Radius Bone Strength in Bending, Compression, and Falling and Its Correlation With Clinical Densitometry at Multiple Sites

Authors

  • Eva-Maria Lochmüller M.D.,

    Corresponding author
    1. Frauenklinik Innenstadt, Ludwig-Maximilians-Universität München, München, Germany
    • Frauenklinik Innenstadt, Ludwig-Maximilians-Universität München, Maistr. 11, D 80337 München, Germany
    Search for more papers by this author
    • Dr. Lochmüller and Dr. Lill share first authorship because they have contributed equally to this study.

  • Christoph A. Lill,

    1. AO Forschungsinstitut, Davos, Switzerland
    2. Department of Orthopedic Surgery, University of Heidelberg, Heidelberg, Germany
    Search for more papers by this author
  • Volker Kuhn,

    1. Frauenklinik Innenstadt, Ludwig-Maximilians-Universität München, München, Germany
    2. Musculoskeletal Research Group, Institute of Anatomy, Ludwig-Maximilians-Universität München, München, Germany
    Search for more papers by this author
  • Erich Schneider,

    1. AO Forschungsinstitut, Davos, Switzerland
    Search for more papers by this author
  • Felix Eckstein

    1. Musculoskeletal Research Group, Institute of Anatomy, Ludwig-Maximilians-Universität München, München, Germany
    Search for more papers by this author

  • The authors have no conflict of interest.

Abstract

This study comprehensively analyzes the ability of site-specific and nonsite-specific clinical densitometric techniques for predicting mechanical strength of the distal radius in different loading configurations. DXA of the distal forearm, spine, femur, and total body and peripheral quantitative computed tomography (pQCT) measurements of the distal radius (4, 20, and 33%) were obtained in situ (with soft tissues) in 129 cadavers, aged 80.16 ± 9.8 years. Spinal QCT and calcaneal quantitative ultrasound (QUS) were performed ex situ in degassed specimens. The left radius was tested in three-point bending and axial compression, and the right forearm was tested in a fall configuration, respectively. Correlation coefficients with radius DXA were r = 0.89, 0.84, and 0.70 for failure in three-point bending, axial compression, and the fall simulation, respectively. The correlation with pQCT (r = 0.75 for multiple regression models with the fall) was not significantly higher than for DXA. Nonsite-specific measurements and calcaneal QUS displayed significantly (p < 0.01) lower correlation coefficients, and QUS did only contribute to the prediction of axial failure stress but not of failure load. We conclude that a combination of pQCT parameters involves only marginal improvement in predicting mechanical strength of the distal radius, nonsite-specific measurements are less accurate for this purpose, and QUS adds only little independent information to site-specific bone mass. Therefore, the noninvasive diagnosis of loss of strength at the distal radius should rely on site-specific measurements with DXA or pQCT and may be the earliest chance to detect individuals at risk of osteoporotic fracture.

INTRODUCTION

Osteoporosis represents a major health problem relevant to most geographic areas and populations around the world. The clinical endpoint of osteoporotic bone loss are fractures, which cause a substantial reduction in the quality of life of the elderly(1) and a heavy burden on public health care systems.(2)

The incidence of distal radius fractures rises steeply after menopause,(3,4) with 20% of all women at the age of 70 years having sustained a clinical fracture at this skeletal site.(5,6) Fractures of the radius represent an early predictive sign of future fractures at other sites such as the proximal femur or spine.(4,7–8) Thus, the noninvasive diagnosis of loss of strength at the distal radius may be the earliest chance to detect individuals at risk of osteoporotic bone loss and potential fracture. Because it is easier to prevent bone loss at an early stage than to restore bone strength at a later stage of the disease, early diagnosis of reduction in strength at the distal radius may be the first chance to identify patients who profit from treatment.(7)

Several methods are available to measure the bone mineral status clinically.(9,10) Currently, DXA is the most widely used, but the surrounding soft tissues may introduce relevant measurement errors.(11,12) In particular, DXA cannot discriminate between cortical and trabecular bone.(9) However, geometric parameters of the cortical shell, as measured with peripheral quantitative computed tomography (pQCT), have been suggested to be important for accurate prediction of bone strength.(13–15) QCT allows for the separate analysis of trabecular and cortical bone and can be performed with conventional clinical CT scanners.(16) However, measurements in the spine may not correlate highly with those at the distal radius,(14,17) and they do not allow for the extraction of geometric parameters of long peripheral bones such as the distal radius. Quantitative ultrasound (QUS) at the calcaneus has been shown to display a relevant relationship with trabecular microstructure and mechanical properties.(18–20) This has stimulated hope that QUS may provide significant independent information for the prediction of bone strength, beyond that of bone mass or density. The attractiveness of this technique lies mainly in the complete lack of ionizing radiation and in the relatively low costs involved.(19,20)

Previous studies have compared mechanical failure loads of the distal radius in small sets of specimens with site-specific densitometry such as single photon absorptiometry (SPA), DXA, or pQCT(13–15,21–29) (Table 1), but these usually have been performed without soft tissues (except for those performed by Wu et al.(28) and Njeh et al.(29)). In recent studies, we have shown that bone failure loads display only moderate (albeit significant) correlation between skeletal sites and that these are predicted more accurately by site-specific DXA than by nonsite-specific DXA.(30) We also have shown that pQCT measurements at the distal radius predict femoral failure loads with the same accuracy as pQCT analysis at the peripheral lower limb but are inferior to pQCT measurements obtained directly at the femoral neck.(31) However, no previous study has compared the failure loads of the distal radius with nonsite-specific measurements other than DXA(30) or phalangeal QUS,(28,29) and no previous study has compared the correlation of densitometric variables of several techniques with both a fall simulation and the major components of the typical distal radius fracture, namely, dorsal bending and axial compression.(15,32)

Table Table 1.. Summary of Important Studies on the Site-specific Correlation of Bone Densitometry (SPA, DXA, and PQCT) with Failure Load of the Distal Radius
original image

The objective of this study was to evaluate comprehensively the ability of clinically available site-specific and nonsite-specific densitometric techniques in predicting bone strength of the distal radius in different loading configurations in a large set of specimens. We specifically tested the hypotheses that (1) pQCT allows for an improvement of the prediction of bone strength at the distal radius beyond that of bone mass, as measured with DXA under in situ conditions; (2) nonsite-specific measurements (DXA, spinal QCT, and calcaneal QUS) display significantly lower correlations with radial failure loads than site-specific DXA or pQCT; (3) calcaneal QUS can, nevertheless, add significant independent information (to site-specific bone mass) in predicting bone strength of the distal radius; and (4) the predictive ability of the various techniques and parameters differs for bending and axial compression of the distal radius and, therefore, contribute differently to the prediction of the strength that is relevant in resisting a distal radius fracture.

MATERIALS AND METHODS

Study sample

One hundred forty specimens were examined from a course of macroscopic dissection, for which the donors had dedicated their body several years before death.(30) Bone biopsy specimens were taken from the left iliac crest and prepared for histomorphometry as described previously.(30) Three specimens with signs of malignancy and seven specimens with signs of renal osteopathy were discarded. One additional specimen was discarded because of an osteolytic lesion in the QCT scans. Of the remaining 129 specimens, 82 were women (age, 82 ± 8.4 years) and 47 were men (age, 77 ± 11.1 years).

Bone densitometry with DXA, pQCT, QCT, and QUS

DXA measurements were performed before the course with intact skin and soft tissues, using a DPX-L scanner (GE Lunar, Madison, WI, USA). The bone mineral content (BMC; g) and the projected areal bone mineral density (BMD; g/cm2) were obtained for the left forearm (distal, 33%; radius, ulna, and forearm = both bones together), the lumbar spine (AP direction), the left femur (neck, trochanter, and total proximal femur), and the total body. We have shown previously that formalin fixation does not critically affect these measurements(12) and that the precision is similar to that reported for in vivo conditions.(30)

pQCT measurements were performed with intact soft tissues at the distal radius (4, 20, and 33% from the wrist joint), using an XCT 2000 scanner (Stratec Medizintechnik, Pforzheim, Germany). Technical details and the precision have been described previously.(31,33) In brief, the bone mineral content (CNT; g/mm), the cross-sectional bone area (CSA), and the bone mineral density (g/cm3) were determined at the 4% site for the entire cross-section as well as for the trabecular (<400 mg/cm3) and subcortical (>400 mg/cm3) compartment. At the 20% and 33% sites, these parameters were determined for the entire cross-section as well as for the cortical compartment (>710 mg/cm3). Additionally, geometric properties of the cortical shell were determined, including the cortical thickness (Crt.Th; mm), the polar moment of inertia (1 mm4), the minimal and maximal (bending) moment of inertia,(14) and the density-weighted polar moment of resistance (R; mm3), which also has been termed the strain-stress index.(15,21) These parameters originate from engineering principals for estimating the torsional and bending strength of structures.(14)

QCT measurements of the lumbar spine (L2-L4) were obtained ex situ with a Somatom Plus 4 CT scanner at 80 kV (Siemens, Erlangen, Germany). The standard settings of the manufacturer were applied (slice thickness, 10 mm; in-plane resolution, 0.42 mm), and a phantom with equivalents of 0 mg/cm3 and 200 mg/cm3 hydroxylapatite was used to convert Hounsfield units into density values. To avoid artifacts from trapped air, the specimens were degassed under water at 50 mbar, sealed in thin polyethylene bags, and measured within an aqueous surrounding. Images were obtained at the midvertebral level, the trabecular and cortical density of L2-L4 being determined automatically by the software. The root mean square (RMS) average CV% of repeated measurements (4 repetitions in 14 specimens with repositioning, obtained at different days) was 2.9% for trabecular and 3.1% for (sub)cortical density. For determination of the CSAs of the vertebral bodies, the precision error was 1.2%.

QUS measurements of the calcaneus were performed ex situ to avoid potential artifacts from gas in the soft tissues of formalin-fixed cadavers. The calcanei were dissected clean of the surrounding tissues and degassed as described previously. Measurements were performed with an Achilles+ scanner (GE Lunar) sealed for underwater usage to determine the speed of sound (SOS), broadband ultrasound attenuation (BUA), and stiffness index (SI).(34) The calcanei were measured in a temperature-controlled water bath (37°C) at an anatomically defined position and were allowed to equilibrate to temperature before measurement. The precision with this system was found to be in the range of in vivo measurements.(34) Although formalin fixation (>1 year) decreased the SOS and SI, the values after fixation were highly correlated with those before embalmment.(34)

Radiographic analysis

The bones were excised after the course and cleaned of muscles and soft tissues. However, in the right forearm, the soft tissues were left intact, including tendons, ligaments, and the interosseous membrane. The wrist joint and hand also were left intact, to allow for the application of loads for mechanical testing. The forearms and the spines were X-rayed in two planes with a Siemens X-ray system (Polyphos 30 M; Siemens), and all other bones were X-rayed with a Faxitron Cabinet X-ray system (model 43885A, Faxitron X-ray systems; Hewlett Packard, McMinnville, OR, USA). The films were evaluated by an experienced radiologist, to exclude fractures or other local bone disorders. Spinal DXA and QCT values were discarded in cases in which fractures or deformities were detected in lumbar vertebrae.

Mechanical testing

A three-point bending test was performed on the left radius, 33% proximal to the wrist joint (Fig. 1A). Five specimens could not be tested because of osteosynthetic material or suspect of previous fracture. To test identical anatomic regions in all specimens, the shaft was fixed at 16.5% and 49.5% of the individual bone length, using two elements that were able to rotate and translate freely (Fig. 1A). From the failure load at 6.5 mm/s and from the individual moment arms, the failure moment (Nm) was determined. Five specimens were excluded after testing because of fracture at a site other than the 33% location. From the distal end of the same radius, a 10-mm planoparallel axial slice was removed adjacent to the joint surface (Fig. 1B), using a diamond band saw (Exact Trennschleifsystem, Norderstedt, Germany). The CSA of both sides of the slice were measured on scaled paper to allow for the calculation of the failure stress after testing (peak load divided by CSA [N/mm2]). An axial compression test was performed at a 0.8-mm/s loading rate (Fig. 1B), recording the initial peak (failure load) of the load-displacement curve with a subsequent drop of >10%.

Figure FIG. 1.

Drawings showing the mechanical testing conditions (A) three-point bending (left radius), (B) axial compression (left radius), and (C) fall simulation (right forearm).

On the right forearm, a fall was simulated.(13,26) The test was performed with an Instron testing machine (model 4302; Instron, High Wycombe, UK). The ulna and radius were embedded in plaster (True Rock; Benzer Dental AG, Zürich, Switzerland), with the hand in 70° dorsiflexion and 10° radial abduction and with the forearm in pronation (Fig. 1C). The load was applied through a wedge (the loading rate, 3.3 mm/s). Failure loads were recorded as the peaks of the load-displacement curves and were only considered if a distal radius fracture occurred that could be classified according to the Arbeitskreis Osteosynthese Fragen (AO) classification.(35) Briefly, fractures were classified into type A (extraarticular), B (partly articular), and C (intraarticular), with increasing complexity from 1 to 3. The type of fracture was determined from two X-rays (two planes) and from surgical preparation of the fracture zone. We observed 42 A2 fractures, 19 type A3, 1 type B1, 14 type B2, 2 type B3, 8 type C1, 14 type C2, and 2 type C3 fractures. Twenty-seven specimens were discarded because they could not be classified as one of these types of fractures (e.g., shaft fractures, ulnar fractures, or isolated scaphoid fractures).

Statistical analysis

Differences between genders were assessed using the unpaired t-test. Simple linear regression analysis was used to determine the correlation of failure strength between different loading configurations and with the densitometric parameters. Fisher z-transformation was used to assess whether certain coefficients were significantly higher than others.

Stepwise multiple regression models (forward mode) were used to identify whether a combination of pQCT parameters from the metaphyseal (4%) and shaft location (20%) allowed for improved prediction of mechanical strength versus single variables. In the first model (model 1) we combined the total CSA and total density at the 4% and 20% site as well as all parameters describing specific properties of the trabecular (4% site) and cortical compartment (20% site). The total, cortical, and trabecular BMC were not entered at this step, because these represent direct products of the density and CSAs. In a second multiple stepwise regression model (model 2), we included the total content (but not the total area and density) at the 4% and 20% site, as well as the parameters describing specific properties of the trabecular and cortical compartment. In this way we determined whether certain parameters could improve the prediction of bone strength beyond that of total (cross-sectional) content alone. Stepwise regression models also were used to determine whether calcaneal QUS was able to contribute significant independent information to the site-specific bone mass (DXA) in predicting failure strength.

RESULTS

The bone failure loads and moments in bending were significantly higher in men than in women, and this also applied for the failure stress in axial compression (Table 2). The correlation of strength in axial compression and three-point bending was r = 0.81 (p < 0.001). Axial compression displayed a somewhat higher correlation coefficient with the fall configuration (r = 0.77; p < 0.001) than bending (r = 0.65; p < 0.001), but the difference was not significant. There were no obvious differences in failure loads in the fall simulation between groups with different fracture types (AO classification).

Table Table 2.. Descriptive Data of the Mechanical Tests with Mean Values (sd) for Men and Women, respectively
original image

DXA of the left distal radius BMC displayed correlation coefficients of r = 0.89 with the failure moment in three-point bending, r = 0.84 with the failure load in axial compression, and r = 0.70 with the failure load in a simulated fall (Table 3; Fig. 2). DXA of the ulna or total forearm displayed slightly lower associations (Table 3).

Figure FIG. 2.

Bar graph showing coefficients of determination (r2) between failure loads in three-point bending, axial compression, and a fall simulation and site-specific versus nonsite-specific densitometric measurements. BMC, bone mineral content (g); CNT, bone mineral content (g/mm); trab, trabecular; cort, cortical; US, calcaneal quantitative ultrasound; stiff. ind., stiffness index.

Table Table 3.. Correlation Between Bone Strength at the Distal Radius in Three-point Bending, Axial Compression, and a Fall Configuration with Densitometric Data
original image

With pQCT, the highest correlation coefficients in three-point bending were for the cortical content at 20% (r = 0.91), in axial compression for the total content at 20% (r = 0.81), and in the fall simulation for the total content at 4% or 20% (r = 0.67). However, these coefficients were not significantly higher than those obtained by DXA. pQCT measurements at the 33% measurement location were highly correlated with those at 20% and did not display higher correlation coefficients than those at 20% (data not shown). Stepwise regression models with variables from the 4% and 20% measurement site produced the following results for the various loading configurations:

  • (1) Three-point bending failure moment: r = 0.90 for model 1 (cortical area at 20%) and r = 0.91 for model 2 (total content at 4% and cortical content at 20%);
  • (2) Axial compression failure load: r = 0.84 for model 1 (cortical area at 20% and polar moment of resistance at 20%) and r = 0.87 for model 2 (total content and trabecular area at 4% as well as cortical area and polar moment of resistance at 20%);
  • (3) Axial compression failure stress: r = 0.70 for models 1 and 2 (trabecular density at 4% and cortical thickness at 20%);
  • (4) Fall simulation failure load: r = 0.72 for model 1 (total density at 4% and polar moment of resistance at 20%) and r = 0.75 for model 2 (total content and trabecular area at 4% and polar moment of resistance at 20%).

Although the combination of various pQCT parameters tended to improve the prediction of failure loads versus single variables, the multiple regression coefficients were not significantly higher than those obtained by DXA (Table 3; Fig. 2).

For three-point-bending moments and axial failure loads, all nonsite-specific measurements displayed significantly (p < 0.01) lower correlation coefficients than site-specific DXA or pQCT (Table 3; Fig. 2). For the fall simulation, nonsite-specific measurements also displayed significantly lower associations (p < 0.05 to 0.01). Calcaneal QUS tended to produce lower correlation coefficients than other nonsite-specific analysis in three-point bending and axial compression but was equivalent to nonsite-specific DXA in the fall simulation. In predicting radial failure stress in axial loading, QUS tended to display a higher correlation than other nonsite-specific measurements, with the coefficients being in the range of those of site-specific analysis (Table 3). The calcaneal SI was able to contribute significant independent information to site-specific DXA in predicting axial failure stress but none of the QUS parameters were part of multiple regression models when predicting failure loads in any of the loading configurations.

The correlation coefficients for site-specific and nonsite-specific measurements were generally highest for three-point bending, followed by the axial compression and eventually by fall simulation (Fig. 2). Only calcaneal QUS displayed a higher correlation with axial compression than with bending.

DISCUSSION

In this study we tested the hypotheses that (1) pQCT allows improvement of the prediction of bone strength at the distal radius beyond that of bone mass (in situ DXA); (2) nonsite-specific measurements (DXA, spinal QCT, and calcaneal QUS) display a significantly lower association with radial failure loads than site-specific DXA or pQCT; (3) calcaneal QUS can, nevertheless, add significant independent information to bone mass in predicting bone strength; and (4) the predictive ability of densitometric techniques and parameters differs for bending and axial compression.

The forearm (distal radius) has been identified as an important site of fracture, with a high likelihood of relevant attribution to risk from osteoporotic bone loss, particularly in elderly women.(1) Distal radius fractures also cause substantial expenditures (385 million US$/year) to the health care system,(2) and their age-related increase in incidence has been shown to precede that in the hip and spine.(36) Experimentally, we have shown that bone failure loads are only moderately, but significantly, associated between the distal radius and other sites.(30) A recent meta-analysis(8) of clinical studies has convincingly shown that patients with fracture generally suffer from an increased risk of subsequent fractures. Specifically, fractures of the forearm were associated with a high relative risk of subsequent fractures at the hip, spine, and other sites.(8) The increase in relative risk of fracture at all sites was similar for radius fractures that had occurred in adolescence compared with those at or after menopause. Cuddihy et al.(7) concluded from a retrospective study into forearm and subsequent fractures that a sentinel forearm fracture should not be ignored in men and that the opportunity for therapeutic intervention (e.g., hormone replacement therapy) at the time of distal radius fracture was underutilized in women. Regarding the predictive ability of densitometric measurements, Marshall et al.(37) maintained, from a meta-analysis of ∼2000 fractures, that all DXA measurement sites (including the distal forearm) had similar predictive abilities, with the exception that spinal measurement was the best predictor for vertebral fracture and femoral measurement was the best for hip fracture. It has been shown also(38) that the long-term prediction of bone densitometry at the distal forearm (up to 11 years before fracture) was similar in magnitude compared with short-term prediction (<3 years before fracture). Taken together, these results suggest that the noninvasive diagnosis of loss of bone mineral at the distal radius may be the earliest chance to detect individuals at risk of osteoporotic fracture at this and other sites. Because it is easier to prevent bone loss at an early stage than to restore bone strength at later stages of the disease, early diagnosis of reduction in strength at the distal radius thus may be the first chance to identify patients who profit from treatment.

We have selected an experimental (biomechanical) study design because bone strength cannot be quantitatively determined in vivo and the bending and compression component of distal radius strength cannot be separated in an analysis of clinical fractures. Moreover, longitudinal studies into the value of different densitometric techniques require long time intervals and are complicated by cumulative X-ray exposure arising from application of multiple measurement techniques in the same subject (usually healthy volunteers). No previous experimental study has determined the strength of the distal radius in a fall simulation as well as in axial compression and bending, has assessed the ability of site-specific (DXA and pQCT) versus nonsite-specific X-ray techniques (DXA, QCT, and QUS), and has analyzed the correlation of calcaneus QUS with distal radius strength in different loading configurations. In particular, no other study has reached a comparable sample size (Table 1). In contrast to specimens obtained from pathological dissection, our study sample did not include a preselection of highly pathological cases and thus may be considered representative for the elderly population in southern Germany.(30) By obtaining X-rays from all skeletal regions and performing a histomorphometric analysis of the iliac crest, we were able to apply exclusion criteria similar to clinical studies. Regarding the fixation, Edmonston et al.(39) and Augat et al.(14) concluded that mechanical strength of entire bones is affected only minimally by formalin solution and that the correlation between BMC and mechanical strength is not altered. In this context, it also must be noted that the distribution of types of distal radius fractures (AO classification) closely resembled that seen clinically. We are not aware of epidemiological studies into the distribution of AO fracture types in elderly subjects, but a recent analysis of snowboard injuries(40) in Japan (99 distal radius fractures) revealed 51% A2, 19% A3, 7% B1, 11% C1, 9% C2, and 3% C3 fractures. Apart from the absence of B2 and B3 fractures in these Japanese snowboarders, this distribution is in high agreement with that produced by our biomechanical test. In addition, we have shown previously that even 10-month formalin fixation does not significantly affect DXA, not even under in situ conditions.(12) Moreover, densitometric measurements of this study are in the range of those reported in vivo,(41) except for QUS. The SOS and SI of the calcaneus have been shown to decrease significantly with fixation, but values were highly correlated with those before fixation in fresh specimens.(34) Thus, the correlation between QUS and bone strength is likely not affected by the use of fixed specimens.

The highest correlation between densitometry and failure loads in our study generally was observed for three-point bending and the lowest was for the fall simulation. This is not surprising because of the more controlled mechanical conditions during bending, in which failure occurs directly at the site of load application. In axial compression, the site of failure is confined to the cross-sectional specimen but may vary in location and bone compartment (trabecular or cortical). Similar correlation coefficients between failure loads in axial compression and bone densitometry also have been reported by other authors.(22–25) With a more complex failure test such as a fall simulation, the site of failure is more variable. Therefore, it is obvious that it is more difficult to predict fracture loads in this loading mode(13–15,26–29) than in the other loading configurations.

The association between failure loads in the fall configuration and radial DXA in our study (r = 0.70) is in the range of that reported previously,(13–15,26–29) although previous authors have performed measurements in the same arm (usually under ex situ conditions), whereas in our case, DXA was performed in the contralateral arm and by maintaining the soft tissue errors involved in clinical measurements. Myers et al.(26,27) developed a multiple-angle scanning technique to determine the moment of inertia of the distal radius from DXA scans. With this method, they were able to improve the correlation coefficients with failure strength versus the BMC in multiple regression models. Today, the moment of inertia and other geometric parameters of the cortical shell can be determined tomographically by pQCT.(9) Several authors have identified geometric parameters of the radial cortex to be important predictors of strength in the distal radius.(13–15,23) Contrary to Augat et al.,(15) however, other investigators(13,24,28,29) reported that these parameters did not attain higher correlations than the BMC from DXA. We observed a trend for multiple regression models from pQCT parameters to slightly improve the prediction of bone strength compared with single pQCT parameters or DXA, but the gain in information was not statistically significant. Therefore, we found no evidence that pQCT is superior to DXA in determining the mechanical strength of the distal radius, even under in situ measurement conditions, including DXA soft tissue artifacts.(11,12) Likely explanations are that soft tissue artifacts of DXA at the distal radius are small, the bone mass measured with DXA in the distal radius is predominantly determined by cortical bone, and the bone mass is distributed in very similar ways throughout the cross-section in most individuals.

We find pQCT parameters at 20% and 33% radial length to be of similar value in predicting mechanical strength, even for the three-point bending test, in which failure occurred specifically at the 33% site. Interestingly, the pQCT parameters from the shaft also displayed higher coefficients with failure of the metaphysis (axial compression) than the parameters from the metaphyseal site itself. Although this may appear paradox, similar results also have been reported by Augat et al.,(14,15) likely because cortical properties can be measured more accurately with pQCT at the shaft than at the metaphysis. However, we found no differences for pQCT parameters from the shaft or metaphyseal location in predicting strength in the fall configuration.

Wu et al.(28) and Njeh et al.(29) reported phalangeal QUS to display similar correlations with failure as DXA or pQCT and to add independent information to site-specific bone mass. In this study, QUS of the calcaneus displayed significantly lower correlation coefficients with failure loads in all loading configurations compared with site-specific measurements and did not contribute to failure load prediction in stepwise regression models. Potential reasons for the different outcomes include the small sample size (Table 1), measurements at the phalanges are closer to the site of interest than those at the calcaneus, or phalangeal QUS properties depend more strongly on cortical bone,(42) whereas QUS of the calcaneus is determined primarily by trabecular properties.(43) Interestingly, however, calcaneal QUS reached similar correlations as DXA for axial failure stress and was an independent predictor, potentially because calcaneal QUS carries less information on bone size than DXA.

We expected densitometric measurements at predominantly cortical sites (e.g., DXA of the femoral neck) to be correlated more highly with the bending component of strength at the radial shaft, and measurements at predominantly trabecular sites (e.g., greater trochanter or spine) to be associated more highly with failure of the radial metaphysis in axial compression. However, this did not become evident in our study. Potential reasons may be that bone properties are very heterogeneous between skeletal sites,(17,33,44,45) and that the compressive strength of the distal metaphysis is more strongly determined by the properties of the (sub)cortical rather than the trabecular compartment.

In conclusion, this experimental study shows that (1) a combination of pQCT parameters (specifically geometric properties of the cortical shell) tends to improve the prediction beyond that of bone mass BMC alone, but correlation coefficients are not significantly higher than those for site-specific DXA; (2) indeed, nonsite-specific measurements display significantly lower correlation coefficients with failure loads compared with site-specific analysis, independent on which measurement technique; (3) calcaneal QUS adds independent information to site-specific bone mass only in predicting axial failure stress in axial compression but not in predicting failure loads in any other loading configuration; (4) calcaneal QUS tends to be a superior predictor of the compressive (rather than the bending) component of distal radius strength, whereas all other (site-specific and nonsite-specific) parameters exhibit no obvious tendency to predict either component of strength with higher accuracy than the other. The noninvasive diagnosis of loss of strength at the distal radius thus should primarily rely on site-specific measurements with either DXA or pQCT. Because loss of strength at the radius has been shown to precede that in the spine and femur, this may be the earliest chance to detect individuals at risk of osteoporotic fracture and to initiate therapeutic treatment.

Acknowledgements

We thank Günther Delling and coworkers (Abteilung Osteopathologie, Universitätsklinikum Eppendorf, Hamburg, Germany) for the histomorphometric analyses, Jan Grimm (Medizinische Physik, Klinik für Diagnostische Radiologie, Kiel, Germany) for reading the spinal X-rays, and Horst Russ (Anatomische Anstalt München) for drawing Fig. 1. Albert Albrecht (AO-Forschungsinstitut, Davos), Markus Bachmeier, Boriana Barth, Hans-Jürgen Becker, Gudrun Goldmann, Oliver Groll, and Nadine Krefting (Musculoskeletal Research Group, Institute of Anatomy München) are acknowledged for their help with radiography, densitometric measurements, and biomechanical testing. This work has been supported by a grant from the Deutsche Forschungsgemeinschaft (DFG LO 730/2-1). Stratec Medizintechnik and Südmedica are to be thanked for further support.

Ancillary