Fixed- versus mobile-bearing unicompartmental knee arthroplasty: a meta-analysis

Unicompartmental knee arthroplasty (UKA) can be either a fixed bearing (FB) or a mobile bearing (MB) construct with controversy as to which design is superior. This question is addressed with a systematic review and meta-analysis. A literature search was performed using PubMed, Embase and the Cochrane Library. Studies were reviewed according to the inclusion and exclusion criteria developed in advance. We compared the differences in clinical and radiological outcomes between the FB and MB UKAs. Analyses were performed with the Review Manager and STATA software. A total of 17 studies involving 2612 knees were included. No significant differences were presented between the FB and MB prostheses in clinical and radiological outcomes. However, it was evident that there were differences in the modes and timing of the failures, bearing dislocation led to earlier failures in the MB prosthesis, while the FB prosthesis failed later due to polyethylene wear. There was no evidence of publication bias using the incidence of revisions. There is no significant difference between the FB and MB UKAs; however, there are differences in the modes and timing of failures.


Clinical outcomes.
Nine studies assessed the Knee Society Score (KSS) 1,[9][10][11][12]15,19,20,22 ; six studies assessed the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) score 11,[16][17][18]20,22 ; three studies assessed the Oxford Knee Score (OKS) 5,15,17 ; one study assessed the Italian Orthopedic UKR's Users Group (GIUM) score 9 ; one study assessed the Bristol knee score 5 ; and one study assessed the International Knee Society score 8 . No significant statistical differences were reported between the FB and MB groups on all knee scores measured in these studies. Although statistical heterogeneity was high in the assessment of the Knee Society Score, data from the meta-analysis did not demonstrate significant statistical differences between the FB and MB groups (p = 0.22; I 2 = 87%) (Fig. 2). Two studies 1,20 found similar results between FB and MB groups in respect to pain (p > 0.05), however, Gleeson et al. 5 reported that the total pain score, as assessed using the Bristol Knee Score, was significantly greater in the FB group at 2 years (p = 0.013). Postoperative ROM was assessed in five studies 9,12,15,19,20 , no significant difference was shown between the FB and MB groups (p = 0.19; WMD = − 1.53; 95% CI − 3.80 to 0.74).

Radiological outcomes.
According to the Kennedy and White classification, which considers correct alignment when the mechanical axis is in Zone 2 or C (central) and malalignment in Zone 1 or Zone 3 23     The incidence of radiolucent lines at the bone-tibial implant interface was examined in four studies. Li et al. 20 reported a greater incidence of radiolucent lines at 2-years postoperatively in the FB group (37% vs. 8%; p = 0.02), while others demonstrated an increased incidence of radiolucent lines in the MB group (10.3% vs. 23.5%, p = 0.229; 69% vs. 24%, p = 0.001) 10,18 . Meanwhile, Andrea et al. 22 revealed that no differences were found in RLLs between FM and MB group. Based on the cumulative data, no significant difference was found in the frequency of RLLs between the FB and MB groups (42.0% vs. 19.0%, OR = 0.56; 95% CI 0.10-3.05; p = 0.51).
Revisions, and survivorship. A total of 58 knees from the FB group experienced failure and underwent subsequent revisions, the time of failure ranged from 4-days to 15-years; meanwhile, a total of 43 knees from the MB group underwent subsequent revisions and the time of failure ranged from 6-weeks to 17.7-years. Metaanalysis revealed no significant difference in the incidence of revisions between the two groups (OR = 0.96; 95% CI 0.63-1.46; p = 0.85) (Fig. 3).
In the FB group, complications consisted of progression of arthritis, polyethylene wear, aseptic loosening, and persistent pain. In the MB group, complications consisted of progression of arthritis, aseptic loosening, bearing dislocation, and persistent pain. No significant differences were found in the incidence for progression of arthritis (p = 0.33), aseptic loosening (p = 0.45), persistent pain (p = 0.84), and overall reoperation (p = 0.32) between the two groups. However, the differences in frequency of polyethylene wear (p = 0.02) and bearing dislocation (p = 0.03) were statistically significant between the two groups. In addition, the failure time for bearing dislocations was 0.6-years in the MB group and it was 8.3-years for polyethylene wear in the FB group after surgery. Additional details about complications are shown in Table 2.
Four studies demonstrated that survivorship at 5-years, 10-year, and 20-years were 95.6%, 90.9% and 83% in the FB group and 97%, 92% and 80% in the MB group respectively, with a mean follow-up ranging from 3.6 to 17.2-years (Table 3)    www.nature.com/scientificreports/ Publication bias. The publication bias was assessed using a funnel plot, which demonstrates the relationship between the size of the study sample and the accuracy of the estimated treatment effect. The assessment of publication bias using the incidence of revisions indicated no substantial evidence of bias (p = 0.893) (Fig. 4).

Discussion
UKA comes in two basic prosthesis designs, FB and MB, with disagreement as to which design has superior results 5,8,9 . We have conducted a meta-analysis of all available studies comparing FB to MB UKA prostheses.
Only comparative studies were included in the current meta-analysis. While many studies showed favorable results with fixed or mobile bearing design, they only reported results of a single implant and were therefore excluded from this study 24,25 . The most important finding from our study is that FB and MB prostheses did not differ in terms of: (1) clinical outcomes; (2) radiographic outcomes; (3) revision rates, or (4) survivorship. However, the differences in modes and timing of failures were evident between the two prostheses, bearing dislocation led to earlier failures in the MB prosthesis, while later failures were related to polyethylene wear in the FB prosthesis group.
Differences in clinical outcomes between FB and MB prostheses has been reviewed with each study using different functional measurement indexes. Some studies 15,19 found better results in knee functional scores with the MB group, while Inoue et al. 12 and Ozcan et al. 11 found higher postoperative KSS in the FB group. Interestingly, Peersman et al. 24 demonstrated that there was a tendency for the KSS to deteriorate more rapidly in MB than FB prostheses between 10 and 15 years. However, none of these studies demonstrated significantly better results for one prosthesis over the other. No significant statistical differences were found between the FB and MB groups on all knee scores measured in this study (Fig. 2).
RLL have previously been reported at almost quintuple the rate in FB UKA's compared to MB prostheses 20 , while other studies 10,18 found the opposite results. In this meta-analysis, the incidence of RLL was higher in the MB group compared to the FB group (42.0% MB vs. 19.0% FB) based on limited sample sizes, no significant difference was found (p = 0.75). In addition, recent studies suggest that the presence of RLLs is irrelevant to knee scores and revision rates in mid-and long-term follow-up 10,16,26,27 . Therefore, longer-term follow-up RCTs  www.nature.com/scientificreports/ are needed to find the true relationship between the incidence of radiolucent lines and survivorship between different prostheses. A previous meta-analysis 28 suggested that the major complications which required reoperation following UKAs consisted of aseptic loosening, progression of arthritis and wear of the polyethylene insert. Some in vitro studies 6,29 based on knee simulators reported lower wear rates of FB prostheses compared to MB prostheses, while Manson et al. 30 found that wear was a complication inherent to the design of FB prosthesis caused by the fatigue and sheer stress-related mechanism, secondary to higher surface deformation and delamination in comparison to MB prosthesis. In comparison, MB prostheses reduce overall polyethylene wear by increasing the contact area and congruency while minimizing constraints and maintaining normal knee motion 31 . In MB prostheses, wear of the polyethylene is mainly due to the abrasive and adhesive mechanism 28 . Bearing dislocation is a unique complication in MB prostheses. Any under correction of the deformity may play a role in the medial compartment stresses contributing to bearing dislocation. Considering this, surgeons may be inclined to choose a tight knee and risk slight overcorrection into valgus 32 . However, overcorrection to valgus can result in greater contact stress and significant load on the lateral compartment, thus accelerating the progression of arthritis. In contrast, FB prostheses could offload the lateral compartment and thereby slow the progression of arthritis 28 . In the current study, a higher incidence of polyethylene wear was clearly demonstrated in the FB group, and bearing dislocation was only present in MB group (Table 2).
Survivorship differed between prosthesis design in most studies. Whittaker et al. 16 reported that the 5-year survival, with revision to TKA as the end point, was 88% for the MB implants and 96% for the FB implants. Reporting on 38 MB and 68 FB knees with a mean follow-up of 14.2-years and 11.5-years, Neufeld et al. 18 demonstrated that the 10-year survivorship of the MB implants was 82.9% and the FB implants was 90.9%. In addition, a comparative study 10 with a minimum 15-year follow-up reported that 20-year survivorship was 83% in the FB group and 80% in the MB group. Similar to our study, differences of survivorship between the two prostheses in these studies were not significant (Table 3). To assess the true differences in the lifespan of UKA implants, future higher quality RCTs are needed.
Several limitations should be presented in our study. First, we included 14 non-randomized controlled trials (nRCTs) of low level of evidence, which increases the risk of bias and other cofounding variables. Second, publication bias might affect the outcomes, as we did not search for unpublished studies. However, the funnel plot demonstrated minimal evidence of publication bias (Fig. 4). Finally, due to the high proportion of missing data and the observational nature of the included studies, the results should be interpreted with caution. In our study, no significant differences in knee scores were found between the two prostheses, which may be due to the limited studies and incomplete data. Therefore, future larger scale and long-term follow-up RCTs with complete data directly comparing the knee functional scores between FB and MB UKAs are needed. Theoretically, different complications mean different revision surgeries, such as replacing polyethene insert, UKA, Bi-KA or TKA, while most of present studies don't mention it. Lacking the specific surgery of revision is another disadvantage of our study, we will analyze the tendency of the type of revision surgeries in these two designs of UKA in future study.
In conclusion, both Fixed Bearing and Mobile Bearing prostheses provide excellent knee scores, radiological outcomes, and survivorship in Unicompartmental Knee Arthroplasties. The superiority of one prosthesis over the other could not be demonstrated and each have their merits and indications. Patient selection and proper technique are required to ensure the Mobile Bearing prosthesis can survive beyond the short-term dislocation complication to enable excellent implant longevity.

Materials and methods
This meta-analysis was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement 33 . Relevant studies published prior to April 21, 2020 were identified by searching PubMed, Embase and the Cochrane Library with no language restrictions. The following search terms, including their variations, were used: "Arthroplasty, Replacement, Knee", "Unicompartmental", "Unicondylar", "Fixed", "Mobile", "Bearing", and Boolean operators (AND, OR) for various combinations. After the initial electronic search, relevant articles and their bibliographies were searched manually.
To minimize any possible selection bias, all studies that satisfied the search strategy were reviewed and included if they met the following criteria: (1) studies which compare FB and MB knee prostheses directly; (2) patients with isolated medial compartment arthritis of the knee; and (3) studies which report clinical or radiographic outcomes, revisions, or survivorship. The exclusion criteria included: (1) review articles, letter or comment; (2) studies which only describe either FB or MB knee prostheses; (3) patients with lateral compartment arthritis or patellofemoral arthritis of the knee; (4) also excluded if the results of comparison were not reported or the data could not be extracted from the published results. Two reviewers (WCZ, JPW) independently assessed each of the studies for eligibility for inclusion. All disagreements were resolved through discussion to reach a final consensus.
All information regarding relevant results were recorded. Data of participants included: number of knees/ patients, demographic data (age, sex, weight, height, etc.), duration of the follow-up, and type of prosthesis. The outcome measures included knee scores, range of motion (ROM), femoral-tibial angle (FTA), hip-knee-ankle angle (HKA), incidence of radiolucent lines (RLLs), incidence and timing of revisions, and survivorship. The levels of evidence were assessed using a published rating system 34 .
Heterogeneity was determined by estimating the proportion of between-study inconsistencies by examining actual differences between studies identified in the data extraction tables. Heterogeneity is expressed as p and I 2 ; if p > 0.10 and I 2 < 50%, a fixed effects model was adopted; otherwise, a random effects model was chosen. Continuous data was assessed using weighted mean difference (WMD) and associated 95% CI (confidence intervals),

Scientific Reports
| (2020) 10:19075 | https://doi.org/10.1038/s41598-020-76124-z www.nature.com/scientificreports/ adopting the Mantel-Haenszel method 35 ; study-specific OR (Odds Ratio) and associated 95% CI were used to determine the value of dichotomous data. Forest plots were used to graphically present the results of individual studies and the respective pooled estimate of effect size. Statistical significance was regarded as a p < 0.05. Publication bias was assessed using a funnel plot of the outcome measure recorded in the largest number of clinical trials 36 . We assessed funnel plot asymmetry using Begg and Egger tests and defined significant publication bias as a p < 0.1. Review Manager (RevMan, version 5.3) for Windows and the Cochrane collaboration were used to perform all statistical analyses. Publication bias was assessed using STATA (version 12.0).

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.