, SRE and SAE occurrence, and route of administration were included.

Results: Estimated ICERs (cost per quality-adjusted life-year [QALY] gained) for denosumab vs ZA in Austria, Belgium, Greece, and Italy were €26,294, €17,737, €6,982, and €27,228, respectively. Using 1–3 times gross domestic product (GDP) per capita per QALY as willingness to pay thresholds, denosumab was 69–94%, 84–96%, 79–96%, and 50–92% likely to be cost-effective vs ZA, respectively.

Limitations: Economic inputs were derived from various sources, and time to event inputs were extrapolated from 20090482 study data.

Conclusions: Denosumab is cost-effective vs ZA for SRE prevention in patients with MM in Austria, Belgium, Greece, and Italy, based on often-adopted World Health Organization thresholds. This conclusion is robust to changes in model parameters and assumptions. Cost-effectiveness estimates varied across the four countries, reflecting differences in healthcare costs and national economic evaluation guidelines.

Keywords:

• Cost-effectiveness
• denosumab
• Europe
• multiple myeloma
• skeletal-related events
• zoledronic acid

JEL classification codes:

• I10
• I19

Introduction

The proliferation of malignant plasma cell clones in patients with multiple myeloma (MM) results in organ dysfunction and bone damage¹. Despite being rare among cancers—representing ∼1.1% of cases in Europe in 2018²—the crude incidence of MM in 2018 in Austria, Belgium, Greece, and Italy per 100,000 people was 6.4, 8.9, 7.3, and 10.2, respectively². Furthermore, MM incidence is projected to increase by 29% by the year 2040, as the population of Europe ages, and life expectancies increase².

Lytic bone lesions are a common feature of MM and are associated with substantial morbidity and mortality^1,3; most patients (∼80–90%) develop lytic bone lesions during the disease course⁴. Lytic bone lesions cause pain and reduce the structural integrity of bone, thereby increasing the risk of skeletal-related events (SREs)^3,5. SREs include pathologic fracture, spinal cord compression, radiation to bone, and surgery to bone^3,5. Most patients with MM (∼70–95%) will experience at least one SRE during their disease course⁶. Importantly, SREs are associated with a poor prognosis, and with substantial morbidity, pain, and reduced quality-of-life^3,7–9. In addition to their negative impact on patients, SREs are associated with increased healthcare resource use (HRU), including emergency department visits, hospitalization, outpatient clinic attendance, and analgesic use^6,10–13; therefore, SREs can increase MM direct healthcare costs by ∼75% vs those without SREs¹². Furthermore, multiple SREs, which patients often experience, are associated with greater HRU and costs than isolated events^11,12.

Given their negative impact on patients and healthcare systems, it is important to prevent SREs using denosumab, or bisphosphonates such as zoledronic acid (ZA)^3,14–16. Both agents are established treatments for patients with bone metastases from solid tumors^17,18; however, ZA was the standard of care in patients with MM before 2018, which is when the approved indication for denosumab was expanded to SRE prevention in adults with advanced malignancies involving bone^3,16–19. Both denosumab and ZA are effective at preventing SREs in patients with MM, although they have different mechanisms of action and routes of administration (subcutaneous injection vs intravenous infusion, respectively)^20–22. For SRE prevention, denosumab (120 mg) is administered every 4 weeks (Q4W) and ZA (4 mg [dose adjusted]) is administered every 3–4 weeks^17,18.

In addition to lytic bone lesions, most patients with MM (up to ∼60%) develop RI²³, and worsening renal function is a common feature of MM that arises owing to both the disease process and its treatment^1,16,23,24. Unlike ZA, denosumab is not cleared by the kidneys¹⁷; ZA requires dose adjustment for use in patients with renal impairment (RI) and it is not recommended for use in cases of severe RI (creatinine clearance [CrCl] < 30 mL/min)¹⁸. Current European clinical practice guidelines, published before the approval of denosumab for use in patients with MM, recommend bisphosphonates for SRE prevention in patients with lytic bone lesions^15,16,19. International guidelines also recommend bisphosphonates for patients without evidence of lytic bone lesions on conventional radiographs³, and recently updated guidelines from the US suggest denosumab as an alternative to bisphosphonates for patients with MM^25,26.

Denosumab is superior to ZA for SRE prevention in patients with bone metastases from solid tumors²⁷. The subsequent approval of denosumab for use in patients with MM¹⁷ was based on the largest study in MM to date: an international, phase 3, double-blind, double-dummy, randomized study (“the 20090482 study”; NCT01345019)²⁰, which found denosumab to be non-inferior to ZA for delaying time to first on-study SRE (primary endpoint) in 1,718 patients with newly-diagnosed MM (hazard ratio [HR] = 0.98 [95% confidence interval (CI) = 0.85–1.14], p = 0.010)²⁰. Superiority of denosumab over ZA for SRE prevention—a secondary endpoint—was not demonstrated²⁰. As most first on-study SREs occurred within the first 3 months (60%), during which time denosumab and ZA may not have been able to take full effect, differences in efficacy of the two agents may not have been fully captured by the primary analysis; therefore, a post hoc landmark analysis was performed to capture differences in efficacy of the two agents over time²⁰. This analysis suggested that denosumab is superior to ZA for preventing first SRE starting at 15 months (HR = 0.66 [95% CI = 0.44–0.98], p = 0.039)²⁰.

Denosumab is a fully human monoclonal antibody targeting the receptor activator of nuclear factor-κB (RANK) ligand (RANKL)^17,22. In addition to its involvement in bone complications, there is evidence to suggest that RANKL–RANK signaling may play a role in MM pathogenesis^20,28. This hypothesis was supported by the 20090482 study, which suggested that, when used in addition to anti-MM treatment, denosumab improved median progression-free survival (PFS) by 10.7 months vs ZA; PFS was a pre-specified, exploratory endpoint (HR = 0.82 [95% CI = 0.68–0.99], p = 0.036)²⁰. Overall survival (OS)—a secondary endpoint—was similar in both arms (HR = 0.90 [95% CI = 0.70–1.16], p = 0.41), although too few deaths had occurred at the time of analysis²⁰.

As modern healthcare systems have competing demands on HRU and spending²⁹, it is increasingly important to demonstrate the value of novel treatments to multiple stakeholders³⁰. Using clinical data from the 20090482 study²⁰, an economic analysis conducted in the US concluded that denosumab was cost-effective vs ZA for SRE prevention in patients with MM from both a payer and societal perspective³¹. However, there are no published European cost-effectiveness analyses of denosumab in this patient population, and cost-effectiveness results cannot be translated from one healthcare setting to another. Given the burden of SREs on patients and healthcare systems, the efficacy of denosumab in SRE prevention, and the potential of denosumab to extend PFS vs ZA in patients with MM without impairing renal function, it is important to evaluate the cost-effectiveness of denosumab vs ZA in this patient population. Using the same economic model structure as Raje et al.³¹ in the US, we estimated the incremental cost-effectiveness ratio (ICER: incremental cost per quality-adjusted life-year [QALY] gained) for denosumab vs ZA from the national healthcare system perspective in four European countries (Austria, Belgium, Greece, and Italy).

Methods

Cost-effectiveness analysis

The cost-effectiveness of denosumab vs ZA in patients with MM was analyzed using a decision analytical model that was based on the XGEVA global economic model (X-GEM) for patients with MM developed by Raje et al.³¹. The X-GEM is a partitioned-survival semi-Markov model. Briefly, the model included five health states, with 28-day cycles over a lifetime horizon (500 cycles, corresponding to ∼ 38 years). It took the same structure for both denosumab and ZA (Figure 1). Clinical inputs were derived from the phase 3, randomized, controlled 20090482 study²⁰, and other published sources, and cost and quality-of-life inputs were derived from various published sources. ICERs were calculated as the difference in total costs divided by the difference in QALYs between denosumab and ZA.

Figure 1. Schematic representation of the cost-effectiveness model structure. Abbreviations. Dmab, denosumab; MM, multiple myeloma; Tx, treatment; ZA, zoledronic acid.

Model population

The model population comprised adults with newly-diagnosed MM from the 20090482 study²⁰. Patients were enrolled over a 45-month period, with a median follow-up of 17.3 and 17.6 months in the denosumab and ZA arms, respectively^20,31. Eligible patients had at least one lytic bone lesion²⁰. As patients with severe RI (CrCl < 30 mL/min) were not included in the 20090482 study²⁰ owing to their ineligibility to receive ZA¹⁸, the base case analysis did not include patients with severe RI. Instead, a scenario analysis comparing denosumab with no treatment was performed to estimate the cost-effectiveness of denosumab in patients with severe RI.

Model clinical inputs

The following clinical outcomes were included in the model: SRE annual rates by treatment status and the distribution of SREs by type; PFS and OS rates; treatment-discontinuation and -compliance rates; and serious adverse event (SAE) rates (Supplementary Table S1)^{20,21,31–33}. The clinical inputs from the 20090482 study²⁰ were used for all four countries.

Skeletal-related events

SREs comprised pathologic fracture, radiation to bone, surgery to bone, and spinal cord compression. Constant annual SRE rates by treatment status were calculated by dividing the total number of SREs experienced by patients in each arm of the 20090482 study²⁰ by the number of patient-years of follow-up³¹. SRE rates observed in the 20090482 study²⁰ were adjusted based on real-world SRE rates^32,33 by multiplying them by a relative rate ratio (2.84)³¹. The annual SRE rate for patients off denosumab/ZA treatment was estimated by dividing the SRE rate in the ZA arm of the 20090482 study²⁰ by a relative rate ratio (0.57) derived from a recently updated, comprehensive, network meta-analysis by Mhaskar et al.²¹; this was not available at the time of the US cost-effectiveness analysis. The cost-effectiveness analysis by Raje et al.³¹ used a similar rate ratio (0.63), based on analysis of a much smaller data set³⁴ published earlier than the meta-analysis by Mhaskar et al.²¹. The annual SRE rate for patients without SRE-preventive treatment was assumed to be the same as for those who had discontinued treatment³¹. The distribution of SRE types was calculated from the 20090482 study²⁰; most SREs were pathologic fractures (81.6%)³¹ (Supplementary Table S1).

Progression-free and overall survival, and treatment discontinuation

PFS, OS, and treatment-discontinuation rates were derived from independent fits of parametric survival distributions to patient-level data from the 20090482 study^20,31. As OS and treatment-discontinuation rates were similar in both arms, data from the two arms were pooled for each of these parameters^20,31. In the 20090482 study²⁰, PFS was different in the two arms in a pre-specified exploratory analysis; therefore, PFS was modeled independently for denosumab and ZA. Of the distributions assessed, the generalized gamma function was selected for PFS and treatment-discontinuation data, and the Weibull function for OS data (Supplementary Figure S1), based on the Akaike Information Criterion, and visual inspection of fit³¹. The long-term extrapolation of OS was corrected using country, gender, and age-specific life tables to ensure that the death rate of patients with MM was at least that observed in the general population of each country^35–38.

Treatment compliance and serious adverse events

Treatment compliance was calculated as the ratio of the number of denosumab/ZA doses received by each patient to the number scheduled up to the last dose received in the 20090482 study^20,31. Constant annual rates for SAEs of special interest (hypocalcemia, osteonecrosis of the jaw, and RI) were calculated from the safety population of the 20090482 study²⁰ using a similar method to the calculation of SRE rates³¹. SAEs were as defined in the 20090482 study²⁰.

Model utility inputs

QALY decrements for the following clinical outcomes were included in the model: SREs by type, MM disease progression, SAEs by type, and route of administration (intravenous vs subcutaneous) (Supplementary Table S2)^31,39–42. QALY decrements for SREs by type³⁹ and by route of administration⁴¹ were derived from time trade-off studies involving participants from the general populations of the UK and Canada. As per Raje et al.³¹, MM disease progression was associated with a 19.5% QALY decrement; the utility levels before and after MM disease progression were 0.800 and 0.644, respectively (based on van Agthoven et al.⁴²). SAE QALY decrements were based on Stopeck et al.⁴⁰.

Model economic inputs

Cost inputs were derived from national sources to reflect differences between the healthcare systems and economic wealth of the four countries: denosumab/ZA acquisition and administration costs, and SRE and SAE management costs (Table 1)^10,43–55. The most recent estimates for drug costs were used (2018); all other costs referred to the year 2017⁵⁶.

Table 1. Model economic inputs by country.

Economic parameter	Austria	Belgium	Greece	Italy
Annual discount rates, %
Costs	5.0	3.0	3.5	3.0
Health outcomes	5.0	1.5	3.5	3.0
Parameter source	Walter and Zehetmayr 2006^43	Cleemput et al. 2012^44	NICE 2013^45^,^a	Capri et al. 2001^46
Drug acquisition and administration costs per administration, €
Dmab acquisition	333.45	287.15	205.68	282.00
ZA acquisition	86.75	100.28	39.62	73.54
Dmab administration	10.98	9.34	0.0	12.91
ZA administration^b	27.46	24.93	89.05	26.48
Parameter source	Warenverzeichnis 2018^47	INAMI–RIZIV 2018^48^,^c	Ministry of Health 2018^49; Ministry of Health Decree, 2011^50^,^d	Codifa 2018^51; Official Gazette of the Italian Republic, 2013^52^,^e
SRE management cost per SRE episode, €
Pathologic fracture	10,305	6,780	4,478	3,565
Radiation to bone	14,603	2,998	9,734	1,961
Surgery to bone	21,496	14,559	7,943	3,075
Spinal cord compression	22,191	17,249	7,538	12,755
Parameter source	Pereira et al. 2016 (Austrian data)^53	Body et al. 2013^54^,^f	Pereira et al. 2016 (Greek data)^53	Ashcroft et al. 2018^10; Official Gazette of the Italian Republic, 2013^52^,^g
SAE management cost per SAE episode, €
ONJ	4,100	1,790	1,144	1,651
Hypocalcemia	2,100	14,272	319	1,758
Renal toxicity	3,459	12,003	283	3,734
Parameter source	Data on file (Amgen)^h	Data on file (Amgen)^f^,^i	Ministry of Health Decree, 2011^50	Official Gazette of the Italian Republic, 2013^52^,^j

^aThere are no official guidelines in Greece; therefore, we considered NICE recommendations, which is common practice in the Greek setting⁴⁵.

^bIncludes renal function laboratory testing costs.

^cDmab administration cost was a weighted average of administration costs according to administration setting (nurse at home, during a GP visit, or in hospital). ZA administration cost was calculated as (2 × Dmab administration cost) + serum creatinine laboratory testing cost.

^dNo cost incurred to payer for Dmab administration.

^eOfficial Gazette of the Italian Republic no. 23, January 28, 2013—general series (supplemento ordinario 8, allegato 3).

^fCosts inflated using the health consumer index (+13%)⁵⁵.

^gMicro-costing approach using HRU reported by Ashcroft et al.¹⁰ and published unit costs (Official Gazette of the Italian Republic no. 23, January 28, 2013—general series [supplemento ordinario 8, allegato 1 e 3]).

^hMicro-costing approach.

ⁱCalculated using the same method as Body et al.⁵⁴ for SREs.

^jMicro-costing approach using unit costs (Official Gazette of the Italian Republic no. 23, January 28, 2013—general series [supplemento ordinario 8, allegato 1 e 3]).

Abbreviations. Dmab, denosumab; GP, General Practice; HRU, healthcare resource use; INAMI–RIZIV, Institut National d’Assurance Maladie-Invalidité–Rijksinstituut voor Ziekte- en Invaliditeitsverzekering (National Institute for Health and Disability Insurance [Belgium]); NICE, National Institute for Health and Care Excellence (UK); ONJ, osteonecrosis of the jaw; SAE, serious adverse event; SRE, skeletal-related event; ZA, zoledronic acid.

Drug acquisition and administration costs

Denosumab and ZA were assumed to be administered Q4W, which is in line with the schedule used in the 20090482 study²⁰ and the product labels^17,18. Denosumab/ZA acquisition costs were based on national listings^47–49,51. The cost of generic ZA was used in Greece, Italy, and Belgium. In Austria, a weighted-average cost was used because the distribution of sales of generic and originator ZA was available (IQVIA: RSÖ, DPMÖ, and DPMÖK databases 2018). Denosumab/ZA administration costs, which included a renal function monitoring cost for each ZA dose, were based on national sources. In Belgium, the cost of ZA administration was assumed to be twice that of denosumab, with the addition of a renal function monitoring cost. This assumption was made because the tariff system for intravenous drug administration in day clinics in Belgium has changed, with the current system not allowing estimation of the true cost of intravenous administration. A factor of two was selected because it was consistent with the ratio observed in Austria and Italy, and because it was a conservative estimate. The situation in Greece is different because payers do not incur a cost for denosumab administration⁵⁰.

Skeletal-related event and serious adverse event management costs

Country-specific SRE management costs by SRE type for Austria and Greece⁵³ were derived from a published study involving both patients with MM and solid tumors, whereas in Belgium⁵⁴ only data on patients with bone metastases from solid tumors were available. As MM-specific HRU data were available for Italy¹⁰, a micro-costing approach was used, applying national unit costs to HRU⁵². Country-specific SAE management costs were derived using a micro-costing approach in Austria, Belgium (data on file [Amgen]), and Italy⁵², and from national sources in Greece⁵⁰.

Discount rates

Country-specific annual discount rates for costs and health outcomes were derived from local health technology assessment (HTA) guidelines^43,44,46, except for Greece, where guidance from the National Institute for Health and Care Excellence (NICE [UK]) was used owing to a lack of local recommendations⁴⁵ (Table 1).

Scenario analyses

To explore the sensitivity of model results to changes in the model parameters and assumptions, we conducted several scenario analyses, as well as one-way deterministic and multivariate probabilistic sensitivity analyses. The main alternative scenario analyses comprised: (1) including first and subsequent line (up to fourth line) anti-MM treatment costs in the model to account for a delay in initiating subsequent treatment lines due to the prolonged PFS observed with denosumab vs ZA (unlike the cost-effectiveness analysis by Raje et al.³¹, our base case scenario did not include anti-MM treatment costs in the model); (2) comparing denosumab with no treatment for patients who may not be able to receive bisphosphonates, such as patients with severe RI¹⁸, by applying a rate ratio of 0.57²¹ to the ZA SRE rate; and (3) using SRE rates from the post hoc 15-month landmark analysis in the 20090482 study, in which time to first SRE was longer in the denosumab arm than in the ZA arm (HR = 0.66 [95% CI = 0.44–0.98], p = 0.039)²⁰. For alternative scenario 3, the model time horizon was split in two: the first 14 months of treatment, and from 15 months after the start of treatment. Annual SRE rates were calculated by dividing the number of SREs observed by the number of patient-years at risk in each arm in each time period in the 20090482 study²⁰ (Supplementary Table S3).

Additional scenarios included those commonly considered in cost-effectiveness analyses, comprising: (4) using shorter time horizons (10 and 20 years); (5) applying 0% discount rates to costs and health outcomes; and (6) considering other parametric distributions for survival models (modeling PFS using a Weibull distribution, applying an HR of 0.82²⁰, instead of using a generalized gamma distribution; a Weibull distribution was the second most plausible parametric fit). The model input parameters used in the scenario analyses are provided in Supplementary Table S3.

Sensitivity analyses

Deterministic sensitivity analyses

Sensitivity analyses of key model inputs were conducted to evaluate model robustness. The univariate deterministic sensitivity analysis involved varying each model parameter in turn to its lower and upper bound. The model input parameters used in the deterministic sensitivity analyses are provided in Supplementary Table S4.

Probabilistic sensitivity analyses

The multivariate probabilistic sensitivity analysis involved attributing probabilistic distributions to each of the parameters, and simulating 2,000 samples drawn from these distributions. Ranges and parameters for the probabilistic distributions were informed by uncertainty measures associated with the point estimates (standard errors or 95% CIs), or were varied by ±20% when no uncertainty measure was available. The model input parameters used in the probabilistic sensitivity analyses are provided in Supplementary Table S5.

Results

Base case

Results from the base case cost-effectiveness analysis by country are shown in Table 2. Discounted life-years ranged from 6.27 in Austria to 7.75 in Belgium, with no difference between denosumab and ZA. Denosumab was estimated to be associated with QALY gains vs ZA in all four countries: incremental QALYs gained were lowest in Austria (0.19) and highest in Belgium (0.24); they were 0.21 in Greece and 0.22 in Italy. Total costs over the time horizon were largely driven by SRE management costs, which ranged from €60,606 for denosumab in Italy to €183,019 for ZA in Austria. Total costs over the time horizon were higher for denosumab than for ZA in all four countries. Comparing denosumab with ZA, incremental costs incurred were lowest in Greece (€1,445) and highest in Italy (€5,858); they were €4,879 in Austria and €4,314 in Belgium. Although drug acquisition costs were higher for denosumab than for ZA (the difference ranged from €5,579 in Greece to €8,024 in Austria), these costs were partially offset by lower SRE management and drug administration costs for denosumab than for ZA (the difference ranged from –€2,668 in Austria to –€824 in Italy for SRE management costs, and from –€2,879 in Greece to –€431 in Italy for drug administration costs). The difference in SAE management costs between denosumab and ZA was negligible (<€100 in all four countries). The ICER (incremental cost per QALY gained) for using denosumab instead of ZA was €26,294, €17,737, €6,982, and €27,228 in Austria, Belgium, Greece, and Italy, respectively.

Table 2. Base case cost-effectiveness analysis (discounted) by country.

	Austria			Belgium			Greece			Italy
	Dmab	ZA	Δ	Dmab	ZA	Δ	Dmab	ZA	Δ	Dmab	ZA	Δ
Costs, €
Total	191,687	186,808	4,879	134,092	129,778	4,314	97,910	96,465	1,445	70,804	64,946	5,858
Drug acquisition	10,730	2,706	8,024	9,696	3,282	6,414	6,860	1,281	5,579	9,522	2,407	7,115
Drug administration	353	857	−503	315	816	−501	0.0	2,879	−2,879	436	867	−431
SRE management	180,351	183,019	−2,668	123,202	124,878	−1,676	91,016	92,282	−1,266	60,606	61,430	−824
SAE management	253	227	26	878	802	76	35	23	11	240	242	−3
Health outcomes
Life-years	6.27	6.27	0.0	7.75	7.75	0.0	6.83	6.83	0.0	7.04	7.04	0.00
QALYs	3.24	3.06	0.19	3.92	3.67	0.24	3.50	3.29	0.21	3.60	3.38	0.22
Cost-effectiveness
ICER (cost [€] per QALY gained)	26,294			17,737			6,982			27,228
1–3 × GDP per capita^a, €	42,100–126,300			38,700–116,100			16,700–50,100^b			28,500–85,500

^a GDP at market prices (2017)⁵⁷. The European Union GDP per capita (2017) was €30,000⁵⁷.

^b Provisional value.

Abbreviations. Δ, increment/decrement (Dmab vs ZA); Dmab, denosumab; GDP, gross domestic product; ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life-year; SAE, serious adverse event; SRE, skeletal-related event; ZA, zoledronic acid.

Alternative scenarios

Results from the alternative scenario cost-effectiveness analyses by country are shown in Table 3. When adding anti-MM treatment costs to the model, or comparing denosumab with no treatment, denosumab dominated in all four countries (i.e. denosumab was associated with fewer total costs and more QALYs than ZA or no treatment). The PFS benefit of denosumab vs ZA²⁰ translated into cost savings for denosumab vs ZA when anti-MM treatment costs were included in the model. These savings ranged from €11,836 (Italy) to €29,661 (Belgium), and were largely driven by lower anti-MM drug acquisition costs for denosumab than for ZA. When comparing denosumab with no treatment, savings associated with denosumab ranged from €3,863 (Italy) to €34,207 (Austria), and were largely driven by lower SRE management costs for denosumab than for no treatment. Using SRE rates from the 15-month landmark analysis²⁰ resulted in lower ICERs than in the base case (ranging from €3,586 [Greece] to €23,905 [Italy] per QALY). Shortening the model time horizon and modeling PFS using a Weibull distribution resulted in higher ICERs than in the base case (ranging from €9,811 [Greece] to €38,874 [Italy] per QALY using a 10-year time horizon; ranging from €10,359 [Greece] to €40,929 [Italy] per QALY when modeling PFS using a Weibull distribution). Finally, applying a 0% discount rate to costs and health outcomes resulted in slightly lower ICERs than in the base case (ranging from €5,681 [Greece] to €22,744 [Italy] per QALY).

Table 3. Alternative scenario cost-effectiveness analyses (discounted) by country.

Base case	Austria			Belgium			Greece			Italy
	ΔCosts	ΔQALYs	ICER	ΔCosts	ΔQALYs	ICER	ΔCosts	ΔQALYs	ICER	ΔCosts	ΔQALYs	ICER
	4,879	0.19	26,294	4,314	0.24	17,737	1,445	0.21	6,982	5,858	0.22	27,228
Anti-MM treatment costs included	−15,402	0.186	Dominant	−29,661	0.243	Dominant	−20,457	0.207	Dominant	−11,836	0.215	Dominant
Dmab vs no treatment	−34,207	0.456	Dominant	−17,713	0.538	Dominant	−14,715	0.487	Dominant	−3,863	0.499	Dominant
15-month landmark analysis for SREs	3,537	0.196	18,038	3,425	0.256	13,392	783	0.218	3,586	5,421	0.227	23,905
Shorter time horizon
10 years	4,810	0.136	35,372	4,241	0.158	26,763	1,422	0.145	9,811	5,759	0.148	38,874
20 years	4,878	0.178	27,451	4,313	0.225	19,178	1,445	0.196	7,384	5,857	0.202	28,936
0% discount rates	5,537	0.278	19,903	4,666	0.278	16,760	1,583	0.279	5,681	6,336	0.279	22,744
PFS modeled with Weibull distribution^a	4,879	0.130	37,601	4,314	0.155	27,837	1,445	0.140	10,359	5,858	0.143	40,929

Costs are €. ICERs are cost (€) per QALY gained. Dominant = Dmab was associated with more QALYs and fewer overall costs than ZA. The model input parameters are provided in Supplementary Table S3.

^a Proportional hazards assumption.

Abbreviations. Δ, increment (Dmab vs ZA); Dmab, denosumab; ICER, incremental cost-effectiveness ratio; MM, multiple myeloma; PFS, progression-free survival; QALY, quality-adjusted life-year; SRE, skeletal-related event; ZA, zoledronic acid.

Sensitivity analyses

Univariate deterministic sensitivity analyses, which accounted for uncertainty around the estimated values of key model inputs, showed that the ICERs were relatively stable and remained within acceptable limits. ICERs varied the most in Austria and the least in Greece. The highest and lowest ICERs were: €56,910 and €4,391 per QALY (Austria); €31,968 and €6,671 per QALY (Belgium); €18,899 per QALY and dominance (Greece); and €43,362 and €18,771 per QALY (Italy). Annual SRE rates and utility levels after MM disease progression were the main drivers of uncertainty (Figure 2). By using 1–3-times the gross domestic product (GDP) per capita of each country per QALY gained as willingness-to-pay (WTP) thresholds (Table 2), the results of the multivariate probabilistic sensitivity analysis, which accounted for uncertainty associated with several key model inputs simultaneously, estimated denosumab as 69–94%, 84–96%, 79–96%, and 50–92% likely to be cost-effective vs ZA in Austria, Belgium, Greece, and Italy, respectively (Figure 3).

Figure 2. Main drivers of uncertainty in cost-effectiveness analysis by country. The ICER is the cost (€) per QALY gained derived from univariate analyses employing the lower and upper bounds of top five key model inputs. Cross-hatching represents situations in which Dmab is dominant vs ZA (Dmab is associated with more QALYs and fewer overall costs than ZA). The ranges for the model parameters are provided in Supplementary Table S4. Abbreviations. Dmab, denosumab; ICER, incremental cost-effectiveness ratio; i.v., intravenous; MM, multiple myeloma; QALY, quality-adjusted life-year; SRE, skeletal-related event; ZA, zoledronic acid.

Figure 3. Cost-effectiveness acceptability curves by country. The model input parameters are provided in Supplementary Table S5. Abbreviations. Dmab, denosumab; QALY, quality-adjusted life-year; WTP, willingness to pay; ZA, zoledronic acid.

Discussion

This cost-effectiveness analysis compared denosumab with ZA for SRE prevention in patients with MM in four European countries (Austria, Belgium, Greece, and Italy) by populating an economic model (the X-GEM)³¹ with clinical inputs from a large, phase 3, randomized, controlled study²⁰. In our base case analysis, the ICER (incremental cost per QALY gained) for using denosumab instead of ZA ranged from €6,982 in Greece to €27,228 in Italy; the ICERs in Austria and Belgium were €26,294 and €17,737, respectively. In the absence of country-specific WTP thresholds⁵⁸, WTP thresholds of 1–3-times GDP per capita per disability-adjusted life-year (DALY) avoided are often used to assess healthcare interventions⁵⁹. These thresholds have been adopted by the World Health Organization (WHO) initiative ‘Choosing Interventions that are Cost-Effective’ (WHO-CHOICE)⁶⁰. Adopting these thresholds, and assuming that the WTP for a QALY is the same as for a DALY, denosumab was very cost-effective vs ZA in all four countries analyzed. Although this assumption is frequently made in the literature, we recognize that it is not without limitations⁶¹. Furthermore, MM is classified as a rare disease in Europe^2,62, and, in some healthcare systems, orphan drugs benefit from more flexible HTA processes to help improve patient access^63–65.

The alternative scenario and sensitivity analyses demonstrated that estimated ICERs were robust to changes in model parameters. Considering the three main alternative scenarios, estimated ICERs were more favorable than in the rather conservative base case. The longer PFS observed with denosumab vs ZA in the 20090482 study²⁰ led in the model to later initiation and shorter duration of subsequent anti-MM treatment lines, which resulted in lower incremental costs, making denosumab dominant vs ZA in this scenario. At first line, however, longer PFS was associated with slightly higher anti-MM treatment costs, although the influence of PFS was limited because regimens most commonly used at first line are typically administered for a fixed number of cycles⁶⁶. Using SRE rates from the post hoc 15-month landmark analysis²⁰ in the model resulted in more favorable ICERs than in the base case because of significantly fewer SREs with denosumab vs ZA from 15 months after the start of treatment²⁰. It was important to compare denosumab with no treatment for SRE prevention because some patients with MM, such as those with severe RI, are unable to receive ZA^18,67,68; denosumab was dominant vs no treatment in this scenario.

As country-specific economic inputs were aligned with local HTA submission guidelines, the model inputs and results of our cost-effectiveness analysis reflected differences between the healthcare systems and economic wealth of the four countries analyzed^58,69,70. Total costs were slightly higher for denosumab than for ZA across all four countries, which were driven by higher acquisition costs for denosumab vs ZA; the latter is now available as a generic version^47–49,51. The higher acquisition costs for denosumab vs ZA were partially offset, however, by lower SRE management and drug administration costs. Healthcare costs and clinical management can vary substantially between countries, therefore justifying the need for country-specific cost-effectiveness analyses⁵⁸. Of note, estimated management costs for SREs and SAEs vary substantially between the four countries; for example, the cost of managing an episode of hypocalcemia ranged from €319 in Greece⁵⁰ to €14,272 in Belgium. A potential explanation for these differences resides in differences in diagnosis-related group classification and costing. Furthermore, in Greece, denosumab can be administered at community pharmacies without cost to payers, therefore leading to a substantially lower administration cost vs ZA^50,71. Another source of variation in cost and quality-of-life estimates across countries resides in the different annual discount rates considered by national health authorities^43–46; for example, annual discount rates for health outcomes vary from 1.5% in Belgium to 5.0% in Austria^43,44.

Because patient and disease characteristics, treatment approaches and effects, and economic model structures and assumptions are different across studies, it is difficult to compare our results directly with published cost-effectiveness studies evaluating denosumab in either the US healthcare setting³¹, or in patients with bone metastases from solid tumors^{33,40,71–81}. There were three main differences between our model and the one used by Raje et al.³¹ in the US: (1) anti-MM treatment costs were included as a scenario analysis, but not in the base case; (2) the main comparator for denosumab was ZA, with a scenario analysis with no treatment as the comparator (a weighted average of the two was used by Raje et al.³¹); and (3) we limited the analysis to the payer perspective only. We elected not to include anti-MM treatment costs in our base case because of the rapidly evolving and heterogeneous treatment landscape for relapsed MM in Europe^66,82, and because of variations between list prices and net prices, the latter not being publicly available.

Both our cost-effectiveness analysis and the one by Raje et al.³¹ have the advantage of assessing lifetime value, using clinical outcome data from a large, well-designed, phase 3, randomized, controlled study in MM that directly compared denosumab with ZA²⁰; furthermore, SRE rates were adjusted using real-world data^32,33. We recognize, however, that there is uncertainty around true SRE rates in the real world due to a lack of robust data in patients with MM. Therefore, we conservatively tested a wide range of values for the SRE real-world adjustment factor in the deterministic sensitivity analysis (values ranging from 1 [no correction] to 4; base case value = 2.84), and found that the model results were relatively robust to this range. However, further research into the real-world incidence of SREs in patients with MM would be warranted to reduce this uncertainty. Our study also considered pertinent healthcare costs resulting from denosumab/ZA administration and RI, therefore providing a comprehensive assessment of real-world value. Additionally, discount rate and cost inputs in our study were country-specific, therefore accounting for variability in treatment approaches across the healthcare settings analyzed. There are, however, limitations inherent to health economic evaluations, which relate to model assumptions and the quality of model inputs⁸³. As explained by Raje et al.³¹, some costs (e.g. SRE management costs in Belgium⁵⁴) and utility data were derived from studies in patients with solid tumors instead of MM, although we do not believe that this will have had a substantial impact on our results. The need to extrapolate survival data using parametric survival models beyond the 20090482 study²⁰ follow-up is also a common source of uncertainty. However, uncertainty surrounding these parameters was explored by performing extensive scenario and sensitivity analyses.

Conclusions

In our study, denosumab was found to be cost-effective vs ZA for SRE prevention in patients with MM in Austria, Belgium, Greece, and Italy due to its efficacy in SRE prevention and potential PFS benefit without compromising renal function. This conclusion was based on often-adopted WHO cost-effectiveness thresholds, and was robust to changes in model parameters and assumptions. Cost-effectiveness estimates varied across the four countries analyzed, reflecting differences in healthcare costs and national HTA guidelines. Denosumab represents an attractive value proposition for SRE prevention in patients with MM in the four countries included in this analysis.

Transparency

Declaration of funding

This study was funded by Amgen.

Declaration of financial/other relationships

ET has received grants, personal fees, and non-financial support from Amgen, Celgene, and Janssen-Cilag; received personal fees and non-financial support from Takeda; and received personal fees from Bristol-Myers Squibb, GlaxoSmithKline, Novartis, and Roche. AJ is an employee of Amgen. AC, MC, and DB are employees of Amgen and hold Amgen stock. LK has consulted for Amgen, and is the founder and chief economist at Epiphany, which has received funding from Amgen. WW is a part-time employee of Oncotyrol and has participated in steering committees for Amgen; participated in advisory boards and consulted for Amgen, Bristol-Myers Squibb, Celgene, CTI, Gilead, Janssen, Merck, Mundipharma, Novartis, Pfizer, Roche, Sandoz, Takeda, and The Binding Site; presented lectures for AbbVie, Amgen, Bristol-Myers Squibb, Celgene, Gilead, Janssen, Multiples Myelom und Lymphom Selbsthilfegruppe Österreich, Mundipharma, Novartis, Roche, Sandoz, Takeda, and The Binding Site; participated in speaker bureaus for Gilead; and received research funding from Amgen, Bristol-Myers Squibb, Bundesland Tirol Programm: “Translational research”, Celgene, European Commission (FP7 – OPTATIO), Janssen, Novartis, Oncotyrol, Roche, and Takeda. Peer reviewers on this manuscript have received an honorarium from JME for their review work, but have no other relevant financial relationships to disclose

Acknowledgements

The authors would like to thank Cornelia Moser (Austria), Valery Fikkert (Belgium), John Relakis (Greece), and Serena de Fazio (Italy) for their assistance with collecting data inputs (all are employees of Amgen). Oxford PharmaGenesis, Oxford, UK, received funding from Amgen for medical writing support, which was provided by Jack Dean.