Background
Approximately 250,000 proximal femoral fractures occur in the USA each year, and 90% of these fractures occur in patients older than 50 years of age [
1,
2]. Fractures of the femoral neck can be categorized into either non-displaced or displaced fractures in order to facilitate appropriate management, particularly in the elderly [
3,
4]. Displaced femoral neck fractures are defined as unstable fractures that can impair blood supply to the femoral head, resulting in avascular necrosis [
5,
6]. These fractures are associated with substantial fracture-related mortality and morbidity [
6,
7]. An additional contributor to femoral head osteonecrosis involves the quality of the reduction or fracture fixation [
8].
Osteosynthesis refers to the percutaneous placement of several parallel cannulated lag screws, and in the younger patient, such internal fixation is the standard treatment for displaced fractures [
8]. Hip arthroplasty, on the other hand, refers to replacement of all or part of the hip joint with a prosthetic implant [
6,
9] and can be divided into either total hip arthroplasty (THA) or hemiarthroplasty (HA). THA involves replacement of both the femoral head and the acetabular articular surface, whereas in HA, only the femoral head is replaced with an artificial implant, while the patient’s own acetabulum is retained [
5,
6,
9].
Although internal fixation is recommended for most non-displaced fractures of the femoral neck, the optimal treatment for displaced fractures of the femoral neck is still controversial [
10‐
12]. HA was once considered the procedure of choice for elderly patients with displaced (Garden stage III or IV) femoral neck fractures [
13], but a Swedish study concluded that THA should be performed for displaced femoral neck fractures in older patients with normal mental function and high function [
14], a conclusion that has been echoed in several more recent publications [
15,
16]. Davison et al., on the other hand, recommended either reduction with internal fixation or cemented HA as alternative treatments for a displaced intracapsular fracture in a mobile and mentally competent patient under 80 years of age [
10].
The literature also contains conflicting evidence regarding rates of mortality, major postoperative complications, and function in elderly patients with displaced femoral neck fractures treated either by internal fixation or arthroplasty [
17]. In fact, the choice of surgical treatment for a displaced intracapsular fracture of the proximal femur in the elderly remains as controversial now as it was over 50 years ago when it was designated as “the unsolved fracture” [
10,
12,
18]. This meta-analysis was designed, therefore, to address this controversy by comparing outcomes after internal fixation, hemiarthroplasty, and THA, with particular reference to mortality and revision rates because until now, few studies have compared these three alternative treatments [
10]. In addition, due to the limited number of studies with head-to-head comparison of HA and total hip replacement, a statistical analysis (comprised of both direct and indirect comparisons) was utilized to achieve this study’s objective.
Discussion
This meta-analysis compared the overall mortality and revision rates between arthroplasty (HA and THA) vs. osteosynthesis for displaced femoral neck fractures in the elderly. Advanced statistical analysis (indirect comparison) was used simultaneously to compare THA and HA in order to resolve the lack of studies with head-to-head comparison between THA and HA. We also compared clinical outcomes of arthroplasty (THA and HA) vs. osteosynthesis (internal fixation) for displaced femoral neck fractures.
This meta-analysis found no significant difference in mortality rates between THA, HA, and OS. In addition, no significant difference in revision rates was found between THA and HA, but osteosynthesis had higher revision rates than either THA or HA. The additional subgroup analysis, using only studies involving elderly subjects without significant cognitive impairment, provided similar results for mortality (i.e., no difference between HA and osteosynthesis) and revision rates (no difference between THA and HA), but OS had higher revision rates than either THA or HA.
One study (two articles) showed the mean survival time of persons who died for both THA (5.3 years, range 1.3 to 9.1 years) and HA (3.8 years, range 0.003 to 7.5 years) [
45,
46]. Two studies (three articles) assessed the mean survival time after intervention. Davison et al. [
10] found that patients who received Thompson unipolar HA, Monk (hard-top) bipolar HA, and reduction/internal fixation had mean survival times of 61, 68, and 79 months, respectively. There was a significant difference in mean survival time between groups (
P = 0.008). Parker et al. [
43,
44] found that the patients who received HA and internal fixation had mean survival times of 2.7 years (95% CI 2.2–3.1) and 3.2 years (95% CI 2.5–3.9), respectively. No significant difference was found between groups. Due to the limitation in the number of available studies, the survival time after interventions (osteosynthesis, HA, and THA) was not included in the meta-analysis.
No meta-analysis was conducted for the functional outcome after interventions since the methods or scales for evaluating hip function were heterogeneous among the included articles (Table
1), including the Oxford Hip Score [
45,
46], Harris Hip Score [
10,
15,
34,
40‐
42,
47‐
49,
52‐
55], hip rating questionnaire [
50,
51], Charnley’s numerical classification [
27‐
29,
31,
33,
43,
44], Matta Scoring System [
32], Stinchfield’s classification system [
26], and Sheperd’s pain and hip mobility score [
39]. In addition, no data for baseline measurement were shown in most of the studies. Therefore, it was impossible to estimate the difference in mean change before and after the intervention between the two groups.
To the best of our knowledge, this is the first meta-analysis to compare three types of interventions for displaced femoral neck fractures in one meta-analysis. Although a meta-analysis comparing THA, HA, and osteosynthesis was reported in 2012 by Gao et al. [
17], the need to revisit this issue (by conducting another meta-analysis) remained because the meta-analysis by Gao et al. compared the outcomes between arthroplasty and internal fixation and thus pooled together the outcomes of HA and total hip replacement [
17]. Fisher et al. [
56], in their review of 3423 cases of ORIF, THA, and HA, found no differences in the 30-day mortality rates among the ORIF, HA, and THA groups, similar to our findings. ORIF and HA also resulted in a lower likelihood of developing respiratory complications than did THA [
56]. A meta-analysis comparing THA and HA was reported by Burgers et al. [
57]. Given the heterogeneity in surgical technique and experience over time, we felt an update of the evidence was necessary. We have updated the search, but our results were consistent with this meta-analysis that no significant difference was found in mortality and revision rates between THA and HA, but they demonstrated that THA may lead to higher dislocation rates compared with HA [
57]. Therefore, it was felt that the optimal choice of arthroplasty (THA or HA) for treating femoral neck fractures had not yet been established.
It appears that the choice between arthroplasty and internal fixation in some studies was based primarily on the survival time of the implant. For example, in Davison et al. [
10], HA was not recommended due to the shorter mean survival time of the implant compared with internal fixation despite the fact that internal fixation was associated with a 30% risk of failure [
10]. They reported a mean patient survival which was significantly higher in the group undergoing reduction and internal fixation (79 months) compared with that with a cemented Thompson HA or a cemented Monk bipolar HA (61 and 68 months, respectively). We also found that the revision rates were lower in arthroplasty compared with internal fixation, but survival was the same among all three types of intervention (HA, THA, and osteosynthesis). These differences are likely related to the type of arthroplasty used. As in our evaluation, there was significant heterogeneity both in the implant used and the technique applied (for example, cementless HA [
31,
32] vs. cemented HA [
10,
30], articulation of metal on ultra-high-molecular-weight polyethylene in THR vs. metal on articular cartilage following HA [
45]).
Nikitovic [
6], on the basis of two systematic reviews evaluating the effectiveness of THA in comparison with HA for treatment of displaced femoral neck fractures, found a significant reduction in revision rates among patients receiving THA in comparison with HA. In addition, his recent study showed a significant improvement in functional outcome among patients receiving THA in comparison with HA, using the Harris Hip Score for the assessment. THA was favored over HA based on improvements in QoL using mobility and pain measures [
6].
No meta-analysis was conducted for the QoL after interventions since the methods or scales for evaluating QoL were heterogeneous among the included articles. SF-36 was used in three studies (four articles), including three articles for THA vs. HA [
45,
46,
52] and one article for HA vs. osteosynthesis [
32]. EQ-5D was used in six studies (ten articles), including four articles for THA vs. HA [
47,
48,
50,
51], three articles for THA vs. osteosynthesis [
27‐
29], and three articles for HA vs. osteosynthesis [
31,
33,
40]. Although we did not evaluate QoL, some studies have placed emphasis on social functioning after intervention. Jónsson et al. evaluated 50 patients with Garden stage 3 and 4 femoral neck fractures randomized for treatment using either osteosynthesis with the Hansson hook pins or THA with the Charnley prosthesis [
25]. The patients were followed for up to 2 years, and their social function was evaluated using a standardized questionnaire. The authors concluded that a patient over 70 years of age who was relatively healthy, mobile, and socially independent should be considered for a primary hip prosthesis even if late complications, such as mechanical loosening, were taken into account. This conclusion was based on the fact that the majority of patients over 70 years of age are less likely to live long enough to develop implant loosening. For very old, frail, or immobile patients, however, osteosynthesis was the preferred treatment [
25]. And, these findings were echoed in a more recent study from Sweden [
24]. Bachrach-Lindström et al. found that a primary THA group performed better than an osteosynthesis group in weight change over time, locomotion, and pain. They also showed that primary THA could be performed safely in the elderly without increasing postoperative mortality [
24].
As part of our study, we performed a sensitivity analysis and tested for homogeneity and quality. Since our analysis showed heterogeneity among the majority of studies, a random-effects model was primarily applied. We also tested for reliability based on sensitivity analysis. The direction and magnitude of the combined estimates did not change markedly with the exclusion of individual studies, indicating that our meta-analysis had good reliability. The results of quality assessment showed that the most potential risk of bias came from performance bias and detection bias because of inadequate blinding of participants and outcome assessors.
We also tested for publication bias. Although significant evidence of publication bias was found regarding differences in survival between THA and HA, we adjusted the effect of publication bias, and the adjusted point estimates of OR on mortality increased to 1.13 (95% CI 0.77 to 1.67, Fig.
2c).
In the 26 studies evaluated as part of our meta-analysis, almost all subjects had freedom of mobility or were capable of independent walking before their injury, and there was a female predominance. This is not surprising, as the incidence of proximal femoral fractures among females is two to three times greater than the incidence of such fractures in males [
1]. Other risk factors for proximal femoral fractures include osteoporosis [
1], a maternal history of hip fractures [
58], excessive alcohol consumption and high caffeine intake [
59], and physical inactivity [
60], to name a few. Owing to our aging population, the risk of sustaining a proximal femoral fracture doubles every 10 years after the age of 50 [
1]. Therefore, this study has clinical relevance in that it is an attempt to identify the best treatment option for these elderly patients.
Our study had several limitations. Potential performance and detection biases might exist in most of the included studies. We also did not assess functional status of the patients after the reconstructive procedures, and this was an inevitable shortcoming of this study. Notably for a patient with severe cognitive dysfunction, the lack of a surgical revision might correlate with a limited capacity for independent ambulation and with an inability to verbally express the features of a potentially symptomatic hip. We did perform an analysis on a subgroup of elderly patients without severe cognitive impairment and found no significant difference in the results regarding mortality and revision rates. But, the ways in which cognitive impairment was defined as significant or severe differed among studies. In addition, several studies only included patients with acute displaced femoral neck fractures with different time periods between fracture occurrence and admission; this ranged from 12 to 96 h. In addition, the studies included different types of femoral neck fractures, and not all studies specified the Garden stages of fractures. For all these reasons, more future studies comparing these three types of interventions are still needed to confirm our findings. Furthermore, there was significant heterogeneity among studies, especially with respect to the types of implants used for HA and THA and the types of screws used for osteosynthesis. The optimal choice of screw or reduction method (open or closed) for osteosynthesis remains unclear. Among included studies, only a few [
15,
25,
45‐
48] chose independent living as a selection criterion, and it is arguable whether living independently or in a nursing home could have an impact on the results. Furthermore, the surgeons’ experiences and the different numbers and types of procedures performed at the various medical centers were possible confounding factors that may have affected the results and influenced the heterogeneity among studies.