Hip fractures are amongst the most severe consequences of osteoporosis and are associated with high morbidity and mortality and a significant reduction in the patient’s quality of life (1). Hip fracture patients have a mortality of 20% within the first year (2) and 10 to 20% of hip fracture individuals can no longer live independently (3). Hip fracture risk depends on the integrity of the proximal femur and the likelihood of experiencing forces that exceed bone strength (4). With aging, the geometrical integrity of the hip is compromised and the risk of falling increases, resulting in older individuals having an increasing risk of hip fracture. Thus, it is important to identify individuals at high risk of fracture. While areal bone mineral density (aBMD) derived from dual X-ray absorptiometry (DXA) is the routine method to evaluate osteoporosis, studies have consistently shown that it has only moderate capability to predict hip fractures (5–11).

The cortical bone of the proximal femur has become a focus of interest leading to the increased application of hip quantitative CT (QCT) in clinical trials (12). However, few studies have assessed the association of cortical bone with hip fractures, and some of these have only applied cross-sectional slice-based cortex measurements (i.e. one slice or the average of several slices) (6, 9, 13–15) instead of 3D segmented methods. Several studies have used femoral QCT to measure bone shape, volumetric BMD distribution and cortical bone thickness (CortThick) distribution (6, 11, 14, 16–18), concluding that smaller cross-sectional area, lower trabecular vBMD and thinner cortical thickness were all associated with increased hip fracture risk. However, parameters that characterize the strength of specific sub regions of bone compartments, such as bending and buckling, up to now were mostly limited to two-dimensional assessments derived from DXA hip structural analysis (HSA) (19–21). Further, DXA HSA variables are not independent of DXA aBMD (12). Assessment of femoral geometry by the QCT MIAF-Femur application (MIAF: medical image analysis framework) and volume-based structural parameters introduced by Engelke may allow for assessment of bone strength indicators in greater detail (22). MIAF-Femur software is based on 3D segmentation of the whole proximal femur, which also allows for assessment of the femoral head in vivo (23).

This cross-sectional case-control study aims to explore the associations of the geometrical parameters such as cortical volume and thickness with acute hip fractures. We also aim to assess differences in femoral head size between female participants with and without hip fracture.

Based on the analysis of 562 participants enrolled in the CASH cross-sectional case–control cohort, our study shows structural differences between elderly women with and without hip fractures, and a combination of selected geometry variables resulted in equivalent discrimination power to the aBMD model reported previously (11, 24). Our study outcomes also confirm observations of previous studies that the addition of bone volume did not significantly improve hip fracture discrimination. However, inclusion of the FH volume may allow improved prediction of hip fracture propensity.

An interesting finding of this study was that elderly women with hip fracture had larger FH but smaller FN size compared to controls. The femoral head connects continuously with the femoral neck. Thus, the head directly participates in the weight-bearing transfer to the femoral neck and the femoral neck and the trochanter are affected by the stresses and strains in the femoral head (23). Therefore, with respect to hip fracture risk prediction, the traditional DXA regions such as the FN, TR and IT may not be fully adequate to capture the risk of hip fracture. However, up to now, only two studies have reported the relationship between bone deterioration of the femoral head and hip fractures. In the European Femur Fracture Study (EFFECT) the femoral head BMD was associated with hip fracture but there was no difference in femoral head volume between participants with and without hip fracture (11). In the other QCT study, loss of FH vBMD was also found to be related to hip fracture (27). Our findings demonstrated that femoral head volume discriminated hip fracture risk with an AUC value of 0.67 after adjusting for age, height and weight, and inclusion of the FH volume improved the power of the model (Figure 3). Associations between geometric features of the proximal femur and hip fracture have been extensively investigated. For example, the strength of the femur is associated with the shape and size of its cross sections, the lengths of its neck and shaft, the neck–shaft angle, etc. (12). Differences in geometry of the proximal femur between women with and without hip fracture (larger head but a smaller neck in fractured subjects) identified by our study offer a new view of the femur strength and may prove useful in the construction of finite element models.

In agreement with three previous QCT studies (11, 13, 26), our results confirmed that with respect to hip fracture discrimination, cortical volume is an inferior parameter compared to cortical thickness. Previous studies have shown the power of cortical bone in resisting fracture and in hip fracture risk prediction (10, 13, 14, 26, 28–31), although the accurate measurement of cortical bone is still challenging due to the partial volume effect (25). One BMD combined with one geometry variable, for example TR vBMD with one structural parameter (e.g. FN cortical thickness), would be the preferred method of discriminating hip fracture risk using hip QCT (12, 26). The ratio of cortical/trabecular bone mass, a measure of the internal distribution of bone, is a superior parameter to cortical thickness in discriminating hip fracture risk across the entire proximal femur.

The combination of selected geometry variables in this study resulted in a similar AUC value (0.805) as the use of aBMD alone (AUC 0.796 or 0.804) reported previously in case-control studies (11, 24). Further, the AUC values of the combination of selected geometry variables in this study were similar to those reported for reference aBMD in prospective studies, ranging from 0.70 to 0.86 (32–38). Although AUC and OR results varied amongst these studies of different datasets, evidence is accumulating for a slight improvement in hip fracture risk assessment. The resulting five best-subset models for discrimination of hip fractures are ordered according to the BIC information criterion of the best-subset procedure, which combines number of variables and goodness of fit of the binary regression model (26). Similar to an earlier study (12), the combination of BMD measures and geometric parameters improved association with hip fracture but results have to be validated in prospective cohort studies. Unfortunately, the radiation dose of QCT scans hampers the application in osteoporosis screening and frailty hip fracture risk assessments. The integration of QCT-based geometry evaluations may trigger a paradigm shift in hip fracture prediction, namely, under certain circumstances, such geometry parameters could be derived from clinical routine CT images and used as predictors of hip fracture risk.

Our study has several limitations. First, due to the cross-sectional design, the analysis was limited to the evaluation of associations with hip fracture instead of prediction. Second, our results were confined to Chinese women, although our findings are consistent with a few Caucasian studies (11, 39). Third, we did not include comparisons with BMD measurements but only focused on geometric parameters. Fourth, we only studied the intact contralateral femur of the hip fracture patients by taking advantage of the anatomical similarity with the fractured side (40) despite the fact that some subjects hips can be surprisingly asymmetric.

In conclusion, there are substantial differences in total and cortical volume as well as cortical thickness between women with and without hip fractures across the entire proximal femur. The combination of geometric variables resulted in similar discrimination power for hip fracture risk as aBMD alone.

The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding authors.

The studies involving human participants were reviewed and approved by Beijing Jishuitan Hospital. The patients/participants provided their written informed consent to participate in this study.

XW and XC had full access to the data, take responsibility for the content, and guarantee the integrity and accuracy of the work undertaken. LW, MY, XW, KE, and XC designed the study. LW and YL led the analysis with input from MY, YG, SZ, KE, and GB. YL and YS did the literature search. YS, YG, and SZ collected the data. YL did the measurements. All authors contributed to data interpretation. LW and YL wrote the manuscript and all authors reviewed the manuscript.

This work is supported in part by Beijing Hospitals Authority Clinical Medicine Development of Special Funding Support (code: ZYLX202107), Beijing Hospitals Authority Youth Programme (code: QMS20200402), and the National Natural Science Foundation of China (Grant Nos. 81901718, 81771831,82072445).

KE is a part-time employee of BioClinica, Inc.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.