Fracture Risk Assessment (FRAX) tools are calibrated from country-specific fracture epidemiology. Although hip fracture data are usually available, data on non-hip fractures for most countries are often lacking. In such cases, rates are often estimated by assuming similar non-hip to hip fracture ratios from historical (1987 to 1996) Swedish data. Evidence that countries share similar fracture ratios is limited. Using data from Manitoba, Canada (2000 to 2007, population 1.2 million), we identified 21,850 incident major osteoporotic fractures (MOF) in men and women aged >50 years. Population-based age- and sex-specific ratios of clinical vertebral, forearm, and humerus fractures to hip fractures were calculated, along with odds ratios (ORs) and 95% confidence intervals (CIs). All ratios showed decreasing trends with increasing age for both men and women. Men and women showed similar vertebral/hip fracture ratios (all p > 0.1, with ORs 0.86 to 1.25). Forearm/hip and humerus/hip fracture ratios were significantly lower among men than women (forearm/hip ratio: p < 0.01 for all age groups, with ORs 0.29 to 0.53; humerus/hip ratio: p < 0.05 for all age groups [except 80 to 84 years] with ORs 0.46 to 0.86). Ratios for any MOF/hip fracture were also significantly lower among men than women in all but two subgroups (p < 0.05 for all age groups [except 80 to 84 and 90+ years] with ORs 0.48 to 0.87). Swedish vertebral/hip fracture ratios were similar to the Canadian fracture ratios (within 7%) but significantly lower for other sites (men and women: 46% and 35% lower for forearm/hip ratios, 19% and 15% lower for humerus/hip ratios, and 19% and 23% lower for any MOF/hip ratios). These differences have implications for updating and calibrating FRAX tools, fracture risk estimation, and intervention rates. Moreover, wherever possible, it is important that countries try to collect accurate non-hip fracture data. © 2014 American Society for Bone and Mineral Research
Osteoporosis is common and accounts for approximately 9 million new fractures each year worldwide. The Fracture Risk Assessment tool (FRAX), which was developed by the World Health Organization Collaborating Centre for Metabolic Bone Disease, provides individual 10-year probabilities for hip fracture and any major osteoporotic fracture (MOF), defined as a clinical vertebral, forearm, humerus, or hip fracture. FRAX integrates multiple clinical risk factors for fracture and the competing risk of death with an optional bone mineral density (BMD) measurement from the femoral neck.
FRAX tools are calibrated for different countries based on regional fracture and mortality epidemiology. Calibrating FRAX models requires, at a minimum, age- and sex-specific annual hip fracture incidence rates and all-cause mortality rates. Of the MOFs, hip fracture data are the easiest to accurately capture given that virtually all patients with hip fractures undergo hospital admission. Ideally, age- and sex-specific data for non-hip MOFs should also be included in the development of country-specific models. However, because hospitalization or surgery is usually not required after these fractures, accurate data on non-hip fractures are much more difficult to collect. Ettinger and colleagues described many challenges in obtaining non-hip fracture incidence rates in order to update the United States FRAX tool. Where non-hip fracture information is lacking or incomplete, fracture rates are often estimated by assuming a similar ratio of non-hip to hip fractures as reported in Sweden using 1987 to 1996 data. Specifically, ratios were calculated from hip fracture rates observed in Sweden in 1996 and non-hip fracture rates observed in Malmo, Sweden (humerus fractures from 1987, vertebral fractures from 1993 to 1994, forearm fractures from 1995). Currently, these Swedish ratios are considered the reference standard for estimating non-hip fracture rates. However, evidence to support the assumption that non-hip/hip ratios are similar between countries is limited. Moreover, secular changes in osteoporotic fracture rates have been reported in many regions worldwide. For example, there is consistent evidence of declining hip fractures in North America and some other developed countries including Sweden since the Swedish fracture data were collected.[7, 8]
The aim of our study was to examine Canadian fracture ratios using population-based data from 2000 to 2007 and to compare these with the Swedish reference standard.
Materials and Methods
Population-based data for this study are from the Canadian province of Manitoba, which like other Canadian provinces provides universal health-care coverage to its residents. Osteoporotic fracture rates in Manitoba are known to be representative of the rest of Canada. A systematic review of hip fracture rates worldwide placed Canada in the moderate risk category and similar to the United States. Fracture data were obtained from the Manitoba Population Health Research Data Repository, which contains health-care information on nearly all residents of Manitoba, Canada (population 1.2 million in 2006). The Repository contains information on a patient's demographics, date and type of service, and diagnoses. Diagnoses in hospital data are recorded using the International Classification of Diseases, 9th revision, Clinical Modification (ICD-9-CM), up to and including the 2003–04 fiscal year and the International Classification of Diseases, 10th revision, Canadian version (ICD-10-CA), for subsequent fiscal years. Physicians submit fee-for-service billing claims to the provincial health ministry; these claims capture almost all outpatient services, including those provided in hospital emergency and outpatient departments. An encrypted personal identifier allows for anonymous linkage across data sets and creation of person-specific longitudinal records of health service utilization. Ethics approval for this research was received from the Manitoba Health Research Ethics Board and permission to access the study data was provided by the Manitoba Health Information Privacy Committee.
Diagnosis codes in the Manitoba hospital and physician database have been validated and used extensively in epidemiologic research.[12-14] Hospital and physician claims data were used to develop case definitions for each MOF; these definitions were shown to provide fracture incidence rates similar to the population-based Canadian Multicentre Osteoporosis Study (CaMos) and were feasible for national osteoporosis surveillance and monitoring.[15, 16] Briefly, an incident hip fracture was identified if the individual had a corresponding hospitalization where hip fracture was coded as the most responsible diagnosis (ICD-9-CM 820.x or ICD-10-CA S72.0–.2). At least one hospitalization or two physician visits within 3 months with the relevant diagnosis codes was required for forearm fracture (ICD-9-CM 813.x or ICD-10-CA S52.x) or humerus fracture (ICD-9-CM 812.x or ICD-10-CA S42.2), provided this was preceded by a 6-month period without any codes for the same diagnosis to avoid double counting the same fracture episode. For clinical vertebral fractures, one hospitalization or one physician visit (ICD-9-CM 805.x, ICD-10-CA S22.0–.1, S32.0, or S32.7–.8) was required preceded by a 6-month period without any codes for the same diagnosis. Traumatic fractures, identified from external injury codes, were excluded.
The number of incident fractures was tabulated for all individuals 50 years of age and older based upon age at the start of each fiscal year from 2000–01 to 2006–07 and stratified by sex and age (5-year age strata from 50 to 54 years to 90+ years). Individuals suffering multiple fractures of the same type in the same year were only counted once. Few individuals sustained multiple fractures of different types in the same year (<0.5%), and therefore no adjustment was made for this occurrence in tabulating any MOF. Ratios of MOF to hip fractures with 95% confidence intervals (CIs) were calculated. Fracture ratios in men and women were compared as age-stratified odds ratios (ORs) with 95% CIs assuming an asymptotic normal distribution. Similarly defined Swedish fracture ratios (5-year age strata from 50 to 54 years to 85 to 89 years) were derived from published data (Swedish National Database and Malmo, Sweden). Canadian ratios for men and women were fitted with nonlinear continuous power functions and were compared with similarly fitted functions for the Swedish fracture ratios. SAS software (version 9.1; SAS Institute, Cary, NC, USA) was used to perform all data linkage, extraction, and tabulation. Statistical significance was evaluated using α = 0.05.
The number of MOFs for men and women are shown in Table 1. During the 7-year observation period, 21,850 MOFs were identified for men/women as follows: 1611/3285 clinical vertebral, 1425/5673 forearm, 844/2699 humerus, 1698/4615 hip. Forearm fractures were the most common osteoporotic fractures in younger age groups, accounting for 48% and 65% of fractures in men and women aged 50 to 54 years, respectively (Fig. 1). However, with increasing age, they only accounted for a minority of MOFs, making up 10% of fractures in men and 17% in women aged 90+ years. The opposite pattern was observed with hip fractures, which accounted for only 7% and 3% of MOFs in men and women aged 50 to 54 years, respectively, whereas hip fractures made up the majority of fractures in men and women aged 90 + , with percentages of 56% and 49%, respectively. Similar patterns were observed in the published Swedish data.
|Age subgroup (years)||Vertebral||Forearm||Humerus||Hip|
All Canadian ratios decreased with increasing age in both men and women. In the youngest subgroup (aged 50 to 54 years), the vertebral/hip ratio was 3.11 (95% CI 2.33–4.16) in men and 3.61 (95% CI 2.55–5.10) in women, and in the oldest subgroup (aged 90+ years) the ratios were 0.36 (95% CI 0.29–0.45) in men and 0.36 (95% CI 0.32–0.40) in women (Fig. 2). The forearm/hip ratios for those aged 50 to 54 years were 5.15 (95% CI 3.91–6.77) for men and 14.76 (95% CI 10.75–20.25) for women compared with a ratio of 0.15 (95% CI 0.11–0.20) and 0.30 (95% CI 0.26–0.34) for those aged 90+ years (Fig. 3). The humerus/hip fracture ratios for those aged 50 to 54 years were 1.75 (95% CI 1.28–2.40) for men and 3.63 (95% CI 2.57–5.13) for women compared with ratios of 0.12 (95% CI 0.08–0.17) and 0.23 (95% CI 0.20–0.26) for aged 90+ years (Fig. 4). A similar pattern was seen for any MOF/hip ratios (Fig. 5).
Table 2 shows the ORs for fractures in men versus women. Men and women show the same pattern of vertebral fractures to hip fractures (all p values >0.1, ORs 0.86 to 1.25). In contrast, the forearm/hip ratios were significantly lower among men than among women in all age subgroups (all p values <0.05, ORs from 0.29 to 0.54), and also significantly lower in 8 of 9 age subgroups for humerus to hip fractures (p value <0.05 in all age groups except age 80 to 84 years, ORs 0.45 to 0.88). A similar pattern was observed for any MOF/hip ratios, which were significantly lower among men versus women in all age groups except for those aged 80 to 84 years and 90+ years (ORs 0.48 to 0.87).
|Age subroup (years)||Vertebral to Hip||Forearm to Hip||Humerus to Hip||Any MOF to Hip|
|OR (95% CI)||OR (95% CI)||OR (95% CI)||OR (95% CI)|
|50–54||0.86 (0.55–1.35)||0.35 (0.23–0.53)*||0.48 (0.30–0.77)*||0.48 (0.32–0.72)*|
|55–59||1.22 (0.83–1.80)||0.39 (0.27–0.55)*||0.46 (0.31–0.68)*||0.56 (0.39–0.78)*|
|60–64||1.02 (0.72–1.44)||0.29 (0.21–0.41)*||0.53 (0.36–0.76)*||0.51 (0.38–0.68)*|
|65–69||1.25 (0.92–1.70)||0.42 (0.32–0.56)*||0.67 (0.48–0.93)*||0.69 (0.54–0.88)*|
|70–74||1.04 (0.80–1.34)||0.40 (0.31–0.52)*||0.67 (0.51–0.89)*||0.70 (0.57–0.86)*|
|75–79||1.04 (0.84–1.29)||0.46 (0.37–0.58)*||0.68 (0.53–0.87)*||0.78 (0.66–0.92)*|
|80–84||1.00 (0.82–1.22)||0.53 (0.42–0.68)*||0.86 (0.68–1.09)*||0.86 (0.32–1.00)|
|85–89||1.00 (0.81–1.23)||0.47 (0.36–0.63)*||0.46 (0.34–0.64)*||0.82 (0.71–0.96)*|
|90+||1.01 (0.79–1.31)||0.49 (0.35–0.70)*||0.52 (0.36–0.77)*||0.87 (0.73–1.02)|
The published fracture data on Swedish men and women showed similar vertebral/hip ratios for men and women, with lower ratios in men compared with women for forearm/hip, humerus/hip, and any MOF/hip fracture ratios (Figs. 2 to 5). The same decreasing trend with advancing age was seen in both the Swedish and Canadian data. Overall, Swedish vertebral/hip fracture ratios were similar to Canadian fracture ratios (7% greater for men and 1% less for women based upon the fitted power functions) but lower for other sites (men and women: 46% and 35% lower for forearm/hip ratios, 19% and 15% lower for humerus/hip ratios, and 19% and 23% lower for any MOF/hip ratios).
Using population-based data, we were able to evaluate site-specific fracture to hip fracture ratios according to age in Canadian men and women. Hip fractures accounted for the largest proportion of fractures in older individuals, whereas younger individuals were much more likely to sustain non-hip fractures, resulting in declining ratios with older age. Forearm fractures were predominant among younger patients, as noted by previous studies. Calculated ratios showed a decreasing trend with advancing age. Vertebral to hip ratios were similar between men and women, but ratios were significantly lower in men than in women for forearm, humerus, and any MOF fracture site.
Compared with Canada, published Swedish fracture data showed many similarities with regard to relative fracture patterns: (i) Forearm fractures made up the majority of MOFs among younger individuals but only a minority of fractures among older individuals; (ii) hip fractures accounted for only a minority of MOFs in younger patients but were the most common fracture site among older individuals; (iii) fracture ratios showed a decreasing trend with advancing age in both men and women; and (iv) vertebral/hip ratios were similar between men and women but were lower in men for the forearm/hip, humerus/hip, and any MOF/hip fracture ratios.
Several factors could contribute to apparent differences in the fracture ratios between Canada in Sweden beyond underlying population and geographic considerations. Temporal decreases in hip fracture rates have been clearly evident in Canada since at least 1985. Age-standardized hip fracture incidence rates in Sweden were stable in 1987 to 1996 but have decreased about 2% per year since 1996. Whether there have been comparable declines in non-hip fractures is less certain, and limited evidence suggests that where this is occurring it is proceeding more slowly than for hip fractures.[7, 19] Factors responsible for declining fracture rates are uncertain, but one recent report found that declining MOF rates from 1996 to 2006 were explained by improvements in BMD rather than greater rates of obesity or osteoporosis treatment. Because BMD shows a steeper gradient of risk for hip fracture than for other fractures, this would be expected to result in greater declines for hip than non-hip fractures. Although hip fractures are generally well coded in hospital data, readmissions for the same fracture event are not infrequent and may lead to slight overestimation in hip fracture rates in the Swedish Hospital Discharge. Because the Swedish fracture non-hip/hip ratios were calculated from hip fracture rates observed in Sweden in 1996, this might contribute to lower ratios than we saw for Canada. The ability to accurately identify nonhospitalized fractures is likely to differ between countries and may also change over time. Vertebral fracture diagnosis is particularly problematic and is an area of controversy even among experts.
We have previously reported that ratios of observed hip fractures to MOFs in Canada showed general agreement with ratios estimated from FRAX. However, this study was based upon on clinical registry of patients referred for BMD testing and may not apply to the general population. Moreover, it only evaluated women and was based upon a much smaller number of fractures. Therefore, the current population-based study is likely to be more robust. We found similar trends but some differences between ratios calculated from Canadian and Swedish data. Vertebral/hip fracture ratios showed a high level of agreement between the data sets. Larger differences were observed for the other non-hip/hip ratios with the largest noted for forearm/hip fracture ratios, which were 46% and 35% lower in Sweden compared with Canada for men and women, respectively. The latter may even be conservative because it relied upon a relatively strict case definition that underestimated the forearm fracture rate seen in CaMos; a more liberal definition would increase the number of forearm fractures and give an even larger difference between the Canadian and Swedish forearm/hip and MOF/hip ratios.
The observed differences between Canadian and Swedish ratios mean that calibration of the Canadian FRAX tool with Swedish ratios would slightly underestimate predicted MOF risk. Our findings are similar to those reported in the Global Longitudinal Study of Osteoporosis in Women (GLOW). Based upon self-reported fracture data in 60,393 women aged 55 years and older, relative hip fracture proportions observed in GLOW were approximately 30% to 50% lower at all age classes older than 60 years compared with Sweden reference data. Using fracture ratios from the GLOW study would translate into higher estimates of MOF risk. These differences could have significant clinical implications because a steep relationship exists between miscalibration in fracture risk prediction and resulting treatment rates. An overestimation of 20% in MOF probability resulted in a 97% and 55% increase in the number of men and women categorized as needing treatment, respectively, whereas a 20% underestimation in fracture risk resulted in a 52% and 50% decrease in the numbers categorized as needing treatment, respectively. Thus, even small miscalibration in FRAX tools may lead to significant differences in intervention rates.
The recalibration of the US FRAX tool with US hospital discharge data from 2006, which included hip fracture rates that were approximately 16% lower than those included in the original model, resulted in lower estimates for MOF risk by 13% to 24% in men and 19% to 24% in women. Notably, the US White FRAX tool gives higher ratios of 10-year MOF probability to 10-year hip fracture probability than Sweden despite using the Swedish vertebral/hip ratios. The Canadian FRAX tool was calibrated using the US MOF/hip fracture ratios and, in two independent cohorts, was shown to accurately predict fracture risk in the Canadian population.[29, 30] Our data would, therefore, support the applicability of the US MOF/hip fracture ratios and also the Swedish vertebral/hip ratios to both the US and Canada. In contrast, a study involving Hong Kong Chinese men and women noted higher vertebral/hip fracture ratios compared with those described for Sweden. The Italian and Turkish FRAX tools were updated to include more recent hip fracture data, which significantly impacted on the risk predictions generated: For Italy, there were lower estimated 10-year fracture probabilities in younger age groups and slightly higher ones in older age groups (differences ranging from 10% to 20%); for Turkey, hip fractures had increased markedly over the last 20 years, resulting in a corresponding increase in estimated 10-year fracture probabilities at all ages. Together, these studies underline the importance of (re)calibrating FRAX tools with hip and non-hip fracture data that is accurate, current, and country-specific.
MOF risk prediction appears to be subject to greater between-country variability than hip fracture risk prediction. In a direct comparison of several FRAX models in Canadian women, three tools calibrated using the Swedish non-hip/hip ratios (Australia, New Zealand, France) showed lower MOF predictions than three other tools (USA, Canada, UK), although all tools gave similar hip fracture predictions. When compared with observed fracture rates in Canadian women, calibration for all six tools was equally good for hip fractures, but the tools relying on Swedish non-hip/hip ratios substantially underestimated MOF risk. Smaller studies from New Zealand and France have reported that FRAX underestimated MOF risk when applied directly to their respective populations.[35, 36] Therefore, if confirmed by other studies, our data may have broader relevance in terms of how FRAX models are calibrated in the absence of non-hip fracture data.
The strengths of our study include the use of large population-based data sources that capture reported fractures and that are not dependent on patient recall. Limitations include fracture ascertainment from administrative data, which is less reliable than direct radiographic and medical chart review. However, we have used definitions that give age- and sex-specific fracture rates concordant with population-based rates from the Canadian Multicentre Osteoporosis Study. Our study examined fractures between 2000 and 2007, and we cannot exclude changes in more recent years, although no consistent trends were seen across these 7 years. Even though our analyses were population based, the number of fracture events in some subgroups yielded wide 95% confidence intervals. Finally, we could not examine ethnicity-specific fracture data and, therefore, cannot exclude ethnic differences in fracture ratios.
The changing epidemiology of osteoporosis should be reflected in estimates of osteoporotic fracture risk by including current data on MOFs. In countries where these are not available, hip fractures are often used to impute other MOF rates using Swedish from 1987 to 1996. Using population-based data, our study provides updated ratios for the estimation of vertebral, forearm, and humerus fracture rates. Small differences in risk estimates have large implications for both patients (eg, initiation of pharmacological treatment) and for policymakers (eg, projecting the burden and cost of fractures). Therefore, wherever possible, it is important that countries collect accurate non-hip fracture data for FRAX calibration and updating.
WDL has served on the speaker bureau for Amgen, Eli Lilly, and Novartis, and has received research grants from Novartis, Amgen, and Genzyme. SNM has served as a consultant to Amgen, Novartis, Eli Lilly, and Merck; has served on the speaker bureau for Amgen and Novartis; and has received research grants from Amgen. All other authors state that they have no conflicts of interest.
The authors are indebted to Manitoba Health for the provision of data (HIPC 2008/2009–16, Theme 2). The results and conclusions are those of the authors, and no official endorsement by Manitoba Health is intended or should be inferred. This study was funded in part through a research grant from Amgen Canada Ltd. The funding source had no role in the study design, no role in data collection, no access to the data before publication, no input into the writing of the manuscript, and no input in the decision to publish the results.
SRM holds the Endowed Chair in Patient Health Management (Faculties of Medicine and Dentistry and Pharmacy and Pharmaceutical Sciences, University of Alberta) and receives salary support as a Health Scholar of the Alberta Heritage Foundation for Medical Research and Alberta Innovates-Health Solutions. SNM is chercheur-boursier des Fonds de Recherche du Québec en Santé. LML is supported by a Manitoba Health Research Chair.
Authors' roles: Conception and design (WDL), analysis (WDL, LML, and MSY), and interpretation of data (all); drafting the article (AL) or revising it critically for important intellectual content (all); and final approval of the version to be published (all). WDL accepts full responsibility for the work and/or the conduct of the study, had access to the data, and controlled the decision to publish.