Locked compression plating versus retrograde intramedullary nailing in the treatment of periprosthetic supracondylar knee fractures: a systematic review and meta-analysis

Background Periprosthetic fractures of the distal femur above a total knee arthroplasty (TKA) have traditionally been managed by locking compression plating (LCP). This technique is technically demanding and is associated with high rates of non-union and revision. More recently, retrograde intramedullary nailing (RIMN) has been proposed as an acceptable alternative. This meta-analysis aims to evaluate clinical outcomes in patients with periprosthetic supracondylar femoral fractures who were treated with LCP and RIMN. Methods An up-to-date literature search was carried out using the pre-defined search strategy. All studies that met the inclusion criteria were assessed for methodological quality with the Cochrane’s collaboration tool. Operative time, functional score, time-to-union, non-union rates and revision rates were all considered. Conclusion Ten studies with a total of 531 periprosthetic fractures were included. This meta-analysis has suggested that there is no significant difference in any of the outcome measures assessed. Further, more extensive literature is required on the subject to draw more robust conclusions.

date, level I evidence in the literature. We will specifically compare operative time, knee function, time-to-union, non-union rate and revision rate between the two groups.

Literature search
The systematic reviews and qualitative analysis were performed in accordance with the Preferred Reporting items for Systematic reviews and Meta-analyses (PRIS MA) [21] and the Cochrane Handbook for Systematic Review of Intervention [22]. We have searched Medline, Embase and Cochrane library databases from Jan 2000 to May 2020. All case reports and narrative reviews were excluded. The search was performed on the following areas: 'Periprosthetic fracture' [Mesh] and Distal femur or 'knee arthroplasty' [Mesh] and 'locking plate' [Mesh], or 'intramedullary nail' [Mesh].

Data extraction
Data were extracted separately by two reviewers; the basic demographics for each study were first author, year of publication, study design, sample size, mean age and the outcome parameters measured. Any disagreements were solved through discussion between the two reviewers, and in the case of conflict, a further reviewer is consulted.

Data synthesis and statistical analysis
Mean difference was used to analyse continuous variables whereas risk difference was used to analyse dichotomous variables. Review Manager 5.3 was used for all data synthesis and analysis. P < 0.05 was considered statistically significant with confidence intervals (CI) set to 95%. The 'random effects model' was applied if significant heterogeneity existed between the compared studies. Final results for each parameter were displayed in a forest plot. Chi square test was used to analyse heterogeneity between the studies, and heterogeneity size was formally determined with I 2 (where 0-25% indicates low heterogeneity, 25-75% indicates moderate heterogeneity, and > 75% suggests high heterogeneity).

Methodological quality assessment
All the studies included in this meta-analysis were retrospective cohort studies. All studies were formally assessed for quality according to the Newcastle Ottawa scale [23]. This scale uses a star system ranging from 0 to 9. High quality studies were those who scored more than six stars.

Literature search results
The initial search of the Medline and Embase databases resulted in 1986 articles, and 2 other articles were added to literature search from other sources. Titles and abstracts of the articles were reviewed, and 15 articles were deemed eligible for screening. Out of 15 articles, 5 studies were excluded, as they do not meet the inclusion criteria. Finally, 10 cohort studies were included in the meta-analysis for qualitative and quantitative assessment. The PRISMA flowchart is shown in Fig. 1.

Quality assessment
The non-randomised studies were assessed for quality using the Newcastle Ottawa score with a subjective score out of 9. All the included studies are of good quality with 2 or 3 stars in the selection domain, 1 or 2 in comparability domain and 2 or 3 in outcome/exposure domain. All scores are displayed in Table 1

Characteristics of studies included
The 10 cohort studies included in the study are reviewed in Table 2. All included studies were published between 2008 and 2019. A total of 531 periprosthetic fracture cases were included in the study, of which 320 were treated with LCP and 211 had RIMN. The follow-up time of the involved studies ranged from 8.4 (mean) to 51.6 months (mean).

Outcome 1: operative time
The operative time was reported in 4 studies with a high level of heterogeneity (I 2 = 84%) [9,11,24,26]. The difference between the LCP and RIMN groups in terms of operative time was not statistically significant (Fig. 2).

Outcome 3: time to union
The time to union was reported in 6 studies (n = 117) with a low level of heterogeneity (I 2 = 12%) [9,11,24,25,27,28]. Results from Kyriakidis et al.'s study were not included in the forest plot because the mean and standard deviation could not be derived from the reported data [26]. The difference between the LCP and RIMN groups in terms of time-to-union was not statistically significant (Fig. 4).

Outcome 4: non-union rate
The non-union rate was reported in 8 studies (n = 185) with a moderate level of heterogeneity (I 2 = 42%) [9-12, 19, 24, 26, 28]. The comparative analysis suggests that no significant difference exists between the LCP and RIMN groups in terms of non-union rate (Fig. 5).

Outcome 5: revision rate
Revision rate was reported in 3 studies (n = 98) with a moderate level of heterogeneity (I 2 = 52%) [9,11,12]. The difference between the LCP and RIMN groups in terms of revision rate was not statistically significant (Fig. 6).

Sensitivity analysis
A sensitivity analysis was performed on all comparisons where both a fixed and random effects model was applied; all results remain unchanged after sensitivity analysis.

Discussion
To our knowledge, this is the most up-to-date metaanalysis to compare locking compression plate with retrograde intra-medullary nail fixation for the treatment of periprosthetic supracondylar knee fractures. The results from the meta-analysis demonstrate that clinical outcomes, including operative time, functional score, time-to-union, non-union rates and revision rates did not differ significantly in patients who underwent LCP or RIMN fixation. It is evident that the treatment of peri-prosthetic supracondylar femur fractures remains challenging. Many complications have been reported in the treatment of periprosthetic supracondylar fractures; the literature suggests that this is attributed to the comparatively older populations and lower levels of experience of surgeon involved [11].
The use of IMNs fixation, at large, provides good biomechanical stability with minimal soft tissue disruption with the aim of preserving local fracture biology and therefore healing potential [10,11,29]. In the context of a periprosthetic TKR injury, it must be noted that they can be only used with open-box design prosthesis, few modern PS designs and non-stemmed femoral components. A closed box design poses challenge to make a nail entry usually requiring drilling with a burr [30]. Fracture comminution may also preclude the use of RIMN fixation. In addition, a proximal intramedullary device or hip prostheses may also prevent the use of a RIMN.
The results from this meta-analysis do not support the theoretical advantage of RIMN over LCP. It is possible that the disruption of the periosteal blood supply is less than originally expected; this may be due to more minimally invasive techniques or the more recent LCP designs that preserve periosteal tissue [31]. Furthermore, RIMN fixation involves indirect fracture reduction, which may result in incomplete primary reduction and the potential for secondary loss of reduction [12]. Large et al. has hypothesised that RIMN may not adequately fill the metaphyseal flare and therefore allow toggling of the distal fragment [19].
Two systematic reviews suggest that LCP fixation is associated with a lower overall complication when compared to RIMN in treating PSF [32,33]. The most commonly quoted complications for LCP were hardware failure (5%), malunion (2.5%) and deep infection (2.2%) where for RIMN, malunion (11.5%), hardware failure (6.3%) and peri-implant fracture (1.9%) were often cited. However, more recent literature, which includes a meta-analysis by Shin et al., suggests no significant difference in the complication rates between the two techniques [13,34]. This is in keeping with the findings of our meta-analysis where there was no significant difference in non-union or reoperation rates.
We acknowledge the limited numbers involved and quality of the included data. We also appreciate that all of the papers included are retrospective cohort  studies that lack randomisation and therefore may lower the quality of the data included. The operating surgeon commonly determined the fixation method and choice of implant; this is likely to be influenced by familiarity with each technique, the trend within each institution and the fracture type. We were unable to delineate any reliable data from the included studies to determine if fracture pattern or bone quality altered the choice of fixation method. Furthermore, the meta-analysis includes subjects across many different departments and surgeons of varying ability and experience due to relatively longer learning curve to perform these procedures in periprosthetic fractures.
Fluoroscopy time was not mentioned in any of the included studies. In addition, full weight-bearing status and return to activity were not included in our comparisons as only a few studies mentioned these outcomes. Successful outcomes were measured only by way of knee functional scores and time to union.
There is no clear advantage of either RIMN or LCP in terms of successful post-op rehabilitation. It is noted that most of the studies allowed 'weight bearing as tolerated' for both treatment groups with no record of when a successful post-operative mobility status was achieved for each treatment choice. Specifically, Large et al. followed a protocol of ROM and quadriceps strengthening from day 1, partial weight bearing at 4 weeks and full bearing in 12 weeks [19]. Interestingly, Meneghini et al. mentioned that time to weight bearing is lower in RIMN (9.1 weeks) when compared to LCP (11.7 weeks) [10]. However, the remainder of included studies has suggested no significant difference in terms of successful rehabilitation between the RIMN and LCP groups, by any measure.
Newer methods of treating peri-prosthetic distal femur fracture that have also been successful are nail-plate technique and distal femur replacement (DFR) [35,36]. Fractures with insufficient distal bone stock, unstable implant and incompetent ligaments are not amenable to treatment with LCP or RIMN. The alternative treatment in such cases is DFR.
Moreover, recent literature on DFR suggests good outcomes in elderly patients where it allows early mobilisation and weight bearing, thereby reducing hospital stay and adverse sequela of prolonged immobilisation [37,38]. Further research on DFR is needed to provide a stronger consensus for this treatment method.

Conclusion
The results of this meta-analysis suggest there is no significant difference between LCP and RIMN in terms of operative time, functional score, time to union, nonunion rates and revision rates for the treatment of periprosthetic supracondylar knee fractures. The results of this study indicate either management option remains an acceptable choice for the treatment of periprosthetic supracondylar fractures. The authors recommend that the chosen fixation method is determined by the surgeon's expertise, where stable fixation can be achieved. Undoubtedly, more extensive literature is required to draw more robust conclusions arrangements), or non-financial interest (such as personal or professional relationships, affiliations, knowledge or beliefs) in the subject matter or materials discussed in this manuscript.