Intramedullary fixation versus arthroplasty in the treatment of geriatric trochanteric fractures: a systematic review and meta-analysis

Abstract

Background

Trochanteric fracture (TF) is of high incidence, especially among geriatric patients with osteoporosis. It imposes a heavy burden on both society and individuals. Surgical operation is the preferred choice for the treatment of TF, and intramedullary fixation and arthroplasty are two mainstream surgical procedures performed in clinical settings. This meta-analysis aims to attempt to explore the superiority of intramedullary fixation and arthroplasty in treating geriatric TF concerning the functional recovery, complications, mortality, and other operative details.

Methods

This review was planned, conducted, and reported following the PRISMA 2020 guidelines. By searching online databases (PubMed, Cochrane Library, Web of Science), related prospective studies from the year of 2000 to April 2024 were included in this meta-analysis. The language was restricted to English and Chinese. Studies were screened by two researchers independently, and the results of interest were analyzed by RevMan (version 5.4) and R (version 4.5.3). Both the conventional DerSimonian-Laird (DL) method and the Hartung-Knapp-Sidik-Jonkman (HKSJ) adjustment were employed to compare results, enhancing methodological transparency, and ensuring a more objective interpretation of therapeutic differences.

Results

A total of 981 studies were screened, of which 12 studies (7 randomized controlled trials (RCTs) and 5 prospective non-randomized controlled trials (non-RCTs)) involving 1,262 patients were finally included. Intramedullary fixation was associated with significantly shorter operative time (WMD -21.47, 95% CI -36.84 to -6.11, P = 0.006), reduced operative blood loss (WMD -176.29, 95% CI -232.32 to -120.27, P < 0.001), and superior long-term functional recovery at 12 months (SMD 0.50, 95% CI 0.16 to 0.84, P = 0.004). Notably, while initial DL estimates suggested that arthroplasty had advantages in short-term functional recovery (within 6 months, P = 0.04) and lower implant-related complications (P = 0.003), these differences lost statistical significance following the more conservative HKSJ adjustment (P = 0.1069 and P = 0.0565, respectively). However, arthroplasty maintained a lower revision rate (P = 0.009) across both models. Subgroup analyses indicated that nail types and fracture stability did not fundamentally alter the comparative trends (P_interaction > 0.05), except for blood loss (P_interaction = 0.0116). Meta-regression revealed that publication year significantly contributed to the heterogeneity of blood loss (P = 0.0133, R² = 51.51%), while mean age did not significantly influence any observed endpoints (all P > 0.05). No significant differences were observed in mortality, full weight-bearing time, general complications, or hospitalization time (P > 0.05).

Conclusions

Intramedullary fixation remains the primary choice for geriatric trochanteric fractures due to shorter operative times, reduced blood loss, and superior long-term function. Arthroplasty may offer potential advantages in short-term recovery, lower revision rates, and fewer implant-related complications, though these benefits were sensitive to the statistical model used. Therefore, primary arthroplasty should be considered a strategic alternative rather than a routine replacement, particularly for patients with poor bone quality or pre-existing hip arthritis.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12877-026-07902-z.

Background

Hip fractures are one of the most concerning issues of public health globally. It typically occurs as a consequence of a fall or stumble, especially among those with decreased bone density and reduced muscle strength [1]. Femur fracture, the most common type of hip fracture, mainly includes TF, subtrochanteric fracture, femur neck fracture, and femur head fracture. There were over 10 million cases per year involving femur fractures other than the femoral neck [2]. It presents a huge challenge and socioeconomic burden to both individuals and the healthcare system due to long hospitalization, loss of independence, decreased productivity, disability, and increased mortality.

Osteoporosis, which occurs in a large amount of the elderly [3, 4] is closely related to the geriatric TF. Accordingly, the treatment strategy may differ between the geriatric patients and the younger patients. Surgical treatment for TF is widely accepted as the optimal choice, especially for the unstable cases. Currently, intramedullary fixation and arthroplasty are predominant surgical strategies for TF among the elderly. The ineffective fixation of the nails and plates system caused by the poor condition of the bone mineral density is a widespread problem among osteoporosis patients, leading some surgeons to prefer arthroplasty. Meanwhile, arthroplasty guarantees an early weight-bearing, which may avoid some severe post-operation complications such as pneumonia, deep venous thrombosis, pressure ulceration, as well as ensure an on-time functional rehabilitation [5]. However, with the development of the intramedullary fixation system, new techniques are coming out to reduce the implant-related complications. Therefore, other researchers maintain that the arthroplasty brings about longer operation duration, more blood loss during the operation, which increases the risk in the operating theatre, and does not necessarily result in a better functional outcome [6–10].

A network meta-analysis of 5,703 patients with femoral neck fractures found that outcomes including the recovery of hip function, quality of life and reoperation risk showed an advantage of arthroplasty compared with internal fixation. Therefore, hip arthroplasty is recommended as the primary surgical method for femoral neck fractures [11]. This conclusion is consistent with the AAOS guideline for the treatment of unstable femoral neck fractures [12]. However, the trochanteric patterns are different from femoral neck and should be considered accordingly.

Despite the lack of evidence in prior research indicating that patients with stable TF may benefit more from intramedullary nails compared to extramedullary devices, intramedullary fixation demonstrated superior biomechanical outcomes in cases of unstable fractures [13, 14]. Therefore, this meta-analysis proceeds to compare intramedullary nail and arthroplasty with the following details. First, do intramedullary fixation and arthroplasty have a similar failure rate or reoperation rate? Second, which procedure may lead to fewer complications and a lower mortality rate? Then, how about the rehabilitation of short and long-term hip function? Finally, we also compared the related operation details and duration of hospitalization.

Considering osteoporosis is a vital factor corresponding to the prognosis in TF, and the age of the patients with high morbidity of osteoporosis is over 50 years old [15], the included age criteria of the subjects in this meta-analysis were 50 and above. With the development of new internal fixation systems since the onset of the twenty-first century, such as proximal femoral nail antirotation (PFNA), PFNA-II, and intertrochanteric antegrade nail (InterTAN), the mainstream surgical procedures, methodologies, and prognoses may have changed. For this reason, publications from the year 2000 were taken into consideration in this study.

Methods

Search strategy

The protocol of this systematic review was registered in PROSPERO (CRD: 420251067328). This study was conducted following the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement and the Cochrane review method [16]. A systematic literature search was conducted in PubMed, Cochrane Library, and Web of Science, filtering publications from the beginning of 2000 to April 2024. The keywords consist of Mesh and entry terms as follows: intertrochanteric fracture, trochanteric fracture; internal fixation, intramedullary fixation, cephalomedullary nail, proximal femoral nail (PFN), PFNA, proximal femoral nail, GAMMA nail, InterTAN; hip replacement, arthroplasty, prosthetic replacement, endoprosthesis, hemiarthroplasty. Manual search of potential correlated sources was also undertaken.

Inclusion and exclusion criteria

Studies were included based on the following criteria: 1. Prospective studies (RCTs and non-RCTs). 2. TF patients confirmed by radiography; 3. Patients over 50 years old; 4. Primary fracture in lower limbs; 5. Comparison of cases treated with any kind of intramedullary fixation and arthroplasty; 6. Studies included the results of interest.

Studies were excluded based on the following criteria: 1. Retrospective studies; 2. Repeated publications; 3. Publications not available for full text; 4. Bilateral TF; 5. Studies concentrated on TF with a certain comorbidity; 6. Studies with insufficient assessment of methodological quality or with obvious low quality; 7. Lack of sufficient outcomes of interest.

Data extraction

All of the primary articles were categorized according to the titles and abstracts by two reviewers independently, and subsequently screened by the full text for eligibility. Whenever necessary, the third reviewer was involved to resolve the discrepancies. The details of included studies were then extracted and analyzed, such as the baseline information, operational data, and relevant outcomes of prognosis.

Quality assessment

The risk of bias was independently assessed by two reviewers according to the standard risk bias assessment forms [17]. The risk of bias of RCTs was assessed by the Cochrane Collaboration`s tool, and non-RCTs were assessed with the Methodological Index for Non-randomized Studies (MINORS) [18]. P < 0.05 was considered statistically significant.

Statistical analysis

Statistical analyses were primarily conducted using RevMan (version 5.4). To ensure the robustness of our conclusions, particularly in the presence of high heterogeneity, we implemented a dual-methodological approach by comparing results from the conventional DerSimonian-Laird (DL) method with those adjusted by the Hartung-Knapp-Sidik-Jonkman (HKSJ) framework using R (version 4.5.3).

Heterogeneity was examined by Cochran`s Q test and I² statistic, with P > 0.10 and I² < 50% indicating low heterogeneity. Between-study variance (tau²) was estimated using the DL estimator with a random-effects model. The fixed-effect model was used in those which have no heterogeneity, and the random-effects model was applied when there is heterogeneity or when the evaluation methods of an outcome are not consistent across different studies.

Furthermore, we utilized the HKSJ adjustment during subgroup analyses to further explore potential sources of heterogeneity and to provide more conservative estimates where conventional methods might be underpowered. This side-by-side comparison of DL and HKSJ results was intended to enhance methodological transparency and facilitate a more objective interpretation of the findings (Supplementary material 1). To further verify the stability of the pooled results, a sensitivity analysis was performed, and a subgroup analysis was conducted to explore the source of heterogeneity whenever necessary.

Results

Baseline information of included studies

A total of 981 studies were yielded and then screened for eligibility. Most included fractures were unstable, except those in the three studies [8, 19, 20]. Twelve studies with a number of 1262 cases were finally included in this study, 7 of which were RCTs, and 5 of which were prospective non-RCTs (Fig. 1). There were 8 studies comparing PFN and hemiarthroplasty, 2 comparisons of PFNA and arthroplasty, 1 comparison of PFNA and femoral head replacement, and 1 comparison of Gamma nail and arthroplasty. The gender of the patients in this analysis has no statistical difference (P > 0.05). The age of included cases shows a significant difference (P < 0.05), and heterogeneity (I² = 82%) existed, which may influence the outcomes of prognosis to some extent. The baseline characteristics are summarized in Table 1.

Fig. 1.

Flow chart for the process of studies screening

Table 1.

Baseline characteristics

Studies	Country	Type of the study	Intervention	Sample size (n)	Side (left/right)	Mean age (year, range)	Gender (male/female)	Type of fracture	Follow-up (month)
Mansukhani et al. 2017 [6]	India	RCT	PFN HA	18 13	/	74(60–92) 74.38(61–92)	12/6 5/8	Evans III, IV, V	18.8 (12–28) 19 (12–30)
Jolly et al. 2019 [21]	India	RCT	PFN HA	50 50	/	81.2 ± 7.8 78.7 ± 8.3	/	unstable ITF	12
Garg et al. 2022 [22]	India	RCT	PFN HA	35 35	18/17 27/8	70.38 ± 7.79 (60–82) 76.24 ± 10.94 (60–94)	10/25 13/22	AO/OTA 31A2, 31A3	12
Desteli et al. 2015 [7]	Turkey	RCT	PFNA HA	42 44	/	67.0 ± 1.21 65.0 ± 1.52	27/15 26/16*	AO/OTA 31A1.3 ~ 31A3.3	24
Sun et al. 2021 [20]	China	PNS	PFNA FHR	52 56	19/33 24/32	68.23 ± 4.67 66.74 ± 4.82	20/32 20/36	Evans-Jensen I ~ IV AO/OTA A2.2 ~ A3.1	6
O ¨ zkayın et al. 2015 [19]	Turkey	RCT	PFN HA	21 33	7/14 13/20	79.57 ± 4.833 (75–91) 83.94 ± 4.924 (75–97)	9/12 10/23	AO/OTA A1 ~ A3	32.33 ± 10.97 31.33 ± 10.65
Öztürk et al. 2018 [8]	Turkey	PNS	PFNA HR	40 40	/	79 ± 10.1 (60–94) 81.4 ± 7.2 (62–93)	23/17 15/25	intertrochanteric fracture	12
Joshi et al. 2023 [9]	India	RCT	PFN HA	30 30	12/18 10/20	72.93 ± 7.9 73.13 ± 7.37	12/18 10/20	AO/OTA 31A2	6
Bonnevialle et al. 2011 [23]	France	PNS	Gamma HR	113 134	/	85.5 85.9	73.5%female 81.6%female	AO/OTA 31 A2.2	6
Kim et al. 2005 [10]	Korea	RCT	PFN HA	29 29	/	81 ± 3.2 82 ± 3.4	8/21 6/23	AO/OTA 31-A2 Evans III, IV	34(24–57) 35(24–58)
Hussain et al. 2017 [24]	India	PNS	PFN HA	30 30	/	at least 70	/	AO/OTA 31A2.2 ~ 31A3.1	6
Ekinci et al. 2020 [25]	Turkey	PNS	PFN HA	152 156	87/65 94/62	79 ± 6.1 77.7 ± 5.9	43/109 38/118	AO/OTA 31A1.2 ~ 31A3.3	24

HA hemiarthroplasty, HR (total) hip replacement, PFN proximal femoral nail, PFNA proximal femoral nail anti-rotation, FHR femoral head replacement, RCT randomized controlled trial, PNS prospective non-randomized study, ITF intertrochanteric fracture

^*In the study by Desteli et al., the gender breakdown for the arthroplasty group sums to 42 patients (26 males and 16 females), which is inconsistent with the explicitly stated group sample size of 44 patients. To maintain data integrity, values are reported exactly as extracted from the primary text

Quality assessment

The quality of included RCTs and non-RCTs was assessed by the Cochrane Collaboration`s tool and MINORS, and the results are shown in Fig. 2 and Table 2, respectively. The quality of eligible studies was acceptable. According to the Oxford Centre for Evidence-Based Medicine (CEBM) levels, this work constitutes a Level II systematic review. Although randomized controlled trials (RCTs) were included, the body of evidence also incorporates non-randomized comparative studies. The inclusion of these study designs, which are more susceptible to bias compared to RCTs, justifies the downgrade from Level I to Level II [26].

Fig. 2.

Quality assessment of RCT

Table 2.

Quality assessment of non-RCT

Items	Öztürk et al	Bonnevialle et al	Sun et al	Hussain et al	Ekinci et al
A clearly stated aim	2	2	2	2	2
Inclusion of consecutive patients	2	2	2	2	2
Prospective collection of data	2	2	2	2	0
End points appropriate to the aim of the study	2	2	2	2	2
Unbiased assessment of the study endpoint	1	1	1	1	1
Follow-up period appropriate to the aim of the study	2	0	0	0	2
Loss to follow-up < 5%	2	2	2	2	2
Prospective calculation of the study size	0	0	0	0	2
An adequate control group	2	2	2	2	2
Contemporary groups	2	2	2	2	2
Baseline equivalence of groups	2	2	2	2	2
Adequate statistical analyses	2	1	1	2	2
Total	21	18	18	19	21

 Outcome measures

Our study demonstrated that arthroplasty has advantages over intramedullary fixation in short-term hip functional recovery (within 6 months), reoperation rates, and implant-related complications, whereas intramedullary nailing is beneficial for shorter operation duration, less intraoperative blood loss, and better long-term hip function (after 12 months). The differences between these two surgical procedures were not significant in mortality rates, time to weight-bearing, and overall complication rates. The secondary outcomes include fundamental information from the included studies (Table 3), regarding the operation time, blood loss, blood transfusion, and hospital stay during admission. The primary outcomes are shown in Table 4 and 5, which represents postoperative assessments. The meta-analysis results are shown in Fig. 3.

Table 3.

Fundamental information during admission

Studies	Operation time (min)	Blood loss (ml)	Blood transfusion (units)	Hospital stay (day)
Mansukhani et al. 2017 [6]	86.7 ± 22.7 106.2 ± 26.31	252 ± 146.0 573 ± 152.2	1.14 ± 0.38 1.6 ± 0.52	17.72 ± 4.14 18.27 ± 4.43
Jolly et al. 2019 [21]	48.4(6.4) 60.4(9.9)	46 ± 13.5 187 ± 38.1	/	/
Garg et al. 2022 [22]	112.80 ± 17.564 115.6 ± 10.92	130.3 ± 48.639 200 ± 34.15	/	8.42 ± 2.542 6.52 ± 1.19
Desteli et al. 2015 [7]	76.0 ± 4.19 72.0 ± 2.89	/	/	7.0 ± 0.83 6.0 ± 0.68
Sun et al. 2021 [20]	53.37 ± 13.83 76.87 ± 17.58	185.38 ± 38.37 257.47 ± 53.27	/	/
O ¨ zkayın et al. 2015 [19]	28.19 ± 12.044 52.33 ± 10.457	/	/	6.76 ± 3.846 5.55 ± 1.922
Öztürk et al. 2018 [8]	5.3 ± 2.5 (1–13) 6.1 ± 3.3 (1–16)	/	0.95 ± 1.4 (0–6) 1.25 ± 1.2 (0–5)	12.3 ± 4 (4–23) 14.6 ± 5.9 (4–32)
Joshi et al. 2023 [9]	86.07 ± 11 144.93 ± 9.76	84.5 ± 15.05 263.67 ± 42.95	/	12.57 ± 1.33 12.83 ± 1.26
Bonnevialle et al. 2011 [23]	/	/	2.47 ± 1.1 2.29 ± 0.9	/
Kim et al. 2005 [10]	60 ± 17 96 ± 26	168 ± 44 511 ± 103	0.8 ± 0.7 1.9 ± 0.8	11 ± 3.1/13 ± 2.6

Table 4.

Postoperative assessment

Studies	Weight-bearing time (days)		Harris or other functional assessment postoperatively		Shortening (cm)
	Partial weight-bearing time (days)	Full weight-bearing time (days)	6 months	at least 12 months
Mansukhani et al. 2017 [6]	/	43.19 ± 5.04 44.03 ± 5.32	/	*6.75 ± 2.3 6.78 ± 2.3	0.69 ± 0.68 0.95 ± 1.12
Jolly et al. 2019 [21]	/	70.7 ± 24.5 22.4 ± 32.9	74.2 ± 14.3 70.4 ± 15.1	86.7 ± 13.1 70.3 ± 18.7
Garg et al. 2022 [22]	14.25 ± 8.34 3	48.20 ± 10.25 5.91 ± 1.76	77.7 ± 9.40 79.90 ± 7.70	86.85 ± 10.52 85.40 ± 7.53	0.50 ± 0.02 (8 cases) 0.4 ± 0.64 (12 cases)
Desteli et al. 2015 [7]	/	/	/	*5.04 ± 0.02 4.9 ± 0.18
Sun et al. 2021 [20]	19.13 ± 6.88 4.12 ± 1.02	/	81.23 ± 6.411 83.78 ± 6.62	/
O ¨ zkayın et al. 2015 [19]	/	/	68.14 ± 6.544 66.91 ± 10.672	75.95 ± 7.324 68.44 ± 10.549	1 case (4.8%) 0 case
Öztürk et al. 2018 [8]	/	/	/	+97.7 ± 28.7 93.1 ± 22.8
Joshi et al. 2023 [9]	/	/	72.83 ± 3.89 79.72 ± 2.53	/
Bonnevialle et al. 2011 [23]	/	/	*4.3 ± 1.5 5.1 ± 1.5	/
Kim et al. 2005 [10]	/	/	/	82 ± 12.4 80 ± 9.7
Ekinci et al. 2020 [25]	(median)3.4 1.8	/	/	/
Hussain et al. 2017 [24]	/	/	90.3 ± 4.4 92.4 ± 3.8	/

⁺functional independence measure (FIM) score

^*Assessed by the mobility score of Parker and Palmer

Table 5.

Postoperative assessment (complications and mortality)

	General complications		Complications related to the implant			Mortality
	Complications	Number of cases	Internal fixation	Arthroplasty	Reoperation
Mansukhani et al. 2017 [6]	SI, DI, BS, UTI, VTE	5 (27.7%) 4 (30.5%)	C/O 2(12.5%) FH-AVN 2(12.5%)	D/L 2(18.2%)	2(12.5%) 0	(1y) 2 (11.1%) 3 (23.1%)
Jolly et al. 2019 [21]	SI, DI, BS, UTI, LRTI, VTE	38 26	FH-AVN 1 C/O 5	10 (20%)	4 4	(1 m) 4(8%)/2(4%) (1y) 10(20%)/12(24%)
Garg et al. 2022 [22]	SI, BS	1 3	/	/	/	/
Desteli et al. 2015 [7]	SI	4 4	0	0	/	/
Sun et al. 2021 [20]	SI, BS, UTI, VTE, LRTI	13 (25%) 6 (10.71%)	FA 1(1.92%)	D/L 2 (3.57%)		0
O ¨ zkayın et al. 2015 [19]	SI, DI	1 (1.9%) 1 (1.9%)	0	0	1 (4.76%) 0	/
Öztürk et al. 2018 [8]			/	/	/	(6 m) 13(32.5%)/14(35%) (12 m) 14(35%)/16(40%)
Joshi et al. 2023 [9]	SI	2 2	C/O 2 (6.67%)	D/L 1 (3.33%)	/	(4 m) 0/1(3.33%)
Bonnevialle et al. 2011 [23]	VTE, CRC, NC, sepsis, infection	15 19	D/A 6 (4.48%)	D/L 2 (1.9%)	12.5%/2.8%	(6 m) 24(21.2%)/28(21%)
Kim et al. 2005 [10]	SI, VTE, RC, CVC, NC	7 11	C/O 2 (6.90%) N/U 3	D/L 1 (3.45%)	1 (3.45%) 0	(before discharge) 0/2(6.90%) (1y) 4(13.79%)/6(20.68%) (1-3y) 1(3.45%)/8(27.59%) overall 5(17.24%)/16(55.17%)
Ekinci et al. 2020 [25]	VTE	3 17	16	2	22 15	8 12
Hussain et al. 2017 [24]	infection, VTE	0 3	Z/E 3 N/U 1	D/L 1	/	/

SI superficial infection, DI deep infection, BS bed sore, UTI urinary tract infection, VTE venous thromboembolism, LRTI lower respiratory tract infection, CRC cardiorespiratory complications, NC neurological complications, CVC cardiovascular complications, RC respiratory complications, D/L dislocation, D/A disassemblies, C/O cut-out, FH-AVN femoral head avascular necrosis, Z/E Z-effect, BK breakage, FA fixation abnormality, N/U non-union

Fig. 3.

Forest plot diagram of the comparison between internal fixation and arthroplasty. A Operation time. B Blood loss. C Hospital stay. D Functional recovery (6 months postoperatively). E Functional recovery (12 months or longer postoperatively). F Full weight-bearing time. G General complications. H Complications related to the implant. I Reoperation. J Mortality

Operation time

There were 8 articles involving 567 cases that provided data on operation duration [6, 7, 9, 10, 19–22]. The pooled results indicated that the duration of operation was significantly shorter in the internal fixation group than in the joint replacement group (WMD −21.47, 95% CI −36.84 to −6.11, P = 0.006) with significant heterogeneity (Tau² = 473.01, Chi² = 617.05, P < 0.05, I² = 99%). The result was proven reliable according to the sensitivity test applied by excluding each study separately (Table 3, Fig. 3A).

Blood loss

Six studies involving 427 patients provided data about the blood loss during the surgery [6, 9, 10, 20–22]. The blood loss is significantly less in the internal fixation group than in the arthroplasty group (WMD −176.29, 95% CI −232.32, −120.27, P < 0.001). The heterogeneity among the six studies was shown in the heterogeneity test (Tau² = 4454.99, Chi² = 238.28, P < 0.05, I² = 98%), and the heterogeneity could not be eliminated after sensitivity analysis (Table 3, Fig. 3B).

Hospital stay

Seven articles reported the hospitalization stay of 439 cases in total [6–10, 19, 22]. The duration of hospital stay in the two groups was no different, with the P-value of 0.87 (WMD 0.08, 95% CI −0.90, 1.06, P = 0.87). The heterogeneity exists when using the random effect model (Tau² = 1.21, Chi² = 38.19, P < 0.05, I² = 84%), and the heterogeneity could not be eliminated in the sensitivity analysis (Table 3, Fig. 3C).

Functional recovery

Results of the Harris score from 7 studies [9, 10, 19–22, 24], results of the mobility score of Parker and Palmer from 3 studies [6, 7, 23], and results of the Functional Independence Measure scores from 1 study [8] were analyzed. The outcomes of functional recovery at 6 months and at least 12 months postoperatively are shown in Table 4. The random effect model was applied in the meta-analysis.

Seven studies evaluated the functional outcomes six months following the surgery, and better functional recovery in the arthroplasty group was presented compared to the internal fixation group. For the reason that the functional outcomes were evaluated by different scales, the standardized mean difference was used (SMD −0.46, 95% CI −0.90, −0.03, P = 0.04), and the outcome was consistent with most of the results presented in the included studies. The heterogeneity analysis showed that there was heterogeneity among the studies (Tau² = 0.29, Chi² = 43.36, P < 0.05, I² = 86%). The main heterogeneity was caused by the studies of Jolly et al. and Joshi et al.. After removing these articles, the I² in the heterogeneity test decreased to 23% with no significance altered (Fig. 3D).

However, the functional recovery in the internal fixation group was significantly better than that in the arthroplasty group at least 12 months after the surgery (SMD 0.50, 95% CI 0.16, 0.84, P = 0.004) according to seven articles. There was high evidence of heterogeneity across the studies (Tau² = 0.15, Chi² = 19.85, P = 0.003, I² = 70%). Excluding the studies of Desteli et al. and Jolly et al. simultaneously caused a low heterogeneity with the I² of 6%, and it did not change the significance (P = 0.04). For the reason that the functional outcomes were evaluated by different scales, the standardized mean difference was used (Fig. 3E).

Notably, the statistical significance of short-term functional recovery was sensitive to the choice of the meta-analysis model. After applying the HKSJ adjustment to account for high heterogeneity, the initial advantage observed in the arthroplasty group was no longer sustained (SMD −0.48, 95% CI −1.06, 0.10, P = 0.1069). In contrast, the superiority of internal fixation at 12 months remained statistically significant even after the HKSJ adjustment (SMD 0.50, 95% CI 0.15, 0.85, P = 0.0046).

Full weight-bearing time

Only three studies provided the data about full weight-bearing time, and no significant difference was found between the two groups (WMD = 29.71, CI −3.81, 63.23, P = 0.08) [6, 21, 22]. However, the heterogeneity was significant (Tau² = 864.35, Chi² = 297.63, P < 0.05, I² = 99%), and it was caused by the study of Mansukhani et al.. In the study of Mansukhani et al., the full weight-bearing time in the internal fixation group was slightly shorter than that in the arthroplasty group, but it showed no statistical significance. In the other two studies, it took a markedly shorter period to bear the full weight in the arthroplasty group (P < 0.05) (Table 4, Fig. 3F).

General complications

Ten studies reported the general complications of 822 cases postoperatively, and the results did not indicate a significant difference (RR = 1.24, 95% CI 0.97, 1.58, P = 0.09) [6, 7, 9, 10, 19–24]. The heterogeneity was not significant (Chi² = 10.70, P = 0.30, I² = 16%), and a fixed effect model was used. After excluding the study of Hussain et al., there was no evidence of heterogeneity (Chi² = 7.57, P = 0.48, I² = 0). The general complications mainly include superficial and deep infections, venous thromboembolism, pneumonia, bed sores, urinary tract infection, and cardiovascular accidents (Table 5, Fig. 3G).

Complications related to the implant

A total of 8 studies reported on the results of implant-related complications [6, 7, 9, 10, 20, 21, 23–25]. Desteli et al. did not find any complications relevant to the implant. Among the other articles, the incidence rate of implant-related complications was significantly higher in the internal fixation group in contrast with the other group (RR = 2.15, 95% CI 1.31, 3.53, P = 0.003). The heterogeneity was acceptable (Chi² = 13.52, P = 0.06, I² = 48%), and the heterogeneity did not exist after removing the study of Jolly et al. (I² = 0). The most common complications were cut-out and dislocation in the osteosynthesis and arthroplasty groups, respectively (Table 5, Fig. 3H).

Furthermore, the HKSJ adjustment was implemented to test the consistency of the findings regarding implant-related complications. Although the conventional DL method showed a significant difference favoring arthroplasty (P = 0.003), the HKSJ-adjusted analysis yielded a non-significant result (RR = 2.18, 95% CI 0.98, 4.84, P = 0.0565).

Reoperation

Six studies with a total of 741 patients reported the information about revisions [6, 10, 19, 21, 23, 25]. The revision rate in the internal fixation group was significantly higher compared to the arthroplasty group (RR = 1.89, 95% CI 1.17, 3.05, P = 0.009), with no heterogeneity (Chi² = 3.66, P = 0.60, I² = 0). The only patient who underwent the reoperation was caused by deep infection in the study of Ozkayin et al. The reasons that led to revision in the study of Bonnevialle et al. were disassemblies and infections in patients who underwent osteosynthesis, and dislocations and sepsis among the patients with arthroplasty. According to the article of Mansukhani et al., two patients treated with PFN went through the proximal screw migration and developed avascular necrosis (AVN) of the femoral head, who were revised with a total hip replacement then. In the study of Kim, three cases in the PFN group had a nonunion. Two were caused by the cutting-out of the hip screw, and they refused a reoperation; another was caused by the concomitant breakage of the anti-rotatory screw and then converted to a calcar-replacement implant. Ekinci et al. reported 22 cases of revision surgery in the PFN group for 5 cases of cut-out and migrations, 2 cases of deep infections, 11 cases of periprosthetic fractures, and 4 cases of insufficient primary osteosynthesis; fifteen cases in the BHA group were performed the revision surgery, with 2 cases for the reason of dislocations, 6 cases for the reason of deep infections, and 7 cases for the reason of periprosthetic fractures. In the study of Jolly et al., four cases of cut-off in the osteosynthesis group and 4 cases of peri-prosthetic fracture in the arthroplasty group underwent revision (Table 5, Fig. 3I).

Mortality

Seven articles reported on the mortality among the 878 patients [6, 8–10, 21, 23, 25]. The heterogeneity was not shown (Chi² = 1.47, P = 0.96, I² = 0), and a fixed effect model was applied. The difference between the internal fixation group and arthroplasty group was not significant (RR = 0.85, 95% CI 0.64, 1.14, P = 0.29) (Table 5, Fig. 3J).

Subgroup analyses

Subgroup analyses stratified by implant types

Subgroup analyses were performed for different intramedullary nail designs (PFN vs. PFNA vs. Gamma) to assess the consistency of the treatment effects. Regarding arthroplasty, subgroup analysis comparing hemiarthroplasty and total hip replacement was not conducted due to the insufficient number of studies available for a meaningful comparison. Consequently, the subgroup analyses focused on nail types.

The results demonstrate that the specific type of implant did not fundamentally alter the comparative trend between intramedullary fixation and arthroplasty across most clinical indicators. The non-significant interaction P-values (P_interaction > 0.05) for the majority of outcomes confirm the robustness of our pooled estimates. However, a significant interaction was still observed for blood loss (P_interaction = 0.0116). Regarding heterogeneity, substantial intra-group heterogeneity persisted despite the stratification. The most notable reduction was observed in the subgroup analysis of functional recovery at 12 months, where the I² for the PFN group decreased from 69.2% to 65.1%. (Supplementary material 2).

Due to the limited available data, a subgroup analysis based on specific AO/OTA fracture classifications was not feasible. Therefore, studies were categorized into "all unstable" and "not all unstable" groups to evaluate the impact of fracture stability. Under the HKSJ adjustment, the P_interaction values for all assessed outcomes were greater than 0.05. (Supplementary material 3).

Meta-regression analyses

Publication year

Meta-regression analysis revealed that publication year significantly contributed to the heterogeneity of blood loss (P = 0.0133, R² = 51.51%). For other outcomes, including hospital stay (P = 0.2608, R² = 4.73%) and full weight-bearing time (P = 0.2724, R² = 9.28%), the publication year explained only a minor proportion of the heterogeneity. Regarding the reoperation rate, although the meta-regression did not reach statistical significance (P = 0.1482), the publication year explained a substantial proportion of the heterogeneity (R² = 51.34%) between included studies. In contrast, for general complications (P = 0.4904), implant-related complications (P = 0.4535), mortality (P = 0.7416), operation time (P = 0.9964), and functional recovery at 6 and 12 months (P = 0.3855 and P = 0.8644, respectively), the R² values were all 0.00% (Fig. 4).

Fig. 4.

Meta-regression Bubble plot of publication year for Blood Loss (left) and Reoperation Rate (right)

Mean age

The results indicated that the weighted total mean age of patients did not significantly influence any of the observed endpoints (all P > 0.05). In terms of operative and perioperative parameters, no significant moderating effect of age was found for operation time (P = 0.4644, R² = 0.00%), blood loss (P = 0.1434, R² = 18.63%), or hospital stay (P = 0.2136, R² = 9.06%). Similarly, age showed minimal impact on functional recovery at 6 months (P = 0.4996, R² = 0.00%) and 12 months (P = 0.2900, R² = 3.03%), as well as full weight-bearing time (P = 0.5728, R² = 0.00%). Regarding safety outcomes and potential sources of heterogeneity, although statistical significance was not reached, mean age accounted for a substantial proportion of the variance in mortality (P = 0.3178, R² = 62.98%) and reoperation rate (P = 0.255, R² = 21.6%). General complications (P = 0.766) and implant-related complications (P = 0.4273) remained stable across the age spectrum, both with R² values of 0.00%.

Discussion

Hip fracture is a serious issue in global public health. It is reported that the age-standardized incidence rate of hip fracture in 2019 was 182.5 (95% UI 141.9 to 230.9) per 100,000 population, and the age-standardized prevalence rate was 298.1 (95% UI 271.6 to 331.6) per 100,000 population in that year. Femur fracture, excluding the femur neck, was the anatomical site that ranked second among the most prevalent fracture sites [2]. It not only imposes a heavy burden on society and families, but also severely impairs the life quality of the patients. More than ninety percent of TF patients are over 65 years old, and they are always combined with comorbidities [27]. Although intramedullary nail is a widely used treatment for TF, it may fail in unstable type, especially for older patients with osteoporosis. Therefore, this meta-analysis was conducted to explore the superiority of two surgical procedures for treating geriatric TF across multiple outcome measures. Main outcomes included recovery of hip function, revision rates, weight-bearing time, complications, and mortality rate; secondary outcomes included operation time, intraoperative blood loss, and hospital stay.

Our study showed that arthroplasty is advantageous in relation to short-term hip function, implant-related issues, and revision rates, whereas intramedullary fixation reduced operation time and blood loss, and improved long-term hip function. However, the differences of short-term functional recovery (within 6 months, P = 0.04) and implant-related complications (P = 0.003) lost statistical significance following the HKSJ adjustment (P = 0.1069 and P = 0.0565, respectively). These shifts indicate that the perceived benefits of arthroplasty in these domains are less robust than previously estimated and should be interpreted with clinical prudence. Meanwhile, this underscores the importance of methodological transparency and suggests that the comparative advantages of either procedure in these specific metrics are not definitive. This meta-analysis also revealed high evidence of heterogeneity in secondary outcomes. These may be caused by the different surgical procedures and competency of the surgeons, various racial groups, and the preoperative health condition of the cases. Regarding blood transfusion, although data were available in four studies, a formal meta-analysis was not performed due to inconsistent reporting formats and unclear patient numbers. Specifically, Öztürk et al. and Mansukhani et al. [8] reported no significant difference in transfusion amounts (P = 0.325 and P = 0.124, respectively), whereas Kim et al. [10] found significantly fewer units transfused in the internal fixation group (P < 0.05). Bonnevialle et al. provided only mean transfusion units (2.47 ± 1.1 vs 2.29 ± 0.9) without further statistical comparison [23]. Given these discrepancies, we concluded that a quantitative synthesis of transfusion data would have limited clinical significance. The secondary outcomes about operation time and intraoperative blood loss are generally consistent with other meta-analyses [28–30]. However, it is worth noting that in the study of Jia-bao Ju et al. [31], intramedullary fixation resulted in less blood loss during the surgery but failed to shorten the surgical duration. This discrepancy might be attributed to the varying technical proficiency of the surgeons.

Meta-regression analysis revealed that while publication year significantly contributed to the heterogeneity of blood loss, the results remained stable regardless of the publication year for many other clinical outcomes. These findings suggest that the results of the comparative efficacy and safety between intramedullary nail and arthroplasty are consistent over time. Beyond the temporal factor, we further explored the potential moderating effect of patient age via meta-regression. Similar to the publication year, the weighted total mean age did not significantly influence any clinical outcome (all P > 0.05). Specifically, while most parameters showed R² values of 0.00%, mean age accounted for a substantial proportion of the variance in mortality (R² = 62.98%) and reoperation (R² = 21.67%). This suggests that while the overall comparative efficacy remains robust, the diversity in postoperative survival and secondary surgery risk across studies is largely driven by the baseline age of the patient populations rather than the era of the study.

In relation to the subgroup analysis of implant type, the results demonstrated that the specific type of implant did not fundamentally alter the comparative trend between intramedullary fixation and arthroplasty across most clinical indicators. The only significant interaction between subgroups was observed for blood loss (P_interaction = 0.0116). However, it should be noted that the PFNA subgroup comprised only a single study, whereas the PFN subgroup included five studies. While this suggests that nail design may influence surgical blood loss, the results should be viewed as exploratory rather than definitive due to the limited sample size. Overall, the consistent findings across most subgroups further reinforce the reliability of our primary conclusions. Exploration of heterogeneity through stratification revealed that while the PFN subgroup in the 12-month functional recovery analysis showed a slight decrease in I² (from 69.2% to 65.1%), substantial variance remained across most categories. This implies that while implant design is a critical modifier for certain functional and trauma-related outcomes, the substantial variance observed in other procedural metrics is likely influenced by additional clinical factors, such as surgical technique and patient baseline characteristics. In addition, in the subgroup analysis concerning fracture stability, the P_interaction values for all assessed outcomes exceeded 0.05, suggesting that the comparative effectiveness of the two surgical interventions remains consistent across different fracture stability profiles.

The outcome parameter of weight-bearing time revealed no significant difference between the two groups. However, there is high evidence of heterogeneity, which might be attributed to variations in surgical proficiency, rehabilitation protocols, and the limited number of included studies. It is known that early commencement of weight-bearing ambulation and short bed-rest time are able to decrease various complications [5], such as pneumonia, VTE, pressure ulceration, and others (Table 4). Partial weight bearing can be quite different between studies according to the treatment pattern of the doctors (Table 4). In general, the time of weight bearing was shorter in the patients with an arthroplasty than those with an osteosynthesis [32]. Patients in the arthroplasty group of Garg et al. all started to bear weight on the third day after surgery, while this figure in other studies varied according to the situation. It is known that early weight bearing promotes the repair of the fracture and prevents myoatrophy. In the study of Bafor et al., earlier weight-bearing ambulation resulted in faster mineralization of the regenerate bone [33]. Additionally, Sherrington et al. manifested that patients with hip fractures treated with weight-bearing exercise developed a better balance and functional ability than those with non-weight-bearing exercise [34]. According to the fracture healing process mediated by angiogenesis and mechano-regulation, Ganadhiepan et al. developed an algorithmic model to establish the optimal weight-bearing scheme on the sheep fracture model to enhance fracture healing [35]. The full weight bearing from 48 h postoperatively was broadly recommended [36, 37], nevertheless, it is hard to apply in the practical clinical settings. The mean bed rest time varies among different countries. For instance, the partial weight-bearing time in China and Japan is later than that in the United States and Europe [38]. In Europe and the United States, patients usually initiate partial weight-bearing on the following day after intramedullary fixation [31]. However, in most medical institutions in China, patients with trochanteric fractures are recommended to be immobilized for at least four weeks following intramedullary nail fixation, whereas this restriction period is reduced to three days for those undergoing hip replacement. In addition, the timing of postoperative weight-bearing varies according to individual bone quality. Patients with osteoporosis typically require prolonged restricted weight-bearing, whereas early mobilization is critical for optimal functional recovery of the hip joint. This may lead to different prognoses related to complications and functional recovery among the patients in various studies, and thus these variations represent key sources of heterogeneity in our meta-analysis.

In the meta-analysis of functional recovery 6 months following the surgery, the mean scores and standardized difference were not given exactly in the study of Sun et al. but were shown by a graph. The figures were extracted by the software AutoCAD 2018, and this might bring about measurement bias. Bonnevialle et al. did not provide the standard deviation (SD) of the movement scores, and the figures were estimated according to the article of Hozo et al. [23, 39]. Except that the long-term Harris score of Kim et al. was evaluated 24 months after the surgery [10], the rest results applied in the meta-analysis were all assessed 12 months postoperatively. Regarding the outcomes for functional recovery 6-month after surgery, which initially showed a significant difference in the primary DL model (P = 0.04), we have conducted a detailed sensitivity analysis to evaluate this finding. First, as estimated data might introduce potential measurement bias, we individually excluded Bonnevialle et al. and Sun et al.. However, their removal resulted in no significant improvement in heterogeneity (I²). Instead, the exclusion of Joshi et al. was found to significantly decrease the heterogeneity from 86 to 63%. Second, our leave-one-out sensitivity analysis demonstrated the unstable nature of the initial significance (P = 0.04). We found that the P-value remained below 0.05 only when Jolly et al. or Özkayin et al. were individually excluded; the removal of any other single study shifted the result to non-significance (P ≥ 0.05). This finding is consistent with results after the HKSJ adjustment (P = 0.1069). After integrating the HKSJ adjustment and sensitivity analysis, the statistical significance of 6-month functional recovery was no longer sustained (P = 0.1069). This shift indicates that the early functional advantage is not robust and is highly dependent on specific study data. Notably, compared to the conventional DL model, the HKSJ adjustment provides a more conservative and reliable estimate by accounting for the uncertainty of heterogeneity among a small number of studies, thereby ensuring a higher level of statistical evidence and preventing potential Type I errors.

Moving to the long-term outcomes, after reviewing the article of Jolly et al., we found that the mean Harris score 12 months postoperatively was slightly lower than that in the 6 months. In the present study, the incidence of implant-related complications was relatively high in relevant to other included studies, with 12% in the internal fixation group and 20% in the arthroplasty group. Concerning the functional recovery at least 12 months after the surgery, the heterogeneity was significant (I² = 70%), and it no longer existed after the studies of Jolly et al. and Desteli et al. were excluded (I² = 6%) [7, 21]. As a result, we assumed that the abnormal prognosis of hip function recovery in the arthroplasty group might be caused by the surgical procedure details, postoperative care, and rehabilitation scheme.

The World Health Organization(WHO) has disclosed and called for attention to the phenomenon that the rehabilitation needs were unmet around the world, especially in underdeveloped and developing countries [40]. By performing rehabilitation after the operation, it is expected to improve or restore the function, or slow down the rate of functional loss [41]. Studies manifested that a longer course of rehabilitation improves the functional restoration [42, 43]. However, the fracture pattern and general condition are different from case to case, and the rehabilitation schedule, including physiotherapy and occupational therapy, should be modified individually. Consequently, both operation methods and rehabilitation plans contribute significantly to divergent functional outcomes in existing studies. While the decline of hip function remains inevitable after TF regardless of therapeutic intervention [8], therefore, we emphasize that the broader injury-prevention initiatives and enhanced accessibility to osteoporosis screening and management in the elderly are essential to relieve the overall burden and preserve the quality of patients` lives [44, 45].

Ryan et al. reported that the time from injury to operation was related to the prognosis [46]. Surgery within the first 24 h would decrease the incidence rate of mortality, non-union, and other complications. As recommended by the British Orthopaedic Association, the surgeries should be performed on the older patients with hip fractures within 36 h of admission [47]. However, only a limited number of articles in this study provided valid data for analysis [8, 23]. Clinically, performing such emergency surgery within 24 h is challenging, and variations in operative timing after the injury may contribute to heterogeneity across various evaluation indices. In a large retrospective cohort study of patients with femoral fractures [48], operation time was identified as a significant independent risk factor for complication and mortality rate after a hemiarthroplasty. Specifically, each additional minute of surgery was associated with a 2.2% increase in the odds ratio (OR) of complications. Furthermore, patients whose procedures exceeded the 86-min threshold faced a 111.8% higher OR for mortality. Additional factors negatively influencing outcomes included advanced patient age, male sex, and less experienced surgeons. Therefore, the pooled estimates regarding the comparation between arthroplasty and internal fixation could be confounded by factors other than implant choice.

In patients with femoral neck fractures, a meta-analysis of RCTs revealed that bipolar hemiarthroplasty via the SuperPATH approach yields better short-term outcome (HHS) compared to conventional approaches [49]. Furthermore, the advantages of this approach extend to significant perioperative benefits, including reduced incision length, blood loss, pain (VAS), hospital stay, as well as faster mobilization. This study reminds us that prognostic determinants extend beyond the choice of implant to include surgical specifics such as the operative approach, use of bone cement, type of hip prosthesis employed, and other details. As such, differences in surgical technique and operation details inevitably contributed to heterogeneity across the included studies.

Numerous studies have investigated the relative superiority of surgical treatments for trochanteric fractures among geriatric patients, and the outcomes vary. A meta-analysis by Dong-peng Tu et al. showed that hemiarthroplasty enables earlier weight-bearing and has lower implant-related complications than internal fixation, despite requiring longer operative time and causing greater blood loss [28]. The outcomes of hip function recovery favored internal fixation within two years but showed no significant difference between the two groups after two years. Consequently, they suggested hemiarthroplasty could be recommended over internal fixation for geriatric patients with unstable intertrochanteric fractures [28]. In this study, while the hip functional recovery was better in the arthroplasty group in the short-term following the surgery, it presented a better outcome in the internal fixation group a year after the operation. This was consistent with the study of Zhaojun Wang et al. [50]. However, the included articles were not eligible for analyzing long-term outcomes beyond 2 years, and hip functional recovery might change or even reverse over time, as reported in the study by Dong-peng Tu [28]. The results of some meta-analyses showed that the Harris hip score, representing hip function, was better in the internal fixation group at the latest follow-up [29, 30]. Nevertheless, some articles reported superior functional outcomes with arthroplasty [51], while others showed no significant difference between the two surgical groups [31]. A network meta-analysis comparing internal fixation and arthroplasty for femoral neck fractures indicated that arthroplasty was superior in several aspects [11]. However, Femoral neck fracture is intracapsular, which has poor blood supply to the femoral head. Thus, there is a high risk of avascular necrosis and nonunion among patients with femoral neck fracture. Unlike femoral neck fracture, trochanteric fracture is extracapsular, which normally remains rich blood supply and carries a relatively low risk of avascular necrosis and nonunion. Moreover, the articular cartilage, labrum, and other structures of the hip joint remain unaffected. Meanwhile, Standard arthroplasty prosthesis stems are designed to be fixed within an intact femoral canal. In a comminuted trochanteric fracture, the prosthesis stem lacks reliable proximal and distal fixation, posing a high risk of postoperative complications such as stem subsidence, loosening, or even catastrophic femoral shaft splitting. While specially designed "long-stem" or "calcar-replacing" prostheses exist, they are costly, involve more extensive surgery. Therefore, the evidence of surgical strategy for femoral neck fracture could have limited reference value for trochanteric fracture. Simultaneously, there were limitations in these meta-analyses, as they failed to carry out an analysis of how hip function changed over time. Our study, therefore, introduced a subgroup analysis, resolving this critical limitation.

According to the guidelines, the American Academy of Orthopaedic Surgeons (AAOS) recommends that older adults with unstable intertrochanteric fractures should be treated with a cephalomedullary device [12]. The National Institute for Health and Care Excellence (NICE) states that extramedullary implants are preferred over intramedullary nails for the treatment of adult trochanteric fractures above and including the lesser trochanter, except reverse oblique fractures, as intramedullary implants are associated with higher reoperation rates. For 31A3 fractures, available evidence remained insufficient to make clinical recommendations [52]. The motivation for early ambulation and concern over fixation failure has driven some surgeons to carry out arthroplasty. However, these guidelines failed to address a direct comparison between intramedullary devices with arthroplasty, so we conducted this study to compare the standard treatment to an off-label treatment approach. While further supporting the established reliability of internal fixation, this study affirms the comparable safety and efficacy of arthroplasty for geriatric trochanteric fractures.

Based on our findings, both surgical techniques demonstrate clinical efficacy, and each possesses distinct advantages and limitations. Notably, surgical outcomes are determined not only by implants selection but by a range of technical and human factors, including the surgical approach, operative details, implant type and even surgeon proficiency. Therefore, the selection should be guided by individualized patient factors, including patient age and expected prosthesis longevity, overall health status and relevant complications, tolerance for revision surgeries, and the personal willingness of the patients. A potential downward trend in the risk ratio for reoperation over time was observed through meta-regression (P = 0.1482), potentially reflecting maturing surgical techniques or improved implant designs in recent years. With the evolution of prosthesis technology and surgical techniques, it is anticipated that not only will surgical injury be reduced, and operation duration shortened, but complication rates and surgery failure rates will also decline, and prognosis will also improve in the foreseeable future.

It is noteworthy that the majority of previous meta-analysis evaluating intrameduallary fixation versus arthroplasty for trochanteric fractures relied exclusively on conventional DL method. In addition to incorporating the most recent clinical evidence, our study provides a direct comparison of both DL and HKSJ results, offering a more conservative and reliable estimation of effect sizes. This methodological transparency enables readers to more objectively interpret the differences between these two surgical interventions across multiple outcomes. These findings further underscore the necessity of employing the HKSJ adjustment in meta-analyses, as it provides a more conservative and reliable statistical framework for evaluating surgical outcomes. By accounting for the uncertainty in heterogeneity, especially critical when dealing with a small sample size of studies, this approach significantly reduces the risk of Type I errors and enhances the statistical robustness of the clinical inferences [53].

Nevertheless, several limitations should be acknowledged. The existence of limitations was acknowledged in this study. First, the included studies were limited, leading to the restricted amounts of valid results of some of interested items, such as blood transfusion, shortening of the limbs, and Harris scores at certain time points. Analysis of hip functional recovery beyond 1 year was not carried out due to insufficient data from the included studies. Specifically, due to the lack of granular data providing detailed fracture classifications and implant designs in the included studies, we were unable to perform a subgroup analysis based on AO/OTA subtypes or specific prosthesis types. This limitation precludes the development of a standardized treatment template, thereby representing a vital direction for future investigation. As more detailed data on fracture sub-types becomes available, research can further transition towards evidence-based, individualized surgical decisions. Second, not all articles included were RCTs, which may result in the bias that patients with poor general and regional conditions had a tendency to undergo arthroplasty. Third, the quality of the trials was barely satisfactory. The unclear or self-contradictory data were noticed in several studies [7, 20, 21, 23]. Then, the publication bias of included studies is inevitable. Finally, the variation in the population included in this study suggests that the conclusions may vary among different ethnicities, owing to the anatomical differences between various racial groups and the disparities in lifestyle habits and socioeconomic conditions across different regions.

Conclusion

For geriatric patients with trochanteric fractures, intramedullary fixation remains the primary surgical choice due to its advantages in reducing operative time, minimizing blood loss, and ensuring better long-term functional recovery. While primary arthroplasty may offer potential benefits in reduced revision rate, accelerated short-term functional recovery, and lower implant-related complications, the statistical robustness of these advantages is limited by their sensitivity to the more conservative HKSJ adjustment. Given this sensitivity, clinicians should interpret the potential benefits of arthroplasty with caution. Therefore, the choice between these two interventions must be highly individualized. Primary arthroplasty should be considered a selective strategic alternative rather than a routine replacement, particularly for patients with ipsilateral hip arthritis, highly unstable fractures, or severe osteoporosis where poor local bone quality may compromise the success of internal fixation.

Abbreviations

TF

Trochanteric

DL

DerSimonian-Laird

HKSJ

Hartung-Knapp-Sidik-Jonkman

RCTs

Randomized controlled trials

PFN

Proximal femoral nail

PFNA

Proximal femoral nail antirotation

InterTAN

Intertrochanteric antegrade nail

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

AVN

Avascular necrosis

VTE

Venous thromboembolism

SD

Standard deviation

WHO

World Health Organization

Authors’ contributions

Zheng LIU designed and conducted this study, and drafted the article. Frankie Ka Li LEUNG reviewed, revised, and approved the final draft for publication. Shuzhuang PENG searched and screened the articles.

Funding

No funding.

Data availability

Data are available upon request to the corresponding author.

Declarations

Ethics approval and consent to participate

Not applicable.

Competing interests

The authors declare no competing interests.