- Top of page
Background: According to several guidelines, the assessment of postmenopausal fracture risk should be based on clinical risk factors (CRFs) and bone density. Because measurement of bone density by dual x-ray absorptiometry (DXA) is quite expensive, there has been increasing interest to estimate fracture risk by CRFs.
Objective: The aim of this study was to determine the cost-effectiveness of osteoporosis screening of CRFs with and without DXA compared with no screening in postmenopausal women in Germany.
Methods: A cost-utility analysis and a budget-impact analysis were performed from the perspective of the statutory health insurance. A Markov model simulated costs and benefits discounted at 3% over lifetime.
Results: Cost-effectiveness of CRFs compared with no screening is €4607, €21,181, and €10,171 per quality-adjusted life-year (QALY) for 60-, 70-, and 80-year-old women, respectively. Cost-effectiveness of DXA plus CRFs compared with CRFs alone is €20,235 for 60-year-old women. In women above the age of 70, DXA plus CRFs dominates CRFs alone. DXA plus CRFs results in annual costs of €175 million, or 0.4% of the statutory health insurance's annual budget.
Conclusion: Funders should be careful in adopting a strategy based on CRFs alone instead of DXA plus CRFs. Only if DXA is not available, assessing CRFs only is an acceptable option in predicting a woman's risk of fracture.
- Top of page
Osteoporosis, a multifactorial disorder resulting in increased bone fragility, occurs in women after menopause and is one of the most important disorders affecting the elderly . Population aging is expected to increase the number of osteoporosis-related fractures such as hip fractures and, hence, the economic burden for society.
Bone mineral density (BMD) is considered an important predictive factor for osteoporotic fractures and is measured by densitometry. Densitometry results are usually reported as a t-score, which is the number of standard deviations between the value of an individual and the mean value of a group of young adults of the same sex . According to the criteria of the World Health Organization (WHO) osteoporosis is defined by a t-score of ≤−2.5 . If bone density is measured by dual x-ray absorptiometry (DXA), the risk of hip fracture (other fractures) increases by a factor between 3.7 (1.2) at age 50 and 1.7 (1.6) at age 90 for each standard deviation decrease in BMD at the femoral neck . This increase in fracture risk for each standard deviation change is called the gradient of risk (GR/SD) .
DXA is expensive, not available everywhere, and to a certain degree unreliable because BMD can vary by up to 20% to 50% around an individual's true BMD . Furthermore, different scanners for bone density used in the same patients vary considerably in the proportion of those who receive a diagnosis of osteoporosis . While, in Germany, the prevalence of osteoporosis in women varies between 7% (age: 55) and 19% (age: 80) , the majority of fractures occur in women who do not have osteoporosis . Thus, the use of DXA in primary or secondary prevention is disputed .
Recently, several risk factors for fracture have received attention . These include prior fragility fractures, a family history of hip fracture, low body mass index (BMI), smoking, alcohol intake, and the use of oral corticosteroids. Combinations of these risk factors were used to develop decision rules for BMD referrals . Although a case finding based on DXA alone (compared with fractures that will occur in the following 10 years) has a specificity (proportion of true-negatives) of approximately 90%, its sensitivity (proportion of true-positives) is only 34% . Different strategies based on CRFs alone, in turn, have shown to exceed a sensitivity of 80% although their specificity is only about 50% (compared with the reference standard low BMD measured by DXA) . Thus, predicting fracture risk based on CRFs in addition to BMD increases the GR for the prediction of hip and other fractures . As a result, the sensitivity and the positive predictive values (proportion of women with positive test results who will have a fracture in the following 10 years) increase [13,14].
To guide treatment based on a combined use of CRFs and BMD, several organizations recently have recommended using CRFs and BMD to assess an individual's absolute 10-year risk of fracture [9,15] or annual incidence of fracture . According to the German osteology umbrella organization, Dachverband Osteologie (DVO) guideline, DXA should be provided for women when there is a 10-year risk of combined vertebral (clinical and morphometrical) and hip fractures of ≥20%. Drug treatment should be provided for women with a combined risk of ≥30% for vertebral and hip fractures . The t-score required to reach this risk threshold varies by age. A 55-year-old woman, for example, receives treatment for a t-score of −4, whereas a 67-year-old woman receives it based on a t-score of −3. Additional risk factors further increase the t-score required to reach the threshold .
While this and other recently developed screen-and-treat strategies agree that CRFs should be given more attention, treatment recommendations still tend to center on DXA [7,15,16]. Although the combined use of DXA and CRFs improves the GR/SD , for women aged >65 years, its sensitivity is only 60% even if a risk threshold of 30% is chosen . This is an increase of 80% compared with DXA alone in women aged 70 to 79 years, but the price to pay for this improvement is a decrease in specificity by 16% [11,13]. The usage of CRFs alone, however, may be of diagnostic value for predicting fracture risk because age-specific GRs are similar to those of BMD alone . Thus, a strategy where fracture risk is calculated by CRFs alone may improve the cost-effectiveness compared with an expensive DXA-centered strategy. The National Institute for Health and Clinical Excellence, for example, recommends bisphosphonates in postmenopausal women aged 75 years and older even without the need for DXA if the clinician considers DXA to be clinically inappropriate or unfeasible .
In the vast majority of cost-effectiveness analyses on postmenopausal osteoporosis treatment, women at increased risk were selected by low BMD . Nevertheless, the use of CRFs as a prescreening tool for DXA (i.e., DXA only for those women with elevated CRFs) has been shown to be cost-effective when compared with mass screening with DXA alone . In addition, there are two modeling studies that analyzed the cost-effectiveness of treatment in postmenopausal women, based on long-term fracture risk rather than on BMD alone [19,20]. In contrast to our analysis, these studies did not consider treatment costs of false positives, selected women at increased risk based on additional risk factors (e.g., BMI or the use of oral glucocorticoids), used a 10-year modeling horizon, and assumed that beyond 10 years, women would have a mortality rate equal to that of an age- and sex-matched population [19,20,21].
The present study investigated the cost-effectiveness of the following strategies: 1) screening based on CRFs alone (without information about BMD) and treatment with alendronate in case of risk of ≥30% (age groups 60–70 and 70–80), or treatment with alendronate for all women (age group: 80–90); 2) screening with DXA plus CRFs (plus alendronate); and 3) no screening (Fig. 1). While different medical treatment options are recommended, we chose alendronate, an antiresorptive biphosphonate, as the sole drug because, in our previous cost-utility analysis, it has shown to be most cost-effective .
Figure 1. Overview of screen-and-treat strategies (clinical risk factors (CRFs) alone, dual x-ray absorptiometry (DXA) plus CRFs, no screening). *Numbers in parentheses indicates the 10-year risk of vertebral and hip fractures in women of this age group. †The true bone mineral density (BMD) of women in the CRFs group is unknown because women in this group can have normal, osteopenic, or osteoporotic BMD.
Download figure to PowerPoint
- Top of page
Treatment should be based on bone density as long as DXA is available for screening women above the age of 70 years. Therefore, health policy should aim to increase the availability of DXA. Only in cases where DXA is not available, the usage of CRFs alone is justified. In women aged 60 to 70 years, where CRFs are superior, there is substantial uncertainty in the results. Because cost-effectiveness acceptability curves (CEACs) do not indicate the costs of making a wrong decision, funders should be careful in adopting CRFs alone instead of DXA plus CRFs.
In this analysis, ICERs were calculated based on an intervention threshold of ≥30%. Lowering this threshold would decrease specificity, and, as a result, treatment costs for false positives would largely increase. Thus, we would expect a lower threshold to be less cost-effective. In contrast, raising the threshold above 30% may improve the ICER in women aged ≥70 years because the number of false positives would decrease. Nevertheless, these assumptions could not be tested because of a lack of data.
Our analysis has several strengths. It considers long-term follow-up costs of hip fractures in Germany including costs of hip implants, revision surgery, transportation services, rehabilitation care, and long-term care . Moreover, in contrast to many economic analyses in the field of osteoporosis prevention, which did not consider the costs of case finding , our analysis includes all screening costs including costs of women being diagnosed incorrectly (false-positives). Furthermore, in contrast to the model by Zethraeus and colleagues , our model is able to consider the occurrence of vertebral and forearm fracture after hip fracture. The no-memory assumption of Markov models was avoided by creating health states that correspond to combined health states .
An important limitation is that the number of women at increased risk in the age group of 60 to 70 years is based solely on prior vertebral fractures. Additional risk factors to increase the number of women at high risk could not be used because their distribution in the German population has not been sufficiently evaluated yet . This results in a lower ICER than that of screening in women aged 70 and 80 years although fracture risk increases with age and the ICERs of screening strategies usually decrease with age. The reason is that the cost-effectiveness ratio is very sensitive to specificity that is, based on this strategy, 95% for screening in women aged <70 years, but only 20% in women aged 70 years, and 0% in women aged 80 years. If it were possible to calculate the risk increase by combining several risk factors, the detection of more women at risk would be inevitably attenuated by a significant decrease of specificity in the identification of nonosteoporotic women, as shown in the evaluation of many decision rules for CRFs in osteoporosis screening [10,12].
There are a number of reasons why the results of our study are rather conservative. First, when calculating the costs of fracture treatment, analgesics were not considered. A recent cost analysis showed that osteoporotic patients receive three times more prescriptions for analgesics than nonosteoporotic patients do  and that those who have received nonsteroidal antiphlogistics were significantly more often hospitalized for peptic ulcer disease than those who have not received nonsteroidal antiphlogistics. Including these and other drug costs in the calculation of treatment costs for vertebral and forearm fractures would improve cost-effectiveness of CRFs with and without DXA.
Second, our analysis incorporated an increased risk of subsequent fractures for women who have suffered a prior fracture. This risk increase was determined only by the last fracture because there were no data available on the relationship between risk increase and the number of prior fractures. Consideration of two or more prior fractures would probably further increase the risk of a subsequent fracture and improve the cost-effectiveness ratio of CRFs with and without DXA. If these data become available, individual patient-level models could be more accurate in simulating the relationship between the number of prior fractures and subsequent fractures than between cohort-based approaches . Moreover, by considering only the last fracture, in women with a prior hip or vertebral fracture who have a forearm fracture, the risk of subsequent fractures decreases. Continuing the higher risk increase of the prior fracture would improve the ICERs of CRFs and DXA plus CRFs.
Finally, the fracture states in this model are the same as in the reference model  except for fractures at “other” sites that were not modeled here. By considering only hip, vertebral, and forearm fractures, our analysis may have underestimated potential cost savings from the prevention of fractures at other skeletal sites. Although studies on alendronate did not show a significant reduction of other fractures , a study on risedronate, which also belongs to the class of bisphosphonates, reported a significant reduction of nonvertebral fractures by 39% (defined as fractures of the clavicle, humerus, wrist, pelvis, hip, or leg) .
On the other hand, there are also several reasons why a screen-and-treat strategy based on CRFs alone may be less cost-effective than that calculated by the base-case analysis. First, there is contradictory evidence whether individuals selected for treatment based on CRFs alone benefit from treatment or not. To date, efficacy data have to be taken from a population selected by the WHO criterion of BMD including individuals with osteopenia . It remains unclear whether these data reflect the true efficacy of a population with a 30% long-term risk based on CRFs. In some efficacy trials, however, pharmacological interventions with bisphosphonates have been shown effective in patients not selected on the basis of low BMD [74–77].
Third, alendronate was assumed to be efficacious in all age groups although there have not been studies in women above the age of 80 .
Finally, for the state “long-term risk of ≥30%,” disutility of women aged 70 to 80 was underestimated. The reason is that an unknown number of women in this group suffer from prior fractures causing disutility. If a lower QoL was assumed, the incremental gain in QoL using CRFs with and without DXA compared with no intervention would be reduced, and, therefore, both strategies would become less cost-effective. Nevertheless, the increase of the ICER would be negligible, as shown in the sensitivity analysis.
The structure of this Markov cohort model is similar to that of an established reference model  although there are also some discrepancies: in our model, a state for other osteoporotic fractures was not modeled, and a woman may suffer a vertebral or forearm fracture after a hip fracture. As recommended there, a lifelong time horizon with a cycle length of 1 year was adopted, effectiveness was assumed to decrease linearly for a given “offset time,” and increased mortality after hip or vertebral fracture was modeled. In addition, for all fracture states, a postfracture state was included to reflect the sustained risk increase for subsequent fractures.
To compare the results of this analysis to those of other models, we used two cost-effectiveness studies for treatment with bisphosphonates in postmenopausal women [19,20], which are also based on long-term fracture risk. In these analyses, intervention thresholds for cost-effectiveness of bisphosphonates in women at different T-scores with or without prior fracture were calculated. Although the 5-year baseline risks of hip fracture were similar (e.g., 7.1% and 12.3% for women with a fracture history aged 70 and 80 years, respectively, compared with 6.7% and 12.1%, respectively, in our analysis) , the results of these analyses are quite different from our results. But again, there are important methodological differences compared with our study: costs of hip fracture are much lower given that costs beyond the first year after fracture were not included, treatment costs of false-positives were not considered, the discount rates used for costs (6%) and benefits (1.5%) differed from our study, and compliance was not considered in the base case. In addition, the selection of women at increased risk was based on several other risk factors such as BMI, history of peripheral fractures, use of oral glucocorticoids, and history of rheumatoid arthritis [19,20]. Consideration of several risk factors is likely to have decreased the specificity of the identification of nonosteoporotic women and thus increased the cost-effectiveness ratio.
If we assumed a societal perspective for our analysis, we would expect similar results. Copayments for drugs may be partly outweighed by savings for copayments for fracture treatment. Costs through loss of productivity are of minor importance because the proportion of women aged >60 years being employed in Germany is below 15% .
We do not know whether our findings are transferable to other countries. The main reason is that, for costs and epidemiological data, German sources were preferred as inputs to the model, so differences in resource consumption and prices may exist.
There are several important areas for future research. First, predicting fracture risk based on CRFs is still not accurate enough. Whereas risk factors such as a prior vertebral fracture and low BMD can be used for precise measurements of the risk increase , other risk factors such as low BMI and immobility only are known to be indicators of low BMD . Thus, their interrelationship has yet to be formalized with more precision.
Second, effectiveness of alendronate has to be shown in clinical trials with patients selected on the basis of CRFs instead of low BMD .
Finally, this analysis compared different screen-and-treat strategies, based on an arbitrary risk threshold of 30% for hip or vertebral fractures. It would be of general interest to determine ICERs for different intervention thresholds of screen-and-treat strategies, using different ratios of hip to vertebral fracture risk or hip to nonhip–fracture risk. In our analysis, the ICERs of a combined risk of exact 30% were calculated for a specific ratio of hip to vertebral fracture risk. Nevertheless, different risk factors have different impacts on fracture sites and, thus, different impacts on QoL, costs, and mortality . For example, if the absolute fracture risk threshold is reached by risk factors that are assumed to have a larger impact on the risk of hip fractures than on vertebral fractures (e.g., use of corticosteroids), treatment is more cost-effective because hip fractures have a higher impact on costs and mortality than fractures at other sites do .
In summary, CRFs are of considerable value for decision-making regarding the treatment of postmenopausal osteoporosis. Nevertheless, until the interrelationships between CRFs have been evaluated more extensively and until treatment with bisphosphonates in women selected by risk factors has shown to be as effective as in women selected by BMD, their usage should be combined with DXA. As recommended also by National Institute for Health and Clinical Excellence, providing treatment without DXA should be limited to older women only if the responsible clinician considers it to be clinically appropriate or unfeasible . As long as DXA is available, the implementation of any approach based on CRFs alone will result in an uncontrolled increase in health expenditures.