Introduction

Economic evaluations are increasingly being used by decision-makers to evaluate the cost-effectiveness of treatments. In Australia and in the Canadian province of Ontario, economic evaluations are mandatory components of the pharmaceutical licensing process [1]. In the UK, the National Institute for Clinical Excellence (NICE) requires an economic evaluation as part of its submission procedures for the appraisal of new technologies (www.nice.org.uk). In the USA, the Academy for Managed Care Pharmacy (AMCP) has developed submission guidelines that require evidence of cost-effectiveness of new treatments for the inclusion on pharmacy formularies (www.amcp.org).

Economic evaluations can be conducted alongside clinical trials where economic data are collected at the same time as clinical data. Often, however, economic evaluations are conducted using models that explicitly combine available information in a formal framework. These models enable the combination of evidence from a variety of sources in order to explore scenarios that for different reasons have not been empirically tested. For example, results from studies with relatively short follow-up periods can be extrapolated to longer time periods that are more relevant to policy-makers. Economic models can also explore the long-term effectiveness and cost-effectiveness of treatments in populations at different risk.

A number of interventions are available to prevent and treat osteoporosis [2, 3]. However, recommendations for the treatment and prevention of osteoporosis have changed radically in the last few years. In 2002, the Women’s Health Initiative trial was stopped early because of the excess risk in cardiovascular events that were observed in women taking estrogen and progestin [4, 5]. The segment of the study investigating the effects of estrogen alone in women without a uterus was also stopped early due to the increase in observed strokes [6]. Following these results, treatment with hormone replacement therapy (HRT) for long-term prevention is no longer recommended. However, newer therapies such as bisphosphonates and raloxifene have been shown to be effective in clinical trials. Less expensive therapies with no known adverse side effects such as vitamin D with or without calcium supplements and hip protectors are also being investigated [7, 8, 9, 10].

A number of economic evaluations have been conducted in the field of osteoporosis and have investigated the cost-effectiveness of interventions in a variety of settings and study populations. A review of models by Zethreaus et al. published in this journal in 2002 described some of these models [11]. Because of the increase in cost-effectiveness studies, and the rapid changes in the use of treatments for osteoporosis, we conducted an updated structured review of economic evaluations of interventions relating to the prevention and treatment of osteoporosis.

Materials and methods

Two reviewers (RF and CI) conducted the search independently according to predefined inclusion and exclusion criteria. The inclusion criteria for the structured review were: economic evaluations of interventions to prevent or treat osteoporosis; economic evaluations could be modeling studies as well as economic evaluations of clinical study data; study populations could include elderly men as well as postmenopausal or elderly women; interventions included any treatment or device used to prevent or treat osteoporosis or osteoporotic fractures, with or without prior screening of bone mineral density (BMD); economic evaluation outcomes were cost per quality-adjusted life year (QALY) and cost per fracture avoided; the languages were English, French and Spanish (languages spoken by the reviewers).

The exclusion criteria for the structured review were: cost of illness and burden of disease studies; cost studies, such as retrospective claims analyses; editorials and reviews; any study describing a model, but not presenting results; economic evaluations of interventions to treat glucocorticoid-induced osteoporosis.

The search for papers was up to December 2004. Four broad search strategies were used to identify economic models of interventions to prevent and treat osteoporosis: a search on MEDLINE, a search of the bibliographies of relevant review articles, a search of the bibliographies of identified economic models of osteoporosis and a request for further references from expert opinion.

Both reviewers (CI and RF) conducted the search independently. The literature search on MEDLINE used combinations of the following Mesh terms: “Osteoporosis,” “Fractures,” “Economics,” “Cost-Benefit Analysis,” “Costs,” “Model” and “Models.” We also used the “Related Articles” function available in MEDLINE to identify any additional studies. Abstracts of papers from these searches were analyzed, and any potential study meeting the inclusion and exclusion criteria was obtained for further investigation. Bibliographies of review articles were analyzed to search for additional references [11, 12, 13, 14, 15, 16]. Bibliographies of identified economic models in osteoporosis were checked for any additional references that could have been missed from the other search strategies. At this stage, both reviewers compared their potential lists of studies to be included, and a single list was created. Any discrepancy was resolved through discussion between the reviewers. Finally, the third author (DT) reviewed our list to identify any potentially missing study.

Relevant information from each included study was extracted according to a predefined grid. Information included: study reference, country to which economic evaluation refers, type of fractures included, type of prevention or treatment (primary, secondary or treatment), interventions and control, study population and authors’ stated conclusions. Primary prevention is used for asymptomatic populations with no apparent osteoporosis or elevated risk of the condition to reduce their risk of onset in later life. Secondary prevention is used for asymptomatic patients who have been shown to have BMD sufficiently low to place them at elevated risk of fracture, to slow down the decline (or restore) BMD and hence reduce the risk of fracture. Treatment is used for patients known to have osteoporosis and who have already experienced one or more fractures to reduce their risk of further fractures [14].

Results

Forty-two (42) economic evaluations of interventions for the prevention or treatment of osteoporosis met the specified inclusion and exclusion criteria. A summary of the models by author, country, type of fracture, level of intervention (primary, secondary or treatment), intervention and control, study population and authors’ stated conclusions is presented in Table 1.

Table 1 Economic evaluations for the prevention and treatment of osteoporosis. BMD bone mineral density, CE cost-effectiveness, CHD coronary heart disease, CRT combined replacement therapy, ERT estrogen replacement therapy, HRT hormone replacement therapy, OHT ovarian hormone therapy, QALY quality-adjusted life year, QoL quality of life
Full size table

Twenty-nine percent (12) of the economic evaluations were set in the UK, 21% (9) were set in the USA and Sweden, respectively, 7% (3) were set in Canada, 5% (2) were set in Australia, Germany, and Spain, respectively, and 2% (1) were set in Italy, Japan, and Denmark, respectively. Three countries were associated with 71% of the economic evaluations, Sweden, the UK and the USA, reflecting the concern over the large burden of osteoporosis in these countries.

The main intervention investigated was HRT (in which we include Estrogen Replacement Therapy or ERT) with 27% (17) of studies investigating its cost-effectiveness. Bisphosphonates represented 17% (11) of the interventions, calcium and/or vitamin D 16% (10), calcitonin 9% (6), hypothetical interventions and hip protectors 8% (5) each, raloxifene 6% (4), and exercise, steroids and other, 3% (2) each. The numbers add to more than 42 as some studies investigate multiple therapies. It should be noted that seven of the HRT/ERT studies were conducted before 1994. In recent years, newer interventions such as bisphosphonates and raloxifene as well as cheaper interventions with no known side-effects such as vitamin D and calcium and hip protectors have been more readily investigated.

In 38% (16) of studies, hip fractures were the sole fracture outcome. Twenty-nine percent, 29% (12) investigated hip, wrist and vertebral fractures, 17% (7) hip, wrist, vertebral and other fractures (often proxied by shoulder fracture), 7% (3) analyzed hip and wrist fractures and 7% (3) vertebral fractures alone. Two percent 2% (1) investigated hip and vertebral fractures. Thirty-one percent (13) of the reviewed studies looked at primary intervention, 17% (7) at secondary intervention and 19% (8) investigated treatment. The remainder, 33% (14), analyzed various combinations of intervention levels. Finally 88% (37) of the studies investigated female populations. Only five studies included men in their analyses (although two of these studies questionably used general rates of fracture for both men and women).

Discussion

This structured review of economic evaluations for the prevention and treatment of osteoporosis identified 42 studies. The studies identified were published between 1980 and 2004 and span 24 years of research in the field of economic evaluation of interventions for the prevention and treatment of osteoporosis. Initially, the area was dominated by variations of a model developed in the USA, investigating the use of ERT and HRT in women with and without hysterectomy. Screening policies were also investigated to evaluate the cost-effectiveness of BMD measurements followed by HRT compared with universal HRT treatment [17, 18, 19, 20, 21, 22]. More recently, a model developed in Sweden has been employed a number of times to investigate various hypothetical interventions, at different costs, effectiveness and offset times [23, 24, 25]. Other models have been developed to investigate the cost-effectiveness of bisphosphonates and raloxifene following the results of clinical trials published in that area (FIT and MORE). Recently, a number of new models have also been published, investigating interventions such as vitamin D and calcium and hip protectors.

Despite the variety of interventions and populations considered, all but one of the reviewed studies employed cost-effectiveness models (one study provided an analysis based on economic data collected alongside clinical data [26]). The use of models is indicated in conditions such as osteoporosis because of the need to model long-term costs and effects that are not always available from trial or even observational data. In the absence of available data, decision-makers need formal frameworks on which to base their decisions, and such models can provide such as a basis when they are well conducted, transparent and explicit [27].

The quality of models in the field is variable [11, 14]. For example, some of the earlier studies made questionable use of cost-effectiveness decision rules [14]. However, the quality of the methodology and the reporting in publications is on average increasing, reflecting the availability of structured guidelines for developing and reporting cost-effectiveness models [28]. Most studies are explicit about the inputs and structure of the model. A number of models now include vertebral and wrist fractures as well as hip fractures. Some studies also include other fractures, sometimes proxied by shoulder fractures. Recently, much effort has been devoted to validating the models by providing explicit internal and external validity checks [29]. In addition, Zethreaus et al. have proposed making their model available to researchers in the field [11].

With the large number of assumptions that such models rely on, quantifying the uncertainty associated with the estimates is essential for the validity of the models. Although single and multi-way sensitivity analysis may be used to investigate the effect of different model parameters, only full probabilistic models allow the exploration of the interaction of different sources of uncertainty present within a model [28]. In these models, each input parameter is assigned an appropriate statistical distribution and a 95% confidence interval, representing a range of plausible values obtained from the literature. A Monte-Carlo simulation is then run to obtain a large number of iterations of the model. These results are used to obtain cost-effectiveness acceptability curves that show the probability that an intervention is cost-effective as a function of the decision-maker’s ceiling cost-effectiveness ratio (this ceiling will vary according to the resources available for health care and is in general unknown to the analyst). An increasing number of economic evaluations are using probabilistic sensitivity analyses and cost-effectiveness acceptability curves to investigate uncertainty in the model parameters and to present this to decision-makers (for example, [26, 30, 31, 32, 33, 34]).

Several specific trends were identified from the analysis of the studies included in the structured review: first, there is a major shift away from investigating the cost-effectiveness of HRT. While the reduction in fractures following HRT use has been established through trials [35], in earlier economic evaluations, a number of assumptions were made on the potential cardioprotective effect of HRT. Because of the high absolute risk of coronary heart disease (CHD), such assumptions had large impacts on the results. However, the results of the Women’s Health Initiative trial have radically challenged this approach [5, 6, 36]. With the current recommendations that HRT be used only for the short-term relief of menopausal symptoms, and the sharp drop in prescription and use of HRT [37], the long-term use of HRT for the prevention of fractures is no longer recommended [38]. Reflecting this major shift, no economic evaluations analyzing HRT have been published since 2002.

Second, the shift away from HRT has brought other therapies to the forefront. Bisphosphonates, such as alendronate, etidronate and risedronate, and raloxifene have been investigated following the results of large clinical trials [39, 40, 41, 42, 43]. Cheaper therapies with no side effects have also been investigated: vitamin D with or without calcium and hip protectors. These interventions have also been investigated in clinical trials, or are currently being investigated [7, 8, 9, 10].

Third, there has been a shift from investigating postmenopausal women only to also investigating the cost-effectiveness of treatments in elderly men. The interest in male osteoporosis has only recently been reflected in economic evaluations, as these all date from 2003 and 2004. It has been reported that actually one-third of hip fractures will occur in men [44]. Two studies investigated male and female populations [31, 45]. One study investigated alendronate in men only [30]. Two studies did not investigate men and women separately, although this may be problematic as hip fracture rates are different in both populations.

Finally, there has been a shift away from using BMD-based measures to predict the risk of fractures to using measures of fracture risk that are age dependent. It has been argued that they more accurately predict the risk of fracture than BMD-based measures [11, 29, 46].

There have been major developments in the treatment of osteoporosis in the last few years with the publication of the results of the Women’s Health Initiative on ERT and HRT, but also a number of clinical trials of other treatments for osteoporosis such as bisphosphonates, raloxifene, vitamin D and calcium and hip protectors. This shift of focus in treatments is reflected in the cost-effectiveness models that have been developed and subsequently published that have been described in this article. While a number of particular assumptions, such as the putative cardioprotective effect of HRT in these models may be obsolete, this does not make the models themselves obsolete. Methodological developments, such as the use of probabilistic sensitivity analysis and cost-effectiveness acceptability curves have also taken place. Such changes are reflected in the studies that were reviewed. If economic evaluations are to be useful decision-making tools, their development should follow an iterative process that incorporates new information, whether clinical or methodological, as it becomes available.

This structured review has provided an update and an extension to the review published by Zethreaus in 2002 [11]. It uses different inclusion and exclusion criteria from that study. For example, it does not include articles that described models without providing results [29, 47]. On the other hand, it includes studies that present cost per fracture as an outcome, while Zethreaus et al. did not. In this study, accepted methodologies for conducting systematic reviews were used. The search strategy was predefined, and data were extracted using pre-defined evidence tables. However, we did not provide a critical appraisal of the quality of theses economic evaluations, which would be necessary for it to qualify as a full systematic review. Future work should build on these initial results to provide a structured appraisal of the quality of the economic evaluations identified. Such work will provide researchers with additional information to build and develop future cost-effectiveness models that will provide economic and clinical evidence to decision-makers on treatment and prevention options for patients with osteoporosis.