Cost effectiveness of teriparatide and PTH(1-84) in the treatment of postmenopausal osteoporosis

Abstract

Objectives:

The purpose was to assess the cost effectiveness from a societal perspective of the recombinant human parathyroid hormones: PTH(1-34) (teriparatide) and PTH(1-84) for patients with osteoporosis with similar characteristics to patients treated in normal clinical practice in Sweden.

Methods:

A Markov model of osteoporosis in postmenopausal women was developed using 6-month cycles and a lifetime horizon. The model was populated with patients similar to the Swedish cohort of the European Forsteo Observational Study (postmenopausal women; mean age: 70 years, total hip T-score: −2.7 and 3.3 previous fractures). The cost effectiveness of both teriparatide and PTH(1-84) was estimated compared to no treatment and each other. Relative effectiveness assumptions were based on efficacy estimates from two phase III clinical trials.

Results:

The cost per QALY gained of teriparatide vs. no treatment was estimated at €43,473 and PTH(1-84) was estimated at €104,396. Teriparatide was indicated to be less costly and associated with more life-years and QALYs than PTH(1-84). When assuming no treatment effect on hip fractures the cost per QALY gained was €88,379. In the sensitivity analysis the cost effectiveness did not alter substantially with changes in the majority of the model parameters except for the residual effect of the treatment after stopping therapy.

Conclusions:

Based on the efficacy estimates from pivotal clinical trials and characteristics of patients treated in clinical practice in Sweden, teriparatide seems to be a more cost-effective option than PTH(1-84) when compared to no treatment. The relative efficacy between the two PTH compounds was based on an indirect comparison from two separate clinical trials which has to be considered when interpreting the results.

Key words::

• cost effectiveness
• osteoporosis
• parathyroid hormone
• teriparatide

Introduction

The weakening of bones and decline in neuromuscular function occurs after menopause in women, leading to an increased risk of fractures. Osteoporosis has clinical and public health importance, as osteoporotic fractures are one of the most common causes of disability and a major contributor to medical care costs in many regions of the world1. The major complications of osteoporosis include vertebral and hip fractures, which are associated with pronounced morbidity and excess mortality. Therefore, the aim of treatment of postmenopausal osteoporosis is to reduce the frequency of vertebral and non-vertebral fractures, which are responsible for the morbidity associated with the disease2,3.

Treatment of osteoporosis with intermittent subcutaneous injections of parathyroid hormone (PTH) stimulates new bone formation that permits the restoration of bone microarchitecture, including improved trabecular connectivity and increased cortical thickness, and reduces fracture risk in postmenopausal women with osteoporosis4. Currently, there are two approved preparations of PTH in Europe: the human recombinant (1-34) N-terminal fragment of PTH (teriparatide; Forsteo, Eli Lilly) and the full-length human recombinant molecule (rhPTH(1-84); Preotact, Nycomed). Teriparatide and PTH 1-84 are both administered by daily subcutaneous injection.

In pivotal phase III trials, both drugs significantly reduced the risk of vertebral fractures in postmenopausal women with osteoporosis compared with placebo5,6. Teriparatide has also been shown to significantly decrease the risk of non-vertebral fractures5. No head-to-head clinical trials have been performed to compare the two drugs.

The cost effectiveness of teriparatide has previously been estimated on women with characteristics similar to those of the patients in the Fracture Prevention Trial7. This analysis did not consider PTH(1-84), which was not approved at the time of the study, and utilised a simulated patient population designed to follow the strict inclusion and ongoing protocol conditions of a clinical trial. By contrast, the present study aims to establish the cost effectiveness of both teriparatide and PTH(1-84) in postmenopausal women with postmenopausal osteoporosis representative of those prescribed teriparatide as administered in routine clinical practice.

Methods

Model overview

A seven health state Markov cohort simulation model was developed to investigate the cost effectiveness of teriparatide and PTH(1-84) at licensed doses in the treatment of postmenopausal osteoporosis relative to each other and placebo (Figure 1). Due to a lack of comparative head-to-head clinical trials comparing the two forms of PTH, the clinical efficacy and safety data were extracted from two global, multi-centre placebo-controlled trials of teriparatide and PTH(1-84). The key outcomes of the model were the lifetime cost, measured under a societal perspective with a cost year of 2007, and the quality-adjusted life-years (QALYs) associated with teriparatide, PTH(1-84) and no treatment. Discounting was applied to costs and outcomes at 3% per year in the base case and varied in the sensitivity analysis8.

Figure 1.  Health state structure of the Markov model. Note: From each of the six health states above, transition to the seventh state (Dead) was possible.

The cost-effectiveness model was a development of models that have previously been used for estimating the cost-effectiveness of other osteoporosis treatments7,9,10 with modifications to include the results of PTH treatments and cost data for 2007. The model was programmed in TreeAge Pro 2007, with short-term (1st year) and long-term (2nd and subsequent years) outcomes estimated using a tunnelling technique11. The model also included a health state which relates to patients sustaining a vertebral fracture after a hip fracture. The cycle length was for 6 months, and all patients were followed through the model from age at treatment initiation until they were 100 years old or dead. Patients began the simulation in the no-fracture health state. The patient cohort was adjusted according to their baseline characteristics, i.e. if the patients had a prevalent fracture before intervention; their quality of life, costs and mortality risk would be adjusted accordingly.

Target patient population

The population simulated in the model had the characteristics of the Swedish cohort of patients from the European Forsteo Observational Study prescribed teriparatide in routine clinical practice12,13. This study covered eight European countries and included a cohort of 219 postmenopausal women from Sweden. At baseline, the Swedish cohort had a mean (SD) age of 69.9 (8.40) years, a mean (SD) T-score of −2.7 (1.08) and −3.5 (1.21) at the total hip and lumbar spine, respectively, and 3.3 (2.0) prevalent fractures (SD).

Clinical data

The model requires assumptions as to the clinical effectiveness of each modelled treatment and the risk of relevant adverse events. The two pivotal phase III trials on which the efficacy assumptions were based were the teriparatide Fracture Prevention Trial (FPT)5 and the Treatment of Osteoporosis with PTH(1-84) (TOP) study6. Effectiveness was initially assumed to be equal to the efficacy results obtained from these two global, multi-centre trials.

Patients were assumed to be given treatment for 18 months, in line with Swedish treatment guidelines, and the effectiveness was assumed to apply at its maximum initial level for the whole treatment period14. Thereafter the effect was assumed to taper to zero over 24 and 30 months (i.e., the offset time of treatment effect) for vertebral and non-vertebral fractures, respectively, based on a follow-up of the FPT7,15,16. In a sensitivity analysis, treatment periods of 12 and 24 months and scenarios assuming no and doubled offset time were explored.

Fracture efficacy

The FPT included 1,637 postmenopausal women with prior vertebral fracture to receive placebo or daily injections of 20 µg or 40 µg teriparatide. Treatment is currently licensed at 20 µg. The primary outcome was the occurrence of new vertebral fractures; the secondary outcome measures were the occurrence of non-vertebral fractures and changes in bone mineral density of the hip and lumbar spine. The risk of new vertebral fractures in the teriparatide arm was reduced by 65% compared with the placebo arm (95% CI: 45–78%) and the risk reduction of non-vertebral fragility was 53% (95% CI: 25–88%)5. These point estimates were used as the base-case model assumptions. To acknowledge the fact that hip fracture reduction alone did not reach statistical significance in the FPT, since only five fragility hip fractures were observed (one case in teriparatide 20 µg/day and four cases in placebo) a sensitivity analysis was conducted assuming no effect on hip fracture risk with teriparatide17.

The TOP study involved 2,532 postmenopausal women with low bone mineral density at the hip or lumbar spine treated with placebo or daily injections of 100 µg PTH(1-84), the licensed dose. The primary outcome measure was the occurrence of new vertebral fractures and the secondary outcome was the occurrence of non-vertebral fractures and changes in bone mineral density and markers of bone metabolism. A reduction in risk of vertebral fracture was observed in the active treatment arm relative to the placebo arm, but point estimates and confidence intervals for this varied according to assumptions made in relation to fracture incidence among those who discontinued the study. The relative risk reduction (RRR) of vertebral fractures was 40% (95% CI: 64–0%) if non-completers were assumed to have had the same incidence as completers in their respective treatment groups, 38% (95% CI: 63–4%) if non-completers were assumed to have had the same incidence as completers in the placebo group, and 58% (95% CI: 76–28%) if non-completers were assumed to have had zero fracture risk subsequent to leaving the trial. There was no statistically significant difference between treatment groups in the risk of non-vertebral fractures. The point estimates used for the base-case model assumptions were the 58% reduction in vertebral fractures (as this estimate is the closest to a standard ITT calculation) and no risk reduction in non-vertebral fractures. The lowest estimated efficacy (38%) was tested in a sensitivity analysis.

Adverse events

In the base-case analysis, no adverse events related to the treatments were considered. Data from the trials reported that the frequency of early post-injection (4–6 hours) hypercalcaemia in the 20 μg teriparatide group was 11 vs. 2% in the placebo arm5, while for full-length PTH(1-84) pre-injection (or 24 hours post previous injection) hypercalcaemia rates were 28% in the PTH(1-84) group with 4.5% in the placebo6. Empirical data on the impact of hypercalcaemia on costs and quality of life is scarce but can be considered to be of low magnitude. In a sensitivity analysis, a cost and quality of life reduction related to hypercalcaemia based on the same frequencies as observed in the clinical trials was applied.

Compliance

Based on returned medications, the compliance for teriparatide in the FPT trial was estimated at approximately 80% in all trial arms (range: 79–83%)5. In the TOP study, no such corresponding estimate was provided6. In the base case, a compliance rate of 80% for both medications was used meaning that the drug costs were reduced by 20%. In a sensitivity analysis, 100% compliance without any adjustment of the treatment cost for both medications was explored.

Discontinuation/withdrawal rates

In the FPT study, after 18 months of treatment, 141 of the 541 (26%) patients in the 20 μg teriparatide arm at baseline had discontinued treatment5. The corresponding estimate in the TOP study was 36% of the patients receiving PTH(1-84)6. In the analysis, these rates were used to model discontinuation of treatment for teriparatide and PTH(1-8). In studies of osteoporotic treatments, it has been shown that a large proportion of those patients withdrawing from treatment do so early in the intervention period18. Due to lack of data on discontinuation rates from long-term studies, it was assumed that patients who discontinued treatment did so during the first 6 months and that all remaining patients continued treatment for the full period of the model. Patients who discontinued did not receive any treatment benefit but was allocated 3 months of intervention costs. Discontinuation rates of 100% and 50% were tested in the sensitivity analysis.

Epidemiological data

Fracture risk

Age-differentiated fracture risks for hip, clinical vertebral and wrist fracture in a female Swedish population were derived from Kanis et al.19. Incidences of other non-spine, non-hip osteoporotic fractures were derived from Kanis et al.20. These risks were mainly based on Swedish data but some were imputed from fracture risk data from the US (Olmsted County; Rochester)21. The population fracture risk needs to be adjusted to reflect the increased fracture risk in the target patient group. Based on an approach previously described in Kanis et al. and De Laet et al.20,22, the relative risk of fracture for the target patient group (i.e. the EFOS Swedish patient group) compared to the population risk was estimated based on their age, T-score and number of prevalent fractures. For the base case, the relative risk of fracture was estimated to be 5.33, 6.08, 3.65 and 3.48 for hip, vertebral, wrist and other osteoporotic fractures. These relative risks vary at different ages, T-scores and fracture prevalence.

A fracture is a strong risk factor for another fracture. The risk of a subsequent fracture is at its highest immediately after a fracture event, and subsequently decreases over time. Risk functions from a Swedish study that estimated the risk of subsequent fractures over a 5-year period after an initial fracture event23 was therefore incorporated into the model. In the model simulations, these risks were used up to 5 years after fracture or until the baseline fracture risk was higher. The same approach was used to calculate the risk of a subsequent fracture in the model by Lundkvist et al.7.

Mortality

Mortality rates for the general female population in Sweden were based on the years 2001–200524. Hip and clinical vertebral fractures lead to an increased mortality compared to the general population25–29. The mortality after a hip fracture was estimated based on all hip fracture patients from the Swedish inpatient register and causes of death register between the years 1997 and 2001. Swedish age differentiated clinical vertebral mortality rates in the first and subsequent years after a fracture event were derived from a study by Johnell et al.30.

The excess mortality after a fracture cannot entirely be related to the fracture event but also to co-morbidity factors among osteoporotic patients31–33. The proportion of deaths related to the actual hip fracture event has been estimated to lie between 15% and 50%32,33. Therefore, in this analysis it was assumed that 30% of the excess mortality (compared to mortality in the general population) after a hip and a vertebral fracture was associated with the fracture event itself. In the sensitivity analysis, the proportion of excess deaths related to fractures was varied from 0% (no excess mortality after fracture) to 100% (all deaths related to fractures). There are no indications that wrist fractures are associated with any excess mortality25,26. Also, because of lack of data it was conservatively assumed that other osteoporotic fracture types were not associated with an increased mortality in the base-case scenario. In the sensitivity analysis, it was assumed that excess mortality after these fractures was half the excess mortality after a hip fracture. This is based on the assumption of the relative utility loss of other osteoporotic fractures relative to hip and wrist fractures made by Kanis et al.34.

Costs

All costs were inflated to mid-2007 prices using the Swedish consumer price index24. Costs are given in euros (€) using the average exchange rate (9.22SEK/€) for the first 6 months of 200735.

Intervention costs

The price of teriparatide in Sweden is €404 (3,728.50 SEK) for 28 doses and for PTH(1-84) the price is €1,074.9 (9911 SEK) for 84 doses36. Therefore, on an annual basis the price for teriparatide will be €5275.2 (48,637 SEK) and €4674.1 (43,095 SEK) for PTH(1-84).

The monitoring of patients while on treatment was assumed to consist of one annual physician visit (€106) and a bone mineral density measurement (BMD) every second year (€126.60). Also, in the treatment guidelines and SPC for PTH(1-84) calcium monitoring visits are recommended at months 1, 3 and 636,37. In a sensitivity analysis, the costs (€412 or 2800 SEK/visit) for these visits were added to the monitoring costs for PTH(1-84) and in another analysis both treatments were ascribed these extra monitoring costs.

Disease costs

Direct and indirect (productivity losses) fracture costs during the first year after a hip, clinical vertebral and wrist fracture were derived from the Swedish KOFOR study which collected resource use and quality of life related to hip, vertebral and wrist fractures during the 18 months after a fracture38,39. Downward adjustments were made to account for deaths, the proportion of patients living at home at the time of fracture and the proportion of vertebral fractures that are hospitalised in clinical practice. The costs of other osteoporotic fractures in the first year after a fracture event were not available for Sweden. To derive an estimate of these costs it was assumed that pelvic, humerus, tibia, fibula and other femoral fractures were associated with the same cost as a hip fracture and rib, clavicle, scapula and sternum were assumed to have the same costs as a wrist fracture. These assumptions are derived from Kanis et al.34 in which this relationship was used for the quality of life loss related to all types of osteoporotic fractures. By weighting the assumed costs with the incidence of each fracture type an average cost of €6,455 was estimated.

Hip fracture costs relating to nursing home care in the second and following years were based on the age-differentiated proportion of patients that come from independent living before fracture and which reside in nursing home 1 year after fracture17. These patients were assumed to remain in a nursing home for the rest of their lives40 at an annual cost of €63,48041. It is also likely that patients that sustain a hip fracture and move to a nursing home would eventually be institutionalised regardless of the fracture. Therefore, in a sensitivity analysis a down adjustment of the long-term costs of hip fracture was made for the annual population incidence of being institutionalised42.

Both hip and vertebral fracture was associated with considerable costs up to 18 months after fracture in the KOFOR study. In the analysis it was assumed that these costs were maintained for the whole second year after fracture. Although, it is likely that there are fracture-related costs beyond the second year after fracture the KOFOR estimated costs were not extrapolated for a longer time period. However, in the sensitivity analysis the costs were used for the remainder of a fractured patient’s lifetime.

The costs related to a wrist fracture in the KOFOR study in the 12–18-month period were minimal and would not impact on the cost effectiveness, therefore, wrist fractures were assumed to have related costs only in the first year after fracture. It was also assumed that other osteoporotic fractures were not associated with any costs beyond the first year after fracture. All fracture related costs are summarised in Table 1.

Table 1.  Annual direct cost in different age groups, € (95% CI).

	Age (years)
Event	50–64	65–74	75–84	85–
	Mean (95% CI)*
1st year after event
Hip fracture1	9161	10,764	10,909	11,319
	(6596–12,459)	(8288–14,100)	(9382–12,546)	(9055–13,696)
Clinical vertebral fracture1	2405		7519
	(1323–3896)		(4136–12,181)
Wrist fracture1	2138
	(1754–2673)
Other osteoporotic fractures2	6455
	(5296–8070)**
2nd year after event
Hip fracture – Nursing home cost3***	3057	6209	10,354	13,123
	(2243–3940)	(5344–7089)	(9570–11,111)	(11,934–14,311)
Hip fracture – excluding living arrangements1	1690	119	153	2934
	(1217–2299)	(919–1563)	(1315–1758)	(2347–3550)
Clinical vertebral fracture1	1266		4624
	(696–2051)		(2543–7490)

*Confidence intervals were obtained by the bias-corrected accelerated (bca) percentile bootstrapping method47.

**Assumed to be same proportional variance as for wrist fracture.

***The cost at 57, 70, 80 and 85 years respectively.

138,39; 234 and own estimations; 317.

Cost related to hypercalcaemia as an adverse event was derived from a US-based analysis43 that estimated the cost as $130 (2003 prices), which is €195 (2007 prices).

Costs in added life-years

The difference between consumption and production varies over the life cycle for any individual. It has been argued that this difference, called costs in added life-years, should be included in cost-effectiveness evaluations44. The base-case analysis is provided with and without costs in added life-years.

Quality of life

The loss in quality of life, measured by the EQ-5D instrument, after a hip fracture, vertebral fracture and wrist fracture was derived from the KOFOR study38,39. The estimated quality of life loss in the second year after a hip and vertebral fracture was assumed to persist for all years beyond the first year after fracture. Wrist fracture was only assumed to have a reduction in quality of life in the first year after fracture. The proportionate quality of life loss in the first year after other osteoporotic fractures was calculated as 0.87 using the same method as described in the cost section above. Other osteoporotic fractures were assumed to have no reduction in quality of life beyond the first year after fracture event. By linking these fracture-related proportional quality of life losses to Swedish population utility values45, age differentiated fracture specific quality of life weights were obtained34,46. All fracture-related quality of life values used in the model are summarised in Table 2.

Table 2.  Fracture related quality-of-life loss.

Event	Mean (95% CI)*
1st year after event
Hip fracture1	0.80
	(0.77–0.85)
Clinical vertebral fracture1	0.65
	(0.56–0.74)
Wrist fracture1	0.94
	(0.92–0.98)
Other osteoporotic fractures2	0.87
	(0.85–0.91)**
2nd year after event
Hip fracture1	0.90
	(0.87–0.95)
Clinical vertebral fracture3	0.93
	(0.84–1.0)***

* Confidence intervals were obtained by the bias-corrected accelerated (bca) percentile bootstrapping method47.

**assumed to be same proportional variance as for wrist fracture.

***assumed to be same proportional variance as for vertebral fracture the first year.

1.38,39; 234 and own estimations; 334.

In the sensitivity analysis the cost effectiveness was estimated assuming no quality of life loss beyond the first year after a vertebral fracture. In another sensitivity analysis the fracture related utility multipliers were varied by ±50% from the mean estimates. Empirical data on the quality of life loss related to hypercalcaemia are not available. Therefore, we assumed a 2% quality of life reduction with hypercalcaemia43.

Univariate sensitivity analysis

The numeric assumptions for certain model parameters were each varied in turn so as to assess the sensitivity of the model results to each of the assumptions made. The sensitivity analyses performed are outlined in Table 3.

Table 3.  Sensitivity analysis.

	Teriparatide		PTH(1-84)
	Cost per QALY gained	% change compared to base case	Cost per QALY gained	% change compared to base case
Base case estimates	43,473	0	104,396	0
Assuming no effect on hip fractures with teriparatide	88,379	103	–	–
Assuming lowest effect on vertebral fracture with PTH(1-84) from the TOP study	–	–	163,576	57
Discount rate 0%	34,465	–21	86,196	−17
Discount rate 5%	49,938	15	117,181	12
Other osteoporotic fractures excluded	47,616	10	104,396	0
12 months treatment duration	32,782	−25	89,477	−14
24 months treatment duration	50,363	16	114,357	10
No offset time	108,067	149	190,370	82
Doubled offset time	21,185	−51	77,367	−26
100% persistence	38,528	−11	92,248	−12
50% persistence	52,611	21	113,817	9
100% compliance	58,915	36	127,946	23
No co-morbidity	28,816	−34	58,883	−44
No excess mortality after fracture	65,876	52	194,823	87
Increased mortality after other osteoporotic fracture	41,629	−4	110,913	6
Hypercalcaemia related adverse events included	45,284	4	126,954	22
Calcium monitoring costs included	54,439	25	124,966	20
Costs related to vertebral fractures beyond the 2nd year after fracture included	34,153	−21	87,160	−17
Costs related to hip fractures other than nursing home beyond the second year after fracture included	42,032	−3	–	–
Costs related to hip and vertebral fractures beyond the second year after fracture included	32,735	−25	87,342	−16
Adjusting long-term nursing home costs related to hip fracture for the population incidence of moving to nursing home	49,120	13	–
No quality of loss beyond the first year after vertebral fracture	49,821	15	137,079	31

Probabilistic sensitivity analysis (PSA)

The collective sensitivity of the model results to assumptions relating to treatment effects, fracture related costs, quality of life loss, mortality after hip fracture and long-term institutionalisation after hip fracture were assessed through a probabilistic sensitivity analysis. The uncertainty for these variables was considered by randomly (using a uniform distribution) drawing bootstrapped mean values from the underlying individual data47. The effects of the drugs were assigned log normal distributions based on the reported confidence intervals in the clinical trials. The PSA was conducted by drawing 2000 samples and were presented in the form of acceptability curves, which show the proportion of samples that fall below different values of willingness to pay for a QALY gained. For the interpretation of the cost-effectiveness ratios a WTP of a QALY of €60,000 was used, which is in the vicinity of the threshold acknowledged by Swedish reimbursement authorities48.

Results

Base-case analysis

The base-case cost-effectiveness results are presented in Table 4 and Table 5. Teriparatide compared to no treatment was associated with a cost per QALY gained of €43,473 and €60,450 when excluding and including costs in added life-years, respectively. The corresponding cost per QALY gained for PTH(1-84) compared to no treatment were €104,396 and €124,897. Teriparatide was associated with less costs and more QALYs and life-years when compared to PTH(1-84).

Table 4.  Simulated average costs and effects per patient in base-case analysis (€).

	No treatment	Teriparatide	PTH(1-84)
Costs
Fracture costs	47,473	45,619	47,251
Treatment costs	0	4,993	4,010
Costs in added  life years	281,252	282,478	281,999
Total costs	328,725	333,090	333,260
Effects
Life years	11.977	12.023	12.005
Quality-adjusted  life-years	8.008	8.080	8.044

Table 5.  Base-case incremental cost-effectiveness analysis.

Comparators	teriparatide vs. no treatment	PTH(1-84) vs. no treatment	teriparatide vs. PTH(1-84)
Incremental cost including CALYs	4,365	4,535	−170
Incremental cost excluding CALYs	3,139	3,787	−649
Life-years gained	0.046	0.028	0.018
QALYs gained	0.072	0.036	0.036
Cost per life-year gained including CALYs	94,409	161,716	−9,345 (teriparatide dominant)
Cost per life-year gained excluding CALYs	43,473	104,396	−18,061 (teriparatide dominant)
Cost per QALY gained including CALYs	60,450	124,987	−4,732 (teriparatide dominant)
Cost per QALY gained excluding CALYs	43,473	104,396	−18,061 (teriparatide dominant)

CALY, costs in added life years.

In Figure 2, the results of the PSA for the base-case scenarios are shown in the form of acceptability curves. Given a willingness to pay (WTP) of €60,000, teriparatide was found to be cost effective in about 80% of the cost-effectiveness simulations compared to no treatment whereas PTH(1-84) was not below this WTP in any of the simulations. Compared to each other teriparatide was cost effective in 97% of the simulations at the assumed threshold.

Figure 2.  Acceptability curves for the base-case scenarios.

Sensitivity analysis

The cost per QALY gained of different starting ages of treatment is shown in Figure 3. Although the risk of vertebral fracture increases with age, the gain in avoiding vertebral fracture decreases which is the reason why the cost effectiveness slightly improves up to an age of 71 years but declines thereafter for PTH(1-84) compared to no treatment. The reason for the steady decrease in the ICER for teriparatide is the applied treatment effect on all fractures, especially hip fracture for which the risk increases exponentially in ages above 70 years. In Table 3, the cost per QALY gained and the percentage change compared to the base case estimates for various different sensitivity analyses is shown. The impact on the ICER was moderate for changes in variables and parameters such as discount rate, exclusion of other osteoporotic fractures, treatment duration, persistence rates, increased mortality after other osteoporotic fractures, including consequences of hypercalcaemia-related events, calcium monitoring costs, extrapolating costs from the KOFOR study for hip and vertebral fracture beyond the 2nd year after fracture and no quality of life reduction related to vertebral fracture after the 1st year. Compared to the base-case estimates the ICER is fairly stable (±30% change) for most of the estimated sensitivity scenarios. A larger impact on the ICER was seen for changes in compliance, excess mortality after fracture and treatment effect on hip fracture risk. The parameter that was the most sensitive was assumptions concerning the offset time of the fracture risk reduction after the treatment period. When assuming that the effect stops immediately after the treatment period, the ICER increases by 149%, whereas by doubling the length of the offset time period the ICER is reduced by 51% for teriparatide. When assuming no effect on hip fractures with teriparatide, the cost-effectiveness ratio increased by 103%. Also, the cost per QALY gained of teriparatide vs. PTH(1-84) was estimated at €47,656 when no effect on hip fracture risk was assumed.

Figure 3.  Cost per QALY gained (€) for different starting ages of treatment.

The sensitivity analysis over the variation of fracture related quality of life loss ±50% from the baseline quality of life estimates are presented in the form of a tornado diagram in Figure 4.

Figure 4.  Sensitivity tornado diagram over variation of fracture related quality of life loss ±50% from baseline estimates based on teriparatide (effect on all fractures) vs. no treatment.

A 50% change in the utility loss after fracture does not lead to any major changes in the ICER compared to the base case. However, it is clearly the quality of life related to vertebral fracture in the 1st year and hip fracture in the 2nd and following years that are the most sensitive (about a 10% change compared to the base case), while changes in the quality of life related to other osteoporotic fractures and wrist fractures are marginal (about 1% compared to the base case).

Discussion

The cost effectiveness of teriparatide and PTH(1-84) in post-menopausal women in Sweden was estimated via a Markov model from a societal perspective using the fracture risk results from the pivotal clinical trials for each drug5,6. Compared to a no-treatment alternative, the cost per QALY gained was estimated as €43,473 for teriparatide and €104,396 for PTH(1-84). Treatment with teriparatide was less costly and associated with more life-years and QALYs than PTH(1-84). The main reason for this was a slightly better risk reduction for vertebral fractures for teriparatide, but more importantly the fact that there is no evidence on reduction of non-vertebral fractures with PTH(1-84). This is an essential part of the analysis. One might argue that there should be a class effect shared by these compounds; nonetheless this was not observed in the phase III trial. In the paper by Greenspan et al.6, it is suggested that the lack of effect of PTH(1-84) is due to lack of power in the TOP study. The incidence of peripheral fractures in the placebo group was of 5.9% in TOP versus 9.7% in the FPT study5, which suggests a lower power. However the power is driven by the numbers of events, and given the fact that treatment groups were considerably larger in the PTH(1-84) trial (n = 1246) versus the teriparatide trial (n = 544) there were a similar number of non-vertebral fractures in the two trials. This was also the case for the number of some of the typical osteoporotic fractures such as forearm and hip. Therefore, the fact that the groups are larger would compensate for the fact that the incidence was lower in the PTH(1-84) trial, and thus the power for showing this effect is similar.

As these two clinical trials estimate the treatment effect compared to placebo, the relative efficacy between the compounds had to be derived by adjusted indirect comparison. That is, the relative effect of the treatments of interest are linked through a common comparator, i.e. placebo49. The preferable approach would have been to derive evidence from a direct head-to-head trial, since an indirect comparison does not consider differences in, for example, study design, patient characteristics or factors that might have impacted the results in the clinical trials. However, based on the current available data the indirect comparison approach is the sole option to derive an estimate of relative efficacy for the use in a cost-effectiveness analysis. Therefore it is important that the results of the comparative cost-effectiveness analysis have to be interpreted in the light of this.

In the FPT, teriparatide was shown to significantly reduce the risk of non-vertebral fractures, which was a secondary endpoint in the study. Separated, the fracture types would not show a significant risk reduction, mainly because the trial was not powered to detect such significant changes. Using non-vertebral fracture as an aggregated outcome has become more common in recent clinical trials due to the difficulty in including patients at sufficiently high risk of fracture. The difficulty of applying a non-vertebral fracture risk reduction in a cost-effectiveness analysis is that the fracture types differ substantially in terms of costs and morbidity, e.g. hip fracture vs. wrist fracture. Using an aggregated measure of efficacy in the cost-effectiveness analysis might lead to underestimations of the potential benefits for some fracture types and overestimations for others. One approach could be to exclude all fracture types that have not shown significant risk reduction independently. This approach would most likely cause all fracture types except vertebral fractures to be excluded from the cost-effectiveness analysis and lead to an apparent underestimation of the fractures avoided with treatment. The other approach is to assume the non-vertebral fracture efficacy for all fracture types (except vertebral fractures). In this study we chose the latter alternative in the base-case, however, we conducted sensitivity analysis excluding a risk reduction of hip fractures with treatment.

Although the effect of treatment is an important parameter in a cost-effectiveness model, there are other variables which might impact on results. In the sensitivity analysis the ICER, however did not alter substantially with changes in the majority of the model parameters. For teriparatide, the cost per QALY gained compared to no treatment was below €70,000 in all sensitivity analyses except when assuming no residual effect on fracture risk after the intervention period. The model was particularly sensitive to the time over which the treatment effect is assumed to act. It must be assumed that the effect on fracture risk ceasing immediately after treatment is stopped is unrealistic, as is assuming that the effect will last indefinitely. Therefore, the offset time periods used in this study (24 and 30 months for vertebral and non-vertebral fractures, respectively) were derived from the follow-up studies of the FPT trial15,16 and should be seen as reasonable assumptions.

A common approach when estimating the cost effectiveness of osteoporotic treatments has been to use the patient characteristics of the clinical trial from which the treatment effect was derived when comparing one treatment with placebo5,6. The advantage with this approach is that the link between the treatment effect in the clinical trial and the patient group in the cost-effectiveness analysis is maintained even though the cost effectiveness is assessed in different geographical settings (internal validity). The disadvantage is that the patients in clinical trials often do not reflect the patients that will be given the treatment in clinical practice making the interpretation of the estimated cost effectiveness and patients in clinical practice unclear (external validity).

The analysis in this study was based on patient characteristics of the subjects in the Swedish cohort of the multinational EFOS study, resulting in an advantage in external validity. Although the link between the fracture-risk efficacy from the clinical trials and the patient group might be weaker, the interpretability of the cost effectiveness for the indication of PTH in Sweden is stronger.

The model also accounted for poor compliance. However, the compliance data used in the cost-effectiveness simulations were derived from the clinical trials. In a clinical setting, compliance is more likely to be poorer than what is observed in a controlled clinical trial, although studies have shown that in clinical settings patient education and follow-up programmes for teriparatide can result in high persistence and compliance rates50–52.

In this study PTH was not compared to other available osteoporotic treatments. This is because in Sweden, PTH is currently indicated for second- and third-line treatment in patients with osteoporosis (T-score < −2.5) with at least one prevalent clinical vertebral fracture and it should be documented that the patient cannot take any other osteoporosis drug due to safety and tolerability reasons. PTH could also be considered as a first-line treatment option for patients with a T-score of less than −3 and have at least two clinical vertebral fractures. Thus, the usage of PTH drugs for the treatment of osteoporosis is restricted to a fairly small segment of the osteoporosis population where no other osteoporotic treatment is a relevant option.

There are no other published cost-effectiveness estimates of PTH(1-84) treatment in Sweden that the authors are aware of. The cost effectiveness of teriparatide in a Swedish setting has previously been assessed by Lundkvist et al.7, who found a cost per QALY gained for treatment of a population of 69-year-old women with a T-score at the femoral neck of −3.0 SD of €20,000–64,000 for patients with a recent or historic vertebral fracture. Although the analysis was not based on the same patient groups, the results from the previous study are similar to the results from this current study. The Lundkvist et al.7 study was conducted before the introduction of PTH(1-84) in Sweden yet suggested that teriparatide might provide clinical and economic benefits.

Conclusions

Based on the efficacy estimates from pivotal clinical trials and characteristics of patients treated in clinical practice in Sweden, teriparatide appears to be cost effective relative to PTH(1-84) or no treatment. Randomised, head-to-head clinical trials of teriparatide and PTH(1-84) would provide further insight into the relative efficacy of these treatments, and thereby the validity of the current study results and conclusions.

Transparency

Declaration of funding

This study was supported by Lilly Europe.

Declaration of financial/other relationships

F.B. and O.S. have disclosed that they are have undertaken health economic analyses in the field of osteoporosis for several agencies and pharmaceutical companies, including Lilly. F.M. has disclosed that he is an employee of Lilly, and A.K. has disclosed that he is a former employee of Lilly. Ö.L. has disclosed that he has been a scientific advisor for Lilly.

Acknowledgements

The authors would like to thank Annabel Barrett, Lilly Europe for her help in preparing this manuscript.

Notes

*Forsteo, Eli Lilly AB, Sweden.

†Preotact, Nycomed AB, Sweden.