Introduction

The incidence of femoral neck fractures (FNF) is approximately 80 in 100,000 [1]. About 67% FNF are displaced, requiring open reduction [2]. The most important risk factors are posed by bone density, neck anatomy, injury mechanism, female gender, alcohol, and corticosteroid abuse [3]. Interestingly, an epidemiologic study estimated that the number of FNF will increase to over six million over the next 50 years [4]. FNF report a high mortality rate [5] and represents a pivotal cause of disability, having a negative impact on patient mobility and physical independence. Furthermore, FNF present a considerable healthcare expenditure [6]. Concerning the treatment, there exists no definite consensus and several different techniques have been described [7]. Total hip arthroplasty (THA) and hip hemiarthroplasty (HHA) represent the treatments of choice for displaced FNF [8]. The discussion regarding hemiarthroplasty versus total arthroplasty began approximately 50 years ago, and is still subject of contentious debates [9]. Two international surveys conducted by Bhandari et al. [8] and the American Association of Hip and Knee Surgeons [10] confirm that 80–85% of surgeons preferred HHA. From the patient’s point of view, a general preference for THA has been reported [11]. Since there still exists an ongoing discussion regarding the pros and cons of either treatment [12,13,14,15], it is essential to review current evidence to help in the decision-making process. Previous meta-analyses were published in 2012 [9, 16, 17] and in 2015 [18]. Recently, RCTs have been published that have not yet been included in any meta-analysis [19,20,21,22]. An update of the literature is necessary. Therefore, a meta-analysis comparing THA versus HHA for displaced femoral fractures in the elderly was conducted, to update and analyse current outcomes and evidence.

Materials and methods

The present study was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [23]. A preliminary guideline protocol was compiled:

  • P (patient): displaced femoral neck fracture;

  • I (intervention): hip hemiarthroplasty (HHA);

  • C (comparison): total hip arthroplasty (THA);

  • O (outcomes): surgical duration, length of stay, clinical scores, complications.

Literature search and study selection

Two authors (FM, JE) independently performed the literature search. In October 2019, the same authors accessed the following databases: Pubmed, Embase, Google Scholar, Scopus. The following keywords were used in combination: displaced femur neck fracture, total hip arthroplasty, hip hemiarthroplasty, hip replacement, Harris hip score, revision, dislocation. The same authors independently screened the titles resulting from the database search for inclusion and accessed the full-text version of the articles of interest. The bibliographies of the full-text articles were checked as well.

Eligibility criteria

All RCTs comparing THA versus HHA for femoral neck fractures were included in the present study. According to the Oxford Centre of Evidence-Based Medicine [24], only articles classified as level of evidence I were included. Articles in English, Italian, Spanish, German, French, were taken into account. Every type of implant or surgical approach was considered for inclusion. Only studies treating patients with HHA or THA for displaced femoral neck fractures were considered, as were studies reporting quantitative data under the outcomes of interest. Only studies treating patients > 70 years old were considered. Articles published before 2000 were excluded, as were those treating animals, biomechanics, cadaveric, or in vitro studies. Studies treating revision surgeries were excluded. Studies evaluating these procedures through the addition of adjuvants, such as stem cells, PRP, or any other infiltrations, were excluded. Disagreements between the authors were debated and mutually solved.

Outcomes of interest

Two independent authors (FM, JE) exported data of interest from each clinical trial. The demographic data exported were the following: author, year of the publication, number of enrolled hips, and duration of the follow-up (months). Successively, the included articles were divided into two study groups: THA, HHA. The following data were extracted: number of enrolled hips, percentage of female patients, mean age of the samples, Harris hip score [25], surgical duration, duration of hospitalization. Further dislocation, acetabular erosion and revision surgeries were collected for each group, along with patient mortality.

Methodological quality assessment

For the methodological quality assessment, the Review Manager Software Version 5.3 (The Nordic Cochrane Centre, Copenhagen) was used. Two authors (FM, JE) independently performed the assessment. The aforementioned tool analyses the included articles with regard to five aspects: selection bias (random sequence generation and allocation concealment), performance bias, detection bias, attrition bias, and reporting bias.

Statistical analysis

The statistical analysis was performed by the senior author (FM). To assess the baseline comparability, the Student T test was performed. For the statistical analysis of quantitative variables, the Review Manager Software version 5.3 (The Nordic Cochrane Centre, Copenhagen) was used. Dichotomous data were analysed using the Mantel–Haenszel method with the odds ratio (OR) effect measure. Continuous data were analysed using the Inverse Variance method with the mean difference (MD). A fixed model method was used for the setup. To assess data heterogeneity, both chi-square (\(\chi\)2) and Higgins (I2) statistical tests were performed. Values of \(\chi\)2 > 0.5 indicate high heterogeneity. Values of I2 of 0–40%, 40–75%, 75–100% indicate low, moderate, and high data heterogeneity, respectively. In case of high heterogeneity, a random model effect was adopted. To analyse the publication bias, a funnel plot of the most commonly reported outcome was generated. For the analysis of the survivorship (mortality), the STATA/MP software version 14.1 (StataCorp, College Station, TX) was used. Survivorship was analysed through the Kaplan–Meier (KM) survivor function. The KM curve was performed according to the Cox-regression through the Breslow method with the hazard ratio (HR) effect measure. The confidence interval (CI) was set at 95% in all analysis. A p value < 0.05 was considered statistically significant.

Results

Search result

The databases searches resulted in 732 papers. First, a total of 305 articles were rejected due to duplication. Of those, only RCTs were considered, which left 77 studies for inclusion. Further 41 articles which did not match the topic were excluded, while other 19 because incompatibility with the eligibility criteria were in articles were excluded as they did not match the eligibility criteria. A further six papers were rejected as it did not report quantitative data under the outcomes of interest. This last operation left 11 RCTs for inclusion. The flow chart of literature search is shown in Fig. 1.

Fig. 1
figure 1

PRISMA flow chart of literature search

Full size image

Methodological quality assessment

Correspondent to the above-mentioned assessment of each risk of bias, a low to moderate risk of selection bias can be reported. Attrition bias can be considered a low risk, and both performance and detection bias were moderate risks. Therefore, the methodological assessment of this work can be judged as very good quality. The review authors’ rating of each risk of bias item according to the Cochrane Collaboration is shown in Fig. 2.

Fig. 2
figure 2

Review authors’ rating of each risk of bias item

Full size image

Risk of publication bias

To detect possible publication bias, a funnel plot for the most commonly reported outcome (revisions) was performed. The plot evidenced an adequate symmetrical distribution of points close to the no-effect line. Moreover, none of the studies were located outside of the acceptability range, thus testifying satisfactory results. Concluding, the risk of publication bias for this study scored low. The funnel plot is shown in Fig. 3.

Fig. 3
figure 3

Funnel plot of the most reported outcome (revision)

Full size image

Patient demographic

Data from 2325 patients were collected. The mean follow-up was 58.12 ± 52.8 months. A total of 1171 samples were part of the HHA group, with a mean of 73.2% female patients and a mean age of 78.2 ± 4.6 years. The THA group accounted for 1154 patients in total with 70.7% of those being female, with a mean age of 79.4 ± 2.8 years. No differences were detected between the two groups regarding age or sex (p = 0.8 and p = 0.4, respectively). Table 1 summarizes the patient demographics.

Table 1 Demographic data of the included studies
Full size table

Outcomes of interest

The overall results are shown in Table 2. Compared to THA, the HHA group reported lower values of the mean Harris hip score (MD: 3.22; 95% CI: −1.32 to 7.76; p = 0.2), surgical duration (MD: 21.75; 95% CI: 14.47 to –29.03; p < 0.0001), length of the hospitalization stay (MD: 0.8; 95% CI: − 0.90 to 2.49; p = 0.4). The HHA group evidenced lower dislocations rate (OR: 1.78; 95% CI: 1.15 to –2.77; p = 0.01, Fig. 4), but higher rate of acetabular erosion (OR: 0.08; 95% CI: 0.02 to –0.33; p = 0.0006).

Table 2 Overall results
Full size table
Fig. 4
figure 4

Forest plot of the comparison dislocation rate

Full size image

At a mean of 58.12 ± 52.8 months follow-up, revisions rate scored reduced in the THA group (OR: 0.76; 95% CI: 0.48 to –1.19; p = 0.2; Fig. 5).

Fig. 5
figure 5

Forest plot of the comparison revision

Full size image

Subgroup analysis of RCTs < 5 years follow-up revealed reduced revision in favour of the HHA group (OR: 2.19; 95% CI: 10.9 to –4.40; p = 0.03), while subgroup analysis of RCTs > 5 years follow-up revealed reduced revision in favour of the THA group (OR: 0.25; 95% CI: 0.12 to –0.53; p = 0.0003).

The Kaplan–Meier curve detected similarity of patients survivorship between the two groups (HR: 1.06; 95% CI: 0.94 to –1.22; p = 0.3; Fig. 6).

Fig. 6
figure 6

Kaplan–Meier curve of survivorship

Full size image

Discussion

The main findings of the present meta-analysis were that for elderly patients with displaced femoral neck fractures, the HHA provide lower dislocations, shorter surgical duration, but higher rate of acetabular erosion compared to the THA. Revisions rate scored similar between HHA and THA. However, the subgroup analysis including only RCTs > 5 years follow-up revealed reduced revision in favour of the THA group. Hospitalization length and Harris hip score reported any statistically significant difference among the two cohorts. Concerning patients’ survivorship, the Kaplan–Meier curve found similarity among the two techniques.

In favour of HHA, decreased mean length of surgical duration was evidenced. This result was expected, since HHA requires less operative installation steps. Indeed, even if there is little variability within-technique operating time, the overall estimated effect was strongly in favour of the HHA group. These results are significantly reliable. Same observations were reported in other two RCTs not included in the comparison [20, 27].

To investigate hip function, the Harris hip score was evaluated. Yu et al. [16] in 2012 reported better Harris hip score in the THA cohort at 1, 3 and 4-year follow-up. Even the study of Burgers et al. [9] found statistically improved Harris hip score in the THA group over 300 patients. Want et al. [18] in 2015 analysed found any differences between HHA and THA within 1, 2, 4 and > 4-year follow-up. The present meta-analysis revealed a greater effect of the Harris hip score in favour of THA. This comparison suffered a high-level of heterogeneity and the final effect was not significant. We concluded, that the two groups were similar in terms of Harris hip score.

In the present meta-analysis, the comparison length of the hospitalization was affected by low grade of heterogeneity and variability, with good distribution of statistical weights. However, the final effect is not statistically significant, attesting similarity among the groups. Similar results were found by Wang et al. [18] in 2015 in a meta-analysis comparing bipolar HHA versus THA over 1014 patients. Woon et al. [32] analyzing 12,757 patients from the US National Hospital Discharge Survey found reduced hospitalisation length in the HHA cohort. Anyway, they evidenced high heterogeneity between region and hospital-size, maybe attributable to the differences in regional training and subspecialist availability. Moreover, in their analysis there was a larger proportion of THA patients was covered by private insurance to increase heterogeneity.

Concerning complications, a statistically significant reduction of the dislocation rate in the HHA group was observed. The forest plot showed variability of the weight distribution, with mostly overlapping CI. Data was in accordance with previous studies in the literature. The study by Burgers et al. [9] analysed dislocations over 800 patients with a reduced risk ratio in favour of HHA of 2.53. Similarly, Yu et al. [16] found a risk ratio of 1.99 in favour of the HHA group. Wang et al. [33] analysed dislocation at 1, 2, 4, and > 4 years follow-up. They found, only at 4 years, a statistically significant minimal reduction in favour of the HHA group (risk ratio 0.2). Zi-Sheng et al. [17] in 2012 analysed dislocations in 1122 patients, founding a statistically significant risk ratio of 0.49 in favour of the HHA group. Although HHA evidenced reduced dislocations rate, on the other hand, exposed the patients to higher risk of acetabular erosion. The analysis of acetabular erosion showed no heterogeneity among studies, attesting to this comparison high reliability and clinical relevance. The rate of erosion is far higher (OR 0.08) compared to the THA group. The higher acetabular erosion can partially explain the higher rate of revision in favour of the THA reported in the subgroup analysis (follow-up > 60 months). The comparison revision showed high concordance among the studies and results, with trustworthy results. Results in previous meta-analysis regarding implant revisions are controversial. Yu et al. [16] found a statistically significant halved risk ratio of revision in the HHA cohort, while Burgers et al. [9] found no difference at 1 year follow-up over 816 patients. Wang et al. [18] in their meta-analysis performed follow-up subgroup analyses. Similarly, they found that longer follow-up provided favourable results for the THA group (risk ratio 3.3 at > 4-year follow-up). Same observations were reported by Zi-Sheng et al. [17] in their follow-up subgroup analyses.

The Kaplan–Meier curve was performed to analyse patients’ survivorship. The result showed distinctly similarity among the two cohorts (HR 1.06, p = 0.3). This result is comparable with previous studies. Results from the US national register by Woon et al. [32] in 2017 found no difference in survivorship over 12,757 patients. Similarly, Hopley et al. [34] in their meta-analysis involving 1023 patients, found no differences between the HHA and THA (risk ratio 0.9, p = 0.8). Comparable results were found by other less recent meta-analyses [9, 16, 18, 35].

The quicker surgical duration and the sparing of healthy structures promoted by the HHA, can clarify the reduction of the total estimated blood loss, the faster recovery, the reduced post-operative attention, and the improved patients collaboration that have been reported in previous studies [5, 8, 12,13,14,15,16, 29, 32, 33, 36,37,38,39]. These features, commit that HHA should been recommend for patients with comorbidities, cognitive impairment, and reduced performance status. However, controversial concerning younger, active and healthier patients remains. Even if more at risk of dislocations, THA reported reduced acetabular erosion and longer survivorship, and may be recommended for healthy and more active patients [32, 40].

This study has several limitations. The most important limitation of the present study was given by the reduced number of included studies and related patients. In consequence of the limited evidence in the literature, no differentiation regarding surgical techniques, approaches, and implants were made. Most of the included studies differed in inclusion and exclusion criteria, thus representing another source of bias. Given these limitations, data from the present study must be interpreted with caution. The most important points of strength of this study were the extensive nature of the literature search, the rigorous methodological quality assessment of the studies, and the strictly eligibility criteria. Furthermore, the optimal baseline comparability of the samples promotes a reduction of heterogeneity and publication bias. Further clinical trials providing long-term follow-up are strongly recommended to establish the best evidence concerning both techniques.

Conclusion

For the elderly population, both HHA and THA are valid solutions to treat displaced femoral neck fractures, with comparable survivorship. HHA detected reduced dislocations, while for THA a lower risk of acetabular erosion and further revision surgeries were reported.