Comparable incidence of periprosthetic tibial fractures in cementless and cemented unicompartmental knee arthroplasty: a systematic review and meta-analysis

Purpose (I) To determine the incidence of periprosthetic tibial fractures in cemented and cementless unicompartmental knee arthroplasty (UKA) and (II) to summarize the existing evidence on characteristics and risk factors of periprosthetic fractures in UKA. Methods Pubmed, Cochrane and Embase databases were comprehensively searched. Any clinical, laboratory or case report study describing information on proportion, characteristics or risk factors of periprosthetic tibial fractures in UKA was included. Proportion meta-analysis was performed to estimate the incidence of fractures only using data from clinical studies. Information on characteristics and risk factors was evaluated and summarized. Results A total of 81 studies were considered to be eligible for inclusion. Based on 41 clinical studies, incidences of fractures were 1.24% (95%CI 0.64–2.41) for cementless and 1.58% (95%CI 1.06–2.36) for cemented UKAs (9451 UKAs). The majority of fractures in the current literature occurred during surgery or presented within 3 months postoperatively (91 of 127; 72%) and were non-traumatic (95 of 113; 84%). Six different fracture types were observed in 21 available radiographs. Laboratory studies revealed that an excessive interference fit (press fit), excessive tibial bone resection, a sagittal cut too deep posteriorly and low bone mineral density (BMD) reduce the force required for a periprosthetic tibial fracture to occur. Clinical studies showed that periprosthetic tibial fractures were associated with increased body mass index and postoperative alignment angles, advanced age, decreased BMD, female gender, and a very overhanging medial tibial condyle. Conclusion Comparable low incidences of periprosthetic tibial fractures in cementless and cemented UKA can be achieved. However, surgeons should be aware that an excessive interference fit in cementless UKAs in combination with an impaction technique may introduce an additional risk, and could therefore be less forgiving to surgical errors and patients who are at higher risk of periprosthetic tibial fractures. Level of evidence V.


Introduction
Unicompartmental knee arthroplasty (UKA) is a well-established treatment for patients with isolated compartmental knee arthritis. Advantages of UKA over total knee arthroplasty (TKA) include reduced morbidity and mortality, preservation of normal knee kinematics and faster recovery [35,49,59]. However, national registry data have shown lower revision rates after TKA in comparison to UKA [49,66]. Reasons for UKA revision include aseptic loosening, malalignment, progression of osteoarthritis, instability, infection and periprosthetic fractures [49,66].
Periprosthetic fractures represent a complex complication with serious consequences in UKA and have been associated with increased mortality and morbidity [26]. The periprosthetic fractures in UKA are most commonly reported on the tibial side (approximately 87%) [66]. Although these periprosthetic tibial fractures are relatively rare compared to other complications in UKA, recent registry-based studies have shown an increased rate of periprosthetic fractures in cementless UKAs compared to cemented UKAs [49,63]. Since the interest of cementless fixation for UKAs is expected to increase, the rate of periprosthetic fractures may increase as well [49,63]. However, registry-based studies may not provide reliable information about all fractures, as some periprosthetic fractures are internally fixed and the components are not revised or are treated conservatively. Another common limitation of registry-based studies is that tibial and femoral periprosthetic fractures are not reported separately. This stresses the need for a thorough evaluation of the incidence of periprosthetic tibial fractures in cemented and cementless UKAs using clinical studies. Furthermore, there is a lack of studies providing an overview of the available evidence on characteristics and risk factors of periprosthetic tibial fractures in UKA to gain a better understanding and awareness. Therefore, the primary study aim was to estimate the incidence of periprosthetic tibial fractures in cemented and cementless UKA using clinical studies. Secondarily, relevant studies were systematically reviewed to summarize characteristics and risk factors of periprosthetic tibial fractures in UKA. Based on earlier large case series of both cemented and cementless UKAs reporting no non-traumatic periprosthetic tibial fractures [62,68], it was hypothesized that comparable low incidences of periprosthetic tibial fractures can be achieved as long as surgeons are aware of factors that could increase the risk.
After duplicates were excluded, titles and abstracts were screened by two independent reviewers (*** & ***). Subsequently, full texts of the potential studies were carefully assessed by the two reviewers to confirm study eligibility. To be eligible, the study needed to contain information on proportion, characteristics and/or risk factors of periprosthetic tibial fractures in UKA. Clinical studies with information on fixation type and proportion were used to estimate incidences. For information regarding characteristics and/or risk factors, any study design was considered eligible, including case reports and laboratory studies. Although case reports and laboratory studies constitute low-level evidence, a systematic review of such studies can provide a better understanding and awareness of tibial plateau fractures in UKA. Studies were excluded if they reported on bicompartmental UKAs, used the same database, were reviews, registry-based studies, commentaries or abstracts. References of the included studies were checked for any missing studies. Any disagreements on study eligibility were resolved through consultation of the third reviewer (***).

Data collection and analysis
Data extraction was entered in predefined spreadsheets by two independent reviewers. First author, publication year and study design were reported for each study. Total number of UKAs, number of fractures and fixation type were collected only from clinical studies for the analysis of incidence. To identify potential risk factors, characteristics of patients with and without periprosthetic tibial fractures were collected from clinical studies and compared. For example, body mass index (BMI) of patients with and without fractures were compared. Both clinical studies and case reports were used to evaluate characteristics of periprosthetic tibial fractures (time of fracture in relation to UKA, fracture mechanism [traumatic or nontraumatic], fracture type, type of treatment). Time of fracture in relation to UKA was classified into the following time-points: during surgery, within 3 months postoperatively, between 4 and 12 months postoperatively and after 1 year postoperatively. Schematic drawings were used to present the fracture types found on available radiographs. Causes of fractures considered by authors from each study 1 3 were evaluated and summarized. Finally, conclusions of laboratory studies were presented.

Methodological quality assessment
Different tools for methodological quality assessment were used depending on study design.
The National Institutes of Health (NIH) checklist was used for all clinical studies [67],The Case Report (CARE) checklist was used for case reports [29],and the Quality Appraisal for Cadaveric Studies (QUACS) checklist [90] was used for cadaveric studies. A score was provided for each article (poor, fair or good). The assessment was performed by two independent reviewers (*** & ***) and disagreements of the level of study quality were resolved through consultation of the third reviewer (***).

Statistical analyses
Incidence of periprosthetic tibial fractures was calculated as the number of fractures divided by the total number of UKAs from each clinical study. These data were combined via proportion meta-analysis [94]. This is a tool to calculate an overall proportion from studies reporting a single proportion. Combined proportions were determined for cementless and cemented UKAs. A subgroup analysis was performed for cementless and cemented Oxford Partial Knee Implants. Effect sizes and 95% Confidence Intervals (CI) were determined using a random-effects model by the back-transformation of the weighted mean of the logit-transformed proportions with Dersimonian weights. Characteristics between patients with and without periprosthetic tibial fractures were compared using the chi-square test for categorical variables and independent t test for continuous variables. All analyses were performed with R version 4.0.0 (R Foundation for Statistical Computing, Vienna, Austria).
Based on information from 167 fractures, 85 (51%) periprosthetic tibial fractures were treated with TKA (with metal augmentation and/or tibial stem extension), 38 (23%) with ORIF, and 44 (26%) with conservative treatment. Authors reported that eight fractures, initially treated conservatively, underwent a subsequent TKA; six fractures, initially treated with ORIF, underwent a subsequent TKA; one fracture, initially treated conservatively, underwent ORIF; and one fracture, initially treated conservatively, underwent ORIF and eventually needed a TKA.

Discussion
The main study finding was that the incidence of periprosthetic tibial fractures in cemented and cementless UKA was comparable. However, experimental evidence showed that excessive interference fit (press fit), excessive resection depth, making the sagittal cut too deep posteriorly, and low BMD reduces the load required for a periprosthetic tibial fracture to occur. Furthermore, clinical studies revealed that patients with fractures were more often female, of older age, exhibited higher BMI and postoperative alignment angles, had lower BMD and had very overhanging medial tibial condyles.
Contrarily to the main finding of this study, two recent registry-based studies showed higher rates of periprosthetic fractures in cementless compared to cemented Oxford Partial Knee implants [49,63], raising some concerns regarding a keel design in cementless techniques. Campi et al. demonstrated that fixation of the cementless mobile-bearing Oxford UKA is ensured by the interference fit [18]. However, an excessive interference increases the assembly load required to push-in the component potentially introducing a splitting force during impaction (type V fracture) [16]. As this interference fit, combined with an impaction technique, could introduce an additional risk factor for fractures, the cementless Oxford Partial Knee implant may be less forgiving to surgical errors and patients who are at higher risk of periprosthetic tibial fractures.
Several surgical errors have been proposed by authors to cause periprosthetic tibial fractures in UKA Table (3). Only a few authors have supported their conclusion with experimental evidence. Laboratory studies showed a vertical saw cut too distal in the posterior tibial cortex and excessive tibial bone resection reduces the load required for a fracture to occur [20,21,39,71]. Additionally, laboratory studies on the role of tibial component alignment suggested valgus alignment and an excessive posterior slope should be avoided [41,42,76]. Other authors based their conclusions on radiographic or intraoperative findings. Radiographs revealed that fracture lines went through multiple pinholes of the extramedullary tibial guide (type II fracture) [15]. One author reported that a fracture occurred due to breaching the posterior cortex while using a tibial gouge for keel preparation in Oxford Partial Knee implants (type V fracture) [82].   angles and a very overhanging medial tibial condyle could contribute to the occurrence of periprosthetic tibial fractures in UKA. The relationship with greater age and osteoporosis is not surprising as fractures have been directly linked to these factors [12]. The higher proportion of periprosthetic tibial fractures in females compared to males may be due to higher rate of osteoporosis [12], the smaller average size of tibial plateaus [97] and the higher likelihood of having very overhanging medial tibial condyles [38,96] in females. The two latter reasons reduce the bone volume to support the tibial component which may increase the risk of fracture. As such, surgeons should avoid large tibial resections as well as peripheral positioning [39], especially in those with already little bone volume to support the tibial component. Further, the relationship of higher BMI and excessive postoperative alignment angles with periprosthetic tibial fractures may be explained by the excessive loads placed on the small tibial surface [40,74,84]. In addition, small medial femoral condyles needing small components might also be a risk factor leading to overload because of smaller contact areas at the medial tibial surface [34].
Despite surgeons should be aware of potential risk factors, current evidence underlines developments in instrumentation and implants can minimize fracture risk. Chang et al. showed a modified technique using a predrilled tunnel through the tibia prior to cutting could avoid extended vertical saw cut errors [19]. Campi et al. suggested the optimal interference fit for good implant stability and minimal risk of fracture is between 0.5 mm and 0.7 mm [16]. Mohammad et al. reported improvements in instrumentation that widen the keel slot could reduce the risk of tibial fractures in cementless Oxford Partial Knee implants without compromising fixation [64]. Some authors suggested to change the depth of the tibial keel in very small cementless Oxford Partial Knee components as the depth of the keel is currently the same in all components, increasing the risk of fracture [38]. Vardi et al. reported that a change was made to the shape and size of the tibial keel of the Alphanorm implant due to high rates of periprosthetic tibial fractures [88].
This study revealed that most of periprosthetic tibial fractures occurred intraoperatively or within 3 months of surgery and were non-traumatic. Studies of intraoperative fractures described that operative damage in combination with the impaction of the tibial component caused the tibial bone to fracture. The postoperative fractures within 3 months may be associated with operative damage and repetitive stress on the bone during daily activities such as walking and stair climbing. Fractures that presented after 3 months were mostly associated with traumatic events, excessive weight,  Furthermore, a classification of periprosthetic tibial fracture types was presented. As only 10% of all fractures could be used in the classification, the incidence and completeness of fracture types in UKA remain unknown. However, presented paths of fractures could explain the high-risk fracture regions. For example, the type I fracture not only suggest that an extended sagittal cut posteriorly can initiate a fracture, but indicate that risk of fracture propagation can be increased by placing pins from the extramedullary tibial guide within fracture line regions.
Some limitations of this study should be noted. First, the pooled estimated incidences of fractures were not adjusted for the follow-up period. However, almost all clinical studies had a minimum follow-up of one year and thus included the period when the majority of fractures occurred. Second, poor reporting on characteristics of fractures may have biased the results. Third, not all risk factors for fractures in UKA mentioned by authors have been verified with clinical data, and therefore might be subjective. Also, it cannot be clarified which risk factors verified with clinical data were independently related to periprosthetic tibial fractures as the findings were based on unadjusted analyses. Fourth, to analyze whether increased BMI and age were related to fracture cases, the weighted mean of the overall UKA population was used with the same standard deviation as those of the periprosthetic tibial fracture cases. Although this approach can be considered a fair approximation, the statistical difference for BMI and age between UKAs with and without fractures may have been underestimated. Finally, this study did not focus on the diagnostics and treatment of periprosthetic tibial fracture in UKA. However, based on the current search, three studies have currently evaluated the management of periprosthetic tibial fractures in UKA [14,80,91]. Treatments of the included fracture cases were reported

Conclusion
The incidence of periprosthetic tibial fractures in cementless UKAs can be similar to those seen in cemented UKAs. However, surgeons should be aware that an excessive interference fit for cementless UKAs in combination with an impaction technique may introduce an additional risk, and may, therefore, be less forgiving to surgical errors and patients who are at higher risk of periprosthetic tibial fractures. While findings of this study raise awareness about periprosthetic tibial fractures in UKA, this study also highlights the importance of improvements in instrumentation and implants to prevent periprosthetic tibial fractures in future practices.