HR-pQCT Derived Forearm Failure Load
Our first objective aimed to compare HR-pQCT FE-based failure load and bone properties obtained from both standard and advanced cortical evaluations between the fixed and the anatomical regions. No significant difference was observed between the means of failure loads derived from the two regions. This might be because of the large overlap between the two regions (5. 8-9 mm), as it has been shown that both sites correspond with the location of forearm fracture (i. e. , resulted from a fall onto the outstretched hand) at their overlap. This finding is in line with that of Bonaretti et al. [3], where they showed that there is no difference between the means of failure loads of the two regions. Observing a smaller failure load for the 4% region than the fixed region was expected because the 4% region was located more proximal to the epiphyseal region, which seems to be weaker [18].
Conversely, most of the bone indices were significantly different between the two regions. Among these parameters cortical pores volume (-19. 1%), trabecular area (17. 4%), cortical porosity (-12. 6%), cortical apparent thickness (-6. 3%), total density (-6. 2%), cortical fine-structured thickness (-5. 4%), total area (4. 9%), and cortical pore diameter (-4. 7%) had largest changes relative to the fixed region, while variations in trabecular structural parameters were small. The large changes in cortical bone properties along a relatively small offset between the two regions suggest that these regions coincide with the transitional zone from diaphyseal to metaphyseal region of the bone. This is consistent with the large variations in cortical bone measurements along the ultra-distal region as showed by Boyd, which is also in line with high precision errors of cortical bone measurements due to variability in operator scan positioning indicated by Bonaretti.
Moreover, our results for cortical indices are in agreement with those of Bonaretti and Shanbhogue where they showed large changes in cortical parameters including cortical thickness and porosity. On the other hand, differences in trabecular micro-architectural parameters were smaller in this study compared to those of Bonaretti, which might be due to motion artifacts present during in vivo imaging, as it is shown that trabecular micro-architectural parameters are sensitive to such artifacts.
According to the result of the first objective implementing the 4% region is recommended when studying bone structural properties, while the fixed region is suitable for failure load estimations. Of note, we anticipate that by using the density-based (E-BMD) FE model, the failure loads of two regions will be different because of two reasons. Firstly, the E-BMD model partially compensates for partial volume effects since it does not need segmentation. Secondly, unlike the current standard model, the E-BMD model differentiates between the cortical and trabecular bones in terms of mineralization level by converting imaged BMD to elastic modulus.
The effect of using E-BMD model can be investigated in future studies. The second objective of this study was to evaluate FE-predictions of forearm failure load acquired from the fixed and the 4% regions in relation to experimental forearm failure load from fall configuration testing. Our results indicated that the fixed and 4% regions explained similar variance in experimental forearm failure load under fall configuration testing, with coefficients of determination of 0. 89 and 0. 87, respectively. This similarity was anticipated due to the similarity of failure loads between the regions, and because both sites correspond with the location of forearm fracture caused by a fall onto the outstretched hand. Our results provided evidence that HR-pQCT based FE models of bone sections can be used to clinically predict fall configuration forearm failure load of a population similar to this study (i. e. , postmenopausal women), without large variations in body size and limb length using either regions. However, in case of cross-sectional studies with large variations in limb length the fixed region might induce bias. The coefficient of determination of this study appeared higher than that reported in the study of Pistoia et. al. , (89% vs 66%). This can be mainly due to using fresh-frozen specimens, while Pistoia et al used embalmed specimens, for which the mechanical properties may change. Further, the coefficient of determination of this study appeared higher than that reported by Mueller et. al. ,(89% vs 73%). One explanation can be that the location of fracture might be different for males and females based on the limb length and might not lie within the region of interest used for FE analysis. Also, the values of failure load might be different between males and females. Therefore, a single regression analysis on both males and females might have reduced the coefficient of determination of the regression [18]. The strengths of this study relate to our analysis of postmenopausal females prone to fracture, mechanical testing protocol and use of fresh-frozen cadavers.
First, our bone specimens included older women, representing clinically relevant population at high risk for bone fragility and fracture. Second, we used intact fresh-frozen female forearms with a mechanical testing setup successful in simulating forearm fracture from falling onto outstretched hand.
Finally, our use of fresh-frozen cadaveric had two benefits. Firstly, we prevented motion artifact typically experienced during in vivo imaging. Secondly, fresh-frozen samples avoided mechanical property variation induced by embalming. In terms of limitation, first, the second objective of this study was limited to 13 specimens with similar sizes, which could be a reason for not noticing a different in FE-derived failure load of the two regions. Second, we measured the length of the radius as a representative measure of limb length whereas Bonaretti et al. measured the length of the ulna. Although this could lead to positional error, the difference between the position of the distal 4% region due to this error is comparable to the error associated with physically measuring the ulna or radius length. Finally, we found that defining the position of the reference line using the “skier” approach was challenging, particularly for samples with blurry scout view images, which can lead to a measurement error. However, accurate reference line placement with the standard method was also somewhat challenging, with reference line positioning precision errors corresponding to 3. 3% of total length of the region of interest.
In conclusion, HR-pQCT derived failure load was not significantly different between the fixed region and 4% region at distal radius, while structural indices were different. Therefore, we recommend the use of the 4% region when studying bone outcomes. Additionally, both regions explained similar variance in experimentally-measured forearm failure load. Thus, our results indicated that either region can be used to non-invasively estimate forearm failure load. However, implementing the anatomically standardized region may be beneficial.
