FRAX underestimates fracture risk in patients with diabetes

Authors


Abstract

The study objective was to determine whether diabetes is a risk factor for incident hip or major osteoporotic fractures independent of the WHO fracture risk assessment tool (FRAX). Men and women with diabetes (n = 3518) and nondiabetics (n = 36,085) aged ≥50 years at the time of bone mineral density (BMD) testing (1990 to 2007) were identified in a large clinical database from Manitoba, Canada. FRAX probabilities were calculated, and fracture outcomes to 2008 were established via linkage with a population-based data repository. Multivariable Cox proportional hazards models were used to determine if diabetes was associated with incident hip fractures or major osteoporotic fractures after controlling for FRAX risk factors. Mean 10-year probabilities of fracture were similar between groups for major fractures (diabetic 11.1 ± 7.2 versus nondiabetic 10.9 ± 7.3, p = 0.116) and hip fractures (diabetic 2.9 ± 4.4 versus nondiabetic 2.8 ± 4.4, p = 0.400). Diabetes was a significant predictor of subsequent major osteoporotic fracture (hazard ratio [HR] = 1.61, 95% confidence interval [CI] 1.42–1.83) after controlling for age, sex, medication use, and FRAX risk factors including BMD. Similar results were seen after adjusting for FRAX probability directly (HR = 1.59, 95% CI 1.40–1.79). Diabetes was also associated with significantly higher risk for hip fractures (p < 0.001). Higher mortality from diabetes attenuated but did not eliminate the excess fracture risk. FRAX underestimated observed major osteoporotic and hip fracture risk in diabetics (adjusted for competing mortality) but demonstrated good concordance with observed fractures for nondiabetics. We conclude that diabetes confers an increased risk of fracture that is independent of FRAX derived with BMD. This suggests that diabetes might be considered for inclusion in future iterations of FRAX. © 2012 American Society for Bone and Mineral Research

Introduction

Osteoporosis is a disease characterized by reduced bone strength and an increased risk of fractures. In the absence of established fragility fractures, the diagnosis of osteoporosis has historically been based on low bone mass, defined by the World Health Organization (WHO) as a femoral neck bone mineral density (BMD) T-score of −2.5 SD or lower.1 Although BMD is an important correlate of fracture risk, only a minority of low-trauma fractures can be attributed to low BMD using this definition.2, 3 More recent fracture risk assessment paradigms combine BMD with important clinical risk factors to improve the prediction of osteoporotic fractures.4–6 The WHO fracture risk assessment tool (FRAX) combines age, sex, body mass index, use of glucocorticoids, current smoking, alcohol intake of three or more units per day, secondary osteoporosis, rheumatoid arthritis, prior fragility fracture, and (optionally) femoral neck BMD T-score to generate estimated probabilities for major osteoporotic and hip fractures over 10 years. FRAX has been shown to improve fracture prediction over BMD T-score alone.4, 7

Identifying additional clinical factors that are independently associated with an increased fracture risk may be one way to further improve fracture risk stratification. Diabetes is a chronic condition affecting 285 million adults worldwide and results in a number of secondary complications, including increased fracture risk.8–11 Both type 1 and type 2 diabetes have been associated with an increased hip fracture risk.9, 10, 12 The increased risk of fractures in individuals with diabetes is evident even after adjustment for BMD, despite the observation that BMD is higher in individuals with type 2 diabetes compared with nondiabetic individuals.13–16 Thus, a growing body of research suggests that diabetes is a clinical indicator of increased fracture risk independent of BMD.

Diabetes status is not among the clinical risk factors in the current formulation of FRAX. It has recently been reported that for a given FRAX score or BMD T-score and age, the risk of fracture among individuals with diabetes is higher than the risk in nondiabetics.17 The primary aim of the current study was to evaluate whether diabetes is associated with incident hip and major osteoporotic fractures independent of derived FRAX probabilities. Secondary aims of the study were to determine if the relationship between diabetes and fractures was modified by age, sex, or femoral neck BMD.

Materials and Methods

Subjects and setting

In the Province of Manitoba, Canada, health services are provided to residents through a single public health care system. Bone density testing with dual-energy X-ray absorptiometry (DXA) has been managed as an integrated program since 1997; criteria and testing rates for this program have been published.18 The database contains all clinical BMD measurements for individuals tested in Manitoba since January 1, 1990. The DXA database has been previously described with completeness and accuracy in excess of 99%.19 The Manitoba Bone Density Program database is linked to the population-based Research Data Repository housed at the Manitoba Centre for Health Policy (MCHP), University of Manitoba, through an anonymous personal identifier. The accuracy and completeness of the Research Data Repository have been described elsewhere.20 The Repository can be used to obtain patient demographics; date and type of health care service(s) received; inpatient, outpatient, and office-based diagnoses and procedures performed; and (since April 1, 1995) prescription medications dispensed to outpatients. Medications are classified according to the Anatomic Therapeutic Chemical (ATC) of the WHO Collaborating Centre for Drug Statistics Methodology.

Study population

We identified men and women who: 1) were over the age of 50 years at baseline BMD assessment; 2) had valid FRAX probability estimates calculated using femoral neck BMD; and 3) had medical coverage from Manitoba Health during the observation period starting in April 1987 and ending March 2008. For individuals with more than one DXA assessment, only the first BMD measurement was included. Subjects were categorized by the presence or absence of diabetes using a previously validated method for identifying individuals with diabetes (administrative data cannot distinguish between type 1 and type 2 diabetes);21 diabetes was ascertained based on the presence of two separate physician claims for diabetes within 2 years or a hospitalization with a diabetes diagnosis (International Classification of Diseases 9th revision Clinical Modification [ICD-9-CM] code 250). The study was approved by the Research Ethics Board for the University of Manitoba. Access to the data was granted by the Manitoba Health Information Privacy Committee.

Fracture ascertainment

Incident fractures were identified as fractures that occurred between the date of BMD testing and March 31, 2008, and were ascertained using ICD-9-CM codes (or equivalent ICD-10 code for hospitalizations after 2004), where two or more hospital or physician ICD-9-CM fracture codes had to be present to confirm a fracture. Fractures associated with trauma codes were excluded from the analysis. Major osteoporotic fractures included: hip with site-specific fracture fixation code; forearm with site-specific fracture fixation or casting code; clinical spine; and proximal humerus. Only a single fracture for a given site within a 180-day period was counted to avoid double-counting health care interactions related to the same injury. For individuals with more than one incident major osteoporotic fracture, analysis was based upon the time to first qualifying fracture.

Dual-energy X-ray absorptiometry

Proximal femur DXA scans were performed and analyzed by technicians according to the manufacturer's guidelines using either pencil-beam DXA (Lunar DPX; GE Lunar, Madison, WI, USA) if the measurement was taken before the year 2000, or fan-beam DXA (Lunar Prodigy; GE Lunar) after the year 2000. We cross-calibrated the instruments and did not identify any clinically significant differences. Densitometers showed stable long-term performance (coefficient of variation [CV] < 0.5%) and satisfactory in vivo precision (CV 1.7% for L1-4 and 1.1% for the total hip).22 Femoral neck (FN) T-scores (number of SDs above or below young adult mean BMD) were calculated based on reference data for US white females from the NHANES III survey.23

FRAX clinical risk factors and medication use

Weight and height were obtained by self-report at the time of the DXA examination before the year 2000. After 2000, height was assessed with a wall-mounted stadiometer and weight was assessed without shoes using a standard floor scale. Body mass index (BMI; in kg/m2) was calculated as weight (kg) divided by height (m) squared. Additional clinical risk factors required for calculating fracture probability with FRAX were assessed through a combination of hospital discharge abstracts (diagnoses and procedures coded using the ICD-9-CM before 2004 and ICD-10-CA thereafter) and physician billing claims (coded using ICD-9-CM).20 We defined prior fragility fracture as fracture that occurred before BMD testing at skeletal sites considered to be associated with osteoporosis (ie, hip, clinical vertebral, forearm, and humerus) that was not associated with external injury codes, as previously described.24 A diagnosis of rheumatoid arthritis testing was taken from physician office visits or hospitalizations with a compatible ICD-9-CM/ICD-10-CA code in a three-year period before BMD testing. Proxies were used for smoking (chronic obstructive pulmonary disease [COPD] diagnosis) and high alcohol intake (alcohol or substance abuse diagnosis) over the same time frame. Prolonged corticosteroid use (more than 90 days dispensed in the year before DXA testing at a mean prednisone-equivalent dose of 7.5 mg per day or greater) was obtained from the provincial pharmacy system.25 We adjusted for the effect of missing parental hip fracture information on FRAX probability estimates before 2005 using age- and sex-specific adjustment factors derived from 2005 to 2008 parental hip fracture responses as previously described.26 We used the provincial pharmacy database to identify use of osteoporosis-related medication (90 days or more of pharmacy dispensations for hormone therapy, bisphosphonate, selective estrogen-receptor modulator, parathyroid hormone or recombinant human parathyroid hormone analog [1-34], calcitonin during the year before the BMD test).

Fracture probability calculations

Ten-year probabilities of a major osteoporotic fracture or hip fracture were calculated for each subject by the WHO Collaborating Centre for Metabolic Bone Diseases based on the Canadian FRAX tool (version 3.1) using the previously defined variables and femoral neck BMD without knowledge of the fracture outcomes. FRAX estimates with the Canadian tool agree closely with observed fracture rates in the Canadian population.26, 27 Ten-year major osteoporotic fracture probabilities from FRAX were categorized as low (<10%), moderate (10% to 20%), or high (>20%) in accordance with national guidelines.28

Statistical analysis

Demographic and clinical characteristics of individuals with and without diabetes at the time of baseline BMD measurement are presented using means and SDs for continuous variables and frequencies and percentages for categorical variables. Independent-sample Student's t tests (continuous variables) and chi-square tests (categorical variables) were used to test for differences in baseline characteristics, fracture probabilities (hip and major osteoporotic) and numbers of fracture between individuals with and without diabetes. Receiver-operating characteristic (ROC) analyses were performed for individuals with and without diabetes to assess the ability of femoral neck T-score and FRAX probabilities to discriminate those with and without major osteoporotic fractures and hip fractures.

The Kaplan-Meier method was used to plot fracture-free survival in diabetics and nondiabetics; statistical testing for differences in age-adjusted fracture-free survival were assessed in Cox proportional hazards models. Incomplete observations were censored at the end of the follow-up period (March 1, 2008), at the time of migration (3.0%), or at the time of death (8.3%). Multivariable Cox proportional hazards regression analyses were performed to assess whether diabetes status was a predictor of time to incident hip or major osteoporotic fracture, adjusted for age, sex, BMI, FN T-score, previous major osteoporotic fracture, chronic obstructive pulmonary disease, alcohol and substance abuse diagnosis, rheumatoid arthritis, prolonged glucocorticoid use and recent osteoporotic medication use. We performed a similar analysis controlling for derived FRAX probabilities. We also used multivariable Cox proportional hazards regression to test the excess mortality attributable to diabetes. FRAX probability was coded as a continuous variable (log-transformed because of a skewed distribution) and as a categorical variable (grouped into risk quintiles). We included interaction terms to evaluate whether age (<65 versus ≥65 years old), sex (male versus female), or FN T-score WHO classification (normal ≥ −1 SD, low bone mass [osteopenia] from −1 to −2.5 SD, or osteoporotic ≤ −2.5 SD) modified the effect of diabetes on risk of hip or major osteoporotic fracture. Hazard ratios (HR) were generated from the regression models with corresponding 95% confidence intervals (CI).

We examined the concordance between observed fracture risk over 10 years and FRAX-predicted 10-year fracture probability estimates for major osteoporotic fracture risk subgroups (low <10%, moderate 10% to 19%, high ≥20%) for both diabetics and nondiabetics, where death was treated as a competing risk.29 The observed 10-year fracture risk was estimated from baseline covariates (ie, age and other variables were not updated to adjust for time dependencies). All statistical analyses were performed using Statistica (Version 10.0, StatSoft, Inc., Tulsa, OK, USA). The criterion for statistical significance was set at a p value of 0.05.

Results

There were significant differences in the characteristics of individuals with diabetes (n = 3518) compared with the nondiabetic individuals (n = 36,085). Individuals with diabetes were slightly older, more likely to be male, had higher mean BMI, and had a higher prevalence of prior fracture but less likely to be receiving osteoporosis treatment. Mean femoral neck T-score was higher among diabetics, but mean FRAX probabilities were similar between groups for major fractures and hip fractures (Table 1).

Table 1. Descriptive Characteristics of Individuals With and Without Diabetes
 DiabeticNondiabeticp Value
(n = 3518)(n = 36,085)
  • COPD = chronic obstructive pulmonary disease; BMD = bone mineral density.

  • a

    Based on 8150 females and 1103 males, respectively. Data are mean ± SD or n (%).

Sex (male)464 (13.2%)2409 (6.7%)<0.001
Age (years)68.0 ± 9.365.7 ± 9.9<0.001
Body mass index (kg/m2)29.7 ± 6.226.5 ± 5.0<0.001
Rheumatoid arthritis147 (4.2%)1383 (3.8%)0.310
COPD402 (11.4%)3047 (8.4%)<0.001
Alcohol or substance abuse diagnosis83 (2.4%)913 (2.5%)0.537
Recent glucocorticoid use312 (8.9%)1864 (5.2%)<0.001
Recent osteoporosis treatment653 (18.6%)9327 (25.8%)<0.001
Prior fracture563 (16.0%)4852 (13.4%)<0.001
Parental hip fracturea125 (11.3%)1071 (13.1%)0.093
Femoral neck T-score−1.3 ± 1.1−1.5 ± 1.0<0.001
Probability of major fracture from FRAX with BMD (%)11.1 ± 7.210.9 ± 7.30.116
Probability of hip fracture from FRAX with BMD (%)2.9 ± 4.42.8 ± 4.40.400

During an observation period of up to 10 years (mean follow-up of 5.4 years), major osteoporotic fractures occurred in 289 (8.2%) and 2254 (6.2%) individuals with and without diabetes, respectively. Hip fractures occurred in 86 (2.4%) and 463 (1.3%) individuals with and without diabetes, respectively. Compared with the no-fracture groups, the proportion of individuals with diabetes was higher in those with major osteoporotic (8.7% versus 11.4%, p < 0.001) and hip fractures (8.8% versus 15.7%, p < 0.001). The 10-year observed risk of major osteoporotic fracture and hip fracture was greater in diabetics compared with nondiabetics (p < 0.001, Fig. 1).

Figure 1.

Kaplan-Meier plots of major osteoporotic fracture (A) and hip fracture (B) in individuals with and without diabetes.

After controlling for age, sex, medication use and FRAX risk factors including BMD, diabetes was a significant predictor of subsequent major osteoporotic fracture (HR = 1.61, 95% CI 1.42–1.83, p < 0.001; Table 2). Similar results were seen after adjusting for FRAX major osteoporotic fracture probability as a continuous measure (HR = 1.59, 95% CI 1.40–1.79, p < 0.001). For major osteoporotic fractures, the interactions of diabetes with age, sex, and femoral neck WHO category were not statistically significant (p > 0.2 for interaction). Identical results were obtained when the FRAX probabilities were recoded as a categorical variable (risk quintiles).

Table 2. Hazard Ratios for Diabetes Status as a Predictor of Time to Incident Major Osteoporotic Fracture or Hip Fracture After Controlling for FRAX Risk Factors or Probability
 Major osteoporotic fractureHip fracture <65 yearsHip fracture ≥65 years
HR95% CIHRa95% CIHRa95% CI
  • HR = hazard ratio; CI = confidence interval.

  • a

    Hazard ratios for hip fracture are age-stratified because of a significant diabetes-age interaction.

Adjusted for FRAX risk factors1.611.42–1.836.273.62–10.872.221.71–2.90
Adjusted for FRAX probability (continuous)1.591.40–1.795.343.14–9.082.061.59–2.66
Adjusted for FRAX probability (categorical)1.591.23–2.055.403.17–9.192.021.56–2.61

After controlling for age, sex, medication use, and FRAX risk factors including BMD, diabetes was a significant predictor of subsequent hip fracture. Age, but not sex or femoral neck WHO category, modified the effect of diabetes on hip fracture risk (p = 0.002 for interaction), where diabetes was a stronger predictor of hip fracture in those aged <65 years (HR = 6.27, 95% CI 3.62–10.87, p < 0.001) than in those ≥65 years (HR = 2.22, 95% CI 1.71–2.90, p < 0.001). Similar results were seen after adjusting for FRAX hip probability as a continuous measure or when recoded as a categorical variable (risk quintiles) (Table 2).

ROC analyses revealed that the discriminatory capacity of femoral neck T-scores was similar in diabetics and nondiabetics (Table 3). FRAX probability measures had slightly better discriminatory capacity in nondiabetics than in diabetics. Higher mortality was observed in diabetics versus nondiabetics after adjustment for individual FRAX risk factors including BMD (HR = 1.99, 95% CI 1.80–2.19], p < 0.001), for major osteoporotic fracture probability (HR = 2.28, 95% CI 2.07–2.51, p < 0.001), and for hip osteoporotic fracture probability (HR = 2.29, 95% CI 2.08–2.52, p < 0.001). The effect of this excess mortality attenuated but did not eliminate the excess fracture risk in diabetics. There was good concordance between FRAX-predicted and observed major osteoporotic and hip fractures for nondiabetics, but FRAX underestimated fracture risk in diabetics (Fig. 2). For each risk category (low <10%, moderate 10% to 20%, or high >20%), the observed 10-year major osteoporotic fracture risk and hip fracture risk (adjusted for competing mortality) were significantly greater among diabetics than nondiabetics (all p < 0.05).

Table 3. Area Under the Curve for Major Osteoporotic and Hip Fracture Discrimination in Individuals Without and With Diabetes
 NondiabeticDiabeticp Value
AUC95% CIAUC95% CI
  1. AUC = area under the curve; CI = confidence interval; BMD = bone mineral density.

Prediction of major osteoporotic fracture
 Probability of major fracture from FRAX with BMD−0.700.69–0.710.670.63–0.700.070
 Femoral neck T-score0.680.67–0.690.660.63–0.700.286
Prediction of hip fracture
 Probability of hip fracture from FRAX with BMD0.840.82–0.860.770.72–0.810.002
 Femoral neck T-score0.810.79–0.830.770.73–0.820.148
Figure 2.

Concordance plots for FRAX major osteoporotic fracture (A) and hip fracture (B) predictions by risk category (low <10%, moderate 10% to 19%, high ≥20%). Error bars are 95% confidence intervals. The dotted line indicates the line of identity (perfect concordance). *p < 0.05, **p < 0.001.

Discussion

The current study provides compelling evidence that FRAX underestimates the risk of major osteoporotic and hip fractures in individuals with diabetes. The risk of hip fracture attributable to diabetes was highest in individuals <65 years of age. FRAX predicted fractures in diabetics but had slightly lower discriminative capacity than in nondiabetics. Our observations that diabetes status was an independent predictor of hip and major osteoporotic fractures after controlling for FRAX risk factors or FRAX probabilities and that FRAX underestimated fracture risk in individuals with diabetes suggest that diabetes status should be evaluated for inclusion in future iterations of fracture prediction models.

Mounting evidence has suggested that diabetics have increased fracture risk despite normal or elevated BMD,13, 15, 30, 31 and our data provide convincing evidence that fracture risk in diabetics is not completely captured by the current FRAX risk factors. Our findings are consistent with a recent study demonstrating that fracture risk was higher among diabetics than nondiabetics for a given FRAX score or BMD T-score.17 The strengths of the current study compared with others are the large study population, the use of a population-based data repository, and the lack of dependence on self-report for determining diabetes status or incident fractures. The Canadian FRAX tool demonstrated good concordance with observed fractures for nondiabetics in our population, as previously reported.26, 27 The effect of diabetes after adjustment for FRAX probability was found to be unchanged using the US white FRAX tool. Similar to previous studies, the ROC analyses revealed that FRAX was a better predictor of hip fracture risk than major osteoporotic fractures.32, 33 Although the ROC analysis showed little difference between FRAX with BMD and femoral neck T-score alone, ROC analysis has low sensitivity for evaluating the benefit of new risk factors because increases in the area under the curve (AUC) are often small even when incremental models lead to a substantial improvement in reclassification.34, 35

In considering diabetes as an independent risk factor for fractures, account should be taken of the increased mortality associated with diabetes that was independent of the FRAX-based probabilities. In this context, the importance of diabetes in predicting hip fracture risk appeared to be stronger among individuals younger than the age of 65 years. It is of interest that a higher HR for hip fracture in younger versus older individuals has been reported for some clinical risk factors used in FRAX, such as history of fragility fracture5 and BMD.36 A higher mortality rate in older diabetics may partially explain the age interaction because accounting for higher mortality attenuated the effect of diabetes on fractures. The complex relationship between age, diabetes, and fractures should be considered in future studies.

The increase in fracture risk in diabetics in our study was not explained by reduced BMD; indeed, femoral neck T-scores were higher than in nondiabetics. In a small case-control study, older women with type 2 diabetes had increased cortical porosity in conjunction with increased trabecular volumetric density, suggesting that changes in the cortex that are not captured with DXA may explain the paradoxical increased fracture incidence despite higher BMD observed in type 2 diabetics.37 FRAX does not incorporate risk factors for falls, and the presence of diabetes-related complications, such as peripheral neuropathy or visual impairment, has been linked to increased fall risk in diabetics,38 which may explain the higher fracture risk for any given FRAX-based probability. However, it has also been suggested that the risk of fracture in diabetics is not accounted for by a history of frequent falls.15 An increase in fracture risk secondary to use of medications such as thiazoledinediones or an accumulation of advanced glycation end products are alternative hypotheses.39, 40 In fact, several diabetes-specific factors have been linked to fracture risk, including diabetic retinopathy, cataracts, renal impairment, insulin treatment, and diabetes duration.17, 41 A feed-forward regulation loop has been described such that insulin signaling in osteoblasts results in increased bone resorption and decarboxylation of osteocalcin, suggesting a link between bone remodeling and glucose metabolism.42 Regardless of the mechanism, the hazard ratio for 10-year fracture risk associated with the presence of diabetes is comparable to or greater than many of the currently used FRAX risk factors,43 suggesting that diabetes should at least be considered as a potential risk factor for inclusion in future fracture prediction models.

Limitations of the current study should be acknowledged. It is expected that fracture risk may be different among individuals with type 1 versus type 2 diabetes, and we were not able to account for these differences. Given the age of our cohort, the great majority of the sample with diabetes would have type 2 diabetes; in the population-based Canadian Multicentre Osteoporosis Study (CaMos), 1.3% of participants over age 50 years had type 1 diabetes and 6.8% had type 2 diabetes.14 A meta-analysis found the risk of hip fracture to be higher among individuals with type 1 diabetes (RR = 6.3, 95% CI 2.6–15.1) than with type 2 diabetes (RR = 1.7, 95% CI 1.3–2.2); type 2 diabetes showed a weaker association with fractures at nonhip sites.9, 12 The larger number of individuals with type 2 diabetes might result in a weaker association between fractures and diabetes status than if only those with type 1 diabetes were studied. However, there is enough evidence that type 2 diabetes is also associated with increased fracture risk to make this an issue of public health importance.13, 15 We were not able to determine the contribution of antihyperglycemic medications to the increased risk of fracture in diabetics. However, a recent study suggested that insulin use does not appear to explain the higher rate of bone loss or increased incidence of fractures observed in diabetics.44 Future research should examine the unique contributions of antihyperglycemic medications to reduced bone strength. Although the interaction between age and the effect of diabetes on hip fracture risk was significant, the number of hip fractures in individuals less than age 65 years was relatively small and this finding should be treated cautiously. The sample represents individuals referred for BMD testing and therefore may be subject to referral bias. Smoking and alcohol use were measured using proxies, which may not accurately account for the risk attributable to these factors. Finally, we chose to estimate 10-year fracture probability using baseline covariates, rather than time-varying covariates, because this was felt to be more relevant for the clinical assessment of fracture risk.

In summary, we have demonstrated that diabetes status is predictive of future hip and major osteoporotic fractures independent of FRAX probability and its associated risk factors including BMD. Age modified the effect of diabetes on hip fracture risk. FRAX underestimated fracture risk in diabetics, suggesting that future fracture prediction algorithms should consider including diabetes as an independent risk factor.

Disclosures

LMG: Unrestricted research grant from Merck Frosst. WDL: Speaker fees and unrestricted research grants from Merck Frosst; unrestricted research grants from Sanofi-Aventis, Procter and Gamble Pharmaceuticals, Novartis, Amgen Pharmaceuticals, Genzyme; advisory boards for Genzyme, Novartis, and Amgen Pharmaceuticals. LML: Unrestricted research grants from Amgen Pharmaceuticals. JAK: Nothing to declare for FRAX and the context of this article but numerous ad hoc consultancies for: industry: Abiogen, Italy; Amgen, USA, Switzerland, and Belgium; Bayer, Germany; Besins-Iscovesco, France; Biosintetica, Brazil; Boehringer Ingelheim, UK; Celtrix, USA; D3A, France; Gador, Argentina; General Electric, USA; GSK, UK, USA; Hologic, Belgium and USA; Kissei, Japan; Leiras, Finland; Leo Pharma, Denmark; Lilly, USA, Canada, Japan, Australia, and UK; Merck Research Labs, USA; Merlin Ventures, UK; MRL, China; Novartis, Switzerland and USA; Novo Nordisk, Denmark; Nycomed, Norway; Ono, UK and Japan; Organon, Holland; Parke-Davis, USA; Pfizer, USA; Pharmexa, Denmark; Procter and Gamble, UK and USA; ProStrakan, UK; Roche, Germany, Australia, Switzerland, and USA; Rotta Research, Italy; Sanofi-Aventis, USA; Schering, Germany and Finland; Servier, France and UK; Shire, UK; Solvay, France and Germany; Strathmann, Germany; Tethys, USA; Teijin, Japan; Teva, Israel; UBS, Belgium; Unigene, USA; Warburg-Pincus, UK; Warner-Lambert, USA; Wyeth, USA; governmental and nongovernmental organizations: National Institute for Health and Clinical Excellence (NICE), UK; International Osteoporosis Foundation, Switzerland; INSERM, France; Ministry of Public Health, China; Ministry of Health, Australia; National Osteoporosis Society, UK; WHO, Switzerland. All the other authors state that they have no conflicts of interest.

Acknowledgements

LMG is supported by a Canadian Institutes of Health Research (CIHR) Randomized Controlled Trials Mentoring Program Award and an Early Researcher Award from the Ontario Ministry of Research and Innovation. LML is supported by a CIHR New Investigator Award and a Centennial Research Chair at the University of Saskatchewan.

We acknowledge Manitoba Health for providing data (HIPC File No. 2007/2008-49). The results and conclusions are those of the authors, and no official endorsement by Manitoba Health is intended or should be inferred. This article has been reviewed and approved by the members of the Manitoba Bone Density Program Committee.

Authors' roles: Study design: LG and WL. Data analysis: WL and LL. Data interpretation: All authors. Drafting manuscript: LG and WL. Revising manuscript content: All authors. Approving final version of manuscript: All authors.

Ancillary