Reliability of 3D quantitative displacement of femoral neck fracture site measured via CT and its correlation with femoral head avascular necrosis: a retrospective cohort study

Introduction

Hip fractures account for 7.5–12% of adult fractures and are a significant clinical challenge, particularly among the elderly population (1,2). Femoral neck fractures (FNFs) represent 50% of hip fractures. The incidence of traumatic and stress FNFs is rising due to increased traffic accidents and physical exercise, as well as the aging population (3-5). Globally, the incidence of hip fractures has doubled in individuals over 50 in recent years, with projections estimating an increase from 1.26 million cases in 1990 to 4.5 million cases by 2050 (6). Femoral head preservation, bone healing, and prevention of avascular necrosis (AVN) are the primary treatment goals for biologically young patients. Conservative treatment has many defects, and closed reduction and internal fixation (CRIF) has become the preferred approach (7-9). However, FNFs treated with CRIF are subjected to an increased risk of complications, such as fixation failure, nonunion, and AVN, resulting in significant mortality and increased economic burden on both patients and society. AVN is the most prevalent complication with an incidence of nearly 16–30% (2,3,10-12). The majority of patients need revision surgery. Studies have shown a correlation between fracture displacement and AVN; thus, fracture displacement has been regarded as a crucial risk factor for AVN of FNFs (7,13,14).

The existing methods for measuring fracture displacement rely on two-dimensional (2D) images (X-ray films) and are limited to capturing angular displacement, which cannot quantitatively measure fracture displacement. For example, the widely used Garden and Arbeitsgemeinschaft für Osteosynthesefragen classification systems effectively characterize the angular displacement at the fracture site and classify FNF into displaced and non-displaced types based on 2D images (15-17). However, they have significant limitations, such as a lack of precise quantification and clear quantification differentiation criteria for various types of fractures, resulting in inaccurate classification with strong subjectivity and limited reliability (16,17).

Computed tomography (CT) scans can more accurately measure displacement. Recently, spatial displacement of FNF measured based on three-dimensional (3D) CT reconstruction and mimics has been developed. However, this technique needs bilateral CT scans of the hip and still focuses on angular displacement rather than actual 3D quantitative displacement of the fracture site (18,19). The actual 3D displacement of FNF site includes three directions: vertical, horizontal and separation displacement. We have developed a novel method of 3D quantitative displacement of FNF site measured via CT. This study aimed to determine the reliability of the novel 3D quantitative CT displacement measurement for FNFs and its association with AVN. We have also developed a prediction model using data from a retrospective cohort of 200 patients. We present this article in accordance with the TRIPOD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1483/rc).

Methods

Participants

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the institutional review board of Wuxi Ninth People’s Hospital Affiliated to Soochow University (No. JY-KT-20240227) and written informed consent was waived due to the retrospective design and anonymity of data. According to literature reports, the sample size for logistic regression analysis should be 10–20 times the number of observed indicators (20). We had 5 variables, calculated 20 times; a total of 200 patients were included, ensuring model robustness (power >0.8, α>0.05).

The inclusion criteria were as follows: (I) consecutive traumatic intracapsular FNF treated with CRIF at the Wuxi Ninth People’s Hospital from January 2018 to January 2022; (II) surgery within one week of injury; (III) preoperative AP/lateral radiographs (X-ray films) and hip CT scans; and (IV) at least 3 years of follow-up. The exclusion criteria were as follows: (I) pathological fractures; (II) age <15 years; (III) unsatisfactory fracture reduction [grades III and IV according to the Garden alignment index previously described by Haidukewych et al. (21)]; (IV) preexisting hip arthritis, diabetes or immune diseases; (V) age >75 years or osteoporosis [bone mineral density (BMD)] T-score ≤−2.5); (VI) alcoholism; and (VII) incomplete follow-up data.

Measurement

2D measurement

2D measurement was defined as the use of X-ray films for measuring the displacement of the fracture site. The Garden’s classification is based on AP and lateral X-rays of the hip. It categorizes FNFs into four types based on the degree of fracture displacement: type I (incomplete fracture), type II (complete fracture, well-aligned without displacement), type III (partial displacement with abduction of the femoral head, mild external rotation, and upward displacement of the femoral neck), and type IV (complete displacement with significant external rotation and upward displacement of the femoral neck) (Figure 1). Types I and II are regarded as non-displaced types, and types III and IV are regarded as displaced types. If there were discrepancies regarding the Garden classification, it was decided by a senior physician.

Figure 1 The Garden classification and corresponding 2D images for Garden types I, II, III, and IV of femoral neck fractures. 2D, two-dimensional.

3D measurement

3D measurement was defined as the use of CT scans for measuring the displacement of the fracture site. The CT scans of 200 patients stored in the mobile disk were connected to the AW Server 4.7 image 3D post-processing workstation (General Electric) of CT (GE Optima CT 660). The measurement protocol was as follows: using 3D reconstruction technology, the 3D reconstruction hip on the affected side (Figure 2A) was rotated until obtaining three ideal processing views (frontal, prone, and supine views of the FNF) based on the following steps: (I) the 3D reconstruction image was rotated to eliminate the anteversion angle of the femoral neck, making the anterior part of the femoral neck parallel to the coronal plane and obtaining the frontal view of the femoral neck (Figure 2B); (II) determining the direction of the fracture line (the line connecting the upper and lower cortical edges of the proximal fracture end), rotating the 3D reconstruction image until the fracture line was parallel to the longitudinal axis of the trunk, and obtaining the front view of the FNF (Figure 2C); (III) rotating the 3D reconstruction image to obtain prone and supine views of the FNF (Figure 2D,2E). We used the built-in ruler in the imaging system to measure the vertical, horizontal, and separation displacements of the fracture site in 3 processed 3D reconstructed images of FNF. The comprehensive displacement was regarded as the sum of the vertical, horizontal, and separation displacements. The vertical displacement of the fracture site is the maximum vertical distance measured from the upper or lower cortical edges on both sides of the fracture in the front view of the processed 3D reconstruction image of the FNF (Figure 3A). In the prone or supine view of the processed 3D reconstruction image of the FNF, the horizontal distance is the maximum vertical distance measured from the anterior or posterior edges on both sides of the fracture (Figure 3B), while the separation displacement is the maximum straight-line distance measured from the anterior cortical edges on both sides of the fracture site (Figure 3C).

Figure 2 The schematic diagram depicting the measurement protocol of the 3D reconstruction images of the hip. (A) Frontal view of the 3D reconstruction of the hip. (B) Frontal view of the femoral neck. (C) Front view of the FNF. (D) Prone view of the FNF. (E) Supine view of the FNF. 3D, three-dimensional; FNF, femoral neck fracture.

Figure 3 The schematic diagram representing the measurement of the 3D quantitative displacements using processed 3D reconstruction images of the FNF. (A) The vertical displacement in the frontal view of the FNF. (B) The horizontal displacement in the supine view of the FNF. (C) The separation displacement in the prone view of the FNF. 3D, three-dimensional; FNF, femoral neck fracture.

Evaluation

Reliability

The X-ray films and CT scans of 200 patients stored in digital imaging and communications in medicine were transmitted to a mobile CD. To assess inter-observer reliability, the 3D displacement and 2D displacement (classified into displaced and non-displaced according to Garden classification) of all images were evaluated by three observers, including one attending orthopedic surgeon, one orthopedic resident surgeon, and one radiology resident. The two measurement methods were explained to all the observers individually, and the observers were allowed to practice the 2D and 3D measurements on 10 separate images 2 weeks before starting the study. After training, three observers who were blinded to the outcomes conducted independent evaluations. Regarding intra-observer reliability, we chose the imaging resident to re-measure displacement after 3 months.

AVN

AVN was diagnosed based on the updated version of the 2019 AVN grading system (22). It could be diagnosed as AVN as long as it met the following criteria: (I) clinical features, such as pain; and (II) MRI demonstrated a limited subchondral linear-shaped low-signal intensity in T1-weighted images or a “double-line sign” in T2-weighted images one year after the surgery.

Statistical analysis

Statistical analyses were conducted using SPSSAU Statistics for Windows, version 21.0 (IBM Corp., Armonk, NY, USA). Intraclass correlation coefficient (ICC) was employed to assess inter-observer and intra-observer reliability for 2D and 3D measurement methods. We used a two-way random-effects model with absolute agreement for this purpose. The inter-observer and intra-observer agreements were interpreted based on convention using the following criteria: 0.21–0.40: fair; 0.41–0.60: moderate, 0.61–0.80: substantial; ≥0.81: almost perfect. Correlation regression was employed to assess the relationship between displacement and AVN. Bivariate correlation analysis was used to screen variables with differences. Thereafter, variables with differences were analyzed using binary logistic regression. Adopting the strategies of collinearity diagnosis and stepwise regression to mitigate bias. The rms and rmda packages of R Studio (3.6.5) were used to set the random seed number to 200 and divide it into a training set and a validation set in a 7:3 ratio. The training set was used to establish a logistic regression model and a nomogram. The validation set was adopted to evaluate and validate the prediction model through receiver operating characteristic (ROC) curve, calibration curve, and draw decision curve analysis (DCA). All the statistical analyses were two-tailed, with the significance level set at α=0.05.

Results

In total, 257 FNF patients were screened, among whom 200 were included in the study (Figure 4), comprising 86 males and 114 females, with ages ranging from 17 to 73 years. BMD T-score varied from –2.5 to 1, corresponding to the following distribution: normal BMD in 57% of patients, and low BMD in 43%. Patients’ demographic data and lesion characteristics are summarized in Table 1.

Figure 4 Study flow diagram of patient enrollment. CRIF, closed reduction and internal fixation; FNFs, femoral neck fractures.

AVN

During the 3.68 years of follow-up, AVN was observed in 55 patients, with an incidence rate of 27.5%.

Reliability

The balance between the training and validation sets was good (Table 2). The ICCs for the inter-observer and intra-observer reliability of 2D measurement were 0.723 (95% confidence interval (CI): 0.639–0.794) and 0.776 (95% CI: 0.685–0.844), respectively. In addition, the ICC for inter-observer and intra-observer reliability of the 3D measurement was 0.963 (95% CI: 0.949–0.974) and 0.962 (95% CI: 0.945–0.975), respectively, suggesting that the agreement of the 3D measurement was higher than that of 2D measurement.

Correlation with AVN

Bivariate correlation exhibited significant differences in 2D and 3D displacements between the AVN and non-AVN groups (P<0.05). The comprehensive displacement was the greatest among the four 3D displacements, followed by vertical and separation displacements. The horizontal displacement was the smallest among the four 3D displacements (Table 3).

Logistic regression

Binary logistic regression showed that, unlike the 3D vertical and horizontal displacements, 3D separation and comprehensive and Garden displacements were independent risk factors for AVN (Table 4). Hosmer-Lemeshow showed good goodness of fit (χ²=8.333, R²=0.581, P=0.119), with a prediction accuracy of 87%, 67.3% for positive results and 94.5% for negative results.

Nomogram of prediction model

Based on the Logistic regression result, the nomogram of AVN prediction model was constructed as shown in Figure 5, indicating that the AVN risk threshold offers substantial clinical net benefits when set between 0.1 and 0.9.

Figure 5 Nomogram for predicting the possibility of AVN after FNFs in 3 years. AVN, avascular necrosis; FNFs, femoral neck fractures.

ROC curves

Table 5 summarizes the area under the curve (AUC), sensitivity, specificity, and Youden index (accuracy) of the three independent risk factors. Among them, the comprehensive displacement had the greatest AUC, followed by the separation displacement, with both P<0.05, whereas the Garden displacement had least AUC with P>0.05, indicating that the comprehensive and separation displacements had better predictive value than Garden displacement.

Model evaluation and validation

The ROC curves in Figure 6 and calibration curves in Figure 7 of the prediction model for the training and validation sets exhibited good consistency. The confusion matrix in Table 6 exhibited high sensitivity, specificity and accuracy of the prediction model. The DCA curves in Figure 8 showed that within the risk threshold probability range of 0.10–0.90, the clinical net benefit of intervention based on the predicted probability of the model was higher than that of no intervention for all (None) and intervention for all (All), indicating that net benefit is generated within the risk threshold range.

Figure 6 ROC curves for the training and validation sets. AUC, area under the curve; CI, confidence interval; ROC, receiver operating characteristic.

Figure 7 Calibration curves for the training set (A) and the validation set (B).

Figure 8 DCA curves for the training set (A) and validation set (B). DCA, decision curve analysis.

Discussion

For the first time, this study introduced a 3D quantitative displacement measurement of the FNF site, demonstrating high reliability, strong correlation between 3D displacement and AVN, and better predictive potential for AVN. The 3D measurement method outperformed the traditional 2D method.

Reliability of the method

The Kappa coefficient for 2D Garden classification of FNFs in previous studies ranged from 0.32 to 0.52, suggesting limited reliability (23-27). Both 2D and currently used 3D measurement techniques for FNF displacement only evaluate angular displacement, rather than providing 3D quantitative measurements at the fracture site. After FNF, the femoral neck loses structural integrity and support, and the affected limb usually shortens and rotates outward due to muscle traction on the greater trochanter, resulting in vertical displacement in the coronal plane, horizontal displacement in the transverse plane, and anterior separation.

The 3D measurement method introduced in this study helps directly quantify fracture site displacement, including vertical, horizontal and separation displacements. This method is not affected by different imaging positions of patients due to pain. Therefore, based on the measurement protocols introduced in this study, more objective and accurate measurements can be conducted on the three standard views of the FNF (frontal, prone, and supine views), with excellent consistency and repeatability. Our comparative analysis revealed inter-observer and intra-observer ICCs of 0.723 (95% CI: 0.639–0.794) and 0.776 (95% CI: 0.685–0.844) for 2D displacement measurement, respectively, while 3D displacement measurement exhibited significantly higher reliability with ICCs of 0.963 (95% CI: 0.949–0.974) and 0.962 (95% CI: 0.945–0.975), respectively.

Correlation between 3D displacement and AVN

AVN after FNFs is caused by impaired blood supply to the femoral head due to fracture or dislocation. Nearly all AVNs occur in intracapsular FNFs, including subcapital and transcervical types. Injuries in this region disrupt the anastomosis between lateral epiphyseal vessels, thereby interrupting blood flow to the femoral head. While Digital Subtraction Angiography, bone scintigraphy (Emission Computed Tomography, Positron Emission Tomography-Computer Tomography), high-selectivity angiography, and MRI are the most direct methods for assessing blood supply and predicting AVN (28-32), their high cost limits their application for suspected AVN among patients with FNF. During the acute phase of FNFs, AVN risk assessment relies on multifactorial analysis, with the lack of early diagnostic tools for AVN. Accurate AVN risk assessment can significantly improve treatment planning for patients with FNF.

Studies have found several factors that affect AVN development after FNF, including female gender, age (<60 years), subcapital fracture location, displaced Garden classification, preoperative traction, high joint capsule tension, injury-to-surgery interval (>10 days), and femoral head posterior inclination angle (>20°) (12-14,28-31). Especially, the fracture location, displacement degree, and joint capsule tension are critical factors affecting the blood flow of the femoral head and key determining factors in assessing the risk of AVN. The greater the displacement, the higher the tension of the joint capsule, indicating more severe damage to the femoral head vascular system and predicting a higher risk of AVN. All 3D displacement types, including vertical, horizontal, and separation, contribute to vascular network damage, increased intracapsular pressure, vascular interruption, and subsequent development of AVN. In our study, the comprehensive displacement had the highest correlation with ANV due to its maximum value than others; Of the vertical, horizontal, separation and Garden displacements, the separation displacement exhibited a moderate correlation with ANV, the Garden displacement exhibited a weak correlation with ANV, and the horizontal and vertical displacements showed no correlation with ANV, indicating that displacement in different directions has different impacts on AVN. However, this article only preliminarily discovered this phenomenon (displacement in different directions has different impacts on AVN), and further experiments are needed to verify the phenomenon and explore the specific reasons. There is currently no study on different impacts of different directional displacements on AVN. This important discovery reminds us that taking timely measures to correct significant displacement, especially the separation displacement, can help prevent the occurrence of AVN.

In this study, the incidence rate of AVN after FNFs was 23.5%, which is consistent with the literature reports (12-14,31).

Model assessment and internal validation

The clinical prediction model is a quantitative tool for risk and benefit assessment. It is a well-known risk score, which has been widely used in the medical field. However, there is no consensus on the risk factors of AVN after FNFs. Wang et al. (33) developed an XGBoost model incorporating six predictors in their comparative analysis of machine learning algorithms for predicting AVN after FNFs treated with CRIF. These variables, including reduction quality, VAS score, Garden classification, operative time, cause of injury, and fracture location, were identified as key factors for assessing the risk of AVN. The model demonstrated strong generalizability, performing reliably on external datasets. In the six-variable XGBoost model, the accuracy, sensitivity, and AUC on the validation set were 0.987, 0.929, and 0.992, respectively. On the external dataset, its accuracy, sensitivity, specificity, and AUC were 0.907, 0.807, 0.935, and 0.933, respectively, supporting its potential utility in clinical settings for estimating the risk of AVN after FNFs.

This study only evaluated the relationship between 3D displacement and AVN, and demonstrated high predictive value of the model, especially the comprehensive displacement for predicting AVN significantly outperforming separation displacement and 2D angular displacement; The curves of ROC and calibration, and confusion matrix of the prediction model showed high consistency for the training and validation sets, with high sensitivity [97% (95% CI: 0.90–1.00)], specificity [71% (95% CI: 0.63–0.80)] and accuracy [76% (95% CI: 0.69–0.83)]; The DCA curves showed the risk score was clinically useful between the risk threshold of 0.10% and 90%. All of these indicate that the model that reflects the 3D displacements has good predictive potential. Notably, this demonstrates the different values of different directional displacements in predicting AVN.

Limitations and future research directions

The limitations of this study include single-center and retrospective design. Secondly, the development of AVN after FNF is multifactorial. This study only evaluated the relationship between displacement and AVN, lacking multi-factor analysis of AVN. Therefore, there were high risks of selection bias, performance bias, detection bias, and reporting bias. The predictive model lacked external validation. It only provides a reference and has not yet been practically used. Future studies should perform multiple-center, prospective studies in larger cohorts from different populations or nations, incorporate additional risk factors and conduct external validation or machine learning to validate these findings. Machine learning plays the main role in the prediction model (13); thus, machine learning combined with 3D displacement is a potential research direction in the ANV prediction model.

Conclusions

Compared to traditional 2D methods, CT-measured 3D quantitative displacement of fracture site exhibited higher reliability, stronger correlation with AVN, and better predictive potential for AVN. The impact of displacement in different directions on AVN varies, but further research is needed.

Acknowledgments

The authors would like to express their gratitude to EditSprings (https://www.editsprings.cn) for the expert linguistic services provided.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1483/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1483/dss

Funding: This study was funded by the Wuxi Municipal Health Commission Suitable Technology Promotion Project (No. T202331).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1483/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the institutional review board of Wuxi Ninth People’s Hospital Affiliated to Soochow University (No. JY-KT-20240227) and written informed consent was waived due to the retrospective design and anonymity of data.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Dong Y, Luo C, Zhang N, You R, Yuan Q, Li W, Zhang Y. Correlation between combined deflection angle classification adduction typing and complications after internal fixation of adduction femoral neck fracture. Zhongguo Xiu Fu Chong Jian Wai Ke Za Zhi 2024;38:405-11. [Crossref] [PubMed]
2. Florschutz AV, Langford JR, Haidukewych GJ, Koval KJ. Femoral neck fractures: current management. J Orthop Trauma 2015;29:121-9. [Crossref] [PubMed]
3. Bernstein EM, Kelsey TJ, Cochran GK, Deafenbaugh BK, Kuhn KM. Femoral Neck Stress Fractures: An Updated Review. J Am Acad Orthop Surg 2022;30:302-11. [Crossref] [PubMed]
4. Kazley JM, Banerjee S, Abousayed MM, Rosenbaum AJ. Classifications in Brief: Garden Classification of Femoral Neck Fractures. Clin Orthop Relat Res 2018;476:441-5. [Crossref] [PubMed]
5. Biz C, Berizzi A, Crimì A, Marcato C, Trovarelli G, Ruggieri P. Management and treatment of femoral neck stress fractures in recreational runners: a report of four cases and review of the literature. Acta Biomed 2017;88:96-106. [Crossref] [PubMed]
6. Veronese N, Maggi S. Epidemiology and social costs of hip fracture. Injury 2018;49:1458-60. [Crossref] [PubMed]
7. Biz C, Fantoni I, Crepaldi N, Zonta F, Buffon L, Corradin M, Lissandron A, Ruggieri P. Clinical practice and nursing management of pre-operative skin or skeletal traction for hip fractures in elderly patients: a cross-sectional three-institution study. Int J Orthop Trauma Nurs 2019;32:32-40. [Crossref] [PubMed]
8. Papakostidis C, Panagiotopoulos A, Piccioli A, Giannoudis PV. Timing of internal fixation of femoral neck fractures. A systematic review and meta-analysis of the final outcome. Injury 2015;46:459-66. [Crossref] [PubMed]
9. Liu H, Li N, Zhang X, He L, Li D, Li Y, Zhao G, Wu X. Internal fixation versus hemiarthroplasty for displaced femoral neck fractures in the elderly: A cost-effectiveness analysis. Injury 2020;51:1346-51. [Crossref] [PubMed]
10. Yuan KX, Yang F, Fu K, Zhu DY, Jiang CY, Jin DX, Wang ZH, Peng XY, Gao YS, Luo PB. Internal fixation using fully threaded cannulated compression screws for fresh femoral neck fractures in adults. J Orthop Surg Res 2022;17:108. [Crossref] [PubMed]
11. Schmidt AH, Swiontkowski MF. Femoral neck fractures. Orthop Clin North Am 2002;33:97-111. viii. [Crossref] [PubMed]
12. Slobogean GP, Sprague SA, Scott T, Bhandari M. Complications following young femoral neck fractures. Injury 2015;46:484-91. [Crossref] [PubMed]
13. Hu Y, Yang Q, Zhang J, Peng Y, Guang Q, Li K. Methods to predict osteonecrosis of femoral head after femoral neck fracture: a systematic review of the literature. J Orthop Surg Res 2023;18:377. [Crossref] [PubMed]
14. Konarski W, Poboży T, Kotela A, Śliwczyński A, Kotela I, Hordowicz M, Krakowiak J. The Risk of Avascular Necrosis Following the Stabilization of Femoral Neck Fractures: A Systematic Review and Meta-Analysis. Int J Environ Res Public Health 2022;19:10050. [Crossref] [PubMed]
15. Du CL, Ma XL, Zhang T, Zhang HF, Wang CG, Zhao F, Ma JX, Fu X, Li ZJ. Reunderstanding of garden type I femoral neck fractures by 3-dimensional reconstruction. Orthopedics 2013;36:820-5. [Crossref] [PubMed]
16. Chen W, Li Z, Su Y, Hou Z, Zhang Q, Zhang Y. Garden type I fractures myth or reality? A prospective study comparing CT scans with X-ray findings in Garden type I femoral neck fractures. Bone 2012;51:929-32. [Crossref] [PubMed]
17. Cazzato G, Oliva MS, Masci G, Vitiello R, Smimmo A, Matrangolo MR, Palmacci O, D'Adamio S, Ziranu A. Femoral neck fracture: the reliability of radiologic classifications. BMC Musculoskelet Disord 2022;22:1063. [Crossref] [PubMed]
18. Wu S, Wang W, Zhang B, Zhang H, Xu X, Li G, Shi H, Mei J. A three-dimensional measurement based on CT for the posterior tilt with ideal inter-and intra-observer reliability in non-displaced femoral neck fractures. Comput Methods Biomech Biomed Engin 2021;24:1854-61. [Crossref] [PubMed]
19. Wu S, Zhu X, Wang W, Zhang Y, Li G, Mei J. Three-dimensional computed tomography mapping and clinical predictive factors for morphologic characterization of displaced femoral neck fractures. Ann Transl Med 2022;10:1096. [Crossref] [PubMed]
20. Maxwell SE, Kelley K, Rausch JR. Sample size planning for statistical power and accuracy in parameter estimation. Annu Rev Psychol 2008;59:537-63. [Crossref] [PubMed]
21. Haidukewych GJ, Rothwell WS, Jacofsky DJ, Torchia ME, Berry DJ. Operative treatment of femoral neck fractures in patients between the ages of fifteen and fifty years. J Bone Joint Surg Am 2004;86:1711-6. [Crossref] [PubMed]
22. Zhao D, Zhang F, Wang B, Liu B, Li L, Kim SY, Goodman SB, Hernigou P, Cui Q, Lineaweaver WC, Xu J, Drescher WR, Qin L. Guidelines for clinical diagnosis and treatment of osteonecrosis of the femoral head in adults (2019 version). J Orthop Translat 2020;21:100-10. [Crossref] [PubMed]
23. Kalsbeek JH, van Walsum ADP, Roerdink WH, van Vugt AB, van de Krol H, Schipper IB. Validation of two methods to measure posterior tilt in femoral neck fractures. Injury 2020;51:380-3. [Crossref] [PubMed]
24. Liu Y, Li M, Zhang M, Sun K, Wang H, Yuan X, Cai L. Femoral neck fractures: prognosis based on a new classification after superselective angiography. J Orthop Sci 2013;18:443-50. [Crossref] [PubMed]
25. Zhang Y, Yuan Q, Wei Q, Dong Y, Liu Y, He W, Jing Z, Zhang L, Wang H, Li W. Combined Deflection Angle Classification: A Novel Typing System of Adult Femoral Neck Fracture. Orthop Surg 2023;15:839-50. [Crossref] [PubMed]
26. Masionis P, Uvarovas V, Mazarevičius G, Popov K, Venckus Š, Baužys K, Porvaneckas N. The reliability of a Garden, AO and simple II stage classifications for intracapsular hip fractures. Orthop Traumatol Surg Res 2019;105:29-33. [Crossref] [PubMed]
27. Van Embden D, Rhemrev SJ, Genelin F, Meylaerts SA, Roukema GR. The reliability of a simplified Garden classification for intracapsular hip fractures. Orthop Traumatol Surg Res 2012;98:405-8. [Crossref] [PubMed]
28. Kim CH, Shin M, Lee D, Choi SJ, Moon DH. Hidden osteonecrosis of the femoral head after healed femoral neck fractures: magnetic resonance imaging study of 58 consecutive patients. Arch Orthop Trauma Surg 2022;142:1443-50. [Crossref] [PubMed]
29. Zhao D, Wang Z, Cheng L, Tian S, Liu B, Yang L, Ma Z. The Significance of Evaluating the Femoral Head Blood Supply after Femoral Neck Fracture: A New Classification for Femoral Neck Fractures. Orthop Surg 2022;14:742-9. [Crossref] [PubMed]
30. Cho MR, Lee SW, Shin DK, Kim SK, Kim SY, Ko SB, Kwun KW. A predictive method for subsequent avascular necrosis of the femoral head (AVNFH) by observation of bleeding from the cannulated screw used for fixation of intracapsular femoral neck fractures. J Orthop Trauma 2007;21:158-64. [Crossref] [PubMed]
31. Wang C, Xu GJ, Han Z, Jiang X, Zhang CB, Dong Q, Ma JX, Ma XL. Correlation Between Residual Displacement and Osteonecrosis of the Femoral Head Following Cannulated Screw Fixation of Femoral Neck Fractures. Medicine (Baltimore) 2015;94:e2139. [Crossref] [PubMed]
32. Wang YX, Griffith JF, Kwok AW, Leung JC, Yeung DK, Ahuja AT, Leung PC. Reduced bone perfusion in proximal femur of subjects with decreased bone mineral density preferentially affects the femoral neck. Bone 2009;45:711-5. [Crossref] [PubMed]
33. Wang H, Wu W, Han C, Zheng J, Cai X, Chang S, Shi J, Xu N, Ai Z. Prediction Model of Osteonecrosis of the Femoral Head After Femoral Neck Fracture: Machine Learning-Based Development and Validation Study. JMIR Med Inform 2021;9:e30079. [Crossref] [PubMed]