1 Introduction

Intertrochanteric hip fractures remain one of the most common and debilitating conditions impacting patients above the age of 65.¹ Indeed, its prevalence is 50% and 1-year mortality is 25%.^1,2 50-60% of these fractures are classified as unstable, whereby the fracture tends to collapse into varus or involve a medial displacement of the shaft secondary to a comminuted posteromedial cortex, reverse oblique nature of the fracture, or subtrochanteric extension.³ Several classification systems including the Orthopaedic Trauma Association (OTA) and the Evans-Jensen classification are used in clinical practice to guide management.^4,5

Traditional teaching states that IT fractures are typically managed with open reduction and internal fixation (ORIF), and to this end, several fixation systems, such as the dynamic hip screw (DHS), proximal femoral nail (PFN) and cephalomedullary nailing have been developed for this purpose.⁶ Recently there has been increasing discussion over the use of hemiarthroplasty as an alternative option to deal with the high mechanical complications and failure rates associated with fixation methods.⁷

Presently, several studies and meta-analyses have been performed, with poorly-defined consensus on the role of hemiarthroplasty in managing unstable IT fractures.^8–10 Some identified indications include severely comminuted fractures, pre-existing degenerative arthritis, and severely osteoporotic bone.¹¹ Thus, we aim to perform a comprehensive updated systematic review and meta-analysis on the relevant existing data in literature. We aim to identify the role of hemiarthroplasty in managing unstable IT fractures, looking for the patient demographic that will benefit most from it over internal fixation. We hypothesise that in the elderly population group, hemiarthroplasty will improve post-operative PROMs and quality-of-life with similar complication rates.

2 Material and methods

2.3

2.3 Data synthesis

Statistical analyses were performed using R 4.1.5. Random-effects meta-analyses (DerSimonian and Laird) were used because substantial clinical and methodological heterogeneity was anticipated across the included studies—particularly differences in fracture classification, implant type, surgical technique, and postoperative protocols—making a fixed-effect assumption of a single true effect size inappropriate. The random-effects approach therefore provides a more conservative and generalisable pooled estimate by incorporating both within-study and between-study variability.

Analyses were conducted using the logit transformation to stabilise variances in proportional data, especially where event rates approached 0 or 1. Corresponding 95% confidence intervals (CIs) were computed using the Clopper-Pearson method, which provides exact interval estimates suitable for small or unevenly distributed event counts.^12–15 Dichotomous outcomes were presented as pooled proportions or pooled risk ratios when comparator groups were available. Continuous outcomes were presented as pooled means or mean differences, each with 95% CIs, using weighted mean differences and pooled relative risk ratios as principal summary measures in keeping with established meta-analytic practice.

Where means and standard deviations were not reported, these values were derived according to the validated method by Wan and colleagues. As inter-study heterogeneity may be over-estimated by I² statistics—particularly in meta-analyses with relatively few studies or varying sample sizes—heterogeneity was additionally assessed using the GRADE framework to account for inconsistency, indirectness, and imprecision of the body of evidence.^16,17 A P-value of <0.05 was considered statistically significant.

Sensitivity analyses excluding studies with moderate or unclear risks of bias were performed to assess the robustness of the pooled estimates. Publication bias was evaluated using Egger's test, which demonstrated no evidence of small-study effects with an intercept of 0.79 and p = 0.45, supported by a symmetrical funnel plot (Fig. 1).

Close

Fig. 1 Funnel plot for Publication Bias.

Export to PPT

The search strategy yielded 187 articles. Following abstract and full-text screening, 37 articles met the inclusion criteria. Of these, 5 were randomised controlled trials, 10 were prospective cohort studies, and 22 were retrospective cohort studies. Most were assessed as low risk of bias; 18 studies exhibited moderate risk of bias, and none were rated high risk (Appendix C). With respect to overlapping patient data, all included studies represented distinct cohorts; therefore, no further adjustments were required.

3 Results and data analyses

3.1

3.1 Primary outcome: Harris Hip Score

HHS at various time points was reported in 20 of 37 articles included. The pooled HHS for patients receiving HA was 72.7 (95% CI: 71.0-74.2) at 3-months, 78.6 (95% CI: 75.3-81.9) at 6-months, 80.5 (95% CI: 75.9-85.1) at 12-months, and 77.7 (95% CI: 73.0-82.3) at 24-months. The pooled HHS for patients receiving IF was 60.0 (95% CI: 57.3-62.7) at 3-months, 73.4 (95% CI: 71.1-75.8) at 6-months, 79.1 (95% CI: 76.8-81.3) at 12-months, and 78.6 (95% CI: 73.8-83.4) at 24-months. The weighted mean difference (WMD) in HHS at 3 and 6-months follow-up was statistically significant, with HA having a higher HHS of 12.5 (95% CI: 8.6-16.4) and 5.3 (95% CI: 2.9-7.7) respectively. The WMD for HHS at 3 and 6-months was seen in Figs. 2 and 3 below.

Close

Fig. 2 Weighted mean difference of Harris hip score at 3-months.

Export to PPT

Close

Fig. 3 Weighted mean difference of Harris hip score at 6-months.

Export to PPT

3.2

3.2 Secondary outcomes

3.2.1

3.2.1 Mortality

Across the pooled total of both treatment arms with reported mortality rates, there was a mortality rate of 12.7% at 1-year, and 19.9% at 2-years. When assessing pooled relative risk at 1-year, the risk was significantly higher in HA at 1.6 (95% CI: 1.2-2.1) compared to IF, with a P-value < 0.05, seen in Fig. 4. Assessing heterogeneity, I² = 0%, τ² = 0, and Cochran's Q test p = 0.72. This suggests that the observed differences in mortality risk were consistent across studies. At 2-years, the difference in relative risk was not statistically significant with HA risk at 0.9 (95% CI: 0.6-1.3).

Close

Fig. 4 Pooled relative risk for mortality at 1-year.

Export to PPT

3.2.2

3.2.2 Operation factors: length of operation and total blood loss

Length of operation for HA was 82.2 min (95% CI: 75.1-89.4) while IF was 69.8 min (95% CI: 63.9-80.4). HA length of operation was statistically significantly longer than IF, with a WMD of 12.2 min (95% CI: 6.6-17.9) and P-value < 0.05, per Fig. 5. Heterogeneity assessment reveals I² = 98%, τ² = 223.4, and Cochran's Q test p < 0.01, suggesting substantial variability across studies. Total blood loss in HA was 334.4 mL (95% CI: 280.6-388.2) while IF was 181.3 mL (95% CI: 140.6-222.0). The total blood loss in HA was statistically significantly greater with a WMD of 152.8 mL (95% CI: 108.1-197.6) and a P-value < 0.05, per Fig. 6. Heterogeneity assessment reveals I² = 100%, τ² = 11718.8 and Cochran's Q test p = 0, suggesting significant variability across studies.

Close

Fig. 5 Weighted mean difference of operation time.

Export to PPT

Close

Fig. 6 Weighted mean difference of blood loss.

Export to PPT

3.2.3

3.2.3 Post-operative complications

When looking at reoperations required, HA had a rate of 5.6% (95% CI: 3.6–9.0) while IF had a rate of 10.2% (95% CI: 7.4-13.9). HA had a lower relative risk at 0.6 (95% CI: 0.3-1.0) compared to IF, and this was statistically significant (P-value < 0.05), per Fig. 7. Heterogeneity assessment shows I² = 42%, τ² = 223.4, and Cochran's Q test p = 0.04, suggesting moderate variability across studies.

Close

Fig. 7 Pooled relative risk for reoperations.

Export to PPT

Post-operative surgical site infections (SSI) had a rate of 6.8% (95% CI: 5.2-8.9) in HA group, and 4.6% (95% CI: 3.3-6.5) in IF group. Heterogeneity was moderate in both with I² = 41% and I² = 47% respectively. HA had a greater risk at 1.4 (95% CI: 1.0-2.1) compared to IF. This difference was statistically significant (P-value < 0.05), with heterogeneity of I² = 0%, τ² = 0.14, and Cochran's Q test p = 0.59, suggesting low variability. (Appendix E).

There were some cases of nonunion of fractures, with HA having a rate of 2.8% (95% CI: 1.4-5.5) and IF group having a rate of 4.3% (95% CI: 2.9-6.2). Both had low heterogeneity at 46% and 56% respectively. HA had a lower relative risk of 0.5 (95% CI: 0.3-0.9) and was statistically significant (P-value < 0.05). Heterogeneity assessment revealed I² = 0%, τ² = 0.022, and Cochran's Q test p = 0.64, indicating low variability (Appendix E).

Prosthetic failures were reported in some studies. HA had a failure rate of 2.5% (95% CI: 1.4-4.3) while IF had a rate of 6.5% (95% CI: 3.6-11.2). There was no heterogeneity in HA arm and moderate heterogeneity at 51% in IF arm. HA had a lower relative risk at 0.4 (95% CI: 0.2-1.0), which was statistically significant (P-value < 0.05). Heterogeneity was I² = 2%, τ² = 0.26, and Cochran's Q test p = 0.42, indicating low variability among studies (Appendix E).

Other complications of limb length deformity, periprosthetic fractures as well as deep vein thrombosis were assessed. There were however no statistically significant differences (Appendix E).

3.3

3.3 Subgroup analyses

3.3.1

3.3.1 Cemented and uncemented implants

Subgroup analysis evaluated whether cementation technique (uncemented or cemented) influenced the 3-month HHS. Six studies of cemented HA (n = 305) reported a pooled mean HHS of 73.3 (95% CI: 70.2–76.3), with substantial heterogeneity (I² = 95%, p < 0.01). Three studies evaluated uncemented hemiarthroplasty (n = 141), demonstrating a pooled mean HHS of 69.5 (95% CI: 65.5–73.5), with considerable heterogeneity (I² = 90%, p < 0.01). There was no statistically significant difference between cemented and uncemented techniques (test for subgroup differences: χ² = 2.14, df = 1, p = 0.14) (Fig. 8).

Close

Fig. 8 Subgroup analysis of cemented vs uncemented HA for 3-month HHS.

Export to PPT

3.3.2

3.3.2 Age

Subgroup analysis evaluated 3-month HHS and 1-year mortality according to age category (>80 versus 65–79 years). For HHS, hemiarthroplasty was favoured in patients aged >80 years (WMD 9.2; 95% CI: 3.1–15.2; I² = 89%) and 65–79 years (WMD 13.8; 95% CI: 9.0–18.6; I² = 96%). There was no significant subgroup effect (p = 0.24), indicative of consistent HHS benefit of HA over IF in both older (>80) and younger-elderly (65–79) patients.

For 1-year mortality, the pooled RR again did not differ significantly between the two age groups. Pooled RR was 2.0 (95% CI: 0.9–4.6; I² = 58%) in patients >80 years and 1.4 (95% CI: 0.9–2.2; I² = 0%) in those aged 65–79 years, with no significant subgroup difference (p = 0.47), suggesting that age did not modify the effect of surgical method on mortality at 1-year (Fig. 9).

Close

Fig. 9 Subgroup analysis of age (>80 or 65-79) for 3-months HHS and mortality at 1-year.

Export to PPT

3.3.3

3.3.3 Internal fixation methods

Subgroup analysis compared types of internal fixation – Proximal Femoral Nail Antirotation (PFNA) versus Dynamic Hip Screw (DHS) – and their influence on 1-year mortality. Pooled mortality was 8.8 per 100 observations (95% CI: 4.3–17.3; I² = 78%) for PFNA and 15.9 per 100 observations (95% CI: 10.8–22.7; I² = 0%) for DHS. There was no significant subgroup difference (p = 0.14) (Fig. 10), indicating that the type of internal fixation did not significantly influence 1-year mortality (Fig. 10).

Close

Fig. 10 Subgroup analysis of type of internal fixation (PFNA or DHS) for mortality at 1-year.

Export to PPT

The choice of outcome for the subgrouping was decided based on the largest sample size for analysis. The remaining postoperative follow up time points had fewer studies included and thus a smaller sample size that would limit meaningful analysis.

4 Discussion

HHS is a measure of dysfunction, assessing the following aspects: Pain, Function, Absence of Deformity and Range of Motion via patient-reported outcomes with inputs by the clinician objectively (total score of 100 points) and is proven to be validated and reliable. Singh et al. defines the minimal clinically important improvement (MCII) threshold for HHS at 15.9-18 points.¹⁸ While the mean differences do not reach the MCII threshold (greatest mean difference at 12.5 points at 3-months), it may be noted that the MCII is derived from a different cohort and MCII is not universal. Singh et al.’s cohort was a younger population (mean of 64 years old) compared to the mean age of 77.4 - 77.8 years old in this review. As older age comes with greater frailty burden and reduces post-operative functional outcomes, a smaller mean difference can be meaningful to the older population.

In this review, HA had improved HHS compared to IF, across all follow-up periods. Notably, HA consistently reported a better HHS in the short term, with the difference between the two groups narrowing at later follow-ups. At the 3 and 6-month follow up, HA HHS had a significantly higher HHS compared to IF with mean difference of 12.5 points.

This is indicative of superior early ambulation post-operatively at 3 to 6-month follow-up compared to IF. This translates into earlier improved functional outcomes, a positive prognosticator in hip surgeries, allowing for greater long-term survival.¹⁹ Therefore, carrying out HA over IF allows for full mobilisation earlier, beneficial for patients who desire early ambulation, aiming to reach preoperative function earlier.

Intraoperative factors — duration of operation and intraoperative blood loss — were assessed due to their effects on surgical outcomes. The duration of operation in HA was significantly greater compared to IF by 12.2 min. Shorter operative times tend to be more protective against adverse outcomes, such as SSI.^20,21 This could partly explain the higher relative risk of SSI, at 1.4 times in the HA group compared to IF. On the contrary, Pollmann et al. found that operative time does not have a significant impact on the risk of SSI, specifically in hip fractures. Literature also suggests other factors to instead be predictors of SSI, namely: number of days from admission to surgery, patient's serum albumin levels and body-mass index.^22–24 Hence, the significance of operation time and its clinical impact on hip fracture patients can be further explored.

Intraoperative blood loss in HA was statistically significantly greater than IF by 152.8 mL, though with significant heterogeneity. Cui et al. suggested that for intertrochanteric fractures, such reported apparent intraoperative blood loss by studies account for a minority of total blood loss, with hidden blood loss — the majority — unable to be accounted for.²⁵ Hence, this intraoperative blood loss has little implication on clinical significance. Nonetheless, with a greater overall blood loss, there is a greater risk of post-operative anaemia that increases mortality.²⁶

In patients who have multiple preoperative comorbidities, HA has been associated with increased morbidity and mortality.²⁷ IF hence becomes an alternative for high-risk patients. It may be noted that selection bias may have existed as surgeons subject patients who have poorer premorbid status to IF over HA, potentially affecting postoperative mortality rates. Without granular data on the premorbid status of patients of the papers assessed, comparison for mortality was challenging.

Nonetheless, in this review of patients with varying comorbidities, at 1-year post-op, HA had a significantly higher relative risk for mortality at 1.6 compared to internal fixation. This suggests that even though patients with poorer preoperative comorbidities may have been selected preferentially for IF, the results still show IF having better mortality rates. Hence, this confounder does not affect the result that IF has lower early mortality rates. Mortality at 2-years showed statistically insignificant difference. Furthermore, subgrouping by age (65-79 vs > 80) was carried out to assess for impact on HHS at 3-months and mortality at 1-year. No significant subgroup differences were noted.

Regarding post-operative complications, HA had a lower risk of fracture nonunion at 0.5 as well as a lower risk of prosthetic failure at 0.4. Nonunion at intertrochanteric sites tend to be less common as it is well vascularised.²⁸ However, IF poses a higher risk for nonunion compared to HA due to its greater demand for callous formation and requires bone healing for a smooth recovery.²⁹ Nonunion of fractures lengthens the healing process, implicating the patient and the healthcare system, resulting in more resources spent.³⁰ Prosthetic failure can be attributed to causes such as mechanical discomfort or prosthesis loosening, potentially requiring reoperations.

HA had 0.6 times the relative risk of reoperations required compared to IF. However, there was no granular data available on the reason for reoperations. Due to the different nature of both operations and its complications, the reasons for reoperations vary. With the lower risk of reoperations, patients will not be subjected to another round of risks of operation and prolonged immobility. Furthermore, reoperations cause a substantial burden to economic resources.³¹

Unstable intertrochanteric fractures’ ideal treatment technique remains in contention. Many factors are considered in this decision-making, from pre-operative, perioperative and to outcomes anticipated. Without a consensus for the ideal technique to treat unstable intertrochanteric fractures, this study consolidated these factors to aid surgeons with their choice.

The present review hence meticulously ensured that the studies included had similar baseline patient factors, robust study design and with minimal bias. Bias was assessed with risk stratification and removed where appropriate. Thus, only studies found to be comparable and suitable were included.

This study recognised some limitations to this meta-analysis. This study had low heterogeneity as it generalises surgical techniques into 2 broad arms — hemiarthroplasty and internal fixation. Subgroup analyses were completed to address this. For HA, 3-month HHS comparison of Cemented and uncemented HA revealed no statistically significant subgroup differences. Within IF group itself comprised multiple techniques, such as dynamic hip screw (DHS), proximal femoral nail anti-rotation (PFNA), Medoff Sliding Plate etc. Subgroups analysis carried out revealed no significant subgroup differences between PFNA and DHS on mortality. Otherwise, subgroup analysis on other IF techniques could not be carried out due to the lack of granular data provided in the articles reviewed. Literature had mixed conclusions on the similarity of outcomes between each specific technique within both HA and IF broad techniques.^32–34 Future meta-analyses may compare hemiarthroplasty against specific surgical internal fixation methods.

Subgrouping patients by premorbid status or Body Mass Index (BMI) yielded insufficient results to conduct subgroup analyses. Subgroup analyses into severity and grade of fracture types were not carried out since this paper aims to group unstable intertrochanteric fractures as a whole.

5 Conclusion

In patients with unstable intertrochanteric fractures, HA provided improved functional outcomes earlier than IF, quantified by HHS. Furthermore, relative risk of post-operative fracture nonunion, prosthesis failure and need for reoperations is lower in HA. However, these advantages must be assessed in conjunction with HA's increased intraoperative burden of greater blood loss and longer operative times. Furthermore, HA has an increased risk of 1-year postoperative mortality rate and risk of SSI compared to IF.

In conclusion, both procedures have their own set of advantages which this review highlights. With advantages assessed in tandem, these results emphasise the need for individualised treatment choice for each patient. In patients who prioritise early post-operative return to baseline function and mobility, HA may be the preferred choice. Conversely, if minimising intraoperative stress and reducing early post-operative morbidity is of concern, IF may be the preferred choice.

Ultimately, the patient should be assessed as a whole and guided by the care team to make the most informed decision.

Consent statement

Consent was not obtained as review articles and details of study characteristics were publicly available.

Ethical statement

Not applicable, since review articles are available publicly.

Credit author statement

Zeremy Tang Conceptualization, Formal Analysis, Investigation, Resources, Data Curation, Writing – Original Draft, Writing – Review and Editing, Visualisation, Project Administration.

Ryan Wai Keong Loke: Conceptualization, Methodology, Validation, Formal Analysis, Investigation, Resources, Writing – Original Draft, Writing – Review and Editing, Visualisation, Project Administration.

Jonathan Jia En Boey: Conceptualization, Software, Validation, Methodology, Formal Analysis, Investigation, Resources, Writing – Review and Editing, Supervision, Project Administration.

Alexander Xi Xuan Ang: Conceptualization, Methodology, Resources, Writing – Original Draft, Project Administration.

Darren Keng Jin Tay: Conceptualization, Methodology, Writing – Review and Editing, Supervision, Project Administration.

Funding statement

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.