Osteoporosis International

, Volume 18, Issue 2, pp 143–151

Which screening strategy using BMD measurements would be most cost effective for hip fracture prevention in elderly women? A decision analysis based on a Markov model

Authors

    • Epidemiology UnitDépartement d’Information Médicale des Hospices Civils de Lyon
    • INSERM U 403Hôpital Edouard Herriot
  • C. Ganne
    • Medico-economic evaluation UnitDépartement d’Information Médicale des Hospices Civils de Lyon
  • D. Hans
    • Nuclear Medicine DivisionGeneva University Hospital
  • G. Monnier
    • Medico-economic evaluation UnitDépartement d’Information Médicale des Hospices Civils de Lyon
  • R. Gauchoux
    • Medico-economic evaluation UnitDépartement d’Information Médicale des Hospices Civils de Lyon
  • M. A. Krieg
    • CHU VaudoisLauzanne University Hospital
  • P. D. Delmas
    • INSERM U 403Hôpital Edouard Herriot
  • P. J. Meunier
    • INSERM U 403Hôpital Edouard Herriot
  • C. Colin
    • Epidemiology UnitDépartement d’Information Médicale des Hospices Civils de Lyon
    • Medico-economic evaluation UnitDépartement d’Information Médicale des Hospices Civils de Lyon
Original Article

DOI: 10.1007/s00198-006-0227-6

Cite this article as:
Schott, A.M., Ganne, C., Hans, D. et al. Osteoporos Int (2007) 18: 143. doi:10.1007/s00198-006-0227-6

Abstract

Introduction

Hip fractures are responsible for excessive mortality, decreasing the 5-year survival rate by about 20%. From an economic perspective, they represent a major source of expense, with direct costs in hospitalization, rehabilitation, and institutionalization. The incidence rate sharply increases after the age of 70, but it can be reduced in women aged 70–80 years by therapeutic interventions. Recent analyses suggest that the most efficient strategy is to implement such interventions in women at the age of 70 years. As several guidelines recommend bone mineral density (BMD) screening of postmenopausal women with clinical risk factors, our objective was to assess the cost-effectiveness of two screening strategies applied to elderly women aged 70 years and older.

Methods

A cost-effectiveness analysis was performed using decision-tree analysis and a Markov model. Two alternative strategies, one measuring BMD of all women, and one measuring BMD only of those having at least one risk factor, were compared with the reference strategy “no screening”. Cost-effectiveness ratios were measured as cost per year gained without hip fracture. Most probabilities were based on data observed in EPIDOS, SEMOF and OFELY cohorts.

Results

In this model, which is mostly based on observed data, the strategy “screen all” was more cost effective than “screen women at risk.” For one woman screened at the age of 70 and followed for 10 years, the incremental (additional) cost-effectiveness ratio of these two strategies compared with the reference was 4,235 euros and 8,290 euros, respectively.

Conclusion

The results of this model, under the assumptions described in the paper, suggest that in women aged 70–80 years, screening all women with dual-energy X-ray absorptiometry (DXA) would be more effective than no screening or screening only women with at least one risk factor. Cost-effectiveness studies based on decision-analysis trees maybe useful tools for helping decision makers, and further models based on different assumptions should be performed to improve the level of evidence on cost-effectiveness ratios of the usual screening strategies for osteoporosis.

Keywords

BMDCost-effectiveness analysisMarkov modelOsteoporosisScreening

Introduction

Osteoporosis can be defined as a systemic skeletal disease characterized by low bone mass and microarchitectural deterioration of bone tissue, with a consequent increase in fracture risk. Since there are no satisfactory clinical tools widely available to assess bone quality, the diagnosis of osteoporosis depends, at present, upon the measurement of bone mineral density (BMD) [1]. From several prospective studies, bone mass decrease is known to increase fracture risk [25]. However, although a low BMD is an important risk factor, other factors contribute to fracture risk, particularly in the elderly [6, 7].

In recent years, a number of approaches to the assessment of osteoporosis have been formulated. In Europe, the approach recommended by the International Osteoporosis Foundation (IOF) is to identify individuals on the basis of strong risk factors, such as prior fragility fractures, corticosteroid use, family history of fracture, low body mass index (BMI), or menopause before 40 years of age [8, 9]. Individuals with such risk factors are assessed thereafter by measurement of bone mineral density (BMD), and treatment is offered in the presence of low BMD. In the approach recommended by the IOF, low BMD is defined by the World Health Organization (WHO) criteria [1], and the intervention threshold is set at the diagnostic threshold, i.e., a t score of −2.5 standard deviation (SD) or less. Possible advantages and disadvantages of this approach have been advocated [10]. However, no comparison of the cost-effectiveness ratio of different screening strategies has been carried out although the choices of screening strategies strongly depend upon economic analyses.

We focused our model on hip fractures in elderly women for several reasons. Hip fracture incidence increases exponentially with age and thus represents a threatening major public health problem in developed countries with the increasing life expectancy [11]. Hip fractures are also responsible for excessive mortality, decreasing the 5-year survival rate by about 20% [12, 13]. From an economic perspective, it represents the major source of expenses, with costs including in- and out-patient hospitalizations, rehabilitation care [13], and costs of institutionalization due to the fact that 30% of the elderly surviving the event become dependent [14]. It has been clearly demonstrated that in women aged 70–80 years, hip fracture rate could be reduced by therapeutic interventions, such as bisphosphonates [15, 16] or calcium and vitamin D [17]. Recent analyses suggested that the most efficient strategy would be to treat women at the age of 70 years [18] and that population screening for osteoporosis would be more judicious in the elderly than at menopause [19]. Decision analysis is a formal approach for structuring and analyzing decision problems [20]. We used this approach for modeling different alternatives for screening elderly women at risk of hip fracture and comparing their respective efficacy and cost-effectiveness. The aim of the study was to compare the incremental cost per year without hip fracture gained of two alternative screening strategies: systematic screening by dual energy X-ray absorptiometry (DXA) and screening by DXA only women with risk factors. Both alternatives were compared with the baseline strategy of “no screening”.

Methods

Decision analysis model

We used two of the most common techniques for modeling in economic evaluations: decision-tree analysis and Markov modeling [21]. Markov state-transition models are useful when a decision problem involves risks that are ongoing over time, such as osteoporotic hip fractures [22]. With the decision analytic model, cost and effectiveness between exclusive strategies were compared so that the additional cost that is required to obtain the additional effectiveness of a strategy could be determined. The lower the value of the ratio, the higher the priority in terms of maximizing the benefits derived from a given health expenditure. The strategies were entered into a decision tree by means of the software TreeAge (DATA 3.6) for Healthcare (TreeAge Software Inc., Williamstown, MA, USA) using a Markov model [23].

Decision alternatives

Two strategies were compared with a baseline reference “no screening” strategy. Strategy 1 was based on the identification of individuals with at least one strong risk factor and further assessment of BMD, with treatment being offered in the presence of BMD t score ≤−2.5, based on the WHO definition of osteoporosis. The definition of risk factors was based on the recommendations published by the French National Agency for Health Technology Assessment (ANAES) (http://www.anaes.fr). These are: a history of fracture after the age of 50 years, menopause before 40 years of age, history of maternal hip fracture, BMI lower than 19 kg/m2, and use of corticosteroids (> 7 mg a day equivalent prednisone) for more than 3 months. Strategy 2 was based on a general systematic screening recommended for all women aged 70–80 years. The decision tree starts with a decision node (squared) from which are initiated three tree fragments, one for each strategy. One tree fragment modeling the “no screening” strategy is detailed in Fig. 1. The average cost of each of the three decision alternatives was obtained by averaging out and folding back, using cost as the measure of outcome. The average number of years without hip fracture was calculated the same way.
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-006-0227-6/MediaObjects/198_2006_227_Fig1_HTML.gif
Fig. 1

Decision-tree fragment showing the strategy “screen women at risk”

Model assumptions

The base case was a postmenopausal 70 year-old woman with no history of hip fracture and followed for 10 years, until the age of 80. Only permanent admissions in nursing home were considered to define the Markov state “institutionalized.” An individual could sustain only one hip fracture per year and at most two hip fracture over 10 years. All women who had a BMD t score ≤−2.5 were assumed to be correctly and effectively treated with therapeutics such as risedronate [15] or alendronate [16]. The treatment effect over the time horizon was approximated on the basis of Kanis estimates, taking into account efficacy, offset, compliance, and discontinuation rate of treatment given for 5 years [24]. The model was based on BMD testing of the femoral neck and on only one measure within 10 years of follow-up. If the 70-year-old woman was institutionalized for another medical reason during the following cycle, she may either stay in the same state, die, or sustain a hip fracture. All assumptions that were made regarding screening, hip fracture risk, and treatment effect, are described in Table 1. Sensitivity and specificity of DXA (with threshold of BMD t score ≤−2.5) regarding the risk of hip fracture are also provided in Table 1.
Table 1

Data sources, costs, and probabilities found in the literature

Events

Age

Values

Sources

Costs of DXA screening

 

75 euros

 

Costs of screening campaign

 

25 euros

Watt 2003 [33]

Costs of hospitalization and rehabilitation after a hip fracture

 

15.540 euros

Maurel 1998 [35]; Treves 1989 [36]; Lévy 1989 [37]

Annual costs of institution

 

18.400 euros

Maurel 1998 [35]; Treves 1989 [36]; Lévy 1989 [37]

Costs for preventive treatments of osteoporosis

 

496 euros

Maurel 1998 [35]; Treves 1989 [36]; Lévy 1989 [37]

Effect treatment (reduction of hip fracture incidence)

 

35%

Kanis 2001, 2002 [29, 30]

Annual probability of institutionalization

70

0.075

INSERM 1996

75

0.125

Annual probability of institutionalization after hip fracture

 

0.20

INSERM 1996; Sernbo I 1993

Annual probability of death from any cause

70

0.0117

INSERM 1997

75

0.0348

Annual probability of death after hip fracture

65

0.10

Cummings 1989; Tosteson 1990; Fisher 1991; Cabases Hita 2000

Annual probability of sustaining a hip fracture in institutionalized women

80

0.038

Baudoin 1996 [26]

Relative risk of death in institution vs. general population

65

1.72

PICAROS 1995; Baudoin 1996 [26]; Graafmans 1996

Strategy: “screen women with at least one risk factor”

  Average sensitivity of t score DXA ≤−2.5 if risk factor to predict hip fracture

 

0.75

EPIDOS

  Average specificity of t score DXA ≤−2.5 if risk factor to predict hip fracture

 

0.54

EPIDOS

Strategy: “screen all women”

  Average sensitivity of t score DXA ≤−2.5 to predict hip fracture

 

0.76

EPIDOS

  Average specificity of t score DXA ≤−2.5 to predict hip fracture

 

0.58

EPIDOS

INSERM Institut national de la santé et de la recherche médicale

Markov model

States of the Markov model

Events were modeled in 1-year cycles. Within the first Markov cycle, the woman was in “good health.” After the first cycle, the woman could either suffer a hip fracture, die, be admitted to a nursing home for other reasons than hip fracture, or suffer none of these events and thus stay in the same Markov state. In the latter case, she started a new cycle, with identical states but different transition probabilities of dying, suffering a hip fracture, or being admitted to a nursing home. These probabilities increased with age. If a hip fracture occurred, she might be in one of three definite states the next year (next cycle): either the fracture was healed and she was back to previous state, or subsequent to the fracture, she died or had to be admitted to an institution for elderly people. Death was considered as an absorbing state in this model.

Probabilities

Probability values of each possible outcome that were found in the literature are displayed in Table 1, together with the sources. Baseline probabilities were estimated from French national registers, when available. Otherwise, scientific papers were used. Most probabilities were calculated from the data of three large European cohorts: EPIDOS and OFELY cohorts in France and SEMOF cohort in Switzerland.

Hip fracture rates

Annual hip fracture probability based on French estimates was implemented in the model [25, 26]. Hip fractures were modeled at a higher rate in nursing home residents [26].

Mortality rates

Age-specific mortality rates in female French population aged 70–80 were based on INSERM data 1996. Increased mortality rates in institutionalized women and excess mortality after a hip fracture were based on French estimates [26, 27]. They are consistent with other published data [28].

Probability of being dependent

Permanent admissions in nursing home were used to define the Markov state “dependence”. The probability of being admitted in a nursing home in the general population was based on French data from a cohort study [26, 27]. Higher probabilities of being admitted in a nursing home after a hip fracture were modeled on the basis of French data from INSERM.

Treatment effect

We chose an average value of 35% reduction of hip fracture incidence over 10 years based on a treatment duration of 5 years as it has been described in other studies [29, 30].

Time horizon

Recently published models are based on 10-year periods [31, 32]. In a recent analysis based on the EPIDOS cohort, modeling for hip-fracture risk showed that after 10 years, the prediction of BMD estimated by the relative risk was no longer significant [7].

Model validation

We confirmed equal hip fracture probability across all strategies when treatment was assumed to confer no reduction in fracture probability.

Main outcomes measures

Effectiveness outcomes

Effectiveness was measured as the number of years without a hip fracture gained over 10 years.

Costs outcomes

In this model, we only considered direct costs. Adopting the healthcare system perspective, the model incorporated the costs of BMD testing (one per woman), the costs of preventive treatments for women with a low BMD, the costs of hospitalization and rehabilitation of patients with a hip fracture, the cost of institutionalization, and the costs of a screening campaign. Nonmedical direct and indirect costs were not included in the analysis.

In France, the price of DXA was not fixed by official rules when this study was conducted. We chose a cost of 75 euros based on the average costs observed in practices, although there is no official data. Costs of the screening campaign were derived from those of the French screening campaign for breast cancer [33]. Costs for preventive treatments was based on risedronate or alendronate (daily or weekly administration) given for 5 years. This represents 1.36 euros per day and 496 euros per year. This is close to the model developed recently by Kanis [34]. Costs of hospitalization after a hip fracture were based on previous French data [3537]. Costs for rehabilitation were included in cost of hospitalization. Direct costs may occur in the future. We chose a 5% discount rate usually used in pharmaco-economic studies. Values of the different costs are displayed in Table 1.

Data sources

Data sources are detailed in Table 1. Most data were based on European prospective cohort studies. For women aged 75–80, data were obtained from the EPIDOS study, a cohort composed of 7,598 women aged 75 and over. For younger women, aged 70–75 years, data were based on unpublished data from two other cohort studies: OFELY and SEMOF.

Expression of cost-effectiveness

Cost-effectiveness may be expressed as the cost incurred per unit of outcome achieved. The incremental (i.e., marginal) cost-effectiveness of one decision alternative to another or to the reference strategy is the extra cost incurred for one incremental unit of outcome. In our case this was the extra cost per additional year without hip fracture gained [38]. When possible, we identified a dominant solution; otherwise, we used the Incremental cost-effectiveness analysis to compare the decision alternatives.

Sensitivity analysis

We evaluated the effect of varying the variables values. Those that strongly influenced the model were tested in two-way sensitivity analyses for evaluating the model for a range of potential costs and success rates of strategies 1 and 2.

Results

Expected effectiveness

The expected clinical effectiveness was estimated for one woman as the average number of years without hip fracture over 10 years. Screening all women was the most effective strategy, with expected efficacy rate estimated at 8.27 years without hip fracture over 10 years (Table 2). For the strategy “screen women at risk,” the expected efficacy rate estimate was 8.03; intermediate between “screen all” and “no screening,” the reference strategy was 7.84. Based on risk factor prevalence observed in the cohorts, at the age of 70 years, approximately 16% of women would be treated under this scenario, at the age of 75, approximately 22% would be treated, and at the age of 80, approximately 28% would be treated.
Table 2

Cost-effectiveness and cost-effectiveness ratios of the two screening strategies and the reference strategy

Strategy

Cost (euros)

Incremental cost (euros)

Effectiveness (years)

Incremental effectiveness (years)

Incremental cost-effectiveness ratios (/year)

No screening

46,800

 

7.836

  

Screen women at risk

48,400

1,600

8.029

0.193

8,290a

Screen all women

48,600

1,800

8.265

0.425

4,235

aExtended dominance

Expected cost

For one woman between the ages of 70 and 80 years, the expected costs of each screening strategy compared with no screening over 10 years are displayed in Table 2. With the strategy “no screening,” the expected costs were 46,800 euros; with the strategy “screen women with at least one risk factor,” they were 48,400 euros; and with the strategy “screen all women at the age of 70 years,” they were 48,600 euros.

Incremental cost-effectiveness ratios

The unit of efficacy in this model was 1 year without hip fracture. The costs to obtain one unit of additional effectiveness, i.e., the incremental cost-effectiveness ratios (ICERs), compared with the reference strategy were as follows: the ICER of the strategy “screen all” was 4,235 euros per year without hip fracture and 8,290 euros for the strategy “screen women at risk.” The strategy “screen all” appeared to be the most cost-effective approach.

Dominance

No strategy dominated the others, as none of them was more effective and less expensive at the same time. The condition of extended dominance occurred when comparing the “screen all” strategy to the “screen women at risk” strategy because it was more expensive to buy an additional life year without hip fracture (i.e., a higher cost-effectiveness ratio) using the strategy “screen women at risk” than the most effective strategy, i.e. “screen all”.

Sensitivity analysis

Sensitivity analysis showed that the choice of strategy was dependent upon several parameters, meaning that uncertainty in those variables had significant effect on the analysis. Those variables were the number of cycles observed in the model (range 5–15), the cost of a nursing home (range 15,000–25,000 euros), the probability of having a BMD below −2.5 (range 0.2–0.6), the cost of treatment (range 194–630 euros), and the cost of DXA (range 50–100) euros. The cost-effectiveness ratio of the strategy “screen all” varied from 216 euros to 12,850 euros per year without hip fracture gained, and that of the strategy “screen women at risk” varied from 3,662 euros to 73,178 euros per year without hip fracture gained. Although very large variations were observed in sensitivity analyses, the hierarchy of cost-effective strategies was not altered, the strategy “screen all” was persistently more cost effective, and thus the condition of extended dominance of the “screen all” strategy compared with the “screen women at risk” strategy was constantly observed.

Discussion

Our decision analysis suggests that in women aged 70–80 years—under the assumptions of the model and assuming results from the observed cohorts are transposable—screening all women with DXA would be more cost effective than no screening or screening only women with at least one risk factor. The overall expected costs for one woman were not very different between strategies. The costs of DXA testing and preventive treatments, which are not very expensive, were compensated by the costs avoided by reducing the hip fracture rate. Although this is a relatively rare event, the cost of each hip fracture is very high compared with the costs of screening and treatments. Differences in expected costs between “screen all” and “screen women at risk” were also explained by the cost of the screening campaign. This model seems to be robust. One-way and two-way sensitivity analyses altered the magnitude of cost-effectiveness ratios but not the hierarchy of cost-effective strategies.

However, much information is still lacking to help decision makers. It is not clear how often the screening test should be repeated, at which age screening should be started, what the real efficiency of the treatments formerly assessed in randomized trials is, or what the participation rate to a screening program will be. Finally, long-term effects of screening and treatments have to be validated.

Worldwide, the incidence of hip fracture is expected to increase from 1.66 million euros per year in 1990 to 6.25 million euros per year by 2050 [11]. Over the last 10 years, many guidelines recommended screening women at high risk with DXA on the basis of clinical risk factors. Possible advantages of this approach have been advocated [10], such as the fact that these guidelines are intuitive to the practice of medicine and that they are conservative in that many individuals identified have a high risk of fracture (higher specificity). Therein also lies a disadvantage, as many individuals at high risk go undetected (lower sensitivity) [6].

In France, the ANAES recently recommended an approach similar to that recommended by the IOF. They recommended focusing BMD assessment in postmenopausal women having at least one strong risk factor (ANAES 2001). However, the efficiency of such strategy has not yet been validated, and recent reports [National Institutes of Health (NIH) 2000 and CCOHTA 2003 (http://www.ccohta.ca)] conclude that available evidence does not support the use of BMD measurements for population screening for asymptomatic individuals. To address this question, clinical studies must be completed, with modeling to perform economic evaluations and [39] to produce information beyond that which is available in clinical studies. The models that we found in the literature were mostly designed to assess and compare the cost-effectiveness of different treatments, such as bisphosphonates or calcium and vitamin D [24, 28, 29, 40]. Recently, Schousboe et al. included in their model the effect of screening and compared the strategy “no intervention” to the strategies “screen all women” and “treat those at high risk” (t score ≤−2.5) [41]. Their results are comparable with ours, with ICERs of similar magnitude, although their method was different. They used quality-adjusted life years (QALY) instead of year without hip fracture, eight health states that are not the same as our four health states, a lifetime horizon, and different age groups. For the age group 75–85 years, they found an ICER of $5,657, which is comparable to our ICER of 4,235 for the strategy “screen all.” We found no model designed for comparing the cost-effectiveness of various screening strategies.

Our model assumptions may be discussed. We chose to run cost-effectiveness analyses based on clinically meaningful outcomes. Several papers recommend using cost-utility analyses based on the assessment of QALYs, mainly for allowing comparison between healthcare programs. Utility values have to be used to calculate loss of QALYs and, until recently, most studies focused on vertebral and Colles’ fracture patients. More recently, papers have been published on utility assessment of hip fracture, and they show that the valuation of health states can be very different according to the method used and the population sample studied [42, 43]. One year gained without hip fracture in elderly women is a clinically and economically relevant outcome. Nevertheless, further studies should be performed in several countries on utility assessment of hip fractures to include this dimension in further models. The only osteoporotic fractures we considered were hip fractures, as they have the most serious consequences on morbidity and health costs in elderly individuals. Other osteoporotic fractures might be included in further models. There were several reasons for targeting elderly women aged 70–80 years. Hip fracture incidence increases exponentially with age after 75 years, and life expectancy continues to increase in developed countries. Recently, Caulin et al. [18] modeled the number of patients needed to treat (NTT) to prevent one fracture. This number decreased markedly with age; NTT was 216 at the age of 50 years, 88 at 60 years, and 36 at 70 years. The most efficient strategy for women with osteoporosis was to treat them at the age of 70 years, with an NTT of 27.

We chose a time horizon restricted to 10 years, as in recent models, a period of 10 years was chosen to obtain more precise estimates [30]. In those models, treatment is usually assumed to be followed for 5 years, with a wear-off over a period of 5 years [29]. This time horizon is also closer to the real situation, as the recommended time between repeated measures in untreated women varies from 1 to 5 years, depending on the baseline BMD [44]. With advancing age the proportion of risk explained by BMD decreases [45], and treatment choice is also dependent on age. Thus, we believe that a specific model should be adapted for each 10-year age strata to obtain meaningful results. The value of 35% reduction of hip fracture incidence over 10 years for a 5-year treatment was based on the Kanis cost-effectiveness study [24]. This was consistent with several randomized trials conducted in elderly women. The trial of risedronate in women aged 70–79 years selected on the basis of a very low BMD showed a reduction of 40% for 3 years of treatment. A previous randomized trial conducted in very elderly institutionalized French women showed a 27% reduction of hip fracture with vitamin D and calcium treatment given for 3 years [46]. Alendronate given for 3 years permitted a 51% reduction of hip fracture risk in the Fracture Intervention Trial (FIT) conducted in 2,027 women aged 55–81 years, with a t score < −1.6 and a previous vertebral fracture [risk ratio (RR) = 0.49 (0.23–0.99)] [16, 47]. In another trial of alendronate, a 47% reduction in nonvertebral fracture incidence was observed [48].

Women at risk were defined as those having at least one risk factor among a short list of significant clinical risk factors recognized world wide. A number of algorithms and various combinations of risk factors have been published in relation to either the prediction of a low BMD or the prediction of fractures. None of these approaches has demonstrated a formal superiority, as none of them had both sensitivity and specificity greater than 50% [4954]. Without objective elements to guide our choice, we turned to the official recommendations of our French National Agency for Technology Assessment and Evaluation in Healthcare. We had data on these risk factors based on a previous cross-sectional analysis of several French cohort studies [55]. Furthermore, these risk factors are available from many cohort studies. The probabilities of events were mostly estimated from one large cohort study of women aged 75 and older. However, for women younger than 75, data came from two cohort studies with limited numbers of women aged 70–75, and the estimates were subject to larger variations. Other underlying hypotheses of the model may be discussed, such as age, strategies being compared, time horizon, discount rate, and outcome measures. Finally, our article was designed for comparing two screening strategies versus no screening strategy, and it does not address the question of treating all elderly women, although several cost-effective models conclude that this strategy may be more cost effective than selective strategies. Thus, we cannot make a conclusion on this particular point.

In conclusion, there is no easy answer for choosing among several screening alternatives. There is a need for quantitative evaluation of the effectiveness and costs of the potential screening programs. In the absence of clinical trials evaluating screening effectiveness, cost-effectiveness models may provide quantitative cost and benefit ratios for different screening alternatives. The results of our decision analysis suggest that in women aged 70 years, screening all women is more effective than no screening or screening only women with at least one risk factor. Although our method differed from that used by Schousboe et al. [41], they found comparable incremental cost-effectiveness ratios in their strategy “screen all and treat” in women aged 75. Further models should be run in other settings, with more locally appropriate risk factors or more comprehensive combinations of risk factors. For example, the cost-effectiveness ratio of the National Osteoporosis Foundation (NOF) prediction model, based on a combination of clinical risk factors and femoral neck BMD, could be compared to other strategies [56]. Also, the algorithm currently being developed by the OMS group for estimating the gradient of hip fracture risk could be used in cost-effectiveness models. Finally, further models should also consider issues of compliance and persistence as well as adverse effects.

Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2006