Osteoporosis International

, Volume 21, Issue 9, pp 1599–1608

Cost-effectiveness of balloon kyphoplasty in patients with symptomatic vertebral compression fractures in a UK setting

Authors

    • i3 Innovus
    • Medical Management CenterKarolinska Institutet
  • C. Leonard
    • Medtronic Ltd
  • D. Marsh
    • Royal National Orthopaedic Hospital
  • C. Cooper
    • MRC Epidemiology Resource CentreUniversity of Southampton
    • Institute of Musculoskeletal SciencesUniversity of Oxford
Original Article

DOI: 10.1007/s00198-009-1096-6

Cite this article as:
Ström, O., Leonard, C., Marsh, D. et al. Osteoporos Int (2010) 21: 1599. doi:10.1007/s00198-009-1096-6

Abstract

Summary

Balloon kyphoplasty (BKP) is a procedure used to treat vertebral compression fractures (VCFs). We developed a cost-effectiveness model to evaluate BKP in United Kingsdom patients with hospitalised VCFs and estimated the cost-effectiveness of BKP compared to non-surgical management. The results indicate that BKP provides a cost-effective alternative for treating these patients.

Introduction

VCFs of osteoporotic patients are associated with chronic pain, a reduction in health-related quality of life (QoL) and high healthcare costs. BKP is a minimally invasive procedure that has resulted in pain relief, vertebral body height-restoration, decreased kyphosis and improved physical functioning in patients with symptomatic VCFs. BKP was shown to improve health-related QoL in a 12-month interim analysis of a randomised phase-III trial.

Methods

The objectives of this study were to develop a Markov cost-effectiveness model to evaluate BKP in patients with painful hospitalised VCFs and to estimate the cost-effectiveness of BKP compared with non-surgical management in a UK setting. It was assumed that QoL-benefits found at 12 months linearly approached zero during another 2 years, and that patients receiving BKP warranted six fewer hospital bed days compared with patients given non-surgical management.

Results

The procedure was associated with quality-adjusted life-years (QALY)-gains of 0.17 and cost/QALY-gains at £8,800. The results were sensitive to assumptions about avoided length of hospital-stay and persistence of kyphoplasty-related QoL-benefits.

Conclusion

In conclusion, the results indicate that BKP provides a cost-effective alternative for treating patients with hospitalised VCFs in a UK-setting.

Keywords

HTAMarkovOsteoporosisQALYUK

Introduction

Osteoporosis is a major cause of morbidity and mortality in the elderly [1]. Because of demographic changes and increasing life expectancy, it is a growing public health concern and the European Commission has forecasted a 57% increase in the prevalence of vertebral fractures in the European Union over a 50-year period [2]. The large number of fractures also makes osteoporosis a growing economic concern and a burden on the health-care systems in Europe. The burden of vertebral compression fractures (VCFs) can be substantial, in particular reflecting chronic pain, a marked reduction in health-related quality of life and high health-care costs [35]. Balloon kyphoplasty (BKP) is a minimally invasive procedure that results in pain relief, vertebral body height restoration, decreased kyphosis and improved physical functioning in patients with symptomatic VCFs [68]. The technique uses an inflatable bone tamp to restore vertebral anatomy. The bone tamp creates a void that can be filled under fine manual control and low pressure with high viscosity bone cement. Vertebral fractures change the biomechanics of the spine [9], and increase the risk of additional vertebral fractures [10]. The effect of BKP on this risk is yet to be proven. Non-randomised trials reported fewer subsequent fractures compared with non-surgical management (NSM) [8, 11, 12], whereas one uncontrolled study has suggested that BKP might increase the risk [13]. The overall safety profile of BKP is good. A meta-analysis by Taylor et al. [12] reported 189 (9%) cement leaks in 2,239 vertebrae that underwent BKP; of which only one leak (.001%) was reported to be symptomatic. The rate of serious adverse events reported with BKP are low, with a perioperative mortality of 0.01% and an overall mortality rate of 3.2%; reflecting the age and co-morbidity status of this patient group, including those patients with cancer.

In order to justify resource allocation and patient selection for new technologies for both the prevention and treatment of osteoporotic fractures, it is becoming increasingly important to determine the cost-effectiveness of all therapies. Cost-effectiveness analysis in health care is used to assess the economic consequences of an intervention and how those consequences relate to the benefit achieved from the intervention. Numerous cost-effectiveness models evaluating strictly preventive therapies such as alendronate, risedronate and strontium ranelate have previously been published [1417]. These therapies can all be evaluated in a rather standardised modelling framework since the indication and population essentially is the same irrespective of the treatment alternative evaluated. The objectives of this study were to develop a cost-effectiveness model for the evaluation of balloon kyphoplasty in osteoporotic patients and to estimate the cost-effectiveness of balloon kyphoplasty compared with NSM in a UK setting.

Methods

The study was performed taking a health care perspective including health care costs. Health effects were measured using quality-adjusted life-years gained (QALYs). We report our main findings as incremental cost-effective ratios (ICERs). Where necessary, costs were inflated to the 2008 price level using the UK consumer price index [18]. Costs and effects were discounted at 3.5% based on current recommendations and guidelines for health economic evaluations in the UK [19].

The FREE trial

Quality of life (QoL) effects of BKP were derived from the FREE trial [20], which is a multicentre randomised trial comparing BKP with NSM in patients hospitalised with vertebral fractures during 24 months from baseline. Three hundred patients with a mean age of 73 years were screened and randomly assigned to BKP (n = 149) or NSM (n = 151). Patients were required to have at least one painful vertebral fracture that had caused oedema assessed by magnetic resonance imaging and at least 15% loss of height. Painful fracture was defined as a response of four or more on a visual analogue scale (VAS) ranging from 0–10. EQ-5D scores derived from the UK utility algorithm [21] were one of the secondary endpoints and were used to model QALY gains (Fig. 1). The FREE trial was not designed nor, indeed, sufficiently powered to measure a difference in the incidence of additional vertebral fractures, and no such difference was found at 12 months in the trial. At this point in time, only 12-month data from the trial have been published and were available for the cost-effectiveness analysis.
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-009-1096-6/MediaObjects/198_2009_1096_Fig1_HTML.gif
Fig. 1

Quality of life results from the FREE trial

Model

The simulation model used in this study was based on Markov cohort methodology [22]. The cycle length was set at 6 months and all patients were followed through the model until they reached 100 years of age, or died. The model was constructed to be able to take additional vertebral fractures into account. This matters because new fractures will affect the mortality of the patient, even though no fracture risk reduction with BKP was assumed. A schematic picture of the model is shown in Fig. 2.
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-009-1096-6/MediaObjects/198_2009_1096_Fig2_HTML.gif
Fig. 2

Schematic picture of transitions and states used in the model

All patients begin with an initial VCF and are either treated with balloon kyphoplasty or non-surgical management. In each 6-month cycle, patients have a probability of sustaining an additional fracture, remaining in their current state (BKP or NSM) or dying. If a patient dies, he/she moves to the absorbing dead health state. If a patient sustains a fracture in addition to the initial fracture, he/she moves to the additional VCF health state. In the additional VCF state, it is only possible to stay, incur additional VCFs or to die. After 6 months in the BKP, NSM, or additional VCF state without incurring a fracture event or dying, the patient moves to the first corresponding sub-state. The sub-states handle the time dimension as patients recover, and fracture-related mortality decreases from the initial level in the first 6 months in each state. It was conservatively assumed that patients treated with BKP loose all QoL-related benefits thereof if they sustain an additional VCF. Due to the good safety profile of BKP, adverse events from the procedure were not included in the model.

Model population

The base-case population were 70-year-old UK women and men in the same proportions as the FREE trial (77% women) with a T-score of −2.5 and one initial vertebral fracture. T-score and age govern the risk of additional fractures. Given that additional fractures are associated with increased mortality, the additional fracture risk impacts the period in which BKP will be associated with QALY gains. In sensitivity analyses, the selected age range was from 60 to 80 years. It was assumed that the BKP procedure was performed on one vertebral level. All patients (BKP and NSM) were assumed to receive analgesics, bed rest, back braces, physiotherapy, rehabilitation programmes and walking aids. Therefore, these costs were not specifically costed nor modelled because they were assumed to be the same irrespective of treatment alternative. However, in clinical practise, the NSM patient cohort may utilise more healthcare resources with respect to VCF management.

Quality of life

Given the current evidence, the principal benefit of BKP resides in QoL gains and the model was directly run on the EQ-5D data estimated in the FREE trial (Fig. 1). Accumulated utility during the first 6 months was estimated at 0.277 and 0.222 for patients treated with BKP and NSM, respectively. Corresponding utility during months 7–12 were 0.314 and 0.259. These figures were estimated in patients with an average age of 73 years, and it should be accounted for that QoL decreases with age. The utility values estimated from the FREE trial were, therefore, assumed to decline with age at the same rate as in the general population [23]. This time-dependent decline was assumed to be the same for all patients in the model irrespective of treatment, and will yield slightly lower absolute QALY gains in older populations.

The FREE trial reported a significant trend in QoL differences, illustrating they were not constant over the year [20]. Because only 12 months of utility data were available, assumptions are necessary for how long differences will persist. Studies indicate that benefits persist well beyond 1 year [24, 25], but long-term randomised data comparing BKP with NSM are not available. In the base case, it was assumed that the QoL differences found at 12 months, thereafter, linearly approached zero during another 2 years (Fig. 3). It was thus assumed in the base case that BKP was associated with QoL effects during 3 years (1 + 2) if no additional fractures were sustained. Different scenarios for persistence of QoL-effects were explored in sensitivity analysis.
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-009-1096-6/MediaObjects/198_2009_1096_Fig3_HTML.gif
Fig. 3

Modelled persistence of utility benefits

Fracture incidence

Patients in the model are at risk of developing an additional clinical vertebral fracture. The incidence of clinical vertebral fractures is not available for the UK and was therefore imputed from UK incidence for hip fractures [26] and Swedish data for clinical vertebral fractures [14, 27]. The imputation was done by assuming that the Swedish ratio between incidence of hip and vertebral fractures was also valid for the UK. Age-specific relative risks of developing a clinical vertebral fracture were calculated with previously described methods [15]. All patients were assumed to be on bisphosphonate therapy reducing fracture risk by 40% for 5 years after the initial VCF. BKP was not assumed to reduce the risk of additional fractures.

Costs

Given that fracture rates after NSM and BKP were assumed to be the same, costs directly associated with the fracture were of little importance. However, to give a more complete overview of the total costs these patients incur, medical costs due to VCFs were included in the model. Vertebral fracture costs, excluding procedure related costs, were obtained by updating the costs after fracture reported by Puffer et al. [28] for patients sustaining their first and second vertebral fracture. A 12-month cost for the initial hospitalised VCF was estimated at £5,974 and any subsequent VCFs at £2,351, assuming 35% of patients are hospitalised [29] with a cost of £450/bed day [30]. With respect to balloon kyphoplasty, the total costs for materials and surgery were based on reference cost data (2007/08) from three National Health Service UK hospitals providing the procedure, estimated at £4,090 (Table 1). A potential source of cost-savings from BKP is avoided hospital bed days/reduced hospital length of stay. Because BKP is a fairly new procedure, it has infrequently been assigned its specific procedural code, making it difficult to directly estimate the length of hospital stay during the first 12 months after the surgery. An extraction of hospital episode statistics in 2006 and 2007 yielded only five osteoporotic BKP patients with an average of 4.0 bed days whilst 19,551 osteoporotic NSM patients had an average of 15 days (11 days difference), which is consistent with length of stay after vertebral fracture reported by Stevenson et al. [29]. Taylor et al. [31] reported differentials of 4.5 and 7.0 days in Italy and Spain, respectively and Greiner et al. (data on file) who reported seven fewer hospital days after BKP in a German, not yet published, observational study. Based on this information, it was assumed that BKP was associated with six fewer bed days during the first 12 months after fracture, including any bed days spent directly after the surgery. Each bed day was costed at £450 [30], and a range of 0–11 days of reduced length of stay was tested in sensitivity analysis.
Table 1

BKP procedure costs (£)

Resource

Cost per hour

No. of hours

Cost (£)

Preliminary phase

 Surgeon

100

0.25

25

 Radiologist

80

0.5

40

 Nurse

15

1

15

 Spine X-ray

72

 MRI

260

 ECG

  

64

 Blood test(s):

20

 Pain therapy

15

 Sum

  

511

Operating phase

 Anaesthetist

100

1

100

 Nurse, anasthesiology

11

1

11

 Drugs

36

 Surgeon

100

1

100

 Radiographer

30

1

30

 Nurse, surgery

16

2

32

 Consumables

  

90

 Devices

  

2,690

 Operating room

  

260

 Sum

  

3,349

Post-operative phase

 Nurse

4.5

24

108

 Surgeon

100

0.5

50

 Spine X-ray

72

 Sum

  

230

 Total sum

  

4,090

MRI magnetic resonance imaging, ECG electrocardiography

Mortality

Vertebral fractures are associated with excess mortality, and since it is necessary to be alive to reap any long-term benefits from the procedure, such data are relevant to include. Detailed mortality data after hospitalised vertebral fractures were not available for the UK, so Swedish data were used instead. The mortality after hospitalised VCF (Table 2) was estimated by linking the Swedish inpatient and cause of death registers [32]. All admissions to Swedish hospitals for vertebral fractures (ICD10, S30 and S32) in patients aged 50 years or more with low energy vertebral fractures between the years 1997–2001 were studied. Five thousand one hundred sixty-four patients (77% women) matched the inclusion criteria and were analysed. In accordance with previous findings post fracture, mortality showed a U-shaped pattern over time [33]. Therefore, mortality for the first year after fracture was estimated with a Poisson model, and the period 2-5 years after the fracture was estimated with a parametric Weibull survival regression. The Weibull distribution is suitable for modelling data with hazard rates that increase or decrease over time and allows for the estimation of probability of an event in different time intervals after the starting point, i.e., the probability of dying in year t after a fracture. Dummy variables were created for sex, and to track if patients had previously sustained one or more vertebral fractures during the last 10 years (n = 708). Equation 1 and Eq. 2 describes the Poisson and Weibull functions respectively (all p values <0.05). If a patient sustained multiple fractures in the Markov model, she/he was assumed to have the same standardised mortality ratio as patients in the database with multiple vertebral fractures. To implement the mortalities in the UK model, the relative risks of death were applied to the UK normal population mortalities.
Table 2

Mortality per 1,000/patients

Years after fracture

Age

60

70

80

90

No previous clinical vertebral fracture

 1

44 (7.1)

75 (4.6)

128 (2.4)

219 (1.3)

 2

34 (5.0)

65 (3.4)

124 (2.1)

228 (1.2)

 3

38 (5.0)

73 (3.4)

138 (2.0)

252 (1.2)

 4

40 (4.6)

76 (3.3)

143 (1.9)

261 (1.1)

 5

41 (4.3)

78 (3.0)

147 (1.7)

268 (1.1)

Previous clinical vertebral fracture

 1

51 (8.2)

87 (5.3)

148 (2.8)

252 (1.5)

 2

42 (6.2)

81 (4.3)

153 (2.5)

277 (1.4)

 3

47 (6.2)

90 (4.3)

169 (2.4)

304 (1.4)

 4

49 (5.7)

94 (4.1)

176 (2.3)

315 (1.4)

 5

51 (5.3)

97 (3.7)

181 (2.1)

223 (1.3)

Relative risk compared to the normal population (77% women) in parenthesis

$$ {y_{{t_1}}} = {e^{ - {\text{6}}{\text{.465}} + {\text{age}} \times 0.533 + {\text{proportion}}\_{\text{men}} \times 0.650 + {\text{prior\_fx}} \times 0.143}} $$
(1)
$$ {y_{{t_{2...5}}}} = 1 - \frac{{{e^{ - \left( {{e^{\left( { - 7.538 + {\text{age}} \times 0.671 + {\text{proportion\_men}} \times 0.631 + {\text{prior\_fx}} \times 0.224} \right)}}{{\left( {t - 1} \right)}^{1.084}}} \right)}}}}{{{e^{ - \left( {{e^{\left( { - 7.538 + a{\text{ge}} \times 0.671 + {\text{proportion\_men}} \times 0.631 + {\text{prior\_fx}} \times 0.224} \right)}}{{\left( {t - 2} \right)}^{1.084}}} \right)}}}} $$
(2)

Results

Base-case analysis

The base-case population were 70-year-old women and men with a T-score of −2.5 SD and at least one painful VCF warranting hospitalisation. BKP was assumed to be associated with a reduction in hospital length of stay of 6 days. The differences in quality of life found in the FREE trial at 12 months were assumed to be followed by a linear recovery of the NSM patients (up to the level of patients treated with BKP) during the following 2 years. Cost/QALY gained amounted to £8,840 (Table 3).
Table 3

Base-case results for 70-year-old patients

 

BKP

NSM

Difference

Procedure cost (£)

4,090

0

4,090

Inpatient and outpatient care costs (£)

6,330

8,926

−2,596

Total cost (£)

10,420

8,926

1,494

Life-years

9.390

9.390

0.00

QALYs

3.842

3.673

0.169

Cost/QALY gained

  

8,840

Sensitivity analyses

Figure 4 shows sensitivity analysis for the persistence of QoL effects. As expected, the results deteriorate when the benefits found in the FREE trial are assumed to dissipate earlier, and are improved when assumed to persist for longer. In the "worst case scenario", where only the 12 month results in the FREE trial were accounted for, a cost/QALY gained of £14,500 was estimated. Sensitivity analysis for the reduction in length of stay showed that results were sensitive to changes in this variable (Fig. 5). Reduction of length of stay was estimated for up to 11 fewer days and BKP dominated NSM in scenarios when the reduction was larger than 9 days (results not shown). The cost/QALY gained was consistently higher in older patients and this was caused by the effect of age on QoL, fracture risk, and mortality. In general, older patients have a lower QoL and utilities were therefore adjusted to reflect this when different age groups were simulated. Furthermore, fracture risk and mortality increase with age, thereby attenuating the benefit of the BKP procedure. Thus, the expected survival of the patient will affect the expected amount of QALYs gained by each patient.
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-009-1096-6/MediaObjects/198_2009_1096_Fig4_HTML.gif
Fig. 4

Sensitivity analysis on the persistence of QoL benefits from BKP

https://static-content.springer.com/image/art%3A10.1007%2Fs00198-009-1096-6/MediaObjects/198_2009_1096_Fig5_HTML.gif
Fig. 5

Sensitivity analysis on length of stay reduction with BKP

Probabilistic sensitivity analysis

To represent the uncertainty around the QoL benefits found in the FREE trial, distributions of the utility gains 0–6 months and 7–12 months after baseline were created with bootstrapping techniques [34] (Fig. 6). Reduction of length of stay associated with BKP was assumed to be normally distributed and the standard error was conservatively assumed to be 50% of the mean (i.e., 3 days). Two thousand samples were drawn from the distributions for the base case, a scenario accounting only for the FREE trial QoL, and one scenario where the FREE trial QoL was followed by linearly declining QoL benefit for 3 years. Figure 7 shows the proportion of samples that fell below different values of willingness to pay for an incremental QALY. The BKP curves never reached zero on the y-axis, and this indicates that BKP dominated NSM in ∼13% of drawn samples.
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-009-1096-6/MediaObjects/198_2009_1096_Fig6_HTML.gif
Fig. 6

Bootstrapped utility distributions from the FREE trial data

https://static-content.springer.com/image/art%3A10.1007%2Fs00198-009-1096-6/MediaObjects/198_2009_1096_Fig7_HTML.gif
Fig. 7

Cost-effectiveness acceptability curves

Discussion

The principal finding of the present study is that treatment of painful vertebral compression fractures with Balloon kyphoplasty is indicated to be a cost-effective intervention. The results are primarily based on the FREE trial, which is the first randomised clinical trial evaluating BKP. The trial evaluated a number of endpoints (e.g., SF-36 PCS, VAS pain, Roland–Morris score and days of restricted activity) but only EQ-5D was used in the present model. The EQ-5D health state results are not useable by themselves but have been transformed into utility values with a commonly used algorithm published by Dolan et al. [21]. BKP could very well be associated with other clinical and economical effects; like effect on the rate of additional fractures and reduced utilisation of community services, but those were not included in the present study. Modelling is often necessary to compile information from various sources to formulate an overview of the clinical and economical consequences of using a particular technology in a pragmatic setting. In turn, drawing conclusions directly from the FREE trial would mean that BKP only affects the patient during the first 12-month period, and the impact of downstream mortality and avoided health care costs would be difficult to include.

In our study, BKP treatment was associated with a QALY gain of 0.17 in the base-case scenario and this should be related to the extra cost associated with the procedure, which was estimated at £1,494, yielding a cost/QALY of ∼£8,800. The willingness to pay for a QALY in the UK lies within the range of £20,000–30,000/QALY [19]. ICERs above 20,000 can be acceptable if the innovation adds demonstrable and distinctive benefits of a substantial nature which may not have been adequately captured in the QALY measure. Given the assumptions and data used in our model, the budgetary impact of treating 1,000 patients with painful hospitalised VCFs would be £1.5 M calculated over a period of 1 year. Our base-case result is, however, contingent on that BKP reduces length of hospital stay. There are a number studies suggesting this [31, 3539], but solid observational data are still scarce. Because BKP is a relatively new and uncommon procedure, the amount of available data describing long-term QoL results, effect on the risk of subsequent fractures, effect on mortality and reduced use of community services and pharmaceuticals are limited. Whilst the FREE trial measured the use of analgesics in a clinical trial setting, it is not ideal when extrapolating the results to clinical practise because the trial programme may drive resource use such as GP visits and prescriptions. The FREE trial did not show any differences in mortality, and was not specifically designed to do so. However, a US registry study by Edidin et al. [40] reported a mortality reduction during 12 months in 40,354 BKP patients compared with 173,200 non-operated patients. Should such a mortality reduction be included in a health economic evaluation, it would have a significant positive impact on the results; but as a conservative measure, we chose not to include this in the present cost-effectiveness analysis. BKP may very well reduce post-fracture mortality, but until more evidence becomes available, a conservative approach to modelling this in health economic evaluations is appropriate.

The present analysis was performed on a mixed sample of women and men chosen to reflect the FREE trial, whereby women and men were not analysed separately. Of the data used in the present analyses, baseline and post-fracture mortality are the only items where differences can be identified. A separate analysis for men would have yielded slightly worse results in our model. Separate cost-effectiveness analyses for women and men might in the future be warranted as the evidence base grows. BKP was only compared with conservative management rather than vertebroplasty (VP), which is an alternative procedure. With VP, the bone cement is directly injected into the fractured vertebra without first inflating a bone tamp, which creates a cavity, to improve height restoration and reduce the kyphotic angle (spinal deformity correction). The three reasons for not comparing BKP with VP in this study were: (1) the only published randomised trial measuring the effects of BKP has NSM as the comparator (standard of care) [20]. (2) Both BKP and VP are relatively uncommon procedures and a first step is to estimate whether it is cost effective versus current standard of care, namely NSM. (3) A comparison with VP would not be feasible given the data scarcity comparing these two interventions with respect to QoL and fracture associated costs following surgery.

To summarise, our results indicate that BKP provides a cost-effective treatment for the treatment of painful hospitalised vertebral fractures in postmenopausal women and men in a UK setting. However, there is a need for more research to fill the data gaps regarding long-term consequences on QoL, down-stream fracture risk, mortality and post-fracture resource use. The cost-effectiveness of balloon kyphoplasty should be revisited as more evidence becomes available.

Acknowledgements

We would like to extend our gratitude to Professors Torgerson and Greiner who generously offered access to data relevant for this study.

Conflicts of interest

Financial support was obtained via an unrestricted grant from Medtronic. Catherine Leonard is an employee of Medtronic Ltd, UK. David Marsh and Cyrus Cooper have received honoraria as expert advisors to Medtronic.

Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2009