Extramedullary versus intramedullary fixation of unstable trochanteric femoral fractures (AO type 31-A2): a systematic review and meta-analysis

Objective The aim of this systematic review was to compare extramedullary fixation and intramedullary fixation for AO type 31-A2 trochanteric fractures in the elderly, with regard to functional outcomes, complications, surgical outcomes, and costs. Methods Embase, Medline, Web of Science, Cochrane Central Register of Controlled Trials, and Google Scholar were searched for randomized controlled trials (RCTs) and observational studies. Effect estimates were pooled across studies using random effects models. Results are presented as weighted risk ratio (RR) or weighted mean difference (MD) with corresponding 95% confidence interval (95% CI). Results Fourteen RCTs (2039 patients) and 13 observational studies (22,123 patients) were included. Statistically superior results in favor of intramedullary fixation were found for Harris Hip Score (MD 4.09, 95% CI 0.91–7.26, p = 0.04), Parker mobility score (MD − 0.67 95% CI − 1.2 to − 0.17, p = 0.009), lower extremity measure (MD − 4.07 95% CI − 7.4 to − 0.8, p = 0.02), time to full weight bearing (MD 1.14 weeks CI 0.92–1.35, p < 0.001), superficial infection (RR 2.06, 95% CI 1.18–3.58, p = 0.01), nonunion (RR 3.67, 95% CI 1.03–13.10, p = 0.05), fixation failure (RR 2.26, 95% CI 1.16–4.44, p = 0.02), leg shortening (MD 2.23 mm, 95% CI 0.81–3.65, p = 0.002), time to radiological bone healing (MD 2.19 months, 95% CI 0.56–3.83, p = 0.009), surgery duration (MD 11.63 min, 95% CI 2.63–20.62, p = 0.01), operative blood loss (MD 134.5 mL, 95% CI 51–218, p = 0.002), and tip-apex distance > 25 mm (RR 1.73, 95% CI 1.10–2.74, p = 0.02). No comparable cost/costs-effectiveness data were available. Conclusion Current literature shows that several functional outcomes, complications, and surgical outcomes were statistically in favor of intramedullary fixation when compared with extramedullary fixation of AO/OTA 31-A2 fractures. However, as several of the differences found appear not to be clinically relevant and for many outcomes data remains sparse or heterogeneous, complete superiority of IM fixation for AO type 31-A2 fractures remains to be confirmed in a detailed cost-effectiveness analysis. Supplementary Information The online version contains supplementary material available at 10.1007/s00402-023-05138-9.


Introduction
Proximal femoral fractures are a major health problem among the elderly worldwide and the incidence rate is rising due to progressive aging.It is expected that the total number of patients with a proximal femoral fracture will rise from 1.66 million in 1990 to 6.26 million worldwide by 2050 [1][2][3].These fractures are not exclusively a problem for public health systems, but they also form a burden for society, due to high disability, costs, and morbidity [4][5][6][7][8].
Current surgical guidelines such as the United Kingdom's National Institute for Health and Care Excellence (NICE) guideline and the Dutch Guideline for treatment of proximal femoral fractures advise the use of extramedullary fixation for both AO type 31-A1 and 31-A2 fractures, mainly due to better cost-effectiveness [13,14].Extramedullary fixation is more cost effective than intramedullary fixation in the majority of cases, largely because of lower implant costs [15].The guidelines do, however, see both fixation strategies as viable treatment options for type 31-A2 fractures and underline the absence of conclusive evidence of superiority for either device.Older literature, mainly utilizing the now obsolete first generation of cephalomedullary nails, discouraged intramedullary devices due to a higher failure and reoperation rate [16].
While clear historic evidence of superiority is missing, in recent years the use of intramedullary fixation is rising to up to 90% in the U.S. [17,18].While more recent studies report improving outcomes for intramedullary fixation, the optimal treatment strategy remains a topic of debate and the increasing trend of intramedullary fixation may be caused by other factors than clinical data alone [17][18][19][20][21].
Therefore, the objective of this study was to compare recent literature assessing the differences in functional outcomes, complications, surgical outcomes, and costs/costeffectiveness between extramedullary and intramedullary fixation using currently available implants in elderly patients with AO type 31-A2 fractures.

Methods
This systematic review and meta-analysis was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines [22].A protocol was developed prior to conducting the current study.This study did not require approval from the local medical research ethics committee.

Search and eligibility criteria
Embase, PubMed/Medline, Web of Science, Cochrane Central Register of Controlled Trials, and Google Scholar were searched on 22 March 2021 and updated on 26 September 2022, including terms related to 'trochanteric fractures,' 'intramedullary treatment,' and 'extramedullary treatment.'An overview of the complete search used is included in Online Resource 1.After deduplication, two reviewers (LHTN and ACP) independently screened all articles for eligibility by title and abstract.Thereafter, independent fulltext analysis for eligibility was conducted.Disagreements were resolved by consensus.
Studies were included when they presented data (a) published after 1990 of (b) acute (c) AO 31-A2 trochanteric fractures, (d) comparing intramedullary (IM) and extramedullary (EM) fixation (e) in patients aged 50 years and older, (f) using currently available devices.Studies were excluded when they (a) presented no original data, (b) did not mention relevant outcomes, (c) were biomechanical, in vitro or cadaveric studies, (d) pathological fractures, (e) bilateral fractures, (f) peri-implant fractures, (g) were case reports, and (h) did not make distinction between types of fracture or treatment.

Quality assessment
The included studies were assessed by two reviewers (LHTN and ACP) independently, using the Cochrane Risk of Bias tool, version 2 (RoB2) for the randomized controlled trials (RCTs) and the methodological index for nonrandomized studies (MINORS) for the observational studies [23,24].RoB2 provides a risk of bias judgement resulting in low, some concerns, or high risk of bias.The MINORS provides a score with a maximum of 16 points for noncomparative studies and 24 points for comparative studies.A higher score indicates higher quality.

Data collection
Data from all included studies were independently extracted by two reviewers (LHTN and ACP) according to a predefined data sheet.The baseline characteristics collected for each study contained the following: first author, year of publication, setting (country), inclusion period, study design, type of fixation device used, follow-up period, and mean age, gender distribution, and total number of patients with AO-OTA 31-A2 fracture in the study population.See Online Resource 2 for specific devices used per included study.
Data were collected on the following outcome measures: Functional outcomes: Harris hip score (HHS), Parker mobility score, lower extremity measure (a modification of the Toronto extremity salvage score (TESS) [25]), recovery to pre-operative walking ability, and time to full weight bearing; Complications: reoperation, superficial wound infection, deep wound infection, nonunion, cut-out/ protrusion (varus collapse of the neck-shaft angle leading to extrusion of the screw), peri-prosthetic fracture, conversion to prosthesis, implant/fixation failure (mechanical loosening or fracturing of the implant), heterotopic ossification, leg shortening, screw migration, femur shaft fracture, and mortality; Surgical outcomes/operation characteristics: mean time to bone healing, radiologic quality of reduction, surgery duration, hospital stay, blood loss, blood transfusion (units per patient), blood transfusion, fluoroscopy time, tip-apex distance (TAD), TAD > 25 mm, femoral neck shortening, and neck-shaft angle (NSA); and Costs/cost-effectiveness.All outcome measures were included and analyzed as defined by the original article.
Authors of studies with missing data (standard deviations) were contacted by email once [26][27][28][29][30].When no data were received, standard deviations were imputed by weighted mean SD of the other included studies, if at least two other studies were available.

Statistical analysis
RevMan version 5.4 was used to analyze data.Binary outcomes were pooled using the Cochran-Mantel-Haenszel statistic and presented as risk ratio (RR) and continuous outcomes were pooled using the inverse variance weighting method and presented as mean difference (MD), both with corresponding 95% confidence intervals (95% CI).All analyses were done separately for each of the different study designs (RCTs and observational studies) and for overall effect and were presented in Forest plots.Random effects models were used in all comparisons because of a high likelihood of heterogeneity between studies due to inclusion of both RCT and observational studies and comparison of multiple types of devices in different countries and clinical settings.Assessment of heterogeneity between studies was done by using the Cochrane Q-test and was quantified using the I 2 statistic.A p-value < 0.05 was considered statistically significant.According to the Cochrane Handbook for Systematic Reviews of Interventions, the level of heterogeneity was found to be unimportant when the I 2 value is between 0 and 40%, moderate between 30 and 60%, substantial between 50 and 90%, and considerable between 75 and 100% [31].Funnel plots were visually inspected to assess publication bias (Online Resource 3).

Study characteristics
The included studies were published from 2006 to 2022 and provided data of 24,232 patients with an AO-OTA 31-A2 trochanteric fracture, of whom 2039 patients were included in RCTs and 22,123 in observational studies (Table 1).Of all included patients, 11,932 were treated with an extramedullary (EM) device and 12,300 with an intramedullary (IM) device.Table S2 in Online resource 4 (OR4) provides an overview of the outcome measures reported in the individual studies.The mean follow-up time was 12 months.

Lower extremity measure (LEM)
Lower extremity measure at one-year follow-up was reported in two RCTs (OR4, Figure S2) [28,33], in which 135 patients were treated with EM devices and 125 were treated with IM devices.The mean LEM was 64.5 for EM devices and 67.0 for IM devices.The LEM significantly favored IM fixation (MD − 4.07, 95% CI − 7.39 to − 0.75, p = 0.02, I 2 = 0%).

Time to full weight bearing (weeks)
Time to full weight bearing was reported in one RCT and one observational study, with 94 patients treated by extramedullary fixation and 67 by intramedullary fixation (OR4, Figure S4) [42,46].Mean time to full weight bearing was 2.5 weeks for EM fixation and 1.4 weeks for IM

Pain scores
Different measurement scales were used for the pain scores, namely, the VAS [27,30,50] and HHS pain score [27].None of the studies reported a significant difference.Meta-analysis of the VAS score was not possible, because two studies out of three did not report SDs, and thus, no imputed SD could be calculated [30,50].

Other functional scores
Multiple other measurement scales are used for the functional outcomes and quality of life, in addition to those previously mentioned: Merle d'Aubigné and Postel score [27,47], EQ-5D [32], Barthel index [32], functional independence measure (FIM) [28], Coval score [50], and HHS [27].None of the included studies reported a significant difference between fixation groups.Due to the high diversity in included functional and quality of life scores and/or missing SDs, no meta-analysis was performed.An overview of functional outcomes is shown in Table 2.

Leg shortening (mm)
Leg shortening was reported in four studies: two RCTs [34,36] and two observational studies (OR4, Figure S5) [47,50], in which 194 patients were treated with an EM device and 194 with an IM device.The mean leg shortening length was 4.3 mm for EM devices and 2.3 mm for devices.Leg shortening was significantly lower for IM fixation (MD 2.23 mm, 95% CI 0.81-3.65,p = 0.002, I 2 = 65%).
An overview of all complications is given in Table 3.

Poor radiological quality of reduction
Poor radiological quality of reduction using the classification by was reported in three studies: one RCT [41] and two observational studies (OR4, Figure S8) [46,48].Poor quality of reduction was reported in 23 out of 255 (9.0%) patients treated with an EM device and in 7 out of 253 (2.8%) patients treated with an IM.Tao et al. [41] reported zero cases.There was no significant difference between fixation groups (RR 2.52, 95% CI 0.71-8.93,p = 0.15, I 2 = 56%).

Femoral neck shortening (mm)
Femoral neck shortening was reported in two studies: one RCT [28] and one observational study (OR4, Figure S16) [47], with 134 patients in the EM fixation group and 141 patients in the IM fixation group.The mean shortening distance was 8.0 mm for EM fixation and 3.1 mm for IM fixation.There was no significant difference between fixation groups (MD 4.00 mm, 95% CI − 3.84-11.84,p = 0.32, I 2 = 96%).

Costs-and cost-effectiveness
No comparable data on costs-or cost-effectiveness could be extracted from any of the included articles.
An overview of all surgical outcomes is given in Table 4.

Discussion
This systematic review and meta-analysis compared functional outcomes, complications, and surgical outcomes for EM versus IM fixation in elderly patients with an AO type 31-A2 fracture.Statistically superior results in favor of IM fixation were found for several outcomes including Harris Hip Score, Parker mobility score, lower extremity measure, time to full weight bearing, superficial infection, nonunion, fixation failure, leg shortening, time to bone healing, and surgery duration.
The most recent Cochrane review, by Lewis et al. [53], on RCTs and 'RCT-like' cohort studies published up to July 2020 compared EM and IM fixation for a combination of AO A1, A2, and A3 fractures.In correspondence with the current review it found an increased risk of several complications including nonunion and implant failure in patients treated with an EM device.However, it found no difference in functional outcomes and found that IM devices were associated with an increased intra-and postoperative periimplant fracture and shorter HLOS.Differences found in comparison with this meta-analysis can be partly explained by its combination of A1, A2, and A3 fractures and inclusion of older studies (before 2005) with a relatively higher rate of complications.The Cochrane review only performed a stratified analysis for stable versus unstable fractures for reoperation (no significant difference) and did not assess surgical outcomes and operation characteristics.
Another recent meta-analysis by Wessels et al. [54] comparing IM nailing with sliding hip screws (SHS) for all combined AO 31-A fractures reported no significant differences between both fixation options for the 31-A2 subgroup in combined major complication rate, infections (superficial and deep infections combined), nonunion, and mortality.Wessels et al. included several articles also included in this analysis, but chose to combine major complications, while the current study explores a wider range of adverse events, biomechanical outcomes, and patient-reported outcomes.Contrary to the current study, they did not find a significant difference for nonunion rate.This review includes nonunion rate from two more recent RCT's not included by Wessels et al., shifting the effect toward IM fixation.A meta-analysis of RCT's published by Zhu et al. [55], compared IM nails with SHS for AO 31-A2 fractures.They also showed statistical superiority of IM nails for intraoperative blood loss, leg shortening, superficial infections, length of hospital stay, days to mobilization, and the Parker mobility score.These results are similar to those found in this meta-analysis.
Results found by older reviews and meta-analyses, demonstrating inferiority of IM fixation based on older studies, featuring mainly first and second generation IM implants, should be considered obsolete nowadays [16,56].A change in paradigm that was already predicted by Bhandari et al. [57].Future meta-analyses should only incorporate implants that are still clinically used.

Interpretation of results
Several points should be considered when interpreting the differences found between EM and IM fixation.In contrast to the meta-analysis of several of the major complications (e.g., to prosthesis, mortality and infections) and their sequelae (e.g., reoperation, conversion to prosthesis), a relative lack of data on functional outcomes in both the number of (prospective) studies and the number of included patients was observed.This study showed a significant mean difference of 4.1 points on the Harris Hip Score, on a scale of 0-100 points, in favor of intramedullary fixation.This number should be considered in the light of the minimally clinical important difference, which is established at 15 points for the HHS [52].While statistically significant, this difference is not expected to be clinically relevant.Similar arguments could be made for the small differences found in the Parker mobility score (MD − 0.67, on 0-9 scale) and LEM (MD − 4.07 on 0-100 scale), although no minimal important change/difference values have been published for these measures.The difference to full weight bearing (MD 1.04 weeks in favor of IM fixation) could in theory be clinically relevant.However, this outcome was reported in only two studies which used radiological union as a starting point for full weight bearing, making this outcome similar to the outcome radiological union mentioned in other studies (favoring IM fixation).Relevant would be a difference in patient-reported time to full weight bearing without restrictions given by the treating surgeon that exceeds the MICD.With regard to complications and operation characteristics several considerations should be taken into account.Most fracture-and implant-related complications are rare and occur at rates under 5% or even 2%.In the two complications with the largest number of included patients, reoperation, and conversion to prosthesis, no significant differences were found.Although reoperation or conversion can be expected to be a result of other complications such as nonunion, infection, or fixation failure, statistically significant differences were found for these three complications.These effects might be exaggerated due to several (randomized) studies with 0 cases in both study arms, that could not be included in a pooled effect measure.The mean follow-up was 12 months; however, the rate of biomechanical complications increases with a longer follow-up time.Therefore, comparing studies with varying follow-up durations might have influenced the meta-analysis.Nearly all operation characteristics or surgical outcomes suffer from very high heterogeneity and variables such as surgery time or measurement of blood loss are often poorly defined by studies.In combination with their relatively low patient numbers and the possible influence of retrospective data, these results should be interpreted with care.
While this review demonstrates that IM fixation for AO 31-A2 trochanteric fractures is no longer inferior to EM fixation, superiority remains questionable.Many differences are below clinically relevant thresholds, hold low quality of evidence, or analyses are underpowered to adequately compare functional outcomes or rare complications.Where clear superiority is missing, costs-effectiveness should also be considered when selecting an EM or IM fixation approach.Currently, EM fixation is considered the most cost-effective approach by the few studies that were conducted on this topic [15,26].This conclusion is also emphasized by the Dutch and U.K. guidelines for treatment of proximal femoral fractures [13,14].This is mainly due to the generally higher IM device cost.However, extensive cost-effectiveness analyses, including broad health care and rehabilitation costs, quality of life, and functional outcomes do not exist at the time of writing.Such an analysis would provide critical data for updating current guidelines.Because of the shifting trend of effectiveness toward IM fixation, due to newer implants and operative strategies, these conclusions may have become outdated.Therefore, this study reaffirms the need for properly powered, largescale comparisons of both fixation strategies, including costs and costs-effectivity.Additionally, due to the relative rarity of major complications and relative lack of functional or patient-reported data, future research should primarily focus on functional outcomes and quality of life, instead of primarily focusing on number of complications.

Limitations
This is the most extensive systematic review and meta-analysis that is restricted to the treatment of only AO 31-A2 proximal femoral fractures, including both observational and RCT data on currently available implants, to date.It includes a substantially larger population than all previous meta-analyses.However, the meta-analysis does have several limitations: There was high heterogeneity for many analyzed variables due to differences in studied implants, study designs, international differences, and duration of studies.In addition to this, many studies can be considered underpowered for often rare outcomes.Large (observational) studies can strongly impact the overall effect.This, together with a high heterogeneity could have resulted in missing or overstated differences between IM and EM fixation.Multiple functional scores and pain measurements were used by the included studies and SDs were often not reported.This made calculation of pooled effects impossible for several functional outcome scores.In addition, not all studies reporting functional scores also included baseline scores, making it difficult to observe potential selection bias.There are sparse data on many of the included variables, as many studies included all three of the subtypes of trochanteric fractures and were only adequately powered for the complete population.While the AO type A2 fracture subgroups could be extracted and included in the meta-analysis, this could result in a lower generalizability and lower quality of evidence for these limited results.Lastly, data for all specific outcomes were included as described by the original articles.Most studies did not provide extensive definitions or definitions of outcomes differed slightly between studies.While articles were reviewed extensively and only comparable data were included, this might have introduced bias, especially in the case of retrospective data.

Conclusion
This review and meta-analysis showed that several functional outcomes, complications, and surgical outcomes were in favor of intramedullary fixation when compared with extramedullary fixation of AO type 31-A2 fractures.Results indicate significantly higher Harris hip score, Parker mobility score, lower extremity measure, and recovery to pre-operative walking ability.No difference was found in reoperation rate or conversion to prosthesis, but studies show a decrease in superficial infections, nonunion, fixation failure, leg shortening, surgery duration, operative blood loss, and increased tip-apex distance, all in favor of intramedullary fixation.Previous meta-analyses describing inferiority of IM fixation in AO type 31-A2 fractures should now be considered obsolete.However, a true superiority of IM fixation for AO type 31-A2 fractures remains questionable as several differences in functional outcomes appear not clinically relevant, data on many outcomes remains sparse or heterogeneous, and a detailed cost(-effectiveness) evaluation of modern IM nails is missing in the literature.As major complications are rare and there is a relative lack of functional, patientreported, and cost data, future research should primarily focus on functional outcomes, quality of life, and costseffectiveness, instead of primarily focusing on the number of complications.

Fig.Fig. 14
Fig. Forest plot of mean to bone healing (weeks) after extramedullary versus intramedullary fixation of AO 31-A2 fractures

Table 1 (
6 months by Zehir et al. and up to 1 year by Xu et al.Recovery to pre-operative walking ability was achieved in 87 out of 145 (60.0%) patients treated with an EM device and in 100 out of 136 (73.5%) patients treated with an IM device.

Table 2
Overview of functional outcomes Statistically significant differences (p < 0.05) are denoted as bold OM, Outcome measurement; EMF, Extramedullary fixation; IMF, Intramedullary fixation; RCT , Randomized controlled trial; OS, Observational studies; RR, Relative risk; MD, Mean difference; N.A.,Not available a Subgroup mean b Cases reported in subgroup Fig. 4 Forest plots of reoperations after extramedullary versus intramedullary fixation of AO 31-A2 fractures.M-H, Mantel-Haenszel; RCT, randomized controlled trial; SD, standard deviation Fig. 5 Forest plots of deep infections after extramedullary versus intramedullary fixation of AO 31- heterotopic ossification occurred in 23 out of 130 (17.7%) patients treated with an EM device and 49 out of 137 (35.8%) patients treated with an IM device (RR 0.51, 95% CI 0.35-0.76,p = 0.008).In respectively, 12 and 35 cases this was Brooker stage 1 heterotopic ossification.No pooled risk ratio for heterotopic ossification could be calculated, because both Garg et al. and Suh et al. [42, 50] reported zero cases.

Table 3
Overview of complications Statistically significant differences (p < 0.05) are denoted as bold OM, Outcome measurement; EMF, Extramedullary fixation; IMF, Intramedullary fixation; TAD, Tip-apex distance; RCT , Randomized controlled trial; OS, Observational studies; RR, Relative risk; MD, Mean difference; N.A.,Not available a Subgroup mean b Cases reported in subgroup