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

Authors

  • Dawn C Mackey MSc,

    Corresponding author
    1. San Francisco Coordinating Center, California Pacific Medical Center Research Institute, San Francisco, California, USA
    • San Francisco Coordinating Center, CA Pacific Medical Center Research Institute, 185 Berry Street, Lobby 4, Suite 5700, San Francisco, CA 94107, USA
    Search for more papers by this author
  • Jean Gaare Eby,

    1. Harvard School of Public Health and Channing Laboratory, Brigham and Women's Hospital, and Harvard Medical School, Boston, Massachusetts, USA
    Search for more papers by this author
  • Fran Harris,

    1. Department of Epidemiology and Biostatistics, University of California, San Francisco, San Francisco, California, USA
    Search for more papers by this author
  • Dennis R Taaffe,

    1. School of Human Movement Studies, Faculty of Health Sciences, University of Queensland, Brisbane, Australia
    Search for more papers by this author
  • Jane A Cauley,

    1. Department of Epidemiology, University of Pittsburgh, Pittsburgh, Pennsylvania, USA
    Search for more papers by this author
    • Dr Cauley receives research support from Merck & Co., Eli Lilly & Co., Pfizer, and Novartis. She also receives honoraria from Merck & Co., Novartis, and Eli Lilly & Co., in addition to serving on the Speaker's Bureau for Merck & Co. All other authors state that they have no conflicts of interest.

  • Frances A Tylavsky,

    1. Department of Preventive Medicine, University of Tennessee Health Science Center, Memphis, Tennessee, USA
    Search for more papers by this author
  • Tamara B Harris,

    1. Laboratory for Epidemiology, Demography and Biometry, National Institute on Aging, Bethesda, Maryland, USA
    Search for more papers by this author
  • Thomas F Lang,

    1. Radiology, University of California, San Francisco, San Francisco, California, USA
    Search for more papers by this author
  • Steven R Cummings

    1. San Francisco Coordinating Center, California Pacific Medical Center Research Institute, San Francisco, California, USA
    Search for more papers by this author

  • Presented in part at the 24th Annual Meeting of the American Society for Bone and Mineral Research, San Antonio, Texas, September 19–24, 2002.

Abstract

In a prospective study of 1446 black and white adults 70–79 yr of age (average follow-up, 6.4 yr), vertebral TrvBMD from QCT predicted non-spine fracture in black and white women and black men, but it was not a stronger predictor than total hip aBMD from DXA. Hip aBMD predicted non-spine fracture in black men.

Introduction: Areal BMD (aBMD) at multiple skeletal sites predicts clinical non-spine fractures in white and black women and white men. The predictive ability of vertebral trabecular volumetric BMD (TrvBMD) for all types of clinical non-spine fractures has never been tested or compared with hip aBMD. Also, the predictive accuracy of hip aBMD has never been tested prospectively for black men.

Materials and Methods: We measured vertebral TrvBMD with QCT and hip aBMD with DXA in 1446 elderly black and white adults (70–79 yr) in the Health, Aging, and Body Composition Study. One hundred fifty-two clinical non-spine fractures were confirmed during an average of 6.4 yr of >95% complete follow-up. We used Cox proportional hazards regression to determine the hazard ratio (HR) and 95% CIs of non-spine fracture per SD reduction in hip aBMD and vertebral TrvBMD.

Results: Vertebral TrvBMD and hip aBMD were both associated with risk of non-spine fracture in black and white women and black men. The age-adjusted HR of fracture per SD decrease in BMD was highest in black men (hip aBMD: HR = 2.04, 95% CI = 1.03, 4.04; vertebral TrvBMD: HR = 3.00, 95% CI = 1.29, 7.00) and lowest in white men (hip aBMD: HR = 1.23, 95% CI = 0.85, 1.78; vertebral TrvBMD: HR = 1.06, 95% CI = 0.73, 1.54). Adjusted for age, sex, and race, each SD decrease in hip aBMD was associated with a 1.67-fold (95% CI = 1.36, 2.07) greater risk of fracture, and each SD decrease in vertebral TrvBMD was associated with a 1.47-fold (95% CI = 1.18, 1.82) greater risk. Combining measurements of hip aBMD and vertebral TrvBMD did not improve fracture prediction.

Conclusions: Low BMD measured by either spine QCT or hip DXA predicts non-spine fracture in older black and white women and black men. Vertebral TrvBMD is not a stronger predictor than hip aBMD of non-spine fracture.

INTRODUCTION

DXA is the current clinical gold standard for assessing BMD and predicting fracture risk. Low areal BMD (aBMD) of appendicular and axial skeletal sites measured by DXA is a strong predictor of increased non-spine fracture risk in white women,(1–6) black women,(7) and white men.(8–11) DXA estimates BMD from a 2D projection of the bone while ignoring depth; therefore, larger bones appear artifactually to have greater density than smaller bones when true volumetric density is equal.(12) QCT shows promise as a clinical technique to improve the prediction of fracture risk.(13–15) QCT captures a 3D image of bone and measures volumetric BMD, which is not biased by the size of the bone.(12,16,17) Furthermore, whereas DXA simultaneously samples both cortical and trabecular bone, QCT can separately assess the size, shape, and relative amounts of bone in the cortical and trabecular compartments.

Cross-sectional studies have shown strong associations between vertebral trabecular volumetric BMD (TrvBMD) and prevalent vertebral fractures, and several studies have found that vertebral TrvBMD is better able to discriminate the presence of vertebral fractures than vertebral or hip aBMD.(12,18–21) One longitudinal study with nonblack women found that both vertebral TrvBMD and hip aBMD were significantly associated with increased risk of new and repeat vertebral fractures, and combining TrvBMD and aBMD measures improved fracture prediction over using a single predictor.(13) However, no prospective study has evaluated the ability of vertebral TrvBMD measurements to predict clinical non-spine fractures or compared the fracture prediction abilities of vertebral TrvBMD and hip aBMD.

Furthermore, data on the association between low aBMD and fracture risk are based primarily on studies of white women. A considerably smaller number of studies have been conducted with white men(8) and black women.(7) To the best of our knowledge, no studies have been conducted with black men. Thus, it is unknown whether aBMD in the axial skeleton is related to fracture risk in older black men.

We analyzed data from the prospective Health, Aging, and Body Composition Study with an average follow-up time of 6.4 yr to determine the association between hip aBMD and vertebral TrvBMD and clinical non-spine fracture in healthy older black and white men and women. We hypothesized that (1) vertebral TrvBMD is related to risk of clinical non-spine fracture in black and white men and women, (2) hip aBMD is related to risk of clinical non-spine fracture in black men, and (3) vertebral TrvBMD is more strongly related to risk of clinical non-spine fracture than hip aBMD.

MATERIALS AND METHODS

Study participants

Participants (N = 1446) were from 3075 well-functioning, community-dwelling black and white men and women 70–79 yr of age participating in the Health, Aging, and Body Composition (Health ABC) Study, a prospective cohort study of healthier older individuals. The study has been described elsewhere.(22) Briefly, volunteer participants were identified from a large mass mailing to a random sample of white Medicare beneficiaries and all age-eligible community-dwelling black residents in designated ZIP code areas surrounding the Pittsburgh, PA, and Memphis, TN, study sites. Eligible participants reported no difficulty walking a quarter of a mile (400 m), climbing 10 steps, or performing basic activities of daily living; they were free of any life-threatening cancer with no active treatment in the past 3 yr; and they did not plan to move from the study area within 3 yr. Potential participants currently enrolled in lifestyle (diet or exercise) intervention trials were not eligible. Participants provided written informed consent, and the institutional review boards at both study sites approved all protocols. Data for these analyses were from the baseline examination in April 1997 to June 1998. We included only the 1446 participants in the Pittsburgh cohort who had technically adequate hip DXA and spine QCT measurements. Participants in the Memphis cohort had hip DXA but not spine QCT measurements taken.

BMD and other measurements

aBMD (g/cm2) of the total hip was assessed by DXA (QDR 4500A, version 9.03; Hologic, Bedford, MA, USA). Identical patient scan protocols were used for all participants. The long-term precision of DXA measurements was assessed by obtaining daily spine and twice-weekly hip phantom scans. A short-term precision study of DXA measurements was not conducted; however, studies of the QDR 4500A system across multiple centers have reported a short-term precision of 0.9–1.2% for total hip aBMD based on duplicate and triplicate measures of women with repositioning.(23,24)

TrvBMD (g/cm3) of the vertebral L3 region was assessed by QCT (CT-9800 Advantage; General Electric, Milwaukee, WI, USA) during the same visit as hip DXA. Scans were made of a single 10-mm slice through the midbody of the L3 vertebrae (80 kVp, 140 mAs). To provide simultaneous calibration, participants were scanned with a calibration phantom (Image Analysis, Columbia, KY, USA) placed under their lower back, which contained calibration objects of calcium hydroxyapatite with known densities of 50, 100, and 200 mg/cm3. The slice was taken at an angle to compensate for lordosis in the vertebral body. Vertebral TrvBMD was measured using a peeled region of interest (ROI) in the anterior vertebral body, an area of trabecular bone.(25) The ROI was localized by orienting the scan of the vertebral body with respect to its anteroposterior axis and identifying two landmarks on this axis: the anterior edge of the vertebral body and the anterior edge of the spinal canal. QCT measurements of vertebral TrvBMD using analogous techniques to this study have interscan precisions reported at ∼2% based on paired repeat samples.(25,26) All QCT technologists were trained and certified, and the QCT data were analyzed with a standardized protocol at the University of California, San Francisco.

The baseline home visit included a detailed interview with participants about their health, medical history, activities, and demographics, and the baseline clinic visit consisted of a variety of physical measurements. Height was assessed using a Harpenden stadiometer (Pembrokeshire, UK). Weight (without shoes) was determined using a calibrated balance scale. Body mass index (BMI) was calculated as weight divided by the square of height (kg/m2).

Fracture ascertainment

Incident non-spine fractures were assessed by self-report at annual clinic visits and at intermediate 6-mo interviews. In addition, participants were asked to notify the clinical center as soon as possible after any fracture. This follow-up was >95% complete. All fractures were confirmed by medical documentation, including review of radiology report (except rib, toe, and finger fractures). All types of non-spine fractures were included in this analysis, including 15 fractures that resulted from major trauma (e.g., motor vehicle accidents).

Statistical analysis

The baseline characteristics of participants in each race-sex subgroup were compared by fracture status using t-tests for continuous variables and χ2 tests for categorical variables.

We used Cox proportional hazards regression models to determine the hazard ratio (HR) and 95% CIs of non-spine fracture per SD reduction in hip aBMD and vertebral TrvBMD. We analyzed the BMD–fracture association separately in each of the four race-sex subgroups for both hip aBMD and vertebral TrvBMD. We tested for homogeneity of the BMD–fracture relation across the four race–sex subgroups. In the absence of a significant interaction (p = 0.38 for vertebral TrvBMD and p = 0.39 for hip aBMD), we pooled the data and used multivariable models adjusted for age, sex, and race. Next, we included both hip aBMD and vertebral TrvBMD measurements (r = 0.63), as well as adjustments for age, sex, and race in a single model. To compare the strength of association between BMD and non-spine fracture for each of the two bone density measurement techniques (DXA and QCT), we calculated the area under the receiver operator characteristic (ROC) curve for each measurement technique and tested for a significant difference in these areas. To derive CIs for the area-under-the-curve values, we used bootstrap procedures(27) and sampled the study population (with replacement) 1000 times.

Finally, to assess the use of measuring BMD with both DXA and QCT, we constructed logistic regression models that were adjusted for age, sex, and race to determine the predicted probability of non-spine fracture for each participant using hip aBMD alone, vertebral TrvBMD alone, and the combination of hip aBMD and vertebral TrvBMD.(28) We used the predicted probabilities to group participants into quartiles of risk and determined the observed rate of non-spine fracture in each quartile. We calculated the HR of non-spine fracture in the highest quartile compared with the three lower quartiles based on hip aBMD alone, vertebral TrvBMD alone, and the combination of hip aBMD and vertebral TrvBMD. Furthermore, to determine whether low measurements of hip aBMD and vertebral TrvBMD identified the same participants at high risk for fracture, we identified the participants with non-spine fracture in the highest quartile of predicted risk based on hip aBMD alone, vertebral TrvBMD alone, and the combination of hip aBMD and vertebral TrvBMD.

RESULTS

A total of 1446 participants had measurements of both hip aBMD by DXA and vertebral TrvBMD by QCT. These included 268 black men, 435 white men, 374 black women, and 369 white women (Table 1). Black men who fractured had lower vertebral TrvBMD (p = 0.008) than black men who did not fracture. White men who fractured were slightly older than white men who did not fracture (p = 0.054). Black women who fractured had lower hip aBMD (p < 0.001) and vertebral TrvBMD (p = 0.035) than black women who did not fracture. Similarly, white women who fractured had lower hip aBMD (p = 0.005) and vertebral TrvBMD (p = 0.022) than white women who did not fracture. There were no significant differences based on fracture status for weight, height, BMI, or self-reported health for any race-sex subgroup. Overall, black adults appeared to report worse health status than white adults, and black women were more likely to have BMI >30 kg/m2 than any other race-sex subgroup.

Table Table 1.. Baseline Characteristics of Health ABC Participants According to Race-Sex Subgroup and Clinical Non-Spine Fracture Status
original image

One hundred fifty-two clinical non-spine fractures were validated during an average of 6.4 yr of follow-up (>95% complete; range, 12 days to 8.2 yr), corresponding to an average incidence rate of 16.6 fractures per 1000 person-years of follow-up (95% CI = 13.9, 19.2; Table 2). Wrist fractures were the most frequently observed fracture, followed by femoral neck and ankle fractures. Fracture incidence was highest among white women and lowest among black men.

Table Table 2.. Location and Number of Clinical Non-Spine Fractures Observed During Follow-Up According to Race-Sex Subgroup
original image

Hip aBMD and vertebral TrvBMD were both associated with risk of clinical non-spine fracture in black men and black and white women (Table 3). For hip aBMD, the age-adjusted HR of fracture for each SD decrease in aBMD ranged from 2.04 (95% CI = 1.03, 4.04) for black men to 1.42 (95% CI = 1.11, 1.80) for white women. For vertebral TrvBMD, the age-adjusted HR of fracture for each SD decrease in TrvBMD ranged from 3.00 (95% CI = 1.29, 7.00) in black men to 1.37 (95% CI = 1.05, 1.78) in white women. Hip aBMD and vertebral TrvBMD were not significantly associated with risk of clinical non-spine fracture in white men.

Table Table 3.. Unadjusted and Age-Adjusted HR and 95% CI of Clinical Non-Spine Fracture per 1 SD Decrease in Total Hip aBMD and Vertebral TrvBMD
original image

Vertebral TrvBMD was not a better predictor of clinical non-spine fracture than hip aBMD in this study population. After pooling data for the four race-sex subgroups and adjusting for age, sex, and race, the HR of fracture for each SD decrease in BMD was 1.47 (95% CI = 1.18, 1.82) for vertebral TrvBMD and 1.67 (95% CI = 1.36, 2.07) for hip aBMD (Table 4). Areas under the ROC curves for vertebral TrvBMD (area = 0.691; 95% CI = 0.648, 0.734) and hip aBMD (area = 0.705; 95% CI = 0.664, 0.747) did not differ significantly (p = 0.168; Fig. 1). Although there was no significant BMD by race-sex subgroup interaction, we repeated the analyses excluding data from black men, because the relationship between BMD and non-spine fracture seemed to be somewhat stronger for black men than the other three race-sex subgroups. The results did not change substantially. Specifically, the HR of fracture for each SD decrease in vertebral TrvBMD was 1.35 (95% CI = 1.10, 1.66), and the HR of fracture for each SD decrease in hip aBMD was 1.59 (95% CI = 1.30, 1.95), controlling for age, sex, and race.

Table Table 4.. Multivariable Models
original image
Figure FIG. 1..

ROC curves showing 1 – specificity (false-positive rate) vs. sensitivity (true positive rate) for total hip aBMD (solid line) and vertebral TrvBMD (dashed line) generated for the range of hip aBMD (0.40–1.55 g/cm2) and vertebral TrvBMD (0.02–0.38 g/cm3) values observed in this study.

After adjustment for vertebral TrvBMD, a low hip aBMD measurement was independently associated with a greater risk of non-spine fracture. Specifically, adjusting for age, sex, race, and vertebral TrvBMD, each SD reduction in hip aBMD was still associated with a 55% higher fracture risk (Table 4). In contrast, after adjustment for hip aBMD, a low vertebral TrvBMD measurement was no longer associated with an increased risk of fracture.

Combining measurements of hip aBMD and vertebral TrvBMD did not improve prediction of clinical non-spine fractures (Fig. 2). Among the 368 participants in the highest quartile of predicted risk for hip aBMD, 74 participants had a clinical non-spine fracture (32.8 fractures per 1000 person-years, 95% CI = 24.7, 40.9). Similarly, among the 368 participants in the highest quartile of predicted risk for vertebral TrvBMD, 68 participants had a clinical non-spine fracture (30.1 fractures per 1000 person-years, 95% CI = 22.4, 37.7). Of these 68 participants, 60 (88%) were also in the highest quartile of predicted risk for hip aBMD. Among the 368 participants in the highest quartile of predicted risk using the combination of hip aBMD and vertebral TrvBMD, 72 had non-spine fractures (31.0 fractures per 1000 person-years, 95%CI = 23.2, 38.7), and 69 (96%) of these participants were also in the highest quartile of predicted risk for hip aBMD alone. For vertebral TrvBMD alone, the HR of fracture comparing the highest risk quartile to the lower three quartiles was 2.57 (95% CI = 1.87, 3.54); for hip aBMD alone the HR was 3.04 (95% CI = 2.21, 4.18); and for the combination of vertebral TrvBMD and hip aBMD the HR was 2.90 (95% CI = 2.11, 3.98).

Figure FIG. 2..

Incidence of clinical non-spine fracture by quartile of expected probability of fracture. Expected probabilities were calculated from logistic regression models based on measurement of vertebral TrvBMD, total hip aBMD, and the combination of vertebral TrvBMD and total hip aBMD.

DISCUSSION

The results of this study support our first hypothesis that vertebral TrvBMD by QCT predicts clinical non-spine fracture risk in older black men and black and white women. In addition, the results of this study also support our second hypothesis that hip aBMD by DXA is related to risk of clinical non-spine fracture in black men. To the best of our knowledge, this is the first study to report an association between low vertebral TrvBMD measured by QCT and risk of all types of clinical non-spine fractures in multiple race and sex subgroups, and it is the first study to show an association between low hip aBMD measured by DXA and risk of non-spine fracture in black men.

The relations between hip aBMD and risk of fracture in our study population were similar to those observed in other cohorts. A meta-analysis of 11 prospective study populations, composed primarily of white and Asian-American women,(29) estimated the relative risk (RR) of any clinical fracture (including clinical spine fracture) per SD decrease in hip aBMD to be 1.6 (95% CI = 1.4, 1.8). A more recent meta-analysis of 12 prospective study populations composed of white and Asian men and women(8) calculated the RR of an osteoporotic fracture per SD decrease in femoral neck aBMD to be 1.60 (95% CI = 1.43, 1.79) for men and 1.53 (95% CI = 1.46, 1.62) for women. Our results for white women were consistent with these reports, although our point estimate was somewhat smaller. Combining non-spine and clinical spine fractures (n = 9) did not change our estimate for white women (HR = 1.46, 95% CI = 1.17, 1.82). We did not observe a significant association between hip aBMD and non-spine fracture risk in white men, possibly because of the small number of non-spine fractures in this subgroup. Other studies have found that hip aBMD is a strong predictor of non-spine fractures in white men.(10) When we combined non-spine and clinical spine fractures (n = 11) for white men, the association was significant (HR = 1.53, 95% CI = 1.10, 2.12) and closer in magnitude to previous reports. Including clinical spine fractures for black women (n = 5) and black men (n = 2) did not change the HR values (results not shown). Among black women, we observed a 91% greater risk of fracture per SD decrease in hip aBMD (95% CI = 1.34, 2.72), which was larger than a previous estimate of 44% (95% CI = 1.12, 1.86).(7) The associations between BMD and non-spine fracture seemed to be stronger for black men than for the other three race-sex subgroups, especially for vertebral TrvBMD. However, this was based on only 11 fractures in black men, so the point estimate was not as stable as for the other subgroups (indicated by the large CIs).

Our results do not support our third hypothesis that vertebral TrvBMD measured by QCT is a stronger predictor of non-spine fracture risk than hip aBMD measured by DXA. There are a number of possible explanations. First, our measurement of vertebral BMD selectively measured metabolically active trabecular bone, whereas most non-spine fracture locations contain a smaller percentage of trabecular bone and are less metabolically active.(18,30) This result implies that trabecular bone cannot account for all non-spine fracture risk, which also involves cortical bone. The advantage of hip aBMD from DXA is that it measures the status of both trabecular and cortical bone. Although the association with hip aBMD was slightly greater, the moderate association between vertebral TrvBMD and non-spine fracture suggests that TrvBMD may be a good surrogate for overall age-related bone loss.

Second, we did not compare the predictive value of QCT and DXA measurements at the same skeletal site. Previous research has shown that the predictive accuracy of aBMD measurements for the risk of all non-spine fractures is either similar across multiple skeletal sites(3,31) or less for vertebral aBMD than hip aBMD.(6) Hip fractures, in particular, are more strongly related to hip aBMD than any other skeletal site.(2,6) Third, we only measured a single slice of the lumbar vertebrae with QCT. In general, QCT has a higher CV than DXA and more measurement error. We would have achieved a more precise measurement of TrvBMD with multiple slices. Therefore, we may have underestimated the true predictive value of vertebral TrvBMD from QCT for non-spine fracture risk. Future research may find that more precise measurements of volumetric BMD obtained from multiple QCT slices at the hip or other nonvertebral sites are more predictive of non-spine fractures than vertebral TrvBMD.

Measurement of vertebral TrvBMD did not add substantially to the predictive value of a single measurement of total hip aBMD. When we determined the risk of clinical non-spine fracture using hip aBMD alone and vertebral TrvBMD alone, we found that the observed fracture rate was slightly greater in participants belonging to the highest quartile of predicted risk based on hip aBMD than based on vertebral TrvBMD. Moreover, there was not substantial benefit to fracture prediction when measurements of hip aBMD and vertebral TrvBMD were combined. The two techniques identified essentially the same women who were at high risk and not a unique group of women who were at particularly high risk.

At this time, DXA offers a number of practical advantages over QCT for clinical prediction of fracture risk, including lower radiation exposure and cost, more widespread availability of scanners, and higher precision and speed of patient throughput. Recent improvements in QCT have increased measurement reproducibility and diagnostic capability. However, the advantages to clinical fracture prediction of assessing volumetric density, bone structure and geometry, the distribution of cortical and trabecular bone, and biomechanical indices remain to be proven.

This study had a number of strengths. It was prospective and included relatively equal numbers of well-characterized, community-dwelling black and white men and women. Many studies lack information on black individuals and white men. This study also had certain limitations in addition to those mentioned already. Subjects were volunteers who were chosen to be healthier than average for their age. Thus, results may have limited applicability to older individuals with profound physiological, cognitive, or neuromuscular deficits. Race was self-reported, and we had no information on ad-mixture. Finally, there were a limited number of fractures in each race and sex subgroup, so we were not able to test whether the relation between BMD and fracture risk differed by sex or race.

In summary, we showed that low vertebral TrvBMD measured by QCT is a risk factor for clinical non-spine fractures in black men and black and white women. Furthermore, we showed that low hip aBMD measured by DXA increases the risk of clinical non-spine fractures in black men. Vertebral TrvBMD was not a better predictor of clinical non-spine fractures than hip aBMD, and it did not add substantially to fracture prediction by hip aBMD alone. Further refinement of QCT measurement and analysis may identify key aspects of bone structure and geometry and biomechanical integrity that add independent information to non-spine fracture prediction beyond DXA.

Acknowledgements

The Health, Aging and Body Composition (Health ABC) Study is supported by National Institutes of Health (NIH) funding, National Institute on Aging Contracts N01-AG-6-2101, N01-AG-6-2103, and N01-AG-6-2106, and in part by the Intramural Research Program of the NIH, National Institute on Aging.

Ancillary