Osteoporosis International

, Volume 18, Issue 8, pp 1047–1061

Cost-effectiveness of alendronate in the treatment of postmenopausal women in 9 European countries - an economic evaluation based on the fracture intervention trial

Authors

    • European Health Economics
  • F. Borgström
    • European Health Economics
    • Medical Management CentreKarolinska Institutet
  • S. S. Sen
    • Outcomes Research, Merck & Co., Inc
  • S. Boonen
    • Leuven University Center for Metabolic Bone Diseases and Division of Geriatric MedicineKatholieke Universiteit Leuven
  • P. Haentjens
    • Department of Orthopaedics and TraumatologyAcademisch Ziekenhuis van de Vrije Universiteit Brussel
  • O. Johnell
    • Department of OrthopaedicsMalmö General Hospital
  • J. A. Kanis
    • Centre for Metabolic Bone Diseases (WHO Collaborating Centre)University of Sheffield Medical School
Original Article

DOI: 10.1007/s00198-007-0349-5

Cite this article as:
Ström, O., Borgström, F., Sen, S.S. et al. Osteoporos Int (2007) 18: 1047. doi:10.1007/s00198-007-0349-5
  • 298 Views

Abstract

Summary

Treatment with alendronate (Fosamax®) has been shown to significantly reduce the risk of fragility fractures. Cost-effectiveness of treatment was assessed in nine European countries in a Markov model and was generally found to be cost effective in women with a previous spine fracture.

Introduction

Treatment with alendronate (Fosamax®) reduces the risk of osteoporotic fractures at the spine, hip and wrist in women with and without prevalent vertebral fracture. Cost-effectiveness estimates in one country may not be applicable elsewhere due to differences in fracture risks, costs and drug prices. The aim of this study was to assess the cost-effectiveness of treating postmenopausal women with alendronate in nine European countries, comprising Belgium, Denmark, France, Germany, Italy, Norway, Spain, Sweden, and the UK.

Methods

A Markov model was populated with data for the nine European populations. Effect of treatment was taken from the Fracture Intervention Trial, which recruited women with low BMD alone or with a prior vertebral fracture.

Results

The cost per QALY gained of treating postmenopausal women with prior vertebral fractures ranged in the base case from “cost saving” in the Scandinavian countries to €15,000 in Italy. Corresponding estimates for women without prior vertebral fractures ranged from “cost saving” to €40,000.

Conclusions

In relation to thresholds generally used, the analysis suggests that alendronate is very cost effective in the treatment of women with previous vertebral fracture, and in women without previous vertebral fracture, cost-effectiveness depends on the country setting, discount rates, and chosen monetary thresholds.

Keywords

AlendronateBisphosphonatesCostEuropeFractureOsteoporosis

Introduction

Osteoporosis is an increasingly important health problem. Approximately two-thirds of all osteoporosis-related fractures occur in women. The number of hip fractures in the world is projected to increase from 1.7 million in 1990 to more than 6 million in 2050 [1]. In addition to having a negative impact on the quality of life of the individual, the disease is also costly for society. The expenditure associated with osteoporosis is expected to increase in the future, because of changes in demographics and improved life expectancy and, in some regions of the world, due to an increasing in the age-specific incidence of fractures.

Alendronate has been shown in clinical trials to reduce the risk of fractures by nearly one half. The largest study, the “Fracture Intervention Trial” (FIT), was conducted at 11 clinical centres in the United States and included 6,459 women with a femoral neck BMD value of 0.68 g/cm2 or less. FIT consisted of two study arms; the Vertebral Fracture Arm (VFA) including 2,027 women with radiographically identified vertebral fractures at baseline; and the Clinical Fracture Arm (CFA) with 4,432 women without vertebral fractures at baseline [24].

Previous studies have investigated the cost-effectiveness of alendronate in the treatment and prevention of osteoporosis in several countries (e.g., Sweden, UK, Denmark and Canada [510]). These analyses have, in various degrees, shown alendronate to be cost-effective in the treatment of osteoporosis, but such data may not be generalisable to other countries since the risk of fractures and their costs may vary markedly. In addition, comparisons of cost-effectiveness between those countries studied are difficult since different modelling frameworks and clinical data sets have been used. The objective of this study was to estimate the cost-effectiveness of alendronate in the treatment and prevention of osteoporosis in a number of European countries, based on the clinical results in the FIT. The countries included in the study were Belgium, Denmark, France, Germany, Italy, Norway, Spain, Sweden, and the UK.

Methods and materials

The cost-effectiveness analysis took a health care perspective. Thus, only direct costs associated with the disease were included and indirect costs and costs in added life years were not considered. Quality-adjusted life years (QALYs) were used to measure health effects.

The cost-effectiveness model

The computer simulation model used to calculate cost-effectiveness was an updated version of a Markov cohort model that has been used previously to estimate the cost-effectiveness of bisphosphonates in Sweden [5, 11]. The main difference is that the model (Fig. 1) can now accommodate the long term effects of vertebral fractures. The cycle length is one year and all patients are followed from start of treatment until they are 100 years old or dead. There is a probability of remaining in the same state or of dying. All the patients begin in the well health state. Each year a patient has a probability of having a fracture, remaining healthy or of dying. If a patient dies, she will move to the dead health state and remain there for the rest of the simulation. If she sustains a fracture she will move to the hip fracture, spine fracture or wrist fracture state. After a hip fracture, a patient cannot sustain any further vertebral or wrist fractures, but it is possible to have multiple hip fractures. If a patient does not sustain an additional hip or vertebral fracture or dies, she goes to a post fracture state where costs and utilities are assigned. After a vertebral fracture it is possible to have a hip fractures but not wrist fractures, which gives a slight underestimation of the number of fractures in the model. The reason for not including these fracture combinations was to avoid the many health states where little or no information is available on costs or utilities.
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-007-0349-5/MediaObjects/198_2007_349_Fig1_HTML.gif
Fig. 1

Note on structure of the model: There is always a probability of remaining in the same state or dying (these transition arrows are excluded from the figure for simplification)

Target patient groups

The patient groups simulated for the base cases in this study were defined from the Fracture Intervention Trial, which included female patients aged 55–80 years treated for three years. The first patient group was based on the Vertebral Fracture Arm of FIT [3], (mean age = 71 years), with a T-score for BMD of −1.6 SD or less at the femoral neck and radiographically identified vertebral fractures at baseline. Mean BMD at the femoral neck was 0.57 g/cm2, which corresponds to a T-score of approximately −2.4 SD. The second patient group was based on a subgroup of patients in the 4-year Clinical Fracture Arm. Among the 4,432 women in the Clinical Fracture Arm, 1,631 met the WHO definition of osteoporosis based on an entry femoral neck T-score of −2.5 SD or less. These patients had a mean age of 69 years and no fractures at baseline. The mean BMD at the femoral neck was 0.53 g/cm2, which corresponds to a T-score of approximately −2.7 SD. This subgroup of the CFA (sCFA) showed a significant reduction in the risk of hip fractures and morphometric vertebral fractures.

Treatment with alendronate was compared to patients without alendronate treatment in the FIT, but all patients were given vitamin D and calcium supplements. In the model, treatment was assumed to start at the mean ages in the VFA (71 years) and the sCFA (69 years). In sensitivity analyses the starting age of treatment was varied between 60 and 80 years.

Effect of treatment

In the VFA of the FIT, treatment with alendronate significantly reduced the risk of hip fracture (RR 0.49, CI 0.23–0.99), clinical vertebral fracture (RR 0.46, CI 0.28–0.75), radiologic vertebral fracture (RR 0.53, CI 0.41–0.68) and wrist fracture (RR 0.52, CI 0.31–0.87). The subgroup of the CFA [2] showed significant risk reductions of hip fracture (RR 0.44, CI 0.18–0.97), any clinical fracture (RR 0.64, CI 0.50–0.82, nonvertebral fractures (RR 0.65, CI 0.50–0.83) and radiologic vertebral fracture (RR 0.51, CI 0.31–0.84). Effect on wrist fractures (RR 0.88, CI 0.55–1.40) or clinical vertebral fractures (RR 0.84, CI 0.38–1.83) was not significant. In the VFA base case analysis, the fracture risk was assumed to be decreased for the hip, clinical vertebral and wrist fractures. In the sCFA base case analysis, the fracture risk was decreased for hip fracture and vertebral fracture. The Vertebral Fracture Arm of the study shows very comparable findings between clinical vertebral fracture and radiologic vertebral fracture in terms of risk reduction. Therefore, the effect on morphometric fractures in the sCFA was used to estimate the risk reduction of clinical vertebral fractures. In a sensitivity analysis it was assumed that alendronate did not reduce the risk of clinical vertebral fractures. There are other, smaller trials that also have assessed the clinical efficacy of alendronate. Risk reduction estimates from a meta-analysis by Stevenson and Davis [12] were used in sensitivity analyses for both sCFA and VFA. Used RR for vertebral, hip and wrist fracture were; 0.56, 0.62 and 0.67. The meta-analysis was based on patients both with and without prevalent vertebral fracture and the same RR was thus used for both arms.

In the base case, patients were assumed to be treated with alendronate for five years. Recent studies have shown that the effect on bone density and suppression of bone turnover of alendronate treatment are maintained after the intervention period [1316]. For this reason, we assumed that the treatment effect on fracture risk would be sustained for two years after stopping treatment and was followed by a three year period, also known as the offset-time, where the effect declined linearly to zero. In previous studies of the cost-effectiveness of bisphosphonates [5, 11, 17], the effect of treatment has been assumed to decline over five years without any sustained effect after treatment was stopped. This scenario was tested in sensitivity analysis. Also tested in sensitivity analysis was a 10-year treatment period and one scenario with no effect of treatment after the intervention period.

Baseline risk of fracture

Age specific hip fracture risks for each country were derived from various sources. [1824]. The hip fracture risks varied quite substantially between the countries as seen in Fig. 2. The incidence data was linearly extrapolated to 100 years of age where data were missing. Age specific clinical vertebral fracture risks were only available for a Swedish population [22]. To assess vertebral fracture risks for the other countries, it was assumed that the ratio of clinical vertebral fracture to hip fracture in Sweden [22] was similar to that of the other countries. That is, the vertebral fracture risk was obtained by multiplying this ratio with hip fracture risk in each country. Age specific wrist fracture risks were available for Denmark, Sweden and the UK. For the other countries the wrist fracture risk was derived using the same method used for the vertebral fracture risk.
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-007-0349-5/MediaObjects/198_2007_349_Fig2_HTML.gif
Fig. 2

Hip fracture incidence (per 10 000) in the female population

Relative risk of fracture

The population fracture incidence rates were adjusted to reflect the risk in the target patient groups. A method to calculate fracture risks for different patient groups relative to the population fracture risks based on bone mass density (BMD) and prevalent fractures is described in Kanis et al. [25] and has also been used in a cost-effectiveness study of risedronate in the UK [26]. Relative risks of fractures were estimated from the gradient of risk (RR of fracture/SD change in BMD) reported in a meta-analysis by Marshall et al. [27]. T-score values were based on the reference values for femoral neck BMD using the NHANES III survey (age 20–29 years) [28]. An assumption in using this method is that BMD levels are similar between countries. The formula for the estimation of the relative risk of fracture based on BMD RRfxBDM can be expressed as:
$$ RRfx_{{BMD}} = \exp {\left( { - Ln{\left( {RRfx_{{sd}} } \right)}{\left( { - Z\_score} \right)} - {\left( {Ln{\left( {RRfx_{{sd}} } \right)}} \right)}\hat{}2 \mathord{\left/ {\vphantom {2 2}} \right. \kern-\nulldelimiterspace} 2} \right)} $$
(1)
where RRfxsd is the increase in age-adjusted relative risk of fracture associated with one standard deviation decrease in bone mineral density (2.6, 1.8 and 1.4 for hip fracture, vertebral fracture and wrist fracture, respectively [27]) and the Z-score is the number of standard deviations a BMD value is below the age matched mean BMD. Additionally, relative risk calculations took into account the relationship between prior vertebral fracture and subsequent fractures found in another meta-analysis by Klotzbuecher et al. [29], and the prevalence of vertebral fractures in the general population. The increased risks of fracture due to prevalent vertebral fracture (RR of hip fracture: 2.3; RR of vertebral fracture: 4.4; RR of wrist fracture: 1.4) are adjusted for age, but not for BMD and were therefore down adjusted by 10% [29, 30].
These relative risks of fractures for the target patient groups must also account for the relationship between prior vertebral fractures, subsequent fractures and the prevalence of vertebral fractures in the general population. The calculation is as follows:
$$ RRfx = RRfx_{{BMD}} *{\left( {{RRprev_{{vfx}} } \mathord{\left/ {\vphantom {{RRprev_{{vfx}} } {{\left( {RRprev_{{vfx}} *Vfx\_prevalence + {\left( {1 - Vfx\_prevalence} \right)}} \right)}}}} \right. \kern-\nulldelimiterspace} {{\left( {RRprev_{{vfx}} *Vfx\_prevalence + {\left( {1 - Vfx\_prevalence} \right)}} \right)}}} \right)} $$
(2)
where RRprevvfx is the increased risk of fracture due to prior vertebral fracture and Vfx_prevalence is the prevalence of vertebral fractures in the population at a given age. The prevalence of vertebral fractures in the general population of Sweden was adjusted by a factor of 0.576 [31] to estimate the prevalence in non-Scandinavian countries.

Using this method, we calculated the age specific relative risk of fractures for the two patient groups.

Mortality

Female country and age-specific normal mortality rates were derived from national sources [3240].

Hip and vertebral fractures are associated with an increase in mortality. In a study by Oden et al. [41] age-differentiated mortality the years following hip and vertebral fractures in Sweden were calculated. These estimates of excess fracture mortality were used for all countries by assuming that the relative risk of mortality the years after a hip and vertebral fracture was the same as in Sweden.

However, not all excess mortality after fracture is causally related to the fracture event. Kanis et al. [42] estimated that 23%–37% of the mortality after hip fracture was associated with the fracture event, similar to the estimate of Parker and Anand [43]. In a review of case notes, 42% of all deaths after a hip fracture were considered to be possibly related and 25% directly related to the hip fracture. For vertebral fractures Kanis et al. [44] have estimated that 28% of deaths are causally related to the fracture. For this analysis, it was assumed that 30% of the total excess mortality could be related to the hip or vertebral fracture event. Wrist fracture was assumed to be associated with no excess mortality after fracture.

Costs

All costs are in year 2004 values. Costs were, when necessary, inflated using country specific consumer price indices. All costs are given in the euro (€) currency and costs were converted to euro (€) at the average annual exchange rates for 2004. Discount rates for costs and effects were based on current recommendations and guidelines for health economic evaluation in the respective countries. Both costs and effects were discounted at a rate of 3% in Sweden, Denmark, France, Italy, Spain and Norway, 5% in Belgium and Germany, and 3.5% in the UK [45]. The ISPOR guidelines for pharmacoeconomic evaluations state that 6% should be used for Spain, but this figure is 11 years old and should be considered as outdated since the long term interest rates are around 3% which suggests that this value is a suitable discount rate [46]. In a sensitivity analysis the cost-effectiveness was estimated using 3% discount rates for all countries in order to simplify comparisons.

Cost of fractures

Costs of a fracture can be divided into short-term costs, which occur the first year following the fracture, and long term costs, which may persist for several years after fracture or even for the remainder of the lifetime of the patient.

There is considerable variability in the amount and quality of cost data available for the countries in the present study. Estimates of hip fracture costs in the first year were available for all countries except France. The French hip fracture cost was derived from the Swedish estimated cost of a hip fracture [47, 48]. To make the cost appropriate for a French environment the difference in resource use and price levels between Sweden and France have to be considered. The average length of stay for an osteoporosis related hospitalisation is 1.05 times longer in France than in Sweden [49] and the general price level in France is 0.92 [50] of that in Sweden. Multiplying these two values (1.05*0.92 ≈ 0.97) gave a factor which was used to convert the Swedish hip fracture cost to a French setting. To obtain vertebral and wrist fracture costs when not available, the first year hip fracture cost has been used as a template together with morbidity equivalents where country specific data were lacking. The cost of hip fracture at the age of 70 years was multiplied with morbidity equivalents for wrist and vertebral fractures presented by Kanis et al. [51]. Sources, conditions and assumptions for cost data in the model are summarised in Table 1. Hip fracture costs the second and following years were based on the age-differentiated proportion of patients who come from own living before fracture, that reside in nursing home 1 year after fracture [52]. These patients were assumed to remain in nursing home for the rest of their lives [53]. The proportions going to long term care after hip fracture and also incurring additional costs at different ages were; 50–69, 7% ; 70–79, 10% ; 80–89, 15% ; over 90, 23%.
Table 1

Costs (€)

 

Belgium

Denmark

France

Germany

Italy

Norway

Spain

Sweden

UK

1st year

         

Hip

16,823 [63]

22,283 [8, 64]

50–64: 8 925, 65–74: 9 716, 75–84: 17 158, 85–:23 983 [47]f

17,399 [24]

17,478 [65]

24,059 [66]

6,922 [62]

50–64: 9 567, 65–74: 10,414, 75–84: 18 392, 85–:25 707 [47]

50–64: 17 424, 65–74: 17 605, 75–84: 20,568, 85–:21 735 [67, 68]

Vertebral

3,721 [63]a

1,163 [64]

3,383 [69]f

50–64: 5 056, 65–74:5 549, 75–84: 6 197, 85–:6 197 [24]

3,866 [65]a

923 [66]

1,531 [62]

3,626 [69]

50–64: 2 833, 65–74: 2,561, 75–84: 2,124, 85–:1,021 [70]

Wrist

1,027 [51, 63]a

1,010 [64]

2,150 [69]f

1,062 [24]a

1,067 [65]a

672 [66]

423 [51, 62]a

2,304 [69]

50–64: 479, 65–74: 479, 75–84: 776, 85–:2,436 [67, 68]

2nd&following years

         

Hip

2,037 [63]

4,086 [71]

5,755 [72]f

3,082c

2,086 [65]

5,371 [66]

1,591 [51, 62]

6,169 [72]

2,968 [73]

Other

         

Physicians visit

21 [74]

169 [75]

25 [76]

17 [77]

32 [65]

91 [78]

76 [79]

130b

41 [80]

BMD

38e

52 [81]

109d

48 [81]

84 [81]

72 [78]

96 [81]

152 [82]

53 [83]

Drug cost

501 [84]

651 [85]

489 [86]

540d

525 [87]

502 [88]

527 [89]

433 [90]

444 [91]

aCosts are calculated as a fraction of hip fracture cost, according to the number of hip fracture morbidity equivalents at age 70 [51].

bBased on the average cost of a physician visit at six different county councils.

cBased on an average cost of three long term care facilities in Germany.

dPersonal communication Shuvayu S. Sen, MSD

ePersonal communication Patrick Haentjens

fThe average length of stay for an osteoporosis related hospitalisation is 1.05 times longer in France than in Sweden [49] and the general price level in France is 0.92 [50] of that in Sweden. Multiplying these two values (1.05*0.92 ≈ 0.97) gives a factor that was used to convert the Swedish vertebral cost to a French setting.

Cost of intervention

In line with previous standard assumptions about the monitoring of osteoporotic treatments, it was assumed that, besides the drug cost, an intervention with alendronate was associated with one yearly physician visit and a bone mineral density measurement every second year. The annual public drug cost of alendronate (Fosamax®, 10 mg tablet packages), physician visit cost and BMD cost in each country are shown in Table 1.

No cost of intervention was incurred after the intervention period. Both the treatment and control groups in the FIT received calcium and vitamin D supplements, which made it possible to exclude the cost of these agents.

Quality of life (QoL)

As for fracture associated costs, the utility loss after a fracture can be divided into short-term (first year after fracture) and long-term effects (second and following years). Patients without fracture were assumed to have the same estimated QoL as the population. Population utility values from the UK (0.82 (50–64 years), 0.78 (65–74 years), 0.72 (75–84) and 0.69 (85 years and above) were used for the UK, Belgium, France, Spain, Italy and Germany [54]. Population utility values from Sweden (0.91 (50–59 years), 0.87 (60–69 years), 0.70 (70–79) and 0.60 (80 years and above) were used for the Sweden, Norway and Denmark [55].

The QoL after fractures was obtained by relating proportionate disutilities (multipliers) to the population values. For hip fracture a multiplier of 0.792 and 0.9 was used the first year and subsequent years, respectively. Corresponding values for a clinical vertebral fracture were 0.626 and 0.93. Wrist fracture was assumed to have a utility loss only the first year after fracture (a multiplier of 0.977). [48, 56].

The reductions in QoL after fractures, mainly derived from a recent Swedish based empirical study [48], differed somewhat from previous estimates of health related QoL after fractures. In particular, vertebral fracture was assumed to be associated with a larger QoL reduction than previously assumed. In a review of utility values associated with osteoporotic fractures Brazier et al. [57] suggested a set of reference health state multipliers (hip fracture: 0.797, vertebral fracture: 0.909 and wrist fracture: 0.981). Since the suggested quality of life loss related to vertebral fracture deviates the most compared to what was used in this study, a sensitivity analysis was carried out assuming a 10% reduction the quality of life the first year after a vertebral fracture and no quality of life loss the second and following years.

The incremental cost-effectiveness ratio

The incremental cost-effectiveness ratio (ICER) was defined as
$$ ICER = \frac{{\Delta C}} {{\Delta E}} = \frac{{C_{1} - C_{0} }} {{E_{1} - E_{0} }} $$
(3)
where ΔC was the difference in total cost between interventions and no intervention, and ΔE was the difference in effectiveness (QALY or life years gained) between intervention with alendronate and no intervention.

Results

Base case analysis

The results of the base case cost-effectiveness estimations are shown in Table 2. The cost per QALY gained for women corresponding to characteristics in the VFA (i.e., 71-year old women with low BMD and previous vertebral fracture) varied from cost-saving (Sweden, Norway and Denmark) to €15,489 per QALY gained (Italy). The cost per QALY gained for women corresponding to characteristics in the sCFA (i.e., 69-year old women with low BMD and no previous vertebral fracture) varied from cost saving (Sweden and Norway) to €39,712 (Spain).
Table 2

Base case analysis (Base case discount rates in parentheses)

 

Belgium (5%)

Denmark (3%)

France (3%)

Germany (5%)

Italy (3%)

Norway (3%)

Spain (3%)

Sweden (3%)

UK (3.5%)

VFA

sCFA

VFA

sCFA

VFA

sCFA

VFA

sCFA

VFA

sCFA

VFA

sCFA

VFA

sCFA

VFA

sCFA

VFA

sCFA

Incremental cost (€)

718

1,361

–159

750

512

1,412

738

1,412

1,415

1,910

–1 907

–871

1,731

2,189

–1 576

–149

188

889

Quality Adjusted Life Years (QALYs) gained

0.111

0.057

0.188

0.121

0.110

0.051

0.096

0.051

0.091

0.048

0.173

0.110

0.131

0.066

0.150

0.091

0.139

0.075

Life years gained

0.071

0.037

0.153

0.099

0.065

0.031

0.059

0.031

0.056

0.030

0.125

0.080

0.082

0.041

0.108

0.065

0.096

0.052

Cost per QALY gained (€)

6,461

23,684

cost saving

6,201

4,670

27,419

7,658

27,821

15,489

39,712

cost saving

cost saving

13,193

32,943

cost saving

cost saving

1,356

11,849

Cost per life year gained (€)

10,090

36,975

cost saving

7,543

7,858

45,625

12,505

45,319

25,263

64,537

cost saving

cost saving

21,048

52,783

cost saving

cost saving

1,963

17,145

Stochastic analysis

Stochastic analyses were performed for both populations and are presented as acceptability curves in Figs. 3 and 4. The effects of alendronate were assigned distributions based on the reported confidence intervals and 4000 samples were drawn for each country and population. The figures show the proportion of samples that fell below different values of willingness to pay for an incremental QALY. A curve that never reaches a zero proportion on the QALY axis indicates that alendronate dominates no treatment in the given proportion of drawn samples (e.g., in the VFA for Norway, alendronate dominates no treatment in 82% of the drawn samples). Since the effect estimates are drawn from a normal distribution, the points where 50% of the samples fall under the given threshold should coincide with the ICERs from the regular cohort simulations (Table 2). It should be noted that the treatment effect was the only variable where distributions were available and possible uncertainty in other inputs are, thus, not represented.
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-007-0349-5/MediaObjects/198_2007_349_Fig3_HTML.gif
Fig. 3

Stochastic analysis for women with prior vertebral fracture (VFA), proportion of simulations cost effective at different theoretical intervention thresholds

https://static-content.springer.com/image/art%3A10.1007%2Fs00198-007-0349-5/MediaObjects/198_2007_349_Fig4_HTML.gif
Fig. 4

Stochastic analysis for women without prior vertebral fracture (sCFA), Proportion of simulations cost effective at different theoretical intervention thresholds

Sensitivity analysis

T-scores

The cost-effectiveness ratio dropped considerably with decreasing T-score values, which can be seen in Figs. 5 and 6. The obvious reason for this is that the risk of fracture increases with decreasing BMD, as does, therefore, the potential benefit of the treatment.
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-007-0349-5/MediaObjects/198_2007_349_Fig5_HTML.gif
Fig. 5

Cost-effectiveness (€/QALY gained) of alendronate at different T-scores in women with prior vertebral fracture (VFA)

https://static-content.springer.com/image/art%3A10.1007%2Fs00198-007-0349-5/MediaObjects/198_2007_349_Fig6_HTML.gif
Fig. 6

Cost-effectiveness (€/QALY gained) of alendronate at different T-scores in women without prior vertebral fracture (sCFA)

Starting age of intervention

Analysis of the effect of starting age on intervention on the cost-effectiveness is shown in Figs. 7 and 8. At the higher ages, the potential gain of avoiding a fracture event decreases because the morbidity in the patient group relative to the population morbidity diminishes with increasing age. This was observed for all the countries studied and reflected as decreasing incremental QALYs at higher starting ages. However, it was only in the Italian setting among women with VFA characteristics (i.e., low BMD and previous vertebral fracture) that this effect led to a slight increase in the cost-effectiveness ratio from the age of 70 years. This is because the considerably lower incidence of hip fracture in Italy makes the QALYs gained decline more steeply than the costs when starting age is increased. In all other scenarios the cost per QALY gained decreased with an increasing starting age of intervention.
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-007-0349-5/MediaObjects/198_2007_349_Fig7_HTML.gif
Fig. 7

Cost-effectiveness (€/QALY gained) of alendronate at different starting ages of treatment in women with prior vertebral fracture (VFA)

https://static-content.springer.com/image/art%3A10.1007%2Fs00198-007-0349-5/MediaObjects/198_2007_349_Fig8_HTML.gif
Fig. 8

Cost-effectiveness (€/QALY gained) of alendronate at different starting ages of treatment in women without prior vertebral fracture (sCFA)

Discounting

Without discounting, the incremental cost-effectiveness ratios were lower compared to the base case scenarios (Tables 3 and 4). The reason is that the effect of treatment has long-term consequences (in terms of quality of life and expected life time) and that most costs (the intervention cost) occur during the treatment period. When discounting the costs but not the effects, the incremental cost-effectiveness ratios were lower for the same reason.
Table 3

Sensitivity analysis: VFA, €/QALY (Base case discount rates in parentheses)

 

Belgium (5%)

Denmark (3%)

France (3%)

Germany (5%)

Italy (3%)

Norway (3%)

Spain (3%)

Sweden (3%)

UK (3.5%)

Base case

6,461

cost saving

4,670

7,658

15,489

cost saving

13,193

cost saving

1,356

Discount rate: costs and effects 3%

4,573

cost saving

4,670

5,081

15,489

cost saving

13,193

cost saving

983

No discount rate (effect and cost)

2, 185

cost saving

480

1,779

10,214

cost saving

9,340

cost saving

cost saving

Discount rate: costs 3% and effects 0%

3,378

cost saving

3,396

3,711

11,182

cost saving

9,772

cost saving

725

Discount rate: costs and effects 5%

6,461

989

7,948

7,658

19,538

cost saving

16,097

cost saving

2,556

No effect on fracture risk after treatment period

20,432

12,819

25,794

23,173

35,024

2,989

29,272

8,821

12,197

5-year treatment followed by a 5 year decline in treatment effect

9,068

1,681

8,640

10,485

18,909

cost saving

16,164

cost saving

3,489

10-year long treatment period

8,249

1 ,657

1,871

13,056

21,886

cost saving

14,682

cost saving

2,224

Efficacy from meta-analysis

12,667

5,763

11,856

15,278

24,017

cost saving

19,561

cost saving

6,506

Utility loss vertebral fracture, 0.9/1.00

8,987

cost saving

7,234

10,938

22,101

cost saving

18,181

cost saving

1,840

Table 4

Sensitivity analysis: sCFA, €/QALY (Base case discount rates in parentheses)

 

Belgium (5%)

Denmark (3%)

France (3%)

Germany (5%)

Italy (3%)

Norway (3%)

Spain (3%)

Sweden (3%)

UK (3.5%)

Base case

23,684

6,201

27,419

27,821

39,712

cost saving

32,943

cost saving

11,849

Discount rate: costs and effects 3%

18,595

6,201

27,419

21,230

39,712

cost saving

32,943

cost saving

10,853

No discount rate (effect and cost)

12,146

1,276

15,997

12,865

27,077

cost saving

23,456

cost saving

5,624

Discount rate: costs 3% and effects 0%

13,138

4,439

19,051

14,843

27,472

cost saving

23,263

cost saving

7,628

Discount rate: costs and effects 5%

23,684

10,151

36,413

27,821

49,603

cost saving

40,290

3,032

15,051

No vertebral fracture risk reduction

52,261

11,947

76,471

62,237

82,686

cost saving

68,332

cost saving

25,605

No effect on fracture risk after treatment period

51,440

28,862

70,705

59,442

80,496

13,883

66,753

30,114

32,710

5-year treatment followed by a 5 year decline in treatment effect

28,272

10,073

34,408

33,079

46,326

cost saving

38,271

3,575

15,346

10-year long treatment period

27,220

10,552

24,480

37,632

50,777

cost saving

33,567

cost saving

15,138

Efficacy from meta-analysis

35,784

17,640

41,250

42,676

56,227

3,688

45,330

10,517

22,057

Utility loss vertebral fracture, 0.9/1.00

29,974

7,504

38,486

35,722

50,857

cost saving

41,522

cost saving

14,710

No effect on vertebral fractures in sCFA

When assuming that alendronate did not reduce the risk of clinical vertebral fractures among women without prior vertebral fractures (i.e., the sCFA patient group) the ICERs increased quite substantially (Table 4). For example, in the Italian setting, the cost per QALY gained rose from €39,712 in the base case simulation to €82,686 when no effect on vertebral fracture risk was assumed.

Offset time and treatment duration

A 2-year period with full effect followed by a 3-year linear decline of the effect was assumed in the base case. Assuming a 5-year linear decline of effect after the intervention period led, as expected, to an increased cost-effectiveness ratio (Tables 3 and 4). When no residual effect of treatment was assumed after stopping treatment, cost-effectiveness ratios were higher still.

The duration of treatment in the FIT was 3 years and the same duration was used, therefore, in the base case. Prolonging the intervention period to 10 years led to modest increases in the cost-effectiveness ratios compared to the base case. Thus, the longer time with effect on fracture risk (15 years instead of 10 years) approximately offset the increased cost of intervention (Tables 3 and 4).

Efficacy estimates from meta-analysis

Intervention appeared less cost-effective when efficacy estimates from a recent meta-analysis by Stevenson and Davis [12] was used instead of estimates from the FIT. Alternative ICERs in the VFA ranged from “cost saving” in Norway and Sweden to €23,682 in Italy. In sCFA ICERs ranged from “cost saving in Norway to €54,921 in Italy.

Quality of life related to vertebral fracture

Compared to the base-case simulations, the cost-effectiveness ratios increased as expected when the utility loss associated with vertebral fractures was decreased to the level suggested by Brazier et al. [57] (Tables 3 and 4).

Discussion

In this paper we assessed the cost-effectiveness of alendronate (Fosamax®) in nine European countries. A strength of the present analysis is that we have used a common methodology to determine cost-effectiveness in the different countries. The model employs, to our knowledge, the best cost and epidemiological data available to compare alendronate to no treatment. Treatment effects and study populations were based on the largest clinical trial investigating alendronate; the Fracture Intervention Trial (FIT). Our results indicate that treatment with alendronate is cost-effective in all countries studied for the treatment of women with low BMD, at least one previous vertebral fracture and similar patient characteristics as the VFA population. Treatment with alendronate in women without prevalent vertebral fractures and with low BMD can be considered cost effective in Scandinavia and the UK whilst cost-effectiveness in other countries largely depends on what monetary intervention threshold is used.

The question of how to determine whether treatments are cost effective in different countries with varying economic conditions and notions of the value of health arises. An increasing number of health-care systems, both public and private, are adopting results from cost-effectiveness (CE) analysis as one of the measures to inform decisions on allocation of resources, although there is no generally accepted international threshold value stating at what cost per QALY a treatment is worthwhile. Indeed, even national thresholds have in many cases not been formulated. An exception is in the UK where guidance on health technologies issued by NICE suggests a threshold value of £30,000 (appr. €44,000) per QALY gained [58], and a further is in Sweden, where the value of a statistical life estimated by the Swedish National Road Administration is used [59]. In Sweden, the value of a QALY is estimated at SEK 600,000 (appr. €66,000). This threshold estimate takes into account cost in added life years. When cost in added life years are excluded, the threshold value has to be adjusted to reflect this. If the target population consumes more than they produce (i.e., most often in older populations), then the threshold value has to be down adjusted and if the target population produce more than they consume then an upward adjustment is necessary. Kanis and Jönsson [60] suggested a threshold value per QALY gained for the evaluation of interventions in osteoporosis of €50,000, when including costs in added life years and €30,000 when excluding such costs in the evaluation. That is a downward adjustment of 40%. A downward adjustment of the latest threshold estimate in Sweden (€66,000) gives a threshold value of €40,000 when excluding the cost in added life years.

A more general estimate comes from the WHO, which suggested that interventions with a cost-effectiveness ratio lower than 3*GDP per capita for each averted disability adjusted life year (DALY) should be considered cost-effective [61]. The thresholds are based on the expected direct and indirect benefits to the national economies. However, it is not specified in the report which costs should be considered when using this threshold value. When all costs are included, the threshold value ought to be down adjusted when not including the cost in added life years for elderly populations in the analysis. Using the same ratio (0.6) for adjustment as used by Kanis and Jönsson [60], as mentioned above, the threshold values excluding cost in added life years would be ranging from €32,000 for Spain up to €51,000 for Norway at the other end of the spectrum. The WHO thresholds are arbitrary but have the advantage of being related to the amount of wealth that the individual countries possess. It should be noted that the GDP per capita threshold concept uses DALY and not QALY as denominator. These two outcome measures are based on different methodologies and are therefore not directly comparable, although studies have indicated that they do not differ substantially [36].

In the vertebral fracture arm (VFA); the patient group with at least one previous radiologically verified vertebral fracture and low BMD (femoral neck T-score of-1.6 or less), treatment in all countries can be considered cost effective. This can be said when using any of the above mentioned thresholds. The incremental cost-effectiveness ratios (ICER) ranged from “cost saving” in the Scandinavian countries to €13,193 and €15,489 in Italy and Spain at the other end of the spectrum (see Table 2). The reasons for higher ICERs in the two Mediterranean countries are not the same. In the case of Italy, it is the low fracture incidence that accounts for much of the difference since a smaller proportion of the treated population can reap the benefits of treatment by avoiding a fracture. In the case of Spain, the low fracture costs reported by Hart et al. [62] is the main reason. Hart et al. most likely underestimate the cost of fractures, not including costs of living or social services, which (in Sweden) accounts for some 40% of costs [47]. If adjusting for this by dividing the Spanish fracture costs by 0.6, the cost per QALY gained was €10,669 compared with €13,193 in the base case.

The estimated fracture risk reduction of alendronate treatment was based on intention to treat (ITT) analyses. That is, the estimated treatment effect also included patients not complying fully with the treatment. The effect of imperfect compliance on fracture risk is difficult derive based on the trial data, and 100% compliance rate was assumed in the cost component of the analysis, i.e., all patients were ascribed intervention costs for the whole treatment duration. Assuming full compliance in the model likely leads to overestimated intervention costs in relation to the assumed effectiveness, which to some extent includes non-compliers. In real clinical practice, the compliance is rarely full, which leads to lower effectiveness and fewer fractures avoided. How this would impact the cost-effectiveness is not entirely clear since treatment drop-outs also result in lower intervention costs.

A limitation of this study is that the data used were gathered from sources that used different methodologies and were of variable quality. In some instances, different types of costs were included in the estimates and assumptions regarding quality of life and mortality after fractures had to be derived from other populations. E.g., admission rates to nursing homes are assumed to be the same in all countries. This is unlikely to be true and is a limitation of this study. Another issue of uncertainty is that no account was taken of informal care. The amount of informal care given by relatives and/or others is likely to vary between countries and cultures. It could thus be an ignored cost item in a study with a societal perspective or a reason for significant cost variations in a study taking a health care perspective, such as this one. The common changes in drug prices are problematic for health economic publications on drug interventions since the drug price often represents a considerable proportion of the incremental cost. The price of alendronate has decreased in several countries with the introduction of generic formulations and increased pressure from reimbursement agencies. The price is likely to decrease even further and these changes will make prescription of alendronate even more cost-effective in a comparison to placebo. The estimates in this paper should be considered conservative with this in mind.

In the subgroup from the sCFA (Clinical Fracture Arm) (no previous fracture and T-score ≤−2.5 SD) cost-effectiveness was not as consistent as in patients with prior vertebral fracture (see Table 2). In the Scandinavian countries treatment could still be considered to be cost effective, whereas in the other countries, it is less evident than in the VFA. This can be attributed to the extra fracture risk associated with a previous fracture, which was an inclusion criterion in the Vertebral Fracture Arm. The lack of assumed effect on wrist fractures had little effect (results not shown). If it was assumed that there was no effect on vertebral fractures in the sCFA, the cost-effectiveness ratio was considerably increased even though treatment still could be considered as cost-effective in many of the countries studied. Even though the fracture intervention trial is the largest trial that has studied the efficacy of alendronate, it is not beyond doubt that it represents the “true” anti-fracture efficacy of alendronate. Trying efficacy estimates from a recent meta-analysis had a considerable impact on the results, implying that the used anti-fracture effect in the model is one of the most important factors governing the results.

The results in both treatment arms are fairly robust when changing oft debated variables such as discount rates or in assuming no period of full effect after stopping treatment. Only marginal change in the results was seen when the proportion of post-fracture mortality that was attributable to the fracture was changed. The most conservative scenario, not assuming any effect at all after stopping therapy was very detrimental to the ICERs, but such a scenario would be rather unlikely considering previous findings concerning the quality and mineral density of bone after treatment [1316]. A general pattern of findings in this study was that cost-effectiveness improved with higher latitudes (Norway, Sweden and Denmark) and consequently deteriorated southwards (Spain and Italy). In the middle were, thus, UK, Germany, France and Belgium. Reasons for the geographical differences are speculative, but the varying pattern of fracture incidences is clearly an important driver. The size of both the incremental cost and effectiveness in the treatment arm are dependent on the incidence of hip and vertebral fractures.

In undertaking a project covering as many as nine countries, data have had to be collected from various sources. Both cost and epidemiological data are of varying quality and are sometimes even lacking. Even though assumptions and data extrapolation between countries creates uncertainty in the results, it is still valuable to examine several countries with a common methodological framework and under an identical set of assumptions.

Conclusion

This study indicates that treatment with alendronate (Fosamax®) is cost-effective in the treatment of women with low BMD, at least one previous vertebral fracture and similar in patient characteristics to the VFA population in all the countries studied. Alendronate treatment in women without prevalent vertebral fractures and with low BMD is variously cost-effective depending on country setting, discount rates, and chosen monetary thresholds.

Conflict of interest statement

JAK and European health economics have received consulting fees from Merck. Dr. Shuvayu S. Sen is an employee of Merck & Co., Inc.

Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2007