The Risk of Avascular Necrosis Following the Stabilization of Femoral Neck Fractures: A Systematic Review and Meta-Analysis

Background: Avascular necrosis (AVN) of the femoral head often requires surgical treatment and is often associated with femoral neck fractures. We conducted a systematic review and meta-analysis of recent research on the risk of AVN following the stabilization of fractured femoral neck with implants in PubMed. We assessed the effect of age on AVN incidence among patients aged > 50 and younger, depending on fracture type, Garden stage, Pouwels degree, Delbet stage, and age category. We followed PRISMA guidelines. Relevant studies were defined as research articles describing real-world studies reporting on the risk of AVN following primary surgical fracture stabilization with implants, published between 1 January 2011 and 22 April 2021. Fifty-two papers met the inclusion criteria, with a total of N = 5930 with surgically managed fractures. The pooled mean AVN incidence was significantly higher among patients with displaced fractures (20.7%; 95% CI: 12.8–28.5%) vs. those with undisplaced fractures (4.7%; 95% CI: 3.4–6.0%). No significant correlation was observed between AVN incidence weighted by sample size and time interval from injury to surgery (p = 0.843, R2 = 0.01). In conclusion, the risk of AVN following femoral neck fractures was generally high, especially in patients with displaced fractures. The time from injury to surgery did not correlate with AVN incidence.


Introduction
Avascular necrosis (AVN) of the femoral head involves osteonecrosis arising from altered blood supplies to the proximal femur [1]. Each year in the US alone, approximately 10,000-20,000 new cases occur [1]. AVN can arise from several causes, which may be classified as traumatic and atraumatic. [1] Femoral fractures can decrease blood flow to the femoral head, placing patients at risk of AVN [2,3]. Indeed, the fracture of the femoral neck or dislocation of the femoral head from the acetabulum is among the most common traumatic causes of AVN of the femoral head [1].
The most common symptom of AVN of the femoral head is radiating groin pain and possessing a limited range of motion with respect to the affected hip, with pain experienced during the forced internal rotation of the hip. Physical activity may aggravate the pain but it is often also present at rest [1]. The diagnosis of AVN is usually made based on both clinical presentation and appropriate imaging (X-ray, radionuclide bone scanning, The review followed PRISMA [6] (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) guidelines and is available as a supplementary file (Supplementary File S1). PubMed was searched on 11 May 2021 to identify relevant studies published between 1 January 2011 and 22 April 2021, using a combination of Medical Subject Heading (MeSH) and free-text terms related to hip fracture (neck of femur fracture and femoral neck fracture), surgery (internal fixation/internal fixators), and outcome of interest (avascular necrosis and avascular necrosis of femur head). Bibliographic details and abstracts for all citations identified through the PubMed search were exported into EndNote version 20 to remove duplicate papers and for screening titles and abstracts.
The entire protocol of the review, including the search strategy, is available in Supplementary Materials (Supplementary File S2). The protocol was not registered with a systematic review database.

Study Selection Process
A single reviewer performed screening, which was checked by a second senior reviewer. Two reviewers performed data extraction from the papers, which met the inclusion criteria. The inclusion and exclusion criteria are described in Table 1. Relevant citations were defined as original research articles describing real-world studies (excluding case reports or case series with ≤10 cases) and reporting on the risk of AVN incidence following the primary surgical stabilization of the fracture with implants (arthroplasty procedures were excluded). We did not exclude studies based on implant types or fracture reduction methods, which may depend on patient condition and surgeon' preference and experience. Studies focused on patients with neglected fractures who underwent surgery several weeks after the causative injury were also excluded.
Of particular interest was comparing the risk of AVN between patients managed conservatively and those whose fractures were stabilized surgically. Nonetheless, studies, in which different surgical approaches were compared were also included. No comparisons between other surgical methods were planned.

The Rationale for the Study
We sought to assess recent research findings on the risk of AVN after stabilising the fractured femoral neck with implants such as screws, plates, or wires in the real-world setting. We focused on research questions in which the analysis of real-world data from unselected patient populations may provide particularly relevant information, such as how the real-world incidence of AVN varies by the type of fracture and by time intervals from injury to surgery.
A previous Cochrane review published in 2001 observed lower AVN incidences with the sliding hip screw compared with different cancellous screws. However, based on the overall body of evidence with regards to all relevant post-surgical outcomes assessed in that review, no overall recommendations on the choice of implant were provided [7]. We therefore did not attempt to conduct comparisons between specific implant types.

Fracture Classification
The incidence of AVN is presented by fracture type (displaced vs. undisplaced) and according to three different fracture classifications: the Garden classification, Pauwels' classification, and the Delbet classification (pediatric studies only). The commonly used Garden classification was first described in 1961 [8] and describes fracture severities using staging (I-IV), based on the degree of fracture displacement as observed on anteroposterior radiographs [8,9]. Pauwels' classification has been in use since 1935 and was the first developed biomechanical classification for femoral neck fractures, although its significance has since somewhat declined [10]. Femoral neck fractures are classified into Pauwels degrees I-III according to the angle of their inclination relative to the horizontal plane and, therefore, the forces acting on the fracture line [10,11]. Delbet's classification is widely used to describe pediatric hip fractures and was first published in 1907 [12]. Delbet's classification defines fracture types I-IV based on the anatomic location of the fracture line [12], which is related to the risk of AVN development [13]. The three classifications are described in Table 2. Table 2. The Garden [8,9], Pauwels [10,11], and Delbet's classification [12,13].

Classification Type Description Other
Garden Stage I incomplete fracture; undisplaced, valgus impacted types or fracture reduction methods, which may depend on patient condition and surgeon' preference and experience. Studies focused on patients with neglected fractures who underwent surgery several weeks after the causative injury were also excluded. Of particular interest was comparing the risk of AVN between patients managed conservatively and those whose fractures were stabilized surgically. Nonetheless, studies, in which different surgical approaches were compared were also included. No comparisons between other surgical methods were planned.

The Rationale for the Study
We sought to assess recent research findings on the risk of AVN after stabilising the fractured femoral neck with implants such as screws, plates, or wires in the real-world setting. We focused on research questions in which the analysis of real-world data from unselected patient populations may provide particularly relevant information, such as how the real-world incidence of AVN varies by the type of fracture and by time intervals from injury to surgery.
A previous Cochrane review published in 2001 observed lower AVN incidences with the sliding hip screw compared with different cancellous screws. However, based on the overall body of evidence with regards to all relevant post-surgical outcomes assessed in that review, no overall recommendations on the choice of implant were provided [7]. We therefore did not attempt to conduct comparisons between specific implant types.

Fracture Classification
The incidence of AVN is presented by fracture type (displaced vs. undisplaced) and according to three different fracture classifications: the Garden classification, Pauwels' classification, and the Delbet classification (pediatric studies only). The commonly used Garden classification was first described in 1961 [8] and describes fracture severities using staging (I-IV), based on the degree of fracture displacement as observed on anteroposterior radiographs [8,9]. Pauwels' classification has been in use since 1935 and was the first developed biomechanical classification for femoral neck fractures, although its significance has since somewhat declined [10]. Femoral neck fractures are classified into Pauwels degrees I-III according to the angle of their inclination relative to the horizontal plane and, therefore, the forces acting on the fracture line [10,11]. Delbet's classification is widely used to describe pediatric hip fractures and was first published in 1907 [12]. Delbet's classification defines fracture types I-IV based on the anatomic location of the fracture line [12], which is related to the risk of AVN development [13]. The three classifications are described in Table 2. Table 2. The Garden [8,9], Pauwels [10,11], and Delbet's classification [12,13]. The effect of age on AVN risk was assessed by analyzing the incidence of AVN in patients aged 20-50 years vs. those aged > 50 years. This age cut-off was selected to assess the possible differential risk of AVN in middle-aged vs. younger patients in light of clinical guidelines recommending arthroplasty or hemiarthroplasty in elderly patients with displaced fractures [14,15]. The effect of age on AVN risk was assessed by analyzing the incidence of AVN in patients aged 20-50 years vs. those aged > 50 years. This age cut-off was selected to assess the possible differential risk of AVN in middle-aged vs. younger patients in light of clinical guidelines recommending arthroplasty or hemiarthroplasty in elderly patients with displaced fractures [14,15].

Meta-Analysis
Due to the apparent heterogeneity of both study designs and their real-world settings, no formal testing for heterogeneity was employed. A random-effects model was used to estimate the pooled incidence of AVN depending on fracture type (displaced vs. undisplaced), Garden stage, Pouwels' degree, Delbet stage, and age category (20-50 years vs. >50 years). The meta-analysis results are presented as pooled means with a 95% confidence interval (CI).

AVN Incidence and Time Interval from Injury to Surgery
The incidence of AVN in patients undergoing fracture stabilization surgery was assessed with respect to the the time interval between injury and surgery by using a general regression model. Mean or median time from injury to surgery, whichever was reported, was used in the analysis. The maximal reported time interval was used for studies that reported all surgeries within a given number of hours from injury.

Study Selection
The search strategy rendered 89 citations, of which 22 were excluded during the title and abstract screening stage so that a total of 67 full-text articles were screened for eligibility. Of these, 15 were excluded because the vast majority (n = 10) were unavailable in English. Overall, 52 articles were included. Outcomes of the selection process are documented in the PRISMA flow diagram ( Figure 1).

Overview of Included Studies
The majority of included studies were retrospective (n = 43). The sample size ranged from 17 to 417 participants. Three studies reported solely on pediatric populations [16][17][18], while nine studies included only participants above 50 years of age [19][20][21][22][23][24][25][26][27]. The studies included those that reported the use of both closed and open fracture-reduction techniques and employed a wide range of implants, most commonly cannulated screws (28 studies). Only one study described conservatively managed patients, and no comparison of AVN risk between surgical and non-surgical treatment could be made [18]. An overview of the included studies is presented in Table 3.
AVN was reported across the included studies in 0-53.4% of patients whose fractures were surgically managed. The timeframe from surgery to AVN detection was broad, ranging from as early as three months [28] up to 5 months [21,29,30] or more [31,32] years post-surgery.

AVN Incidence by Pauwels' Degree
Data on the incidence of AVN by fracture degree according to Pauwel's classification were derived from studies reporting both the number of patients with a given degree and AVN incidence for that degree. Only two studies reported data for Pauwel's degrees I and II [21,42]. Data from 5 studies were available for degree III fractures [21,42,49,60,64]. The pooled mean incidence of AVN was 21.8% (95% CI: 0-70.4%) for Pauwel's degree I fractures, 10.3% (95% CI: 0-25.9%) for degree II fractures, and 5.5% (95% CI: 0-12.4%) for degree III fractures, with no significant differences between different Pauwels' degrees ( Figure 2). However, given the small number of studies contributing data, particularly for degrees I and II, conclusions should be cautiously drawn.

AVN Incidence by Pauwels' Degree
Data on the incidence of AVN by fracture degree according to Pauwel's classification were derived from studies reporting both the number of patients with a given degree and AVN incidence for that degree. Only two studies reported data for Pauwel's degrees I and II [21,42]. Data from 5 studies were available for degree III fractures [21,42,49,60,64]. The pooled mean incidence of AVN was 21.8% (95% CI: 0-70.4%) for Pauwel's degree I fractures, 10.3% (95% CI: 0-25.9%) for degree II fractures, and 5.5% (95% CI: 0-12.4%) for degree III fractures, with no significant differences between different Pauwels' degrees (Figure 2). However, given the small number of studies contributing data, particularly for degrees I and II, conclusions should be cautiously drawn.

AVN Incidence by Delbet's Type
Delbet's classification of fractures is used in pediatrics and was represented by a limited number of studies in this review. Only studies reporting the number of patients with a given fracture type and AVN incidence for that type were included. Although three pediatric studies were identified in the review [16][17][18], Ju et al. [16] did not report the incidence of AVN by fracture type; thus, the study was not included in the meta-analysis. This resulted in two studies contributing data for Delbet's types I, II, and III [17,18] and single research contributing data for Delbet's type IV [18] due to the lack of patients with type IV fractures in the study by Wu et al. [17]. Increasing Delbet stages represents decreasing AVN risk. This trend could also be observed in the present review ( Figure 2). However, the differences in AVN incidence between Delbet stages were not statistically significant. The results should be interpreted with extreme caution due to the small number of pediatric studies.

AVN Incidence by Delbet's Type
Delbet's classification of fractures is used in pediatrics and was represented by a limited number of studies in this review. Only studies reporting the number of patients with a given fracture type and AVN incidence for that type were included. Although three pediatric studies were identified in the review [16][17][18], Ju et al. [16] did not report the incidence of AVN by fracture type; thus, the study was not included in the meta-analysis. This resulted in two studies contributing data for Delbet's types I, II, and III [17,18] and single research contributing data for Delbet's type IV [18] due to the lack of patients with type IV fractures in the study by Wu et al. [17]. Increasing Delbet stages represents decreasing AVN risk. This trend could also be observed in the present review ( Figure 2). However, the differences in AVN incidence between Delbet stages were not statistically significant. The results should be interpreted with extreme caution due to the small number of pediatric studies.

Discussion
The present review summarised the risk of AVN following surgery to stabilize femoral neck fractures. In the meta-analysis of AVN incidences by fracture displacement, the incidence of AVN was significantly higher in patients with displaced (corresponding to Garden III-IV) compared with undisplaced (corresponding to Garden I-II) fractures. However, when the four Garden stages were assessed individually, no significant differences in AVN incidence between locations were identified. Similarly, no significant differences in AVN incidence were observed between different Pauwels' angles and Delbet stages. However, due to the small number of studies using clinical classifications, the results should be interpreted with caution. Regarding age, the incidence of AVN was not significantly different between patients aged 20-50 years and those aged > 50 years; however, numerically lower AVN incidence was observed in the latter group. Finally, AVN incidence did not correlate with the time between injury and surgery.
The findings of this review confirm previous reports of increased AVN risk in displaced compared with undisplaced fractures. A meta-analysis by Xu et al., including 2065 patients from 17 case-control studies, demonstrated that the risk of AVN after internal fixation was 0.4-fold higher in patients with displaced fractures (Garden III-IV) than in patients without displacements (Garden I-II) [68].
Regarding the relationship between AVN incidence and the time interval between surgery and injury, the present study's results are similar to two previously published meta-analyses [68,69]. In a meta-analysis specifically assessing the effect of injury-tooperation interval on the development of late complications, no significant differences in the risk of AVN were detected when comparing surgery performed within and outside of 6, 12, and 24 h intervals from injury and when comparing AVN risk following surgery conducted within 6 h post-injury with surgery performed > 24 h post-injury [69]. Similarly, Xu et al.'s meta-analysis detected no statistically significant difference in AVN risk based on an injury-to-operation interval of ≤24 h vs. >24 h [68].
The observed lack of increased AVN risk in older patients is also consistent with Xu et al., who reported no significant difference in the risk of AVN between patients aged ≤ 60 years and >60 years [68]. It should be noted that the meta-analysis by Xu et al. used a different age categorization compared to the present study in which patients were classified as ≤50 or >50 years old, and the incidence of AVN was numerically higher in the younger patient group. However, both the European Society of Trauma and Emergency Surgery (ESTES) [14] and American Academy of Orthopaedic Surgeons (AAOS) [15] guidelines recommend arthroplasty or hemiarthroplasty in elderly patients with displaced fractures, and these procedures were excluded from the present review. Therefore, it is likely that patients aged > 50 years included in the present review experienced fractures associated with an improved prognosis than younger patients, which influenced the relative incidence of AVN in these groups. Indeed, among the nine studies reporting only on patients aged > 50 years [19][20][21][22][23][24][25][26][27], six studies included only patients with undisplaced fractures [19,[22][23][24][25][26]. In contrast, among four studies reporting only patients aged 20-50 years [36,47,56,67], all included only patients with displaced fractures.
The strengths of this review include a substantial number of studies captured and a broad scope of the review in terms of study settings, fracture types, patient age, and the type of implant used. The real-world nature of the included studies supports the applicability of the review findings to everyday clinical practice. Regarding the limitations of this review, no conclusions could be derived on AVN incidence following the stabilization of different fracture types according to Pauwels' and Delbet's classifications due to the small number of studies rendering relevant data. Furthermore, while both the Ficat classification and the Steinberg classification are often used to classify the severity of AVN [4], only one of the included studies reported AVN severity according to the Fi-cat classification [54]. This paucity of data on AVN severity meant that it could not be accounted for in the review and meta-analysis. Another important limitation is the heterogeneity of the reviewed studies, which may be a source for bias and does not allow the analysis of AVN risks in specific populations; therefore, more analyses will be planned in order to deepen the knowledge regarding the possible differences in incidence and severity of AVN in selected age groups and after different types of interventions.

Conclusions
The risk of AVN following femoral neck fractures was substantial and significantly higher for displaced (Garden III-IV) than undisplaced (Garden I-II) fractures. The time interval from injury to surgery did not correlate with AVN incidence. The review results highlight the substantial long-term risk of AVN, particularly in patients with displaced fractures, and call for prolonged post-surgical follow-up for patients with femoral neck fractures.

Conflicts of Interest:
The authors declare no conflict of interest.