Cost-effectiveness of denosumab in the treatment of postmenopausal osteoporosis in Canada

Abstract

Objective:

Denosumab is a novel biologic agent approved in Canada for treatment of post-menopausal osteoporosis (PMO) in women at high risk for fracture or who have failed or are intolerant to other osteoporosis therapies. This study estimated cost-effectiveness of denosumab vs usual care from the perspective of the Ontario public payer.

Methods

A previously published PMO Markov cohort model was adapted for Canada to estimate cost-effectiveness of denosumab. The primary analysis included women with demographic characteristics similar to those from the pivotal phase III denosumab PMO trial (FREEDOM; age 72 years, femoral neck BMD T-score −2.16 SD, vertebral fracture prevalence 23.6%). Three additional scenario sub-groups were examined including women: (1) at high fracture risk, defined in FREEDOM as having at least two of three risk factors (age 70+; T-score ≤ −3.0 SD at lumbar spine, total hip, or femoral neck; prevalent vertebral fracture); (2) age 75+; and (3) intolerant or contraindicated to oral bisphosphonates (BPs). Analyses were conducted over a lifetime horizon comparing denosumab to usual care (‘no therapy’, alendronate, risedronate, or raloxifene [sub-group 3 only]). The analysis considered treatment-specific persistence and post-discontinuation residual efficacy, as well as treatment-specific adverse events. Both deterministic and probabilistic sensitivity analyses were conducted.

Results:

The multi-therapy comparisons resulted in incremental cost-effectiveness ratios for denosumab vs alendronate of $60,266 (2010 CDN$) (primary analysis) and $27,287 per quality-adjusted life year gained for scenario sub-group 1. Denosumab dominated all therapies in the remaining scenarios.

Limitations:

Key limitations include a lack of long-term, real-world, Canadian data on persistence with denosumab as well as an absence of head-to-head clinical data, leaving one to rely on meta-analyses based on trials comparing treatment to placebo.

Conclusions:

Denosumab may be cost-effective compared to oral PMO treatments for women at high risk of fractures and those who are intolerant and/or contraindicated to oral BPs.

Key words::

• Osteoporosis
• Cost-effectiveness
• Denosumab
• Fracture
• Canada
• Persistence

Introduction

Osteoporosis is a systemic skeletal disease characterized by reduced bone strength and a consequent increase in fracture risk. The clinical importance of osteoporosis lies in the osteoporosis-related fractures and their consequences. All osteoporosis-related fractures are associated with substantial physical and emotional burden for patients, as well as considerable financial impacts to the healthcare system. The burden associated with hip fracture extends beyond the initial hospitalization and leads to an increased risk of all-cause mortality1 and institutionalization after the fracture2. Similarly, while only approximately one-third of all vertebral fractures are clinically diagnosed3, they also appear to be associated with increased mortality1,4–7.

The Canadian consensus guidelines on osteoporosis8 currently recommend therapy with bisphosphonates (BPs; alendronate, risedronate, or zoledronic acid) and denosumab as first-line treatments to decrease the incidence of vertebral, non-vertebral, and hip fractures. Also, the selective estrogen receptor modulator (SERM), raloxifene, is recommended for prevention of vertebral fractures in asymptomatic menopausal women with osteoporosis. Unfortunately, the effectiveness of current oral treatments for osteoporosis is limited by poor compliance. Many patients discontinue oral bisphosphonate treatment soon after treatment initiation, with 30% of patients discontinuing therapy within the first 6 months of initiation, and 50% within 1 year9–13.

Denosumab is a fully human monoclonal antibody that specifically and with high affinity binds and neutralizes RANK ligand, a key mediator of the resorption phase of bone remodeling14. The efficacy of denosumab was demonstrated in the pivotal FREEDOM trial, where denosumab 60 mg (subcutaneous injection every 6 months [Q6M]) was found to significantly reduce the risk of new fractures of the vertebra, hip, and non-vertebral fractures15. This agent is approved for use in Canada for the treatment of post-menopausal women with osteoporosis at high risk for fracture, defined as a history of osteoporotic fracture, or multiple risk factors for fracture; or patients who have failed or are intolerant to other available osteoporosis therapy16. Because denosumab is administered via a 6-monthly subcutaneous injection, both persistence and compliance rates in patients receiving the agent have been found to be significantly higher than in patients receiving a weekly oral bisphosphonate, alendronate17. This improvement, in turn, may increase effectiveness in clinical practice, resulting in more fractures avoided and, potentially, improved cost-effectiveness.

The objective of this study was to estimate the cost-effectiveness of subcutaneous denosumab 60 mg Q6M in women with post-menopausal osteoporosis (PMO) compared to usual care from the perspective of the public payer in Ontario. Secondary objectives included determining the cost-effectiveness of denosumab in various scenario sub-groups of women with PMO.

Methods

Cost-effectiveness model

The analysis was conducted using a previously developed health economic model of PMO treatments. Details regarding the development of this model have been published based on an analysis for Sweden18. Here we provide a general overview of the model structure along with a detailed description of the adaptations made to the model for the Canadian setting.

Briefly, the cohort Markov model consists of eight health states (Figure 1). All patients begin in the well health state. In each 6-month cycle, a patient has a probability of sustaining a fracture, remaining healthy or dying. If a patient dies, she moves to the dead state and remains there for the duration of the time horizon (i.e. this is an absorbing state). If a patient incurs a fracture, she moves to the corresponding fracture health state. After 6 months in any fracture state, the patient has a risk of sustaining a new fracture or death; for patients who experience a wrist or other osteoporotic fracture, there is an additional option to return to the well health state. After 1 year in the hip or clinical vertebral fracture state, that patient moves to the corresponding post-fracture state where she will automatically remain if she does not die or experience a new fracture. While in a fracture or post-fracture state, patients accrue fracture-specific costs and utility decrements associated with the corresponding fracture type.

Figure 1.  Structure of the denosumab health economic cohort model of osteoporosis. Patients in all health states can move to the dead state; arrows have been omitted for simplification.

The baseline fracture risk of the cohort under study (compared to the general population risk) is estimated based on prognostic factors including bone mineral density (BMD) T-score at the femoral neck, age, and prevalence of radiographic vertebral fractures using previously described methods19–21. Movements between health states are based on Canadian-specific general population epidemiologic data, adjusted for baseline risk of the cohort. Treatment efficacy is modeled as a reduction in fracture risk, based on relative risk data obtained from clinical studies. Further details regarding fracture risk calculations, allowed transitions and assumptions regarding assignment of health state costs and utilities have been described explicitly in the previous publication18.

Along with parameterizing the model with Canadian-specific data (which is detailed in the sections that follow), additional changes were made to the previously published model structure to allow for comparisons to be conducted against raloxifene and addressing key adverse events (AEs) associated with each treatment.

As trials of raloxifene have shown the drug to decrease the risk of breast cancer but increase the risk of venous thromboembolism (VTE)22,23, additional health states were integrated that are only ‘activated’ when raloxifene is selected as a comparator. When the VTE state is active, patients can enter this state from the well, wrist fracture, or other fracture health states. Patients who enter the VTE state can suffer any type of event afterwards or go back to the well state. Unlike VTE, entry into the breast cancer state can occur from any state and, once a patient enters the breast cancer state, she can never return to the well state. Following 1 year in the breast cancer state, patients move to the post-breast cancer state where they can either remain, return to the breast cancer state (i.e. incur breast cancer again), or die. Similar to the fracture states, patients accrue event-specific costs and utility decrements while in these states, and epidemiologic and treatment-specific relative risk data are used to predict movements between these states.

The model also considers the adverse impacts of gastrointestinal adverse events (GIAEs) caused by oral BPs, as well as cellulitis associated with denosumab injections. Unlike VTE, however, specific health states were not developed for these AEs; rather all patients compliant with oral BPs or denosumab in each model cycle were at risk of experiencing the corresponding AE according to event rates observed in clinical trials and, for those with the AE, an AE-specific cost and utility decrement was applied.

Patient populations

Population baseline characteristics from the FREEDOM trial were used to model the target population for the primary analysis. Three patient scenario sub-group analyses were also conducted to better assess the cost-effectiveness of denosumab in patients with varying fracture risk (Table 1). The comparators selected for each population reflect usual care in that population, considering therapies reimbursed by the Ontario Public Drug Programs (OPDP) at the time of the analysis.

Table 1.  Target population characteristics.

Population	Start age (years)	T-score at femoral neck	Radiographic vertebral fracture prevalence	Comparators (usual care)
Women with PMO (FREEDOM)40	72	−2.16	23.6%	Once weekly generic alendronate, once weekly generic risedronate, ‘no therapy’
Scenario Sub-group 1: High-risk^a 41	75	−2.40	45.0%	Once weekly generic alendronate, once weekly generic risedronate, ‘no therapy’
Scenario Sub-group 2: Age 75+^b 40	78	−2.40	28.0%	Once weekly generic alendronate, once weekly generic risedronate, ‘no therapy’
Scenario Sub-group 3: Women with PMO who are intolerant or contraindicated to oral BPs^c	72	−2.16	23.6%	Generic raloxifene or ‘no therapy’

^aReflects pre-specified high-risk sub-group from FREEDOM trial. Study participants met at least two of the following three criteria: (1) age >70 years; (2) BMD T-score ≤−3.0 SD at one of the sites; (3) previous vertebral fracture(s).

^bReflects average baseline characteristics of FREEDOM trial subjects aged 75+ years.

^cAverage baseline characteristics of all FREEDOM trial subjects assumed for Scenario sub-group 3 since risk profile data were not available specifically for patients intolerant or contraindicated to oral BPs.

PMO, post-menopausal osteoporosis; BPs, bisphosphonates.

Treatment efficacy and persistence data

Fracture efficacy data for each therapy is shown in Table 2. The modeled treatment duration was capped at a maximum of 5 years, consistent with the original model publication as well as previously published Canadian economic evaluations of PMO therapies24,25. A linear decline to zero was assumed to model residual efficacy following termination of therapy, occurring over a time period equivalent to the time on therapy, but to a maximum of 2 years (which is consistent with a previous evaluation of PMO therapies published by the Canadian Agency for Drugs and Technologies in Health [CADTH]26).

Table 2.  Treatment efficacy data.

	Analysis	Relative risk of fracture relative to placebo (95% CI)				Reference
		Hip fractures	Vertebral fractures	Wrist fractures	Other fractures
Alendronate	Primary, Scenario sub-group 1	0.62 (0.40–0.98)	0.56 (0.46–0.67)	0.85 (0.67–1.09)	0.82 (0.73–0.93)	36
Alendronate	Scenario sub-group 2	0.62 (0.40–0.98)	0.62 (0.41–0.94)	0.85 (0.67–1.09)	0.82 (0.73–0.93)	36,42
Denosumab	Primary, Scenario sub-group 3	0.60 (0.37–0.97)	0.32 (0.26–0.41)	0.80 (0.67–0.95)	0.80 (0.67–0.95)	15
Denosumab	Scenario sub-group 1	0.52 (0.29–0.91)	0.35 (0.26–0.47)	0.88 (0.70–1.11)	0.88 (0.70–1.11)	41
Denosumab	Scenario sub-group 2	0.38 (0.18–0.79)	0.36 (0.25–0.53)	0.84 (0.63–1.12)	0.84 (0.63–1.12)	31,40
Raloxifene^a	Scenario sub-group 3	1.00 (1.00–1.00)^b	0.64 (0.54–0.78)	0.88 (0.68–1.14)	0.91 (0.78–1.05)	36
Risedronate	Primary, Scenario sub-group 1	0.74 (0.59–0.93)	0.64 (0.52–0.78)	0.68 (0.43–1.07)	0.80 (0.72–0.90)	36
Risedronate	Scenario sub-group 2	0.74 (0.59–0.93)	0.56 (0.39–0.81)	0.68 (0.43–1.07)	0.80 (0.72–0.90)	36,42

^aRelative risk of breast cancer and VTE for patients receiving raloxifene was 0.38 (95% CI = 0.24–0.58)23 and 3.4 (95% CI = 1.5–8.0)43, respectively; values for all other therapies were 1.00 since it was assumed that oral BPs and denosumab have no effect on these outcomes.

^bActual data reported as 1.12 (95% CI = 0.64–1.94), however, we conservatively assumed that raloxifene does not increase the risk of hip fracture since this result was non-significant vs placebo.

CI, confidence interval.

Persistence, defined as time on treatment (without consideration of compliance), was modeled. Persistence inputs for the comparator treatments were based on weighted average data from several published analyses of publicly-funded claims databases in Ontario and Quebec. In these studies, persistence was defined as continual prescription refill based on days of drug supply considering the prescription length plus a grace period of either 30 days or 1.5-times the prescription length, depending on the study. The proportion of women remaining on comparator treatment at each 6-month time point between treatment initiation and 5 years following initiation is shown in Table 39–11,13. To model persistence with denosumab, the predicted, time-specific drop out incidence data for the comparators from 6 months to 5 years were multiplied by 0.54, which is the relative risk of discontinuation observed at 1 year in the Denosumab Adherence Preference Satisfaction study (DAPS17); the resultant values are shown for denosumab in Table 3.

Table 3.  Treatment persistence data.

Time from treatment initiation (years)	Proportion of women remaining on therapy
	Oral bisphosphonates and Raloxifene9–11,13	Denosumab*
0	100%	100%
0.5	66%	82%
1.0	48%	70%
1.5	44%	66%
2.0	40%	63%
2.5	35%	59%
3.0	31%	55%
3.5	29%	53%
4.0	26%	51%
4.5	24%	49%
5.0	24%	49%

*Estimated by multiplying the time-specific drop-out incidence data for the comparators by 0.54, which is the relative risk of discontinuation observed at 1 year in the Denosumab Adherence Preference Satisfaction study17; the persistence benefit observed at 1 year was assumed to continue to 5 years.

Epidemiologic data

To the extent possible, epidemiologic data in the model were derived from Canadian sources. Where Canadian data were not available in the format required for the model, data from the US (VTE incidence, and relative mortality risk following VTE and other fractures) and from Sweden (relative risk of death following hip and vertebral fractures) were utilized based on the best available and most applicable evidence. A summary of data inputs and references is shown in Table 4.

Table 4.  Summary of age-specific epidemiologic data.

Age (years)	Morphometric vertebral fracture prevalence (%)	Event rates^a (annual probabilities × 10^−4)						Relative risk of mortality^c
		Hip fracture	Vertebral fracture	Wrist fracture	Other fracture	Breast cancer	VTE	Hip 1st year	Hip 2+ years	Vertebral 1st year	Vertebral 2+ years	Other 1st year	Breast cancer 1st year	VTE 1st year
50–54	11	1.3	7.1	17.8	105.6	20.0	20.7	9.3	3.5	14.6	9.6	1.2	1.1	108.7
55–59	11	3.4	10.8	23.3	116.9	24.0	20.7	8.0	3.2	11.1	7.3	1.2	1.1	66.5
60–64	21	6.1	16.5	32.6	129.4	30.0	20.7	6.8	2.8	8.2	5.4	1.2	1.1	42.6
65–69	21	14.0	25.1	37.8	143.3	34.9	33.5	6.2	2.7	7.0	4.6	1.2	1.1	26.5
70–74	33	26.0	38.4	44.0	158.8	39.9	33.5	4.9	2.2	5.0	3.3	1.2	1.1	16.5
75–79	33	61.8	58.7	49.5	175.9	40.9	60.0	3.7	1.7	3.8	2.5	1.2	1.1	9.8
80–84	45	112.3	89.9	58.1	195.0	39.4	60.0	2.6	1.3	2.4	1.6	1.2	1.2	5.6
85–89	45	198.8	138.0	63.9	216.2	34.9	60.0	1.9	1.0	1.6	1.1	1.2	1.2	3.1
90–94	45	262.2	212.7	74.2	239.8	33.1	60.0	1.5	1.0	1.2	1.0	1.2	1.2	1.9
95+	45	300.0	330.1	80.4	266.1	33.1	60.0	1.4	1.0	1.0	1.0	1.2	1.2	1.3
Ref	44	45	^b	45	45	46	22	18	18	18	18	47	48	22

^aGeneral female population mortality data were obtained from Statistics Canada for years 2000–200249; data not shown.

^bUnpublished data from the Canadian Multicentre Osteoporosis Study (CaMos).

^cThe model assumed that only 30% of the excess mortality shown in the table following hip, vertebral, and other fractures was associated with the fracture event19,50. The duration of increased risk following hip or vertebral fracture was 8 years51,52.

VTE, venous thromboembolism; Ref, reference.

Cost data

All costs were obtained from Canadian sources and are expressed in 2010 Canadian dollars, inflated as necessary using the health and personal care component of the Consumer Price Index (CPI)27. Consistent with health technology assessment guidelines in Canada28, the analysis was conducted from the perspective of a public payer and applied a discount rate of 5% per year to costs and outcomes. A summary of the economic data is provided in Table 5, where costs are shown for all events, denosumab and its comparators, treatments for therapy-specific AEs, and costs for general osteoporosis care.

Table 5.  Summary of costs.

Fracture costs ($)
Age (years)	Hip fracture 1st year	Hip fracture 2+ years	Vertebral fracture 1st year	Vertebral fracture 2+ years	Wrist fracture 1st year	Other fracture^a 1st year
50–59	19,173	4499	10,005	205	1359	2927
60–69	18,277	4499	13,432	205	1927	3565
70–79	24,088	4499	16,285	205	4964	8835
80–89	26,076	4499	20,736	205	12,407	14,509
90+	23,988	4499	24,814	205	17,038	18,544
Reference	53	24	53	24	53	53
Drug costs ($/year including wholesaler mark-up (8%) and a $7 dispensing fee per 100 days)29
Generic Alendronate (70 mg once weekly)		275		Generic Raloxifene (60 mg once daily)		387
Denosumab (60 mg every 6 months)		727		Generic Risedronate (35 mg once weekly)		299
Other costs ($)
Breast cancer, year 124		12,930		Gastrointestinal adverse event treatment (for 1 month)29 ^d		23
Breast cancer, year 2+24		1,057		Nurse visit (for denosumab administration)54		18
Cellulitis treatment (per episode)^c		924		Physician visit (for therapy monitoring)54		32
Daily cost of long-term Care^e 53		149		VTE, year 1 ^b		9992
DXA (per scan)54		106

^aThe cost of humerus fracture was used as a proxy for ‘other’ fractures.

^bCosts obtained from the Ontario Case Costing Initiative (OCCI) for fiscal year 2009/10 (ICD-10 codes for pulmonary embolism [PE]: I26.0 and I26.9; deep vein thrombosis [DVT]: I80.2; weighted according to the PE:DVT ratio from the Worcester population study22); OCCI costing analysis tool available at http://www.occp.com/mainPage.htm; accessed December 5, 2011.

^cCost obtained from the OCCI (2009/10) using ICD-10 codes L03.0–L03.9. OCCI costing analysis tool accessed December 5, 2011.

^dWeighted cost assuming 30-day prescription for generic omeprazole 20 mg OD55 and generic ranitidine 150 mg BID (20%).

^eAssigned to patients entering a long-term care facility following hip fracture (for the remainder of the time horizon) according to age-specific rates from Morin et al.56: 0% age 50–59; 3.7% age 60–74; 24.3% age 75+.

VTE, venous thromboembolism; DXA, dual-energy X-ray absorptiometry.

Costs for comparator therapies were obtained from the Ontario Drug Benefit Program Formulary that was current at the time of the analysis, assuming generic pricing29. A cost of $330 per 60 mg injection for denosumab was provided by Amgen Canada Inc. It was assumed that a nurse visit would be required to complete each of two denosumab injections required per year. For all patients actively receiving osteoporosis treatment, it was assumed that one physician visit annually would be required for treatment monitoring and that one dual-energy X-ray absorptiometry (DXA) per 2-year period would be ordered8.

Women receiving alendronate or risedronate were assumed to be at an increased risk of upper gastrointestinal problems that would require general practitioner (GP) consultation and, conservatively, a prescription for a 1-month course of pharmacologic therapy. The GP consultation rate (over and above a background level of average women) was assumed to be 2.35% in the initial treatment month and 0.35% in subsequent months30. In the absence of treatment-specific Canadian data, we conservatively assumed that 80% of women would receive an H2-receptor antagonist while 20% would be prescribed a proton pump inhibitor. Similarly, women receiving denosumab were at increased risk of experiencing cellulitis. The annual rate of cellulitis based on the 3-year FREEDOM trial was calculated as 0.0009 events per year31. Each episode was assumed to require a hospitalization with an average duration of 7.7 days and cost of $924 per day, as obtained from the Ontario Case Costing Initiative (OCCI) (Table 5).

Health utility data

Normal population utility values gathered using the Health Utilities Index Mark 3 (HUI3) questionnaire were obtained from the HUInc website32 as collected in the Joint Canada/USA Survey of Health 2002/3 (JCUSH 2002/3). Utility values for Canadian non-institutionalized females were 0.849, 0.824, 0.806, and 0.689 for ages 50–59 years, 60–69 years, 70–79 years, and 80–85 years, respectively.

Event-specific multipliers applied in the model to patients experiencing each event are shown in Table 6, along with their references. The majority of fracture-specific utility data were obtained from a systematic review and meta-analyses published by Peasgood et al.33 based on the EQ-5D. Since this review did not provide an estimate of utility following an incident ‘other’ fracture and because estimates for years 2+ following a clinical vertebral fracture were inconclusive, an analysis of data from the Canadian Multicentre Osteoporosis Study (CaMos) (which uses the HUI3 for utility collection) was commissioned. Utility values for all other non-fracture events were derived using data obtained from a variety of published sources as shown.

Table 6.  Summary of utility multipliers.

Health state / Event	Mean	95% CI	Reference
Hip Fracture, 1st Year	0.700	0.640–0.770	33
Hip Fracture, Years 2+	0.800	0.680–0.960	33
Vertebral Fracture, 1st Year	0.590	0.460–0.830	33
Vertebral Fracture, Years 2+	0.828	0.782–0.873	^a
Wrist Fracture, 1st Year	0.956	0.860–1.000	33
Other Fracture, 1st Year	0.843	0.808–0.877	^a
Breast Cancer, 1st Year	0.838	0.713–0.963	57
Breast Cancer, Years 2+	0.865	0.740–0.990	58
VTE, 1st Year	0.865	0.740–0.990	59
Gastrointestinal adverse event	0.910 ^b	n/a	60
Cellulitis	0.820 ^c	0.790–0.850	61

^aUnpublished data from the Canadian Multicentre Osteoporosis Study (CaMos).

^bApplied for 1 month.

^cValue is specific to patient with an infected diabetic foot ulcer; applied for 1 month.

VTE, venous thromboembolism; n/a, not available.

Cost-effectiveness analysis

The model quantifies both the costs and effects of the different therapeutic alternatives over a lifetime time horizon. The main outcome measure is the incremental cost-effectiveness ratio (ICER); however, the model is also able to estimate total costs, total life years, total quality-adjusted life years (QALYs), and 10-year risk of each fracture type, breast cancer, and VTE. In addition to conducting base case analyses for each patient population, a number of deterministic sensitivity analyses were conducted to explore the sensitivity of the model to parameter variability and uncertainty within plausible ranges and also to assess the impact of alternate assumptions on the outcomes of the analysis. Further, a probabilistic sensitivity analysis (PSA) was conducted with variable mean uncertainty sampled for a number of inputs including denosumab and comparator efficacy (log normal distribution assumed), denosumab’s persistence advantage (relative risk of 0.54; log normal distribution), first and subsequent year event costs34, and first and subsequent year event utility multipliers34; 1000 iterations were required to achieve stable results.

Results

Primary analysis

Clinical and economic results for the primary analysis conducted on a population of women with characteristics similar to those enrolled in the FREEDOM trial are shown in Table 7. Denosumab was the most costly treatment, but also resulted in the lowest number of hip, vertebral, and other fractures and the highest number of life years and QALYs. The relatively large difference in intervention cost between denosumab and its comparators is due not only to the acquisition cost itself, but also to the additional nurse visits required for its administration and the longer duration for which patients remain persistent with denosumab.

Table 7.  Costs and clinical outcomes for primary (FREEDOM) population.

	No treatment	Alendronate	Risedronate	Denosumab
Costs/patient
Fracture treatment cost	$15,831	$15,383	$15,463	$14,999
Total intervention cost	$0	$721	$769	$2423
Total cost	$15,831	$16,104	$16,232	$17,422
Outcomes
10-year fracture risks/1000 patients
Hip fractures	77	71	73	68
Vertebral fractures	88	79	80	67
Wrist fractures	51	50	47	48
Other fractures	164	159	158	156
Life years and QALYs/patient
Life years (discounted)	10.073	10.080	10.079	10.087
QALYs	7.429	7.448	7.445	7.470

QALYs, quality-adjusted life years.

The cost-effectiveness results for the primary analysis are shown in Table 8. When conducting a cost-effectiveness analysis (CEA) of multiple therapies, all therapies are first ordered from least to most costly, and then the incremental costs and effects are calculated for each therapy relative to the next less costly alternative. A therapy is said to be dominated when it is both more costly and less effective than its comparator, as in the case of risedronate vs alendronate. Once the dominated strategy is eliminated, incremental cost-effectiveness ratios are calculated for each therapy based on the remaining strategies. As noted in Table 8, following removal of risedronate from consideration an ICER of $60,266 per QALY gained resulted for denosumab relative to alendronate.

Table 8.  Multi-therapy cost-effectiveness results.

Treatment	Average cost per patient ($)	Average QALYs per patient	Incremental cost ($)	Incremental QALYs	ICER ($/QALY)
Primary analysis: FREEDOM population
No treatment	$15,831	7.429
Alendronate	$16,104	7.448	$273	0.019	$14,708
Risedronate	$16,232	7.445	$129	−0.002	Dominated (Simple)
Denosumab	$17,422	7.470	$1318	0.022	$60,266
Scenario sub-group 1: High-risk
Alendronate	$19,741	6.486
No treatment	$19,887	6.460	$146	−0.025	Dominated (Simple)
Risedronate	$19,968	6.482	$227	−0.003	Dominated (Simple)
Denosumab	$20,534	6.515	$793	0.029	$27,287
Scenario sub-group 2: Age 75+ years
Denosumab	$18,694	5.580
Alendronate	$18,900	5.549	$206	−0.031	Dominated (Simple)
Risedronate	$19,193	5.548	$499	−0.032	Dominated (Simple)
No treatment	$19,423	5.529	$730	−0.051	Dominated (Simple)
Scenario sub-group 3: Women with PMO who are intolerant or contraindicated to oral BPs
No treatment	$16,071	7.397
Raloxifene	$17,092	7.397	$1021	0.001	Dominated (Extended)^a
Denosumab	$17,704	7.436	$1633	0.039	$41,822

^aExtended dominance exists when the incremental cost-effectiveness ratio for a given treatment alternative is higher than that of the next, more effective alternative (i.e. the first therapy is weakly dominated by the next).

QALY, quality-adjusted life year; ICER, incremental cost-effectiveness ratio; PMO, post-menopausal osteoporosis; BPs, bisphosphonates.

A considerable number of deterministic sensitivity analyses were conducted on the base case scenario for the primary analysis population. Figure 2 shows the results of key analyses conducted on the ICER obtained for denosumab vs alendronate. The model was most sensitive to the denosumab persistence advantage being modeled (‘on’, in which case treatment-specific persistence data were applied to all patients) vs assuming equivalent persistence among therapies (‘off’, in which case all patients receiving all therapies were assumed to remain on therapy for the entire 5-year treatment duration). The ICERs ranged from as low as $41,000 to as high as $110,000 per QALY gained. The model was not sensitive to reasonable variations in a number of parameters which are not shown, for brevity, in Figure 2 including hip fracture incidence, fracture-related costs, treatment duration, duration of denosumab persistence advantage, risk of mortality following hip or vertebral fracture, rate of entry into long-term care following hip fracture, daily long-term care cost, exclusion of treatment-related adverse events, wrist and other fracture utility mulipliers, inclusion of societal costs, and duration of excess mortality following hip and vertebral fracture.

Figure 2.  Deterministic sensitivity analysis results for primary population. Alternate source of vertebral fracture incidence is Leslie et al.45.

Results of the PSA comparing denosumab, alendronate, and ‘no therapy’ demonstrated that if one’s willingness-to-pay (WTP) threshold is $50,000, $61,000, and $100,000, the probability that denosumab is the cost-effective therapy was 31%, 50%, and 85%, respectively (data not shown).

Scenario sub-group analyses

The cost-effectiveness results for the scenario sub-groups are shown in Table 8. In the high risk scenario (Sub-group 1), ‘no treatment’ and risedronate were dominated by alendronate, while an ICER of $27,287 per QALY gained resulted for denosumab compared to alendronate. In women with PMO aged 75+ years (Scenario sub-group 2), denosumab was the least costly and most effective therapy, thereby dominating all other treatments. In the final scenario, women with PMO who are intolerant and/or contrindicated to oral BPs (Scenario sub-group 3), raloxifene was dominated through extended dominance by denosumab, and an ICER of $41,822 per QALY gained resulted for denosumab compared to ‘no treatment’.

Discussion

The goal of this economic evaluation was to assess the value, from the perspective of the public payer in Ontario, of denosumab vs standard of care in women with PMO. The primary analysis population was chosen to reflect women in the FREEDOM trial, having baseline characteristics including age, BMD, and prevalence of vertebral fracture similar to women enrolled in the study. In this patient group, our findings demonstrated that denosumab was the most costly treatment, but also resulted in the lowest number of hip, vertebral, and other fractures and highest number of life years and QALYs over the lifetime of patients. Our multi-way cost-effectiveness comparison showed that treating with denosumab yielded an ICER of $60,266 per QALY gained relative to alendronate. The incremental cost-effectiveness ratios for denosumab were more favourable when PMO sub-groups at greater risk of fracture were considered. In the sub-group analysis of women with baseline characteristics reflective of the pre-specified high risk criteria from FREEDOM (Sub-group 1), the multi-way comparison showed that treatment with denosumab yielded an ICER of $27,287 per QALY gained relative to alendronate, while denosumab dominated all other therapies (i.e. was the least costly and most effective) in the sub-group of women aged 75+ years (Sub-group 2).

Our deterministic sensitivity analyses identified that our model was highly sensitive to assumptions regarding treatment persistence. It has been established that lack of persistence to therapy is a major problem in the management of osteoporosis and, therefore, the implications of poor persistence on treatment efficacy and resultant cost-effectiveness estimates are important to consider. The claims data used in our analysis to model persistence for comparator treatments suggest that only 48% of patients remain on therapy at 1 year, while at 5 years post-initiation, this is reduced to ∼24%9–11,13. Given that denosumab is administered via 6-monthly injections and is not associated with adverse gastrointestinal effects, it was hypothesized that persistence to therapy would be improved. This was confirmed in the DAPS study, which was a 2-year, randomized, open-label, cross-over study involving centers in the US and Canada. Relative to once-weekly alendronate, treatment with denosumab reduced discontinuation rates by 46% (p = 0.026) during the first 12 months of treatment; after cross-over, this reduction in discontinuations was even greater, at 80% (p < 0.001)17. It should be noted that the absolute 1-year discontinuation rate observed for alendronate in the DAPS study of 23.4% was lower than our estimate from the Canadian claims data, which may be attributable to the impact of the trial setting. Since, at the time of our analysis, it was not yet possible to derive long-term persistence estimates for denosumab in a real-life clinical setting, 1-year data from the DAPS study represented the best information available, which, notably, are consistent with the high compliance rates (89% at 6 months, 82% at 18 months) published for another injectable treatment, teriparatide (once daily parathyroid hormone), in patients with osteoporosis who fractured while on or were intolerant to anti-resorptive treatment35.

As denosumab is now approved for use in Canada, real-world utilization data are beginning to come to light. A national support program has been launched to support patient initiation and adherence to denosumab. When patients receive a prescription for denosumab, they also receive a written invitation and instructions to enroll in the program. Once enrolled, patients receive an initial introductory telephone call followed by a reminder call 1 month pre-injection and follow-up call 1 month post-injection for the first and all subsequent treatments. During the telephone discussions, data are collected on demographics, medical and medication history, healthcare provider information, fractures, and denosumab injections. To date, in excess of 5500 patients have enrolled in the program. Of these patients, the confirmed persistency rate for the second injection is 91%, while this rate is 95% for the third injection, supporting observations made in the DAPS study.

In our third patient scenario sub-group, we examined the impact of denosumab used in women with PMO who were intolerant to or contraindicated to oral BPs. For this analysis, data were lacking on the baseline characteristics specific to this scenario sub-group and, hence, we assumed the same mean age, T-score, and vertebral fracture prevalence as was observed in the full FREEDOM trial population. The only difference, therefore, in this analysis vs our primary analysis was the comparator selection. In Ontario, the only treatment funded by the healthcare system for women who cannot take oral BPs is raloxifene; ‘no therapy’ was also included as women may choose to go untreated. Our multi-way cost-effectiveness analysis showed that raloxifene was dominated by denosumab through extended dominance, while treatment with denosumab yielded an ICER of $41,822 per QALY gained relative to ‘no therapy’. Although data are not available to confirm, clinical opinion would suggest women in this sub-group tend to be older and have more severe osteoporosis since older individuals tend to be more pre-disposed to GI problems and are on a greater number of medications that may contribute to intolerance, making their risk profile more similar to that of the high risk group (scenario sub-group 1). Therefore, we believe the results of this scenario sub-group analysis are likely to be conservative.

Several additional limitations are worth noting. First, there are no head-to-head studies comparing rates of fracture reduction associated with denosumab and its comparators, nor are efficacy data available for the comparators that match exactly the baseline characteristics of the FREEDOM population or those of our scenario sub-groups. Therefore, the efficacy data for our comparators were taken from meta-analyses of trials comparing to placebo36. The baseline characteristics of patients in these trials differ widely between studies; for example, in the FREEDOM trial, ∼24% of patients had a vertebral fracture at baseline, mean femoral neck T-score was −2.16 SD, and average age was 72 years, whereas vertebral fracture prevalence was 100% in one of the Fracture Intervention Trials (FIT) of alendronate37, mean femoral neck BMD was 0.57 g/cm² (T-score not published), and average age was 71 years. In the key randomized controlled hip fracture trial of risedronate, corresponding baseline characteristics were 39%, −2.8 SD, and 74 years, respectively38; it is noted that these characteristics are actually closer in similarity to the FREEDOM population pre-specified high risk sub-group. Data suggest that the efficacy of treatment varies depending on the fracture risk of patients; in the FIT studies (which included a previous vertebral fracture trial and a non-vertebral fracture trial)37,39 and the hip fracture trial of risedronate (which included separate younger and older cohorts)38, the relative fracture risk reductions in the active treatment arms varied between studies of the same therapy. Since it was not possible to adjust for the heterogeneity of the baseline risk of the populations across the comparator trials, nor was it possible to match comparator data to the exact baseline characteristics of the FREEDOM trial population and scenario sub-groups, the efficacy data should, therefore, be viewed with these considerations in mind.

To the extent possible, Canadian data were used to populate the cost-effectiveness model. However, it was necessary to use data from other countries to estimate a relatively small number of parameters. This included data from the US (VTE incidence and relative mortality risk following VTE and other fractures), Sweden (relative risk of death following hip and vertebral fractures), and the UK (GP consultation rate for upper gastrointestinal problems due to BPs). Additionally, utility multiplier data were taken from a meta-analysis that included a number of non-Canadian studies. Although the use of non-Canadian data represents a limitation of our analysis, deterministic sensitivity analyses were conducted to assess the sensitivity of the model to these input parameters. With the exception of the utility multipliers (to which the model was moderately sensitive), the model was not sensitive to reasonable variations in the remaining non-Canadian parameters (data not shown).

Finally, it is important to mention that, following completion of our analysis, the Ontario Drug Benefits Program reduced considerably the price of all generic drugs listed on its formulary. This resulted in a reduction of ∼39% and 45% in the annual cost of alendronate and risedronate, respectively, relative to the costs used in our analysis. When these lower prices were substituted in our model, the ICERs for denosumab vs alendronate increased from $60,000 to ∼$70,000 (primary analysis) and from $27,000 to $35,000 (Scenario sub-group 1) per QALY gained, while the results for Scenario sub-group 2 were essentially unchanged (data not shown). Since the price of generic raloxifene was not impacted by these changes implemented by the Ontario Drug Benefits Program, the results for Scenario sub-group 3 were not affected. Although the absolute ICERs increased in some populations, the key message that denosumab may be cost-effective in women at high risk of fracture remains valid, noting also that generic prices of the bisphosphonates in Ontario are the lowest of all the Canadian provinces. It is also important to note that we did not adjust comparator persistence rates in these analyses to reflect the fact that persistence with generic alendronate is lower than with branded alendronate; Sheehy et al.13 demonstrated that the risk of discontinuation doubled (hazard ratio = 2.08 [95% CI = 1.89–2.28]) in patients initiated with generic vs branded therapy.

Denosumab has already been recommended by Osteoporosis Canada, based on the clinical evidence, as a first-line therapy for prevention of hip, non-vertebral, and vertebral fractures in menopausal women requiring treatment of osteoporosis8. The evidence provided here suggests that, due to its fracture protection and persistence profiles, denosumab may offer good value for money, particularly in women with osteoporosis who are at high risk of fracture or who are unable to tolerate oral BPs. It will next be up to reimbursement decision-makers in Canada to decide whether or not to make denosumab available to its drug plan members.

Conclusions

Lack of persistence to therapy is a major problem in the management of osteoporosis as patients who are indicated for treatment are not receiving therapeutic benefits. Denosumab represents an important, new treatment option with demonstrated fracture efficacy and an improved persistence profile due to its twice yearly dosing regimen and lack of gastrointestinal side-effects. Considering all of these benefits collectively, denosumab may be a cost-effective treatment alternative in Canada, particularly in women at high risk of fracture and in those for whom oral BPs are not an option.

Transparency

Declaration of funding

This study was sponsored by Amgen Canada Inc. Both Amgen Canada Inc. and Amgen Inc. reviewed and provided comment on the draft versions of the manuscript and also approved the final version.

Declaration of financial/other relationships

DC is an employee of Amgen Inc. and has received stocks and stock options from Amgen Inc. DB and MC are employees of OptumInsight. OptumInsight was funded by Amgen Canada Inc. to adapt the model and provide editorial support for the manuscript. The authors from OptumInsight did not receive individual payments from Amgen Canada Inc. for their involvement in this study. JDA has received funding from the following companies to act as a consultant/speaker: Amgen Inc, Eli Lilly, GSK, Merck, Novartis, Pfizer, Procter & Gamble, Roche, Sanofi Aventis, and Warner Chilcott. Dr Adachi has also participated in clinical trials funded by Amgen Inc, Bristol-Myers Squibb, Eli Lilly, Merck, Novartis, Pfizer, Procter & Gamble, Sanofi Aventis, Roche, and Warner Chilcott. RG received funding from Amgen Canada Inc. to participate as a member of the Canadian Denosumab Advisory Board which helped guide the adaptation of the international model for Canada. GI has no relationships or conflicts to declare.