Abstract
Summary
This prospective study aimed to determine the risk factors and the 10-year probability of osteoporotic fracture in Southern Chinese men. The findings show substantial population differences in fracture incidence and risk prediction compared to the FRAXTM model, and the addition of BMD information to clinical risk factor assessment improved fracture risk prediction in Chinese men.
Introduction
Clinical risk factors with or without bone mineral density (BMD) measurements are increasingly recognized as reliable predictors of fracture risk. Prospective data on fracture incidence in Asian men remain sparse. This prospective study aimed to determine the risk factors and the 10-year absolute fracture risk in Southern Chinese men.
Methods
This is a part of the Hong Kong Osteoporosis Study. One thousand eight hundred ten (1,810) community-dwelling, treatment-naive men aged 50 years or above were evaluated. Baseline demographic characteristics, clinical risk factors and BMD were recorded. Ten-year risk of osteoporotic fracture was calculated using Cox proportional hazards models.
Results
The mean age of subjects was 68.0 ± 10.3 years. After a mean follow-up period of 3.5±2.9 years (range 1 to 14 years), 37 incident low-trauma fractures were recorded. The incidence for all osteoporotic fractures and hip fractures was 635/100,000 and 123/100,000 person-years, respectively. The most significant predictors of osteoporotic fracture were history of fall (RR 14.5), femoral neck BMD T-score < −2.5 (RR 13.8) and history of fracture (RR 4.4). Each SD reduction in BMD was associated with a 1.8 to 2.6-fold increase in fracture risk. Subjects with seven clinical risk factors and BMD T-score of −1 had an absolute 10-year risk of osteoporotic fracture of 8.9%, but this increased to 22.7% if they also had a femoral neck BMD T-score of −2.5.
Conclusions
These findings show substantial population differences in fracture incidence and risk prediction. The addition of BMD information to clinical risk factor assessment improved fracture risk prediction in Chinese men.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
Introduction
Osteoporosis has become a major public health concern in the past decade and the burden placed on the community and health care agencies is expected to rise with the aging global population. Global epidemiological data indicate that Asia will carry the greatest burden of osteoporotic fractures over the coming decades. Although it is well documented that the risk and incidence of fractures vary widely between populations [1], the absolute rate of fractures among Asian men remains unclear.
A recent update of the worldwide prevalence of osteoporotic fractures using data from published sources reveals that of the annual incidence of nine million fractures, 39% occur in men [2]. Although men suffer fewer fractures than women, they have a significantly higher morbidity and mortality [2]. It is projected that by 2050, 50% of hip fractures will occur in Asia, with the majority occurring in China [1]. With the growing size of the aging population in Asia and increasing longevity, osteoporosis in men will soon become a significant burden on society and healthcare systems in Asia. Epidemiological studies suggest that the increased incidence of fractures among developing countries may be related to urbanization and altered lifestyle, although evidence to support this is scanty.
Identification of subjects at risk of fracture to enable early treatment in developing countries with limited resources remains a challenge. There is increasing evidence that bone mineral density (BMD) alone is inadequate to detect all individuals at risk of osteoporotic fractures, and factors other than BMD are important for predicting future fracture risk [3, 4]. The World Health Organization (WHO) FRAX™ algorithm for fracture risk assessment utilizes a set of clinical risk factors with or without BMD information to predict the 10-year absolute fracture risk in different populations, including Chinese in mainland China and Hong Kong [5]. Interestingly, the clinical risk factors included in FRAX differ to those identified from prospective population studies. In our previous study of Chinese postmenopausal women [5], we identified eight clinical risk factors that contribute to increasing fracture risk, including the use of walking aids; history of one or more falls in 12 months; being housebound; dietary calcium intake < 400 mg/day; age > 65 years; previous fracture; body mass index (BMI) < 19 kg/cm2; and physical activity < 30 min/day. These findings suggest that population-specific characteristics may need to be taken into consideration when evaluating fracture risk; e.g., other than the common risk factors such as age, BMI and BMD, the Dubbo Osteoporosis Epidemiology Study of Australia took into account of quadriceps strength, body sway, and thiazide use [6]. The QFractureScores algorithm developed for Caucasian population in the UK includes concomitant diseases and medication use as major risk factors for fracture prediction [7].
Although a number of cross-sectional studies and population studies have demonstrated lower BMD values and fracture incidence in Asian men compared with Caucasian men, information on fracture outcome derived from prospective studies in Asian male cohorts is scarce. The objective of this prospective study was to report the incidence of osteoporotic fracture in Southern Chinese men, to evaluate the clinical risk factors associated with fracture risk, and to compare the model build on these population-specific risk factors and the WHO FRAX risk calculator in fracture prediction.
Methods
Study population and design
This was a part of the prospective population-based Hong Kong Osteoporosis Study in which community-dwelling ambulatory Southern Chinese men aged 50 years or above were recruited from different districts of Hong Kong between 1995 and 2009 during health fairs and road shows on osteoporosis. Subjects already prescribed osteoporosis treatment were excluded. All participants were invited to the Osteoporosis Centre at Queen Mary Hospital for evaluation of bone health. X-rays of the thoracolumbar spine were obtained at baseline to identify the presence of morphometric vertebral fracture using Genant's semiquantitative assessment method [8].
Baseline demographic data and information on clinical risk factors were collected including anthropometric measurements, socioeconomic status, education level, low-trauma fracture history after the age of 45 years (both personal and family), history of fall, and medical history including current medication, history of low back pain, prior use of glucocorticoids, and secondary causes of osteoporosis. Information on lifestyle habits including smoking, alcohol consumption, and physical activity were also obtained. Dietary intake of calcium was determined using a semiquantitative food frequency questionnaire [5]. These data were collected from interviews conducted by trained research assistants using a structured questionnaire.
All subjects were followed annually via telephone interviews using a structured questionnaire for assessment of the clinical outcome of incident fractures, falls, hospitalization, living status and functional status. Medical history and incident fractures were verified with the computerized patient information system of the Hospital Authority of the Hong Kong Government. Fractures of the skull, fingers and toes, as well as traumatic fractures were excluded from analysis. Subjects who commenced anti-osteoporosis medication prior to the occurrence of a primary fracture were also excluded. The study was approved by the Institutional Review Board of the University of Hong Kong and the Hong Kong West Clusters Hospital of the Hospital Authority.
BMD evaluation
BMD was assessed at the L1–4 lumbar spine, femoral neck, and total hip using the same dual-energy X-ray absorptiometry machine (Hologic QDR 4500, Waltham, Mass., USA). BMD T-scores were determined according to the local Southern Chinese normative database [9]. The in vivo precision of BMD at the lumbar spine, femoral neck, and total hip was 0.8%, 0.9% and 0.7%, respectively. All DXA measurements were performed by two licensed technologists who had completed training by the equipment manufacturers and were accredited by the International Society for Clinical Densitometry. The least significant change for lumbar spine, femoral neck, and total hip was 2.41%, 3.82% and 2.62%, respectively. BMD was expressed both as an absolute value in gram per square centimeter and T-score.
Statistical methods
The Cox proportional hazards models were used to identify potential independent risk factors for osteoporotic fracture. Time to all incident fractures was calculated according to the date of X-ray reports or physician's consultations when diagnosis was made. Results were reported as relative risks (RR) with 95% confidence intervals (CI). The significance level was set at p < 0.05. The risk of osteoporotic fracture was optimally expressed as a fixed-term absolute risk, that is, the probability of fracture over a given period of time. Predicted 10-year fracture risk adjusted by competing risk of death [10], as well as the relationship between fracture risk and age, BMD T-score and number of risk factor were identified using one minus Kaplan–Meier survival functions. Individual 10-year risk of major osteoporotic fracture was also obtained from the FRAX for Hong Kong website (http://www.shef.ac.uk/FRAX/) for comparison. Receiver operative characteristic curve (ROC) analysis was used to determine the predictive value of ethnic-specific clinical risk factors with or without BMD and FRAX. All statistical analyses were conducted using SPSS for Windows version 15.0 (SPSS, Chicago, IL, USA) and R for Windows version 2.11.1 (R Development Core Team, Auckland, New Zealand) statistical software.
Results
One thousand eight hundred and ten subjects were included in this analysis. The average follow-up period was 3.5±2.9 years (range 1 to 14 years), with a total follow-up of 5,669 patient-years. The mean age at baseline was 68.0 ± 10.3 years (range 50 to 99 years); 33.6% were aged from 45 to 64 years and 63.4% were aged 65 years or older. One hundred and seventy subjects had died at the time of analysis and 19 patients received anti-osteoporosis medication after sustaining a fracture during the follow-up period. The data for these subjects were analyzed up to their last contact time-point or time of treatment initiation. Baseline demographic information and BMD characteristics of the subjects are shown in Table 1. 7.2% of all subjects had osteoporosis with a BMD T-score ≤ −2.5 at any one site. 9.1% of men aged 65 years or above were osteoporotic compared with 4.1% in the 50- to 64-year age group. Prevalence of osteopenia (BMD T-score between −1 and −2.5 at any one site) was 47.2% in men above age 65 years and 39.2% in men aged 50 to 64 years: combined, this accounted for 44.1% of all subjects.
During the follow-up period, 37 new low-trauma fractures were reported, of which seven (22%) were clinical vertebral fractures, seven (22%) were hip fractures, two (6%) were proximal humerus fractures; nine (25%) were distal forearm fractures; and 12 (33%) were fractures at other peripheral sites. The incidence for all low-trauma fractures was 635/100,000 person-years, for non-spine fractures was 528/100,000 person-years, and for hip fracture was 123/100,000 person-years. For subjects aged 65 years and above, the incidence for all fractures was 839/100,000 person-years, for non-spine fractures was 769/100,000 person-years and for hip fracture was 201/100,000 person-years.
Predicted 10-year osteoporotic fracture risk from risk factor assessment
In multivariate Cox regression analysis, seven independent clinical risks factors associated with increased risk of osteoporotic fracture were identified (Table 2). In terms of individual risk factors, history of fall and history of fragility fracture were associated with the highest predicted 10-year risk of fracture: the relative risk was 14.5 (95% CI 6.5–32.2) and 4.4 (95% CI 2.0–9.4), respectively. Other risk factors listed in decreasing order of impact on fracture risk were: outdoor activity < 60 min/day, BMI < 20 kg/m2, difficulty bending forward, use of walking aid, and age ≥ 65 years (p value < 0.05, Table 2). Although a 10-year increase in age accounted for only a 5.8% increase in 10-year osteoporotic fracture risk, older men aged 65 years or above had a 2.7-fold higher risk of fracture compared with younger men. Figure 1 shows the fracture risk in different age groups that was adjusted for competing risk of death along the study period. The interaction of age and other risk factors is shown in Fig. 2a. The combination of older age and history of fall was associated with a twofold increase in 10-year fracture risk after adjusting for competing death risk. Men aged 65 years or older with one or more falls per year had a 10-year risk of fracture of 31.9% compared with 15.8% for those younger than 65 years old.
Predicted 10-year osteoporotic fracture risk from BMD and number of risk factors
While 48% of all incidence fractures occurred in subjects in whom BMD fell in the osteopenic range, only 26% of fracture cases occurred in osteoporotic subjects. Aside from history of fall, low BMD at the femoral neck (T-score ≤ −2.5) had the second highest impact on fracture risk in men (RR = 13.8), and each SD reduction in BMD at the lumbar spine, femoral neck or total hip was associated with a 1.8 to 2.6-fold increase in osteoporotic fracture risk (Table 2). The addition of hip BMD information to risk factor assessment improves osteoporotic fracture risk prediction. Regardless of the risk factor studied, subjects with femoral neck BMD T-score ≤ −2.5 had a 1.7 to 7.8-fold increase in 10-year fracture risk prediction (Fig. 2b). Figure 3 shows the 10-year absolute risk of osteoporotic fracture according to age and femoral neck BMD T-score. The interaction between BMD and age on fracture risk was similar to that observed in women [5], i.e., at the same BMD T-score, older men had a much higher fracture risk than younger men.
It has been reported that approximately half of the fractures in Caucasian men are idiopathic. Nonetheless in this Chinese cohort, all subjects with incident fractures had at least one clinical risk factor other than low BMD. The risk of major osteoporotic fractures (i.e., clinical spine, hip, proximal humerus and distal radius) increased with increasing number of risk factors and decreasing femoral neck BMD T-score. At low T-scores, fracture probability increased markedly with more risk factors. For example, at a femoral neck BMD T-score of −3, the 10-year probability of major osteoporotic fracture increased from 1.9% in those with no risk factor to 31% in those with seven risk factors, (Fig. 4).
ROC analysis showed that clinical risk factors plus BMD information offers better predictive power than clinical risk factors alone in predicting 10-year probability of fracture (area under the curve 0.82 ± 0.04 vs. 0.74 ± 0.04, p < 0.001). We noted that although the percentage of subjects with low BMD was small, those with co-existent multiple risk factors had a very high risk of fracture. For example, 35% of men with a femoral neck BMD T-score ≤ −2.5 had five to seven clinical risk factors and their absolute 10-year fracture risk was 27.6%; 15.9% of men aged 65 years and above had one or more falls per year and their 10-year risk of fracture was 31.9%.
The model based on this prospective study with adjustment of competing death risk was compared to the WHO FRAX risk calculator for Hong Kong with the inclusion of BMD information. The 10-year risk for major osteoporotic fractures differed markedly with these two models: men with seven clinical risks have an absolute 10-year fracture risk of 17.6% with the present model, but the predicted fracture risk by FRAX is only 11% (Fig. 5). Contrary to this, an individual with no risk has only 0.7% risk based on the present model while the predicted risk by FRAX is 2.3%. ROC analysis also showed our model with BMD has significant improvement on overall predictive power compared to FRAX with BMD (area under the curve 0.87 ± 0.02 vs. 0.72 ± 0.05, p < 0.001).
Discussion
This prospective study provided evidence that Asian men have a low fracture rate compared to their Caucasian counterparts. Nevertheless, Hong Kong is a city that has undergone urbanization and the population-based age-adjusted fracture rate in men has doubled between 1966 and 1995 [11]. It was projected that a change in lifestyle associated with urbanization may increase the fracture rate in Orientals to the higher levels seen in Western populations. This prospective study revealed that the fracture rate in Southern Chinese men in Hong Kong remains low. The non-spine fracture incidence for Chinese men aged 65 years or above was 769/100,000 person-years as compared to 1,146/100,000 person-years reported in the MrOS study which involved predominantly US Caucasians [12].
Although this study used a convenient rather than a random community sample and might have biased towards recruiting healthier subjects, the prevalence of osteoporosis at the femoral neck of 3.2% in this cohort was similar to the 2.0% reported in Caucasian White subjects in the population-based NHANES 2005–2006 survey in the US [13]. The prevalence of osteoporosis at the spine and hip were similar in this cohort. Similar to other populations, fractures of the hip, forearm, vertebrae, and humerus were among the most frequent sites of incident fractures in men. In comparison with postmenopausal women in the same population [5], the absolute fracture incidence was lower in men. The reason for this difference in the US population was postulated to be related to an increased frequency of falls in women [14, 15], and fracture risk after a fall was 2.2 times higher in women than men [16]. The relation of fracture risk after a fall in the two sexes was nonetheless reversed in Chinese. Although falls were recorded more often in women [5], the relative risk of fracture in subjects with one or more falls in 12 months was 14.5 for Chinese men and 4.0 for Chinese women.
This study also identified the clinical risk factors for fracture in Chinese men and the interaction between risk factors and BMD. These risk factors partly overlap with those reported for Caucasian population of the MrOs study which are the use of tricyclic antidepressant, history of fracture, inability to complete a narrow walk trial, falls in previous year, age ≥ 80 years, depressed mood and decreased total hip BMD [12]. The risk factors for Chinese men are also slightly different from those identified by the Dubbo study which includes increasing age, decreased femoral neck BMD, quadriceps strength, body sway, previous falls, previous fractures, weight, height, alcohol use, physical activity index and thiazide use [6]. Similar to previous observations of other ethnic groups [17, 18], each SD reduction in BMD T-score is associated with a 1.8 to 2.6-fold increased risk of osteoporotic fractures in Chinese men. The relative risk prediction for osteoporotic fracture was better with BMD measurement at the hip than the spine: this concurs with the findings in Caucasian populations [6, 19]. However, subjects with a femoral neck BMD T-score < −2.5 had a 13.8-fold increased risk of fracture.
The WHO FRAX model utilizes ten clinical risk factors with or without BMD for fracture risk prediction. In areas where BMD measurements are not available, WHO proposes to use BMI to replace BMD as it provides a similar risk profile for fracture prediction. Interestingly, our data revealed that addition of BMD information to clinical risk factors enhanced fracture prediction in this male cohort. This observation concurs with other US Caucasian male studies [20]. Hence identifying subjects with low BMD and multiple risk factors, and targeting them for preventive measures is likely to be a cost-effective approach. With the exception of age, BMI, and previous fractures, the clinical risk factors identified in this present study differ significantly to those included in the FRAX model. The latter shows that risk factors for fracture and fracture risk prediction likely vary between different ethnic groups. The FRAX model also does not take into account the impacts on fracture risk of history of fall, physical ability and mobility, which are important risk factors for fracture as shown in this and other studies [6, 7, 21]. Our model using ethnic-specific risk factors and incorporated fall risk had a significantly better predictive power when compared to FRAX. It would also be interesting to compare other population-specific models such as the Dubbo Study and MrOs Study which have also incorporated history of fall and physical activity as risk factors. It is also likely that FRAX underestimates the risk for osteoporotic fractures, especially vertebral fractures in Asian populations. Although the risk of hip fractures is much lower in Chinese than in Caucasians, the risk of vertebral fractures is similar between the two ethnic groups [22, 23]. There has been a concern that a model that presumes a ratio of vertebral fractures to non-vertebral fractures in a Swedish population might underestimate the risk of vertebral fractures in Asians.
This study has some limitations. The sample size and the number of fractures recorded were small and this study may have underestimated the absolute risk in the general population. Although it is of our interest to compare the 10-year hip fracture risk of our model with the risk predicted by FRAX, such analysis was limited by the low incidence of hip fractures in our sample and we could only compare the two models on prediction of major osteoporotic fractures. The low recruitment rate also reflected the lack of interest of Asian men in health-related activities. Spine X-rays were not obtained in all patients during follow-up, thus the incidence of morphometric spine fractures was not included in the analysis. All participants received a clear explanation of their BMD report and were educated about the importance of risk prevention in osteoporosis. The consequences of this intervention were not quantified. Thus, the actual risk of the general male population in Hong Kong that has not received any advice about osteoporosis prevention or been informed about BMD status is likely to be greater than that reported for the study population. As with other studies, the 10-year fracture risk of this study was predicted using the Cox proportional hazard model based on results generated from a mean follow-up period of 3.5 years. Actual 10-year follow-up information for every subject was not available. As fracture risk increases with age and frailty, the predicted model may well underestimate the actual fracture risk of the population.
In conclusion, this prospective study has determined the incidence of osteoporotic fracture and hip fracture in Southern Chinese men and identified the major clinical risk factors associated with fracture risk. These data highlight the importance of ethnic/population-specific characteristics in better discrimination of individuals at high risk of fracture and targeting of intervention.
References
Cooper C, Melton LJ (1992) Hip fractures in the elderly: a world-wide projection. Osteoporos Int 2:285–289
Johnell O, Kanis JA (2006) An estimate of the worldwide prevalence and disability associated with osteoporotic fractures. Osteoporos Int 17:1726–1733
Siris ES, Miller PD, Barrett-Connor E, Faulkner KG, Wehren LE, Abbott TA, Berger ML, Santora AC, Sherwood LM (2001) Identification and fracture outcomes of undiagnosed low bone mineral density in postmenopausal women. JAMA 286:2815–2822
Kanis JA, on behalf of the World Health Organization Scientific Group (2008) Assessment of osteoporosis at the primary healthcare level. Technical report. WHO Collaborating Centre, University of Sheffield, UK
Kung AW, Lee KK, Ho AY, Tang G, Luk KD (2007) Ten-year risk of osteoporotic fractures in postmenopausal Chinese women according to clinical risk factors and BMD T-Scores: a prospective study. J Bone Miner Res 22:1080–1087
Nguyen TV, Eisman JA, Kelly PJ, Sambrook PN (1996) Risk factors for osteoporotic fractures in elderly men. Am J Epidemiol 144:255–263
Hippisley-Cox J, Coupland C (2009) Predicting risk of osteoporotic fracture in men and women in England and Wales: prospective derivation and validation of QFractureScores. BMJ 339:b4229
Grigoryan M, Guermazi A, Roemer FW, Delmas PD, Genant HK (2003) Recognizing and reporting osteoporotic vertebral fractures. Eur Spine J 12(suppl 2):S104–S112
Kung AWC, Luk KDK, Chu LW, Tang GWK (1999) Quantitative ultrasound and symptomatic vertebral fracture risk in Chinese women. Osteoporos Int 10:456–461
Scrucca L, Santucci AF (2007) Competing risk analysis using R: an easy guide for clinicans. Bone Marrow Transplant 40:381–387
Lau EM (2001) Epidemiology of osteoporosis. Best Pract Res Clin Rheumatol 15(3):335–344
Lewis CE, Ewing SK, Taylor BC, Shikany JM, Fink HA, Ensrud KE, Barrett-Connor E, Cummings SR, Orwoll E (2007) Predictors of non-spine fracture in elderly men: the MrOS study. J Bone Miner Res 22:211–219
Looker AC, Melton LJ 3rd, Harris TB, Borrud LG, Shepherd JA (2010) Prevalence and trends in low femur bone density among older US adults: NHANES 2005–2006 compared with NHANES III. J Bone Miner Res 25(1):64–71
Sattin RW, Lambert Huber DA, De Vito CA, Rodriguez JG, Ros A, Bacchelli S, Stevens JA, Waxweiler RJ (1990) The incidence of fall injury events among the elderly in a defined population. Am J Epidemiol 131:1028–1037
Winner SJ, Morgan CA, Evans JG (1989) Perimenopausal risk of falling and incidence of distal forearm fracture. BMJ 298:1486–1488
Stevens JA, Sogolow ED (2005) Gender differences for non-fatal unintentional fall related injuries among older adults. Inj Prev 11:115–119
Marshall D, Johnell O, Wedel H (1996) Meta-analysis of how well measures of bone mineral density predicts occurrence of osteoporotic fractures. BMJ 312:1254–1259
Cumming SR, Cawthon PM, Ensrud KE, Cauley JA, Fink HA, Orwoll ES (2006) BMD and risk of hip and nonvertebral fractures in older men: a prospective study and comparison with older women. J Bone Miner Res 21:1550–1556
Cummings SR, Cawthon PM, Ensrud KE, Cauley JA, Fink HA, Orwoll ES (2006) BMD and risk of hip and nonvertebral fractures in older men: a prospective study and comparison with older women. J Bone Miner Res 21:1550–1556
Khosla S, Amin S, Orwoll E (2008) Osteoporosis in men. Endocr Rev 29(4):441–464
Mackey DC, Eby JG, Harris F, Taaffe DR, Cauley JA, Tylavsky FA, Harris TB, Lang TF, Cummings SR (2007) Prediction of clinical non-spine fractures in older black and white men and women with volumetric BMD of the spine and areal BMD of the hip: the health, aging, and body composition study. J Bone Miner Res 22:1862–1868
Lau EM, Chan HH, Woo J, Lin F, Black D, Nevitt M, Leung PC (1996) Normal ranges for vertebral height ratios and prevalence of vertebral fracture in Hong Kong Chinese: a comparison with American Caucasions. J Bone Miner Res 11(9):1364–1368
Kung AW (2004) Epidemiology and diagnostic approaches to vertebral fractures in Asia. J Bone Miner Metab 22:170–175
Acknowledgments
This study was supported by the Bone Health Fund of the Hong Kong University Foundation and Osteoporosis Research Fund of the University of Hong Kong.
Conflicts of Interest
None.
Open Access
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
Open Access This is an open access article distributed under the terms of the Creative Commons Attribution Noncommercial License (https://creativecommons.org/licenses/by-nc/2.0), which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
About this article
Cite this article
Bow, C.H., Tsang, S.W.Y., Loong, C.H.N. et al. Bone mineral density enhances use of clinical risk factors in predicting ten-year risk of osteoporotic fractures in Chinese men: the Hong Kong Osteoporosis Study. Osteoporos Int 22, 2799–2807 (2011). https://doi.org/10.1007/s00198-010-1490-0
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00198-010-1490-0