Predicting pathological fracture in femoral metastases using a clinical CT scan based algorithm: A case-control study

We assessed whether there was a difference in attenuation measurements (in Hounsfield units - HU) and geometric distribution of HU between femora with metastatic lesions that fracture, and metastatic lesions that did not fracture nor underwent prophylactic fixation. Nine patients with femoral metastases who underwent CT and developed a pathological fracture were matched to controls. All femora were delineated in axial CT slices using a region of interest (ROI) tool; the HU within these ROIs were used to calculate: (1) the cumulative HU of the affected over the nonaffected side per slice and presented as a percentage, and (2) the cumulative HU accounting for geometric distribution (polar moment of HU). We repeated the analyses including cortical bone only (HU of 600 and above). CT-based calculations did not differ between patients with a lesion that fractured and those that did not fracture nor underwent prophylactic fixation when analyzing all tissue. However, when including cortical bone only, the pathological fracture group had a lower cumulative HU value compared to the no fracture and no fixation group for the weakest cross-sectional CT image (pathological fracture group, mean: 71, SD: 23 and no fracture and no prophylactic fixation group, mean: 85, SD: 18, p = 0.042) and the complete lesion analysis (pathological fracture group, mean: 78, SD: 21 and no fracture and no prophylactic fixation group, mean: 92, SD: 15, p = 0.032). The demonstrated CT-based algorithms can be useful for predicting pathological fractures in metastatic lesion