DXA and pQCT predict pertrochanteric and not femoral neck fracture load in a human side-impact fracture model

Authors


Abstract

The validity of dual energy X-ray absorptiometry (DXA) and peripheral quantitative computed tomography (pQCT) measurements as predictors of pertrochanteric and femoral neck fracture loads was compared in an experimental simulation of a fall on the greater trochanter. 65 proximal femora were harvested from patients at autopsy. All specimens were scanned with use of DXA for areal bone mineral density and pQCT for volumetric densities at selected sites of the proximal femur. A three-point bending test simulating a side-impact was performed to determine fracture load and resulted in 16 femoral neck and 49 pertrochanteric fractures. Regression analysis revealed that DXA BMD trochanter was the best variable at predicting fracture load of pertrochanteric fractures with an adjusted R2 of 0.824 (p < 0.0001). There was no correlation between densitometric parameters and the fracture load of femoral neck fractures. A significant correlation further was found between body weight, height, femoral head diameter, and neck length on the one side and fracture load on the other side, irrespective of the fracture type. Clinically, the DXA BMD trochanter should be favored and integrated routinely as well as biometric and geometric parameters, particularly in elderly people with known osteoporosis at risk for falls. © 2013 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 32:31–38, 2014.

Hip fractures are a cause of significant morbidity and mortality worldwide and are a substantial economic burden.[1, 2] About 54% of women aged 50 years or older will have an osteoporotic fracture during their lifetime.[3] Furthermore, approximately 4% of the patients older than 50 years with a hip fracture die in the hospital and 24% die within 1-year after the hip fracture.[4]

Thus, the prevention of osteoporotic femur fractures is of increased significance to reduce morbidity and mortality on the one hand and socio-economic burden on the other hand.

At present, areal bone mineral density (aBMD) as obtained from dual energy X-ray absorptiometry (DXA) is the best established clinical parameter for diagnosis and follow-up of osteoporosis.[5, 6] It is widely used as a predictor of fracture risk of the hip although there is considerable overlap in the densitometry measurements of individuals who have fractured and those who have not.[7] In addition, over the past years the measurement of volumetric bone mineral density (vBMD) by peripheral quantitative computed tomography (pQCT) gained more and more interest in the field of basic research. Several ex vivo studies indicate that vBMD measured by pQCT has the highest predictive value for mechanical strength of cancellous bone of the proximal and diaphyseal femur.[8-10]

In this context, there is little data available comparing the predictive value of areal and volumetric BMD measuring techniques on the mechanical strength of the proximal human femur. Moreover, recent publications comparing the predictive value of different bone density measuring techniques on the femoral fracture strength significantly vary for the mechanical testing setup (axial load vs. side impact), the resulting mean fracture load (range from 3.7 to 8.5 kN), the fracture location (neck vs. trochanteric), and the age and processing of the specimens.[11-15]

Therefore, to better understand the pathogenesis of proximal femoral fractures and therefore contribute to an improvement of fracture prevention, the aim of this study was to compare the predictive value of areal and volumetric BMD measuring techniques at different sites of the proximal femur on the mechanical strength and by this to detect the best predictor of proximal femoral fracture strength.

METHODS

Specimens

Sixty-five human cadaveric femora (28 male and 37 female, mean age 67.5 ± 14.5 years, mean body weight 73.7 ± 19.9 kg, mean height 166.6 ± 9.2 cm) were retrieved 24 h post mortem. Anteroposterior and axial radiographs were obtained and examined by two orthopaedic surgeons (MG, JB) to exclude previous fracture, local area of lucency or pathology (e.g., arthritis, metastases). Multiple biometric parameters were measured ((body weight, body height, femoral head diameter, femoral neck length, neck-shaft-angle (CCD)), and the specimens were wrapped into saline soaked gauze and stored at −30°C in tightly sealed plastic bags until further processing.

Areal and Volumetric Bone Mineral Density Measurements Dual Energy X-Ray Absorptiometry (DXA)

Each femur was scanned once with a GE Lunar Prodigy Scanner (GE Lunar Corporation, Madison, WI). On the basis of this scan the four standard ROIs according to the International Committee for Standards in Bone Measurement Standardization ((1) trochanteric region, (2) femoral neck, (3) Ward's region and (4) total region as shown in Fig. 1) were analyzed for each femur.[16, 17] We used the GE Lunar Prodigy Scanner and GE's standard femur analysis protocol with the scanner software version enCORE 11.x. The analysis was fully automated. We did not correct the automatic placements of the ROIs nor vary the size of the ROI-boxes manually.

Figure 1.

DXA (left) and pQCT (right) image of representative specimens prior to mechanical testing. pQCT-sections of the neck and the trochanter were performed along the black dotted line as demonstrated on the native proximal femur image (middle).

To simulate the surrounding soft tissue, all specimens were placed in a vessel filled with water up to 15 cm.[14, 18]

Peripheral Quantitative Computed Tomography (pQCT)

To evaluate vBMD in (mg/cm3) for all 65 proximal femur specimens, pQCT sectional images were taken and evaluated using an automated computer program (XCT960, Stratec, Pforzheim, Germany). By this method the cortical bone area (CBA) was defined using a threshold algorithm that detected bone with high density in a defined region of interest (ROI). All voxels with a lower attenuation coefficient than the given threshold were eliminated. The threshold levels were kept constant for all patients (trabecular bone: 200–600 mg/cm3; cortical bone >600 mg/cm3).

Following generation of a scout view, sectional images were taken perpendicular to the femoral neck axis through (a) the point of minimum femoral-neck cross-sectional area and (b) the middle of the intertrochanteric region (Fig. 1). On the basis of the two measurement scans of the entire proximal femur, (1) trabecular, (2) cortical, and (3) total bone mineral density were calculated for both regions.

Mechanical Testing

For biomechanical testing, the femoral condyles were removed and the shaft uniformely shortened to a length 18 cm below the tip of the Tuberculum innominatum of the Trochanter major using an oscillating saw. According to the previously described method,[19-21] the distal femoral shaft was fixed with an intramedullary rod as well as an antirotational screw and embedded in acrylic cement (Moldastone, Heraeus Medical GmbH, Wehrheim, Germany) in a metal cylinder (leaving 12 cm of femur shaft above the cement) and then oriented in the Universalprüfmaschine inspekt desk (50 kN) (Hegewald Peschke, Nossen, Germany). Analogous to previously described studies[20-23] an impact loading was applied, simulating a fall on the greater trochanter (Hayes-fall with 15° of internal rotation and 10° from the horizontal plane (Fig. 2). Load was applied to the femoral head by a non-deformable silicone-based acetabulum attached to the actuator of the test machine, with the trochanter supported on a non-deformable silicone-based pad. An actuator displacement rate of 2 mm/s was applied and the load–displacement behavior was recorded. The fracture load (maximum load applied until fracture occured, [N]), the displacement to failure (the way the silicone-based acetabulum moves downward during each trial) and the energy to failure (area under the curve to the point of maximum load, [Nm]) were calculated. The fracture type was defined according to the AO-classification system.

Figure 2.

Positioning of the femur in the universal testing machine simulating the fall on the greater trochanter (Hayes-fall).

Statistical Analysis

Regression analysis was performed using Pearson's Correlation. The statistical software used was R 2.15, R Development Core Team.

RESULTS

All specimens fractured at the proximal femur, no femoral shaft fracture occured. Subclassified according to the AO-classification system there were 49 pertrochanteric (AO 31A) and 16 femoral neck (AO 31B) fractures. The mean fracture load for all proximal femoral fractures was 3146 ± 1146 N (Fig. 3). Mean fracture load for all pertrochanteric fractures was 3107 ± 1066 N and did not significantly differ from the mean fracture load for all femoral neck fractures with 3206 ± 1264 N. In addition to the values for fracture load, Table 1 shows the values for stiffness as well as energy to failure.

Figure 3.

Exemplary force-displacement curves, (a) femoral neck fracture, (b) pertrochanteric fracture.

Table 1. Mean Values and Standard Deviations of Fracture Load, Displacement to Failure, and Energy to Failure (n = 65 Specimens)
 Proximal Femoral Fractures (n = 65)Pertrochanteric Fractures 31 A (n = 49)Femoral Neck Fractures 31 B (n = 16)
Fracture load (N)3146 ± 11463107±10663206 ± 1264
Displacement to failure (mm)30.0 ± 6.630.5 ± 6.129.5 ± 6.8
Energy to failure (Nm)11 8 ± 11.012.1 ± 10.610.8 ± 10.2

Values obtained from areal and vBMD measurements as well as for the biometric and geometric parameters are shown in Table 2. Correlation analysis showed that there was a significant correlation between DXA and pQCT-values at the femoral neck (0.854, p < 0.0001) as well as between DXA and pQCT-values at the trochanter (R2 = 0.831, p < 0.0001, Table 3).

Table 2. Mean Values and Standard Deviations of all Densitometric, Biometric, and Geometric Parameters. The Resulting 65 Proximal Femoral Fractures were Further Subdivided into 49 Pertrochanteric and 16 Femoral Neck Fractures
 Proximal Femoral Fractures (n = 65)Pertrochanteric Fractures 31. A (n = 49) 28 Female, 21 MaleFemoral Neck Fractures 31.B (n = 16) 9 Female, 7 Male
DXA BMD neck (g/cm2)0.773 ± 0.1650.784 ± 0.1570.740 ± 0.118
DXA BMD trochanter (g/cm2)0.714 ± 0.1630.722 ± 0.1500.689 ± 0.156
DXA BMD ward (g/m2)0.574 ± 0.1710.590 ± 0.1800.525 ± 0.130
DXA BMD shaft (g/cm2)1.032 ± 0.2211.070 ± 0.1900.914 ± 0.191
DXA BMD total (g/m2)0.854 ± 0.1810.869± 0.1600.808 ± 0.154
pQCT BMD neck cortical (mg/cm3)1251 ± 151256±151240 ± 14
pQCT BMD neck trabecular (mg/cm3)335 ± 70348 ± 64295 ± 46
pQCT BMD neck total (mg/cm3)364 ± 77370 ± 69346 ± 56
pQCT BMD trochanter cortical (mg/cm3)1228 ± 241235 ± 261206 ± 43
pQCT BMD trochanter trabecular (mg/cm3)299 ± 60304 ± 51283 ± 48
pQCT BMD trochanter total (mg/cm3)308 ± 62321 ± 53289 ± 50
Age67.5 ± 14.567.3 ± 15.069.7 ± 14.2
Body weight (kg)73.7 ± 19.971.0 ± 21.477.4 ± 16.8
Body height (cm)166.6 ± 9.2165.5 ± 8.2168.3 ± 12.5
Femoral head diameter (mm)44.3 ± 3.944.0 ± 3,645.1 ± 4.8
Femoral neck length (mm)30.7 ± 3.430.5 ± 3.231.6 ± 3.9
Neck-shaft-angle (CCD(°))128 ± 2.7128 ± 2.5127 ± 3.3
Table 3. Correlation Matrix for Selected Areal Versus Volumetric Bone Mineral Density Measurements
 pQCT BMD Neck Total (mg/cm3)pQCT BMD Trochanter Total (mg/cm3)
  1. Values represent Pearson correlation coefficients as obtained from linear regression analysis (*p < 0.01; **p < 0.0001).
DXA BMD neck (g/cm2)0.854**0.723**
DXA BMD trochanter (g/cm2)0.624**0.831**
DXA BMD total (g/m2)0.777**0.812**

Correlation analysis for densitometric parameters versus fracture load (Table 4) revealed a highly significant correlation between several densitometric parameters and the fracture load of pertrochanteric fractures. In this regard DXA BMD trochanter (g/cm2) emerged to be the best predictor of pertrochanteric fracture load (R2 = 0.824, p < 0.0001, Fig. 4).

Table 4. Correlation Matrix for Densitometric, Biometric, and Geometric Parameters Versus Fracture Load
 Proximal Femoral Fractures (n = 65)Pertrochanteric Fractures 31. A (n = 49)Femoral Neck Fractures 31. B (n = 16)
  1. Values represent Pearson correlation coefficients as obtained from linear regression analysis (*p < 0.01; **p < 0.0001). The resulting 65 proximal femoral fractures were further subdivided into 49 pertrochanteric and 16 femoral neck fractures.
DXA BMD neck (g/m2)0.688**0.773**0.448
DXA BMD trochanter (g/cm2)0.737**0.824**0.468
DXA BMD ward (g/cm2)0.606**0.683**0.436
DXABMD shaft (g/m2)0.621**0.702**0.483
DXA BMD total (g/m2)0.695**0.781**0.482
pQCT BMD neck cortical (mg/cm3)0.0620.0750.049
pQCTBMD neck trabecular (mg/cm3)0.477**0.528**0.394
pQCT BMD neck total (mg/cm3)0.492**0.584**0.290
pQCT BMD trochanter cortical (mg/cm3)0.1420.1470.226
pQCT BMD trochanter trabecular (mg/cm3)0.557**0.641**0.408
pQCT BMD trochanter total (mg/cm3)0.544**0.636**0.370
Age−0.226−0.206−0.342
Body weight (kg)0.469**0.464**0.442
Body height (m)0.455**0.429**0.523*
Femoral head diameter (mm)0.310*0.237*0.443*
Femoral neck length (mm)0.263*0.245*0.266*
Neck-shaft-angle (CCD (°))0.0800.0280.253
Figure 4.

Correlation between areal BMD (g/cm2) and volumetric BMD (mg/cm3) and fracture load (N) of 49 tested pertrochanteric femur fractures. (a) DXA BMD trochanter (g/cm2) versus fracture load (N), (b) pQCT BMD trochanter total (mg/cm3) versus fracture load (N).

In contrast, a significant correlation between densitometric parameters and the fracture load of femoral neck fractures was not found (Fig. 5).

Figure 5.

Correlation between areal BMD (g/cm2) and volumetric BMD (mg/cm3) and fracture load (N) of 16 tested femoral neck fractures. (a) DXA BMD neck (g/cm2) versus fracture load (N), (b) pQCT BMD neck total (mg/cm3) versus fracture load (N).

Correlation analysis for biometric and geometric parameters detected a significant correlation between body weight, body height, femoral head diameter, and femoral neck length on the one side and fracture load on the other side. There was no correlation between neck-shaft-angle and fracture load. These findings were irrespective of the fracture type (Table 4).

DISCUSSION

Comparing the predictive value of biometric and geometric parameters as well as areal and volumetric BMD measuring techniques on the mechanical strength of the proximal human femur is of significant importance in the prevention of osteoporotic femur fractures.

The correlation analysis for densitometric parameters versus fracture load revealed a highly significant correlation between several densitometric parameters and the fracture load of pertrochanteric fractures. In this regard DXA BMD trochanter (g/cm2) emerged to be the best predictor of pertrochanteric fracture load with R2 = 0.824, however, a correlation between densitometric parameters and the fracture load of femoral neck fractures was not found.

The predictive nature of a low BMD has been reported for femoral neck fractures,[23-25] but could be confirmed by Heini[19] just for the trochanteric region and the simulated fall on the greater trochanter. We can confirm this in the presented distinct larger cohort.

In contrast to several ex vivo studies indicating that vBMD measured by pQCT has the highest predictive value for mechanical strength of cancellous bone of the proximal and diaphyseal femur,[8-10, 13] we recommend that DXA BMD trochanter should be or remain the focus clinically which, however, is already widely used in standard testing setups. This counts particularly for those elderly osteoporotic people with medical and neurological disorders who are prone for falls and dominantly obtain pertrochanteric and lateral femoral neck fractures.

Wachter et al.[8] confirmed the high predictive value of BMD for the mechanical competence of cancellous bone of the intertrochanteric region. They described that quantification of cancellous bone structure by image analysis of CT scans may provide additional qualitative information for the analysis of bone strength. However, these findings are based on cylindrical biopsies from the intertrochanteric region. Nevertheless, it remains to speculate on, why DXA measures are more accurate predictors of pertrochanteric fractures. This encourages the multivariant condition of fractures and the method itself. Aside from better three-dimensional qualitative statements regarding structural information in CT measurements, the two-dimensional nature of the scan captures an element of bone size that has an independent contribution of bone strength.[26]

Determining a threshold of a critical BMD would be of great interest. This, however, is hard to determine particularly from the experimental situation, as BMD is only one factor of a hip fracture which represents a multivariant condition (BMD, geometric parameters, reactivity, soft tissue, floor etc.) as mentioned above. Similarly and acknowledged in current practical guidelines, decisions to treat should be based on a comprehensive assessment of the risk of fracture rather than the fulfilment of a single diagnostic criterion.[26]

Furthermore, we could again demonstrate a correlation for biometric and geometric parameters and fracture load, irrespective of the fracture type. This is in accordance to the literature. The neck-shaft angle, the diameter of the neck and the hip axis length have been cited as indicators of fracture risk[11, 12, 27-29] as well as an inverse correlation between neck-shaft angle and failure load, reflecting the increased mechanical lever-arm in our previous studies.[19-21] These findings seem to be unaffected by different mechanical testings and processing of the specimens (e.g., formalin-fixation). For example, Bonnaire et al.[11] used a mechanical stance arrangement with vertical loadings up to 12000 N necessary to fracture the bones. The simulation of this stance condition, however, is harder to mimic experimentally and fracture patterns did not show uniformity, as compared to the simulated fall.[19] Faulkner et al.[28] evaluated DXA scans of non fractured and previously fractured bones without a testing. Pinilla et al.[27] also used the testing arrangement referring to Courtney, however, with a distinct higher displacement rate and varying degrees of femoral rotation with a distinct higher rate of variation of fracture types. There are several studies evaluating femoral bone strength by finite element analysis, too.[30-33] Their advantages concerning accuracy and clinical transferability remain uncertain though.

With the mechanical arrangement described we obtained 75% pertrochanteric and 25% femoral neck fractures. Regarding the ratio between pertrochanteric and femoral neck fractures of the femur, Smektala et al.[34] showed an almost equal distribution between femoral neck (n = 1466) and pertrochanteric (n = 1450) fractures in a retrospective study on 2916 proximal femoral fractures. Skála-Rosenbaum[35] examined the epidemiological data of 3683 patients with hip fractures between 1997 and 2007. In this study trochanteric fractures accounted for 54.7% and femoral neck fractures for 45.3% of all proximal femoral fractures. Thus, by predominatly creating pertrochanteric fractures (pertrochanteric 49: 16 neck), the distribution of proximal femoral fractures in our biomechanical study differs from the clinical situation. Furthermore, we acknowledge the limitations of applying the results of our ex vivo study in vivo. We assessed proximal femur failure load using a protocol that simulated one type of fall and displacement rate. A lot of additional factors such as fall mechanics, muscular strength, and overlying soft tissue, as well as bone biometric and geometric parameters, influence an individual's risk of fracture in-vivo.[21, 29, 36-38] And concerning the heterogeneity typically associated with mechanical testing of human bones, heterogeneous patterns of load–displacement curves are also not unexpected.

A methodological limitation of our study is attributed to the technology of pQCT scanners. This technique pretends to analyse BMD on a volumetric basis, although this calculation is only performed on a few selected 2D slices which are acquired. In contrast to pQCT, true volumetric techniques as QCT which uses a high-resolution helical CT system for densitometry, compute the densitometry over the entire femoral volume.[39, 40] Furthermore, Hansen et al.[41] found that both cortical and trabecular structural parameters obtained from pQCT images hold information on bone strength complementary to that of BMD. Another limitation of our study is that pQCT is not a feasible technique to measure BMD in the human proximal femur in vivo. But apart from it's clinical use at the peripheral skeleton (distal radius), pQCT can also be used as an alternative technique to measure BMD at different skeletal sites in basic research. A number of recently published studies have performed pQCT scans to estimate femoral BMD on retrieved femur specimens.[41-44] Thus we wanted to integrate the pQCT-techique as an additional technique and to compare it to the well established method of DXA.

A second methodological limitation of our study is the use of a displacement-controlled mechanical testing device with just one displacement rate of 2 mm/s. To create proximal femoral fractures we used the testing setup simulating a fall on the greater trochanter described by Courtney et al. (Hayes-fall, 15). This is a well established fall condition and it has been approved by a number of biomechanical studies on proximal femoral fractures.[19-21, 23] However, one has to keep in mind the limitations of a displacement-controlled setup with a single displacement rate of 2 mm/s, as bone mechanical behavior is sensitive to strain rates, not to displacement rates.[31] This effect has to be considered as a potential source of variable fracture loads in our mechanical testing setup.

Of course, our results also cannot be transferred to the insufficiency fracture situation as in the stance condition.[19] Spontaneous fractures of the proximal femur are rare and approximately 90% of hip fractures are associated with a fall.[27] Furthermore, literature reveals different failure modes for proximal femoral fractures: Fractures that initiate under tension from the infero-medial neck, fractures that initiate form the superior aspect near the root of the femoral neck, and fractures that involve a complete collapse of the greater trochanter.[1, 45, 46]

The use of not solely severe osteoporotic bones of both male and female with resulting managing cohort size each might be another limitation. However, the cohort represents a normal distribution of the age group and therefore might better be generalized without bias. As described by Bauer et al.[14] male donors had larger femora even after adjustment for body size and height, but no differences in trabecular structure were found between males and females. In this context, the cut of all specimens into the same length might also represent a methodological limitation. The described standardized biomechanical Hayes-fall set-up, however, creates a better comparability with focus on the proximal femoral region and without confounders as leg length with lever-arm or angle.

To restore the mechanical strength of the proximal femur and to prevent patients at high risk from osteoporotic femoral fractures, a number of recent ex vivo studies have evaluated the potential role of femoroplasty as a new technique to effectively strengthen the proximal femur by cemented augmentation.[19-21, 47] Despite promising experimental results with this technique, clinical trials are lacking up to date. Moreover, medical treatment is often initiated too late and acting too slow to prevent fractures[48] and hip pads or an energy-absorbing floor[49, 50] fail due to low compliance and unclear effectiveness.[51]

There is a correlation between several densitometric parameters and the fracture load of pertrochanteric fractures with DXA BMD trochanter (g/cm2) being the best predictor, while there is no correlation between densitometric parameters and the fracture load of femoral neck fractures. Furthermore, there is a correlation for biometric and geometric parameters and fracture load irrespective of the fracture type as well. Clinically, the DXA BMD trochanter should therefore be favored and integrated routinely as well as biometric and geometric parameters, particularly in elderly people with known osteoporosis at risk for falls.

ACKNOWLEDGMENTS

All authors disclose any financial, professional, and personal relationships with other people or organisations that could have biased the presentation. There are no conflicts of interest.

Ancillary