Simplified System for Absolute Fracture Risk Assessment: Clinical Validation in Canadian Women

Authors


  • Dr Leslie receives honoraria or speaker's fees from Merck Frosst Canada, Sanofi-Aventis, and Genzyme Canada; research support from Merck Frosst Canada; and unrestricted educational grants from Procter & Gamble Pharmaceuticals, Novartis Pharmaceuticals Canada, and Amgen Canada. All other authors state that they have no conflicts of interest.

Abstract

Absolute 10-yr fracture risk based on multiple factors is the preferred method for risk assessment. A simplified risk assessment system from sex, age, DXA, and two clinical risk factors (CRFs)—prior fracture and systemic corticosteroid (CS) use-has been used in Canada since 2005. This study was undertaken to evaluate this system in the Canadian female population. A total of 16,205 women ≥50 yr of age at the time of baseline BMD (1998–2002) were identified in a database containing all clinical DXA test results for the Province of Manitoba, Canada. Basal 10-yr fracture risk from age and minimum T-score (lumbar spine, femur neck, trochanter, total hip) was categorized as low (<10%), moderate (10–20%), or high (>20%). Health service records since 1987 were assessed for prior fracture codes (N = 5224), recent major CS use (N = 616), and fracture codes after BMD testing (mean, 3.1 yr of follow-up) for the hip, vertebrae, forearm, or humerus (designated osteoporotic, N = 757). Fracture risk predicted from age and minimum T-score alone showed a significant gradient in observed fracture rates (low 5.1 [95% CI, 4.1–6.4], moderate 11.5 [95% CI, 10.1–13.0], high 25.4 [95% CI, 23.2–27.9] per 1000 person-years; p-for-trend <0.0001). There was an incremental increase in incident fracture rates from a prior fracture (13.9 [95% CI, 11.3–16.4] per 1000 person-years) or major CS use (11.2 [95% CI, 4.1–18.2] per 1000 person-years). This simplified fracture risk assessment system provides an assessment of fracture risk that is consistent with observed fracture rates.

INTRODUCTION

Osteoporosis is a common condition in Canada, affecting up to 16% of women and 7% of men >50 yr of age.math image The presence of osteoporosis is a major risk factor for the development of fractures of the hip, proximal humerus, vertebra, and forearm. Worldwide, the number of fracture sufferers in 2000 was estimated at 56 million, with ∼9 million new osteoporotic fractures each year.math image The global burden of osteoporosis is projected to increase markedly over the next few decades as the number of elderly individuals increases.math image The number of hip fractures in Canada was 23,375 in 1993–1994 and is predicted to increase to 88,124 by 2041, with a parallel increase in associated hospital days from 465,000 to 1.8 million patient-days.math image Moreover, the case-fatality rate for hip fractures can exceed 20%,math image,math image and all osteoporosis-related fractures can lead to significant long-term disability and decreased quality of life.math image,math image

The ability to accurately gauge fracture risk is critical in identifying cost-effective thresholds for intervention.math image,math image The WHO Collaborating Center has recently identified a comprehensive set of seven clinical risk factors (prior fragility fracture, a parental history of hip fracture, smoking, use of systemic corticosteroids, excess alcohol intake, body mass index, and rheumatoid arthritis), which in addition to age and sex, contribute to fracture risk independently of BMD.math image,math image A simplified (semiquantitative) approach suggested by Osteoporosis Canada incorporates age, sex, prior fragility fracture, and systemic steroid use together with BMD to define an absolute fracture risk.math image The Canadian recommendations were issued in June 2005 and have altered the fundamental approach to BMD utilization that had been in use over the previous decade, shifting the emphasis from the relative risk conferred by WHO T-score categories to absolute fracture risk based on the 10-yr absolute fracture risk as estimated from an individual's BMD, age, sex, fracture history, and systemic corticosteroid use.

This study was undertaken to evaluate the Osteoporosis Canada simplified fracture risk assessment system. The clinical database of the Manitoba Bone Density Program was linked to other provincial health databases to estimate osteoporotic fracture rates in relation to the patient's risk categorization using the simplified methodology.

MATERIALS AND METHODS

Patient population

The study population for this retrospective cohort study consisted of all women of white ethnicity ≥50 yr of age at the time of baseline DXA. Women were required to have results for the lumbar spine and proximal femur (total hip, femur neck, and trochanter sites) between May 1998 and October 2002 with medical coverage from Manitoba Health and Healthy Living during the observation period starting January 1997 and ending March 2004. Because earlier software versions before May 1998 did not provide total hip measurements, any records before this date were not included in the analysis. For women with more than one eligible set of measurements, only the first record was included. The final study population consisted of 16,205 women. The study was approved by the Research Ethics Board for the University of Manitoba and the Health Information Privacy Committee of Manitoba Health and Healthy Living.

In the Province of Manitoba, Canada, health services are provided to virtually all residents through a single public health care system. BMD testing with DXA has been managed as an integrated program since 1997 and uses targeted case-finding rather than population screening.math image Criteria and testing rates for this program have been published.math image The program maintains a database of all DXA results, which can be linked with other population-based computerized health databases through an anonymous personal identifier.math image The DXA database has been previously described with completeness and accuracy in excess of 99%. Fractures can be assessed through a combination of hospital discharge abstracts (diagnoses and procedures coded using the International Classification of Disease-9-Clinical Modification [ICD-9-CM] system) and physician billing claims (in-patient, out-patient, and office-based).math image Use of systemic corticosteroids can be obtained by linkage to the provincial Drug Program Information Network (DPIN) database. With ∼20 million transactions annually, DPIN captures information about pharmaceutical dispensations in real time for all Manitoba residents.

BMD measurements

DXA scans were performed and analyzed in accordance with manufacturer recommendations. Lumbar spine T-scores (number of SDs above or below young adult mean BMD) and Z-scores (number of SDs above or below age-matched mean BMD) were calculated using the manufacturer U.S. white female reference values. Hip T-scores and Z-scores were calculated from the revised NHANES III white female reference values (Prodigy version 8.8).math image,math image Vertebral levels affected by artifact were excluded by experienced physicians using conventional criteria.math image Before 2000, DXA measurements were performed with a pencil-beam instrument (Lunar DPX; GE Lunar, Madison, WI, USA), and after this date a fan-beam instrument was used (Lunar Prodigy; GE Lunar). Instruments were cross-calibrated using anthropomorphic phantoms and 59 volunteers. No clinically significant differences were identified (T-score differences < 0.2). Therefore, all analyses are based on the unadjusted numerical results provided by the instrument. Densitometers showed stable long-term performance (CV < 0.5%) and satisfactory in vivo precision (CV = 1.7% for L1–L4 and 1.1% for the total hip).math image

Simplified fracture risk system

Under the simplified (semiquantitative) system, an individual's 10-yr absolute fracture risk (combined risk for fractures of the proximal femur, vertebrae, forearm, and proximal humerus) is stratified into three 10-yr absolute fracture risk zones designated low risk (<10%), moderate risk (10–20%), and high risk (>20%). Similar absolute risk categories have been used for cardiovascular risk assessment.math image,math image

Osteoporotic fracture rates as a function of age, sex, and measured BMD were initially taken from the Malmö, Sweden population.math image These rates have subsequently been validated in Canadian women.math image Certain clinical factors increase fracture risk independently of BMD. The most important are the following: fragility fractures (especially vertebral compression fractures)math image and prolonged systemic corticosteroid use (e.g., 3 mo duration or longer).math image The presence of either of these factors substantially elevates fracture risk. This was operationalized by increasing the risk categorization to the next level: from low risk to moderate risk or from moderate risk to high risk. When both factors are present (i.e., fragility fractures and prolonged systemic corticosteroid use), the patient is considered to be at high fracture risk regardless of the BMD result.

This simplified system for estimating 10-yr absolute fracture risk does not require any detailed calculations or access to a computer. All data required to use the system can be summarized on a pocket-sized laminated card (available on request from the author).

Fracture prediction and outcomes

Each subject in the study population was assigned a basal fracture risk category according to the simplified system (low, <10%; moderate risk, 10–20%; high risk, >20%) based on BMD (minimum T-score for the lumbar spine, total hip, femoral neck, and trochanter) and age. Additional analyses were undertaken in which basal fracture risk was based on a single measurement site (femoral neck or total hip) because hip DXA was used for modeling osteoporotic fracture rates in the source populationmath image and has been designated by the WHO as the diagnostic reference site for osteoporosis diagnosis.math image

Basal risk subgroups were substratified according to presence of additional risk factors: any prior fracture code (from 1987 to the date of BMD testing) and/or recent systemic corticosteroid use. Longitudinal health service records were assessed for the presence of noncraniofacial fracture codes before and after BMD testing that were not associated with trauma codes.math image Hip, clinical vertebral, forearm, and humerus fractures were collectively designated as “osteoporotic” fractures because they are the basis for the 10-yr absolute fracture risk estimates published by Kanis et al.math image,math image We required that hip fractures and forearm fractures be accompanied by a site-specific fracture reduction, fixation, or casting code, which enhances the diagnostic and temporal specificity for an acute fracture. All other fractures were collectively referred to as “minor” fractures (approximately one half of these involve the digits). Major systemic corticosteroid use was defined over the year before BMD testing as at least 90-day possession with a mean dose of prednisone (or equivalent) of 7.5 mg daily or greater. All other recent corticosteroid use was denoted as minor (median duration, 27 days; median dose, 13 mg/d). The outcome of interest was any osteoporotic fracture (according to the previous definition) that occurred after BMD testing.

Statistics

Osteoporotic fracture rates were expressed per 1000 person-years, where the denominator was the number of years of observation (adjusted for loss of follow-up caused by migration out of province before March 2004). The simplified system attempts to predict 10-yr risk of osteoporotic fracture as a percent. To directly compare rates and 10-yr risk, it was assumed that rate per 1000 person-years was equivalent to 10-yr percent risk. This assumption was tested in another Manitoba cohort (20,579 women ≥47.5 yr of age at the time of baseline femoral neck BMD testing) with up to 10 yr of follow-up in which 10-yr fracture risk was derived from Kaplan-Meier curves and also as a rate per 1000 person-years.math image There was a high level of agreement between the two methods of expressing incident fracture data (r = 0.997) with a slope that did not significantly differ from unity (0.961 ± 0.023) and an intercept that did not significantly differ from zero (0.4 ± 0.6).

Rates with 95% CIs were stratified according to basal fracture risk category (low, moderate, or high) and presence of additional risk factors (any prior fracture and/or major systemic corticosteroid use). The simplified approach assumes that the presence of each clinical risk factor confers an incremental risk roughly equivalent to a change of the risk category (i.e., an increase in 10-yr probability of fracture of 10% or ∼10 fractures per 1000 person-years).

Cox proportional hazards models were used to model time to first osteoporotic fracture. A likelihood ratio test for the single-site model versus the multisite model was used to assess the incremental value of combined-site BMD measurement.math image The likelihood ratio χ2 statistic from the Cox proportional hazards model provides a global measure of model fit, and the difference between χ2 values is used to test the improvement in model fit. p < 0.05 indicates a statistically significant improvement in fracture prediction using the combined assessment, whereas p > 0.05 indicates that there is no incremental benefit in the combined assessment. Statistical analyses were performed with Statistica (Version 6.1; StatSoft, Tulsa, OK, USA).

RESULTS

Population

The baseline characteristics of the 16,205 women in the study population are summarized in Table 1. The mean age was 65 ± 9 (SD) yr. Mean T-scores for each of the measurement sites fell within the WHO osteopenic (low bone mass) category. Mean Z-scores were within 0.1 SD of zero, implying that the clinical population had BMD measurements comparable to those from the manufacturer's reference population. Measurement sites differed in the proportion of women categorized as being osteoporotic, with the highest prevalence at the lumbar spine (26%) and the lowest prevalence in the total hip (11%). When diagnosis was based on the lowest T-score, 5419 (33%) of the women met the densitometric criterion for osteoporosis.

Table Table 1.. Study Population Baseline Characteristics (N = 16,205)
original image

Basal risk categorization and fractures

When basal fracture risk category was determined from the minimum site T-score, 28.6% were designated as low risk, 38.7% as moderate risk, and 32.6% as high risk (Table 2).

Table Table 2.. Osteoporotic Fractures After BMD Testing in Relation to Basal Fracture Risk Category Defined From Age and T-Score
original image

The number of women with new osteoporotic fractures during a mean 3.1 yr of follow-up was 75 (1.6%) in the low-risk group, 238 (3.8%) in the moderate-risk group, and 444 (8.4%) in the high-risk group. This translated into 5.1 (95% CI, 4.1–6.4) osteoporotic fractures per 1000 person-years of follow-up for the low-risk category, 11.5 (95% CI, 10.1–13.0) for the moderate-risk category, and 25.4 (95% CI, 23.2–27.9) for the high-risk category. There was a significant gradient in fracture occurrence in relation to risk category (p < 0.0001). Between-group differences were statistically significant (low versus moderate and moderate versus high, p < 0.0001). The preceding analysis was based on all subjects and combined those with and without additional clinical risk factors. In the 10,553 women without clinical risk factors, fracture rates were slightly lower, but a similar risk gradient was seen across the basal fracture risk categories (low 4.1 [95% CI, 3.0–5.4], moderate 8.4 [95% CI, 7.0–10.0], high 17.1 [95% CI, 14.7–19.9] per 1000 person-years; p for trend < 0.0001).

When basal risk category was based on the femoral neck rather than the minimum site, a smaller number of women was categorized as high risk (20.4%) but their actual fracture risk was greater (31.2 [95% CI, 28.1–34.6] per 1000 person-years). When basal risk category was based on the total hip, there was an even smaller number of women categorized as high risk (15.0%) with even greater fracture risk (37.5 [95% CI 33.5–41.9] per 1000 person-years).

Incremental effect of prior fracture and recent corticosteroid use

The overall effect of clinical risk factors on incident osteoporotic fractures across all basal risk categories is summarized in Fig. 1. The presence of any prior fracture (without major corticosteroid use) conferred an incremental risk for subsequent osteoporotic fractures of 13.9 (95% CI, 11.3–16.4) per 1000 person-years. Major corticosteroid use (without prior fracture) also led to a significant incremental risk for subsequent fractures of 11.2 (95% CI, 4.1–18.2) per 1000 person-years. The presence of both risk factors was associated with even higher incremental risk of 18.1 (95% CI, 5.9–30.4) per 1000 person-years. When prior fracture status was stratified as osteoporotic or minor, there was a much greater incremental effect from the former than the latter (25.9 [95% CI, 21.4–30.4] versus 5.5 [95% CI, 2.9–8.1] per 1000 person-years). Minor corticosteroid use in the preceding year had no incremental effect on fracture risk (0.0 [95% CI, −42 to 4.3] per 1000 person-years).

Figure Figure 1.

Incremental effect of clinical risk factors on osteoporotic fracture rates after BMD testing. Reference groups: (1) no risk factors, (2) no prior fracture, and (3) no recent systemic CS use. 95% CI bars are shown.

When fracture rates were stratified by basal fracture risk (as defined by age, sex, and the lowest T-score) and the number of additional clinical risk factors (Table 3), there was a consistently strong incremental effect of prior fracture history that was seen at all basal fracture risk levels and for all BMD measurement sites. For low basal fracture risk from the minimum T-score, a prior fracture increased the risk for future osteoporotic fractures by 5.0 (95% CI, 1.4–8.5) per 1000 person-years. The incremental risk was 8.4 (95% CI, 4.8–11.9) for moderate basal fracture risk and 18.6 (95% CI, 13.5–23.6) for high basal fracture risk. Major corticosteroid use had a significant incremental effect for moderate basal fracture risk (12. 9 [95% CI, 1.8–24.0]) and for high basal fracture risk (18.6 [95% CI, 13.5–23.6]) but not for low basal fracture risk (0.4 [95% CI, −5.9 to 6.8]). The presence of both clinical risk factors (prior fracture and major corticosteroid use) showed an incremental increase in the point estimates for subsequent osteoporotic fractures (range, 6.0–18.1 per 1000 person-years), but confidence intervals were wide and these were not statistically significant. Once again, in the absence of additional clinical risk factors, there was a significant risk gradient for osteoporotic fractures in relation to basal risk category (low: 4.1 [95% CI, 3.0–5.4]) per 1000 person-years versus moderate: 8.4 [95% CI, 7.0–10.0] versus high: 17.1 [95% CI, 14.7–19.9]). Broadly similar results were seen when basal fracture risk was based on the femoral neck or total hip T-score.

Table Table 3.. Osteoporotic Fracture Rates (per 1000 Person-Years) After BMD Testing Stratified by Clinical Risk Factors
original image

Cox proportional hazards models

Time to first osteoporotic fracture was studied in Cox proportional hazards models (Table 4). When basal fracture risk category was based on the minimum site T-score, moderate risk (HR: 2.04 [95% CI, 1.57–2.64]) and high risk (HR: 4.11 [95% CI, 3.20–5.27]) were associated with significantly higher risk than the low basal risk category (reference). Prior minor fracture (HR: 1.42 [95% CI, 1.17–1.71]), prior osteoporotic fracture (HR: 2.92 [95% CI, 2.47–3.44]), and recent major corticosteroid use (HR: 1.62 [95% CI, 1.22–2.15]) were all independently associated with higher fracture risk. Recent minor systemic corticosteroid use was not a risk factor for incident osteoporotic fractures (HR: 0.85 [95% CI, 0.67–1.08]). The HRs associated with prior fracture history and recent systemic corticosteroid use were almost identical when basal fracture risk category was based on femoral neck T-score, total hip T-score, or age alone (without BMD).

Table Table 4.. Hazard Ratios (95% CIs) for New Osteoporotic Fracture According to Prior Risk Factors
original image

The improvement in the Cox proportional hazards model fit when additional clinical risk factor information (prior fracture and/or recent major corticosteroid use) was added to a basal fracture risk model (based on a 10-yr absolute fracture risk prediction model from age and BMD alone) is shown in Fig. 2. There was significant improvement in fracture prediction with the inclusion of additional clinical risk factor information in all models assessed (p < 0.0001).

Figure Figure 2.

Incremental improvement in osteoporotic fracture prediction for Cox proportional hazards models that include additional clinical risk factors (solid bars) over fracture risk models based on from age and BMD alone (cross-hatched bars). p values are for the change in model fit when prior fracture and major CS use are added to a prediction model using basal 10-yr fracture risk.

DISCUSSION

We found that a simplified (semiquantitative) system developed by Osteoporosis Canada for absolute fracture risk assessment gave results that were consistent with observed fracture rates in a large cohort of women ≥50 yr of age undergoing a baseline DXA assessment of BMD. In particular, the inclusion of only two clinical risk factors (prior risk fracture and recent major corticosteroid use) significantly improved fracture prediction compared with models based on age and BMD only. This is consistent with a recent evaluation with the WHO fracture risk assessment system (FRAX), which also found incremental benefit in the inclusion of multiple clinical risk factors (a superset of those considered here).math image,math image We were unable to compare the simplified fracture risk assessment and the WHO quantitative fracture risk system in this study. Whether there is incremental benefit with the additional clinical risk factors of the FRAX system is currently unclear and warrants further study. A previous analysis from the population-based Canadian Multicentre Osteoporosis Study (CaMos) found minimal differences in risk prediction and the rate of high risk categorization when comparing the simplified Osteoporosis Canada model used in this study and the expanded set used in the WHO model.math image

Although similar results were obtained when fracture risk was based on the minimum site, the femoral neck, or the total hip T-scores, some differences were also noted. As expected, the minimum T-score categorized a larger proportion of women as being at high risk (32.6%) compared with either of the hip measurements (femoral neck 20.4% and total hip 15.0%). The minimum site T-score was frequently determined by the lumbar spine, and this site is more strongly associated with the vertebral fracture risk than hip measurements even when global fracture risk is unchanged.math image,math image On the other hand, the group designated as high risk by the minimum site T-score had lower fracture rates than the groups designated as high risk based on the hip T-scores. As a result, in women without additional clinical risk factors, the observed fracture rate per 1000 person-years (equivalent to 10-yr percent fracture risk) was slightly below the nominal 10-yr fracture risk range in women without additional risk factors for the moderate- and high-risk categories. Only the total hip site gave predictions that were within the nominal 10-yr fracture risk range at all levels of risk. This suggests that different sites and methods for BMD and measurement can not be used interchangeably in fracture risk assessment models. The best approach for including information on the lumbar spine in these fracture risk assessment models needs to be better defined.

The finding that prior fractures and significant systemic corticosteroid use are risk factors for osteoporotic fractures is not surprising. These risk fractures have been strongly associated with fractures in numerous analyses, including the WHO meta-analysis.math image,math image We confirmed that the HRs associated with these risk factors are largely independent of BMD, with virtually identical values when age-adjusted (without BMD) or when adjusted using different BMD measurements. The incremental effect of major corticosteroid use (without prior fracture) on the rate of incident osteoporotic fractures (11.2 per 1000 person-years) was close to the weight assumed in the simplified system (one category grade roughly equal to 10% 10-yr risk or 10 per 1000 person-years). Minor corticosteroid use (without prior fracture) had no effect on osteoporotic fracture rates. Previous studies have shown that continuous low-dose corticosteroid use down to 2.5 mg daily is associated with significantly increased fracture rates.math image There were insufficient numbers of continuous low-dose corticosteroid users in our cohort for separate analyses, and therefore, our results should not be taken to indicate that continuous use of prednisone <7.5 mg daily does not affect fracture risk. The importance of stratification by fracture site that we observed has not been incorporated into current absolute fracture risk models. We found that prior fractures of the hip, vertebrae, forearm, and humerus carried a much greater risk for future osteoporotic fractures (25.9 per 1000 person-years) than other fracture sites (5.5 per 1000 person-years). Under the simplified fracture risk assessment system, prior fracture of these major sites seems to be sufficient for a designation of high risk.

Significant limitations to this study are acknowledged. Fracture rates are known to vary by more than an order of magnitude worldwide.math image The simplified osteoporosis fracture risk system was based on an assumption of similar fracture rates in Canada as in Sweden, which is consistent with available data.math image The Canadian system may not be appropriately calibrated for use in other countries, particularly those with substantially different fracture rates. The composition of our cohort also prevents extrapolation to men, women <50 yr of age, and nonwhites.

In summary, we showed that a relatively simple fracture risk system based on only two clinical risk factors (in addition to age and DXA-derived BMD) provides an assessment of absolute fracture risk that is consistent with observed fracture rates in a large clinical cohort of older women. Based on this analysis, future refinements to the system could be considered. The risk estimates and/or category cut-offs may need some adjustment to provide better calibration for the use of minimum site T-score (versus femoral neck T-score as used in the derivation population). Differential weighting of prior fracture history according to site would be beneficial to better reflect the importance of major osteoporotic fractures.

Acknowledgements

We are indebted to Manitoba Health and Healthy Living for providing data. The results and conclusions are those of the authors, and no official endorsement by Manitoba Health and Healthy Living is intended or should be inferred. This article has been reviewed and approved by the members of the Manitoba Bone Density Program Committee. This study was funded in part by an unrestricted educational grant from the CHAR/GE Healthcare Development Awards Programme.

Ancillary