Introduction

Osteoporotic fractures, such as a hip fracture, are associated with increased morbidity and mortality and impose a huge financial burden on healthcare systems. According to a nationwide survey of hip fractures in Japan, the total number of patients who experienced a hip fracture in 2012 was 175,700 (male 37,600; female 138,100), which is higher than the number in 2007 (total 148,100; male 31,300; female 116,800) [1, 2]. The annual costs for medical and nursing care associated with osteoporotic fractures have been estimated to be between JPY 797.4 and 989.5 billion (US$ 6.645 and 8.246 billion; JPY120 = US$1) in Japan and are expected to increase with the increase of osteoporotic fractures. Osteoporosis screening and treatment for patients at high risk of fractures are expected to prevent osteoporotic fractures and to reduce the medical costs associated with fracture treatments [3]. An osteoporosis-screening program for Japanese women has been implemented as a health promotion project funded by the local government under the Health Promotion Act. However, the osteoporosis screening rate remains at a low level compared with those of other screening programs because the local government may not prioritize it because of budget constraints [4].

Previous economic evaluations from Germany and the USA have reported that osteoporosis screening and drug therapy are cost-effective for women under 65 years old [5–7]. Additionally, cost-effectiveness analyses from the USA, Canada, and Switzerland have concluded that osteoporosis screening and treatment for osteoporosis are recommended for women aged 65 years and older in terms of cost-effectiveness [8–10]. Although the cost-effectiveness of drug therapy for postmenopausal Japanese women with osteopenia was evaluated in a previous study of ours [11], the cost-effectiveness of osteoporosis screening and drug therapy for postmenopausal Japanese women has yet to be elucidated fully. Although the prevalence and incidence of vertebral fracture in Japan are higher than those in the USA and European countries [12, 13], the incidence rate of hip fracture, which may cause a greater burden than other osteoporotic fractures, is lower in Japan than in Western countries [14, 15]. In addition, Japan’s healthcare system is different from that of Western countries. Therefore, the results from previous economic evaluations cannot be directly generalized to the Japanese population. The purpose of this study was to estimate the cost-effectiveness of osteoporosis screening and drug therapy for osteoporosis in postmenopausal women with no history of fracture from the perspective of the Japanese healthcare system.

Materials and methods

Model development

An analysis was performed to evaluate the cost-effectiveness of screening and alendronate therapy for osteoporosis compared with no screening and no therapy from the perspective of the Japanese healthcare system. A patient-level, state transition model was developed to estimate the long-term costs and quality adjusted life years (QALYs) associated with each group (Fig. 1a, b). The modeled population was Japanese women who did not have a history of fragility fracture. We divided the population into six groups by age (50–54, 55–59, 60–64, 65–69, 70–74, and 75–79 years) and used those aged 65–69 years as the base case. At the start of the simulation, age and bone mineral density (BMD) of each individual were randomly determined from the probabilistic distributions as shown in Supplemental Table 1 [16].

Fig. 1
figure 1

Model structure. a Short-term model. b Long-term state transition model. Patients indicated by gray in the health state were assumed to receive lifetime alendronate therapy for secondary prevention of fracture

Full size image

In the current Japanese clinical guidelines, the drug treatment for osteoporotic fracture prevention is recommended if BMD is less than 70 % of the young adult mean (YAM) [17]. Therefore, we assumed that women in the screening group experienced BMD screening with dual energy X-ray absorptiometry (DXA) and received 5 years of alendronate therapy if their femoral neck BMD was less than 70 % of the YAM (Fig. 1a). In contrast, women with higher BMD (≥70 % of YAM) were assumed to receive no preventive therapy, but they are to be re-screened every 5 years (Fig. 1a). The value of YAM in Japanese women (0.787 g/cm2) was derived from an epidemiological survey in Japan [16]. In this model, 70 % of the YAM BMD was 0.551 g/cm2 and was equivalent to a T-score of −2.56, when using the reference mean of 0.858 g/cm2 for non-Hispanic, white women aged 20 to 29 years, derived from the National Health and Nutrition Examination Survey (NHANES) III data [18]. It was assumed that women in the no screening group do not receive osteoporosis screening or preventive drug therapy for primary fracture prevention (Fig. 1a).

The long-term model consisted of ten health states, namely, “no previous fracture,” “postfracture” (composed of seven states), “bed ridden,” and “death” as shown in Fig. 1b. In the simulation, individuals started in the state of no previous fracture and faced different risks of fracture depending on their age, femoral neck BMD, and the treatment received. The cycle length of the model was set at 1 year. During each cycle of the analysis, patients experienced one of the following clinical events: “no event”; “hip fracture”; “vertebral fracture”; “other fractures”; “hip and vertebral fracture”; “hip and other fractures”; “vertebral and other fractures”; “hip, vertebral, and other fractures”; and death. If patients experienced a fracture event, they moved to the postfracture state and had an increased risk of subsequent fractures. Once patients had a fragility fracture, they were assumed to receive continuous alendronate therapy with alendronate to prevent secondary fractures. We assumed that a certain proportion of patients that had a hip fracture moved to the bed ridden state. The model was developed and analyzed using TreeAge Pro 2014 Suite (TreeAge Software, Williamstown, MA, USA).

Transition probabilities

A transition probability (p) of an event occurring over a time interval (t) was calculated using an incidence rate (r) according to the following formula: p = 1 − exp(−rt) [19]. We developed equations for age and the femoral neck BMD-specific fracture rate by applying a series of methods previously proposed by De Laet et al. [20] to epidemiological data derived from published studies in Japan (see Supplemental materials). The equations for the age-dependent fracture rate for vertebral fracture (with or without a previous fracture), hip fracture, and other fractures (i.e., humeral fracture or wrist fracture) were constructed by using curve fitting on the basis of Japanese data [13, 21, 22]. The fit of the curve was determined by the Akaike information criterion (AIC) and clinical plausibility. An exponential curve was used to fit the age-dependent fracture rate for hip fracture and vertebral fracture. For other fractures, a sigmoid curve was used to model the age-dependent fracture rates. We then estimated age- and femoral neck BMD-specific fracture rates by combining the equations for age-dependent fracture rate, distribution of femoral neck BMD by age group, and the relative risks (RRs) of fracture per 1 SD reduction in BMD (see Supplemental Table 2) [13, 23].

Incident rates of subsequent fractures for those with a previous fracture were calculated by multiplying the age and BMD-specific fracture rate by the RR of subsequent fractures (see Supplemental Table 3) [24]. Fracture rates for those with additional BMD-independent clinical risk factors (i.e., a family history of hip fracture, high alcohol intake, and current smoking status) were calculated in the same manner by multiplying by the RR of each clinical risk factor and were used in the scenario sensitivity analyses (see Supplemental Table 3) [25–27]. The RRs of BMD-independent clinical risk factor were down adjusted for the prevalence of the risk factor in the general population by using following formula: \( \mathrm{Down}\kern0.3em \mathrm{djusted}\kern0.3em \mathrm{R}\mathrm{R}=\frac{RR}{RR\times \mathrm{prevalence}+\left(1-\mathrm{prevalence}\right)} \) [28]. The prevalence of each risk factor was derived from the published sources (see Supplemental Table 3) [26, 29]. In this study, we modeled clinical vertebral fractures and did not consider asymptomatic, morphometric vertebral fractures. The proportion of clinical fractures among all vertebral fractures was assumed to be 30 % (see Supplemental Table 3) [30]. In the model, the probabilities of multiple fracture events were calculated as the conditional probabilities by using the probability of each osteoporotic fracture.

An age-dependent mortality rate was obtained from the life table reported by the Ministry of Health, Labour, and Welfare in Japan [31]. The rate ratio between the mortality in women with hip fracture and the mortality in the general population by age group was estimated by using the data derived from previous study on mortality following hip fracture in Sweden (see Supplemental Table 4) [32]. The mortality rate for patients who experienced a hip fracture was calculated by multiplying the age-dependent mortality and the rate ratio of death following hip fracture (Supplemental Table 4; Supplemental Fig. 1). The probability of becoming bed-ridden after hip fracture was derived from a published source in Japan (see Supplemental Table 3) [33].

Clinical efficacy

In the screening and treatment group, women with a BMD less than 70 % of the YAM were assumed to receive 35 mg of alendronate once a week for 5 years to prevent a primary fracture. The RR of hip, vertebral, and other fractures was derived from a network meta-analysis of three randomized, controlled trials of alendronate therapy for postmenopausal women with osteopenia or osteoporosis conducted in countries other than Japan (Supplemental Table 5) [34]. We modeled residual effects of alendronate, assuming a linear decline in efficacy over 5 years, after 5 years of treatment (Supplemental Table 5) [35]. In addition, we assumed that patients who had experienced a fracture in each arm receive 5 years alendronate therapy with 5 years residual effect to prevent subsequent fractures. In this model, partial compliance with alendronate therapy was assumed by using a method previously reported [36]. Loss of efficacy due to adherence was assumed to 10 % for the base case scenario and considered using the following formula: efficacy in patients with partial adherence = (1 − [1 − RR of fracture with alendronate] × [100 % − percent reduction in efficacy] (Supplemental Table 5) .

Costs

In this analysis, we considered direct cost of medical and nursing care from the perspective of the Japanese healthcare system. Cost parameters and input values are summarized as shown in Supplemental Table 5. All costs were estimated in Japanese yen and converted to US dollars with a currency exchange rate of US$1 = JPY120. The screening cost was estimated on the basis of Japanese tariff, assuming the DXA scans measured BMD at the spine and hip. The annual drug cost of alendronate was calculated by using a Japanese drug price standard, assuming a dosage of 35 mg per week [37]. The annual medical costs for daily medical practice, including consultation, clinical laboratory testing, and radiography, was calculated using the Japanese tariff, assuming standard clinical practices [33, 38]. Treatment costs associated with an osteoporotic fracture event were obtained from previous studies in Japan [39–41]. In the simulation, treatment costs associated with multiple fracture events were calculated in additive manner. The annual cost of nursing care for a bedridden patient was estimated using the Nursing Care Insurance Scheme in Japan, assuming that the bedridden are unable to perform daily life activities without assistance (nursing care level 5 under the scheme) [42].

Utilities

Parameter inputs for utilities are summarized in Supplemental Table 5. In this analysis, we assumed that the utility value for the event-free health state decreased with age. The age-dependent utility function was developed on the basis of published data measured by EuroQol 5D (EQ-5D) for the general elderly Japanese population by using regression techniques [43]. The utilities of patients who experienced osteoporotic fracture events were calculated by multiplying the event-free utility by the disutility associated with various fracture events. The utilities for multiple fracture events were estimated by multiplying the event-free utility by the disutility associated with each fracture event. The disutility due to osteoporotic fractures for the first year and subsequent years was calculated on the basis of a prospective study in Japan [44]. For bedridden patients, we used the published EQ-5D utility values for Japanese patients provided with nursing care level 5 [45].

Base case analysis

For the base case analysis, we performed a first-order Monte Carlo simulation (individual simulation) using 1000 Japanese women aged 65–69 years with 1000 iterations to obtain the point estimate of lifetime costs and QALYs associated with screening strategy and no intervention. An annual discount rate of 3 % for both costs and QALYs was applied. The incremental cost-effectiveness ratio (ICER) was estimated by the following formula: ICER = (Costscreening − Costno screening) / (QALYscreening − QALYno screening). We also estimated the ICER in other age groups (50–54, 55–59, 60–64, 70–74, and 75–79 years). In this study, we used a willingness to pay threshold of $50,000 per QALY gained [46].

Sensitivity analysis

Deterministic sensitivity analyses were performed to assess the influence of various key parameters on the base case result in the Japanese women aged 65–69 years. The parameters assessed and their ranges are shown in Supplemental Tables 2, 3, and 5. The plausible ranges for each parameter were determined based on the basis of reported values, such as 95 % confidence intervals in published sources, or expert opinions. Probabilistic sensitivity analyses were performed with a second-order Monte Carlo simulation using 1000 iterations to examine the influence of parameter uncertainty on the cost-effectiveness of osteoporosis screening and treatment. Probabilistic distributions of parameters used in the probabilistic sensitivity analyses are shown in Supplemental Tables 3 and 5. In each of the 1000 iterations, the value for each model input was randomly selected from its distribution. From the results of the second-order Monte Carlo simulation, we constructed cost-effectiveness acceptability curves and estimated the proportion of iterations in which secondary prevention with alendronate would be preferred in terms of cost-effectiveness, assuming a willingness to pay of $50,000 per QALY gained. We also conducted scenario analyses using different combinations of BMD-independent clinical risk factors (a family history of hip fracture, high alcohol intake, and current smoking status).

Results

Base case analysis

The cost-effectiveness of osteoporosis screening and treatment compared with no screening are summarized in Table 1 for the base case. For the Japanese women aged 65–69 years, the screening strategy and subsequent alendronate therapy resulted in an additional lifetime cost of $1140 per person and conferred an additional 0.041 QALY, thus resulting in an incremental cost-effectiveness ratio of $27,697 per QALY gained. For other age groups (50–54, 55–59, 60–64, 70–74, and 75–79 years), the ICER was estimated to be $89,242, $64,010, $40,596, $17,027, and $9771 per QALY gained, respectively (Fig. 2). The ICER for the Japanese women aged 60 years and over was less than $50,000 per QALY gained.

Table 1 Base case results in 65- to 69-year-old women
Full size table
Fig. 2
figure 2

The cost-effectiveness of osteoporosis screening and treatment by different age group

Full size image

Sensitivity analysis

The results of the deterministic sensitivity analyses in patients aged 65–69 years are summarized in a tornado diagram, as shown in Fig. 3. The deterministic sensitivity analyses showed that the disutility due to vertebral fracture had a relatively strong effect on the estimated ICER ($20,491 to $42,843 per QALY gained). Additionally, the ICER was strongly influenced by other parameters such as annual drug cost of alendronate ($16,927 to $38,466 per QALY gained), discount rate ($18,623 to $34,959 per QALY gained), residual effect if alendronate ($27,697 to $43,758 per QALY gained), the RR of hip fracture by alendronate ($17,129 to $32,878 per QALY gained), and incidence rate of hip fracture ($24,137 to $31,999 per QALY gained). However, over the full range of model parameters, the ICERs remained less than $50,000 per QALY gained.

Fig. 3
figure 3

The results of deterministic sensitivity analyses for the base case

Full size image

Probabilistic sensitivity analyses were performed for patients aged 50–54, 55–59, 60–64, 65–69, 70–74, and 75–79 years. The cost-effectiveness acceptability curves, constructed on the basis of probabilistic sensitivity analyses, are shown in Fig. 4. Applying a willingness to pay threshold of $50,000 per QALY gained, the probability that the screening strategy became cost-effective compared with no intervention was estimated to be 44.5, 48.5, 50.9, 56.3, 59.1, and 64.7 % for those aged 50–54, 55–59, 60–64, 65–69, 70–74, and 75–79 years, respectively.

Fig. 4
figure 4

The cost-effectiveness acceptability curve by different age groups

Full size image

The results of the scenario sensitivity analysis for patients aged 65–69 years are shown in Fig. 5. The ICERs of screening and treatment in the Japanese women with different combinations of clinical risk factors were explored. In the women aged 60–64 years with one of three risk factors (current smoking status, high alcohol intake, or family history of hip fractures), the ICERs ranged from $17,323 to $23,128 per QALY gained. In patients aged 55–59 years with one of three risk factors, the ICERs of screening and treatment ranged from to $32,243 to $39,336 per QALY gained and became less than $50,000 per QALY gained.

Fig. 5
figure 5

The results of scenario analyses by different combinations of age and clinical risk factors

Full size image

Discussion

In this study, we evaluated the cost-effectiveness of osteoporosis screening and treatment for postmenopausal women from the perspective of the medical and nursing care system in Japan. Using a societal willingness to pay threshold of $50,000 per QALY gained, osteoporosis screening and treatment were cost-effective for women aged 60 years and over compared with no screening or treatment. These results were consistent with those of previous studies. A study from the USA has reported that screening and treatment for osteoporosis are recommended and cost-effective for women aged 65 years and older [10]. However, it should be noted that the study estimated the ICER to be $73,000 per QALY gained and had applied a willingness to pay threshold of $75,000 per QALY gained. A study from Canada has reported that a program promoting physical activity is the most cost-effective option for women 40 to 64 years old [8]. They also have concluded that BMD screening and pharmacological treatment might be considered a reasonable option for women aged 65 years and over (ICER $CAD 55,300/QALY), assuming a willingness to pay of $CAD 50,000/QALY [8]. A simulation study from Switzerland has concluded that population-based DXA screening, followed by an alendronate treatment, in osteoporosis patients is cost-effective for postmenopausal women aged 70 years and over (ICER CHF 28,170–35,412/QALY), using a willingness to pay of CHF 50,000/QALY [9]. Although the present study is different from previous studies conducted in Western countries in terms of the healthcare system, epidemiological characteristics, and the willingness to pay threshold, this study reported a lower cost-effective age for DXA screening and pharmacological treatment for osteoporosis in Japan compared with the results for Western countries.

In this study, we also explored the cost-effectiveness of osteoporosis screening and treatment for postmenopausal women who have an additional clinical risk factor. We found that screening and treatment became cost-effective for women aged 55–59 years with one of three risk factors (current smoking status, high alcohol intake, or family history of hip fractures). These findings indicated that a cost-effective screening program should be developed with considerations for not only the age but also the presence of other clinical risk factors.

The novelty of this study was in clarifying the cost-effectiveness of osteoporosis screening and treatment for Japanese women with different combinations of age and clinical risk factors. These findings support the development of an efficient osteoporosis screening program in Japan. Although the fracture risk assessment tool (FRAX) developed by the WHO is considered to be valid and reliable, its algorithm is not open to the public and was, thus, unavailable for our simulation model [47]. Therefore, we developed a risk equations for the age- and femoral neck BMD-specific incidence rate of fracture using epidemiological data from the Japanese population and combined them with a state transition model. In this study, major osteoporotic fractures, such as hip fracture, clinical vertebral fracture, proximal humeral fracture, and wrist fracture were modeled. Our simulation model predicted osteoporotic fractures over 10 years, and we compared them to those derived from the FRAX. The probabilities of hip fracture and major fracture in our model were almost similar to those of the FRAX (see Supplementary Figs. 2 and 3). These findings support the validity of our model in this economic evaluation.

Sensitivity analyses showed that the cost-effectiveness of osteoporosis screening strategy was sensitive to variations in several parameters, such as disutility due to vertebral fracture, drug cost, discount rate, residual effect of alendronate, the RR of hip fracture by alendronate, and incidence rate of hip fracture with previous fracture. These results suggested that further research on health-related QOL associated with osteoporotic fractures and long-term effectiveness of alendronate on the Japanese should be conducted. In this study, we applied an annual discount rate of 3 % for both costs and QALYs, according to Recommendations of the Panel on Cost-effectiveness in Health and Medicine [48]. However, if a decision maker were to use a higher value such as 5 %, our results might change. A sensitivity analysis for the price of alendronate indicated that the cost-effectiveness of screening and treatment might improve if the generic alendronate were used instead of the brand alendronate. On the basis of previous studies, we modeled residual effects of alendronate, assuming a linear decline in efficacy over 5 years, after 5 years of treatment [35]. However, if the residual effects were not assumed, osteoporosis screening and treatment become less preferable in terms of cost-effectiveness. Nonetheless, the ICERs remained lower than $50,000 per QALY gained over the full range of model parameters. Our results were robust to changes in key input parameters, including variables that vary from country to country. According to the results from the probabilistic sensitivity analyses, the probability that osteoporosis screening and treatment became cost-effective for women aged 60–64 years was estimated to be 50.9 %, assuming a societal willingness to pay of $50,000 per QALY gained. This result indicated that the influence of parameter uncertainty on the base case result is not necessarily small.

The main limitation of this study was the uncertainty of parameters used in our model. In particular, parameter estimations related with disutility due to vertebral fracture, drug efficacy, and incidence rate of osteoporotic fracture had a significant influence on the cost-effectiveness of osteoporosis and treatment strategy. Therefore, further clinical studies on the health-related QOL of osteoporotic fracture, the risk of osteoporotic fracture and the long-term effectiveness of alendronate in the Japanese population are needed to elucidate the cost-effectiveness of osteoporosis screening and treatment in Japan. Secondly, our model did not consider excess mortality for vertebral or other fractures. Although the excess mortality for hip fracture has been reported in the previous epidemiological study in Japan [49], there is no good evidence to suggest other osteoporotic fractures than hip fracture affect mortality in Japanese patients. Therefore, we assumed that other fractures than hip fracture did not affect mortality in our model. Our assumption may lead to underestimation of the cost-effectiveness of osteoporosis screening in Japan. Finally, the cost-effectiveness of alendronate in osteopenic women varies depending on the willingness to pay thresholds for each additional QALY gained. Although the willingness to pay threshold is commonly used in the USA ($50,000 per QALY) and in the UK (£20,000 to £30,000 per QALY), it varies among countries [46]. Therefore, the cost-effectiveness of osteoporosis screening may vary depending on the acceptability level of the ICER in Japan.

In conclusion, DXA screening and alendronate therapy for osteoporosis would be cost-effective for postmenopausal Japanese women over 60 years and those aged 55 to 59 years with at least one clinical risk factor. In terms of cost-effectiveness, the target population for osteoporosis screening should be determined after considering age and clinical risk factors.