Introduction

Unicompartmental knee replacement (UKR) is an effective well-established treatment for anteromedial knee osteoarthritis which has failed to respond to conservative management [66]. Whilst UKR offers substantial benefits over total knee replacement (TKR) [36, 44, 67], it has a higher revision rate, particularly for aseptic loosening [7, 48, 62].

The two main types of fixation used to implant components are cemented and cementless techniques. Cemented components rely on bone cement to fix the components to surrounding trabecular bone; whereas, cementless components rely on the principle of press-fit fixation and osseointegration [39, 40, 63]. The current gold standard for knee replacements is cemented fixation [7, 48, 62] given the poor results of the first cementless knee replacements [5, 10].

There has been a recent increase in interest in cementless fixation given the need for fixation to last a patient’s lifetime with rising life expectancies [32]. Additionally, the merits of a more natural biological fixation, avoidance of cementation errors, a reduction in radiolucent lines and pain are certainly attractive [10, 27, 44].

There is currently no consensus on how the overall long-term clinical outcomes of cemented UKRs compare to cementless UKRs across the world and for different UKR types. Such a comparison would need to investigate not only the revision rate, but also the functional outcomes achieved from both fixation types. This systematic review addresses this question by comparing cemented and cementless UKR results published globally by comparing: revision rates, revision indications and PROMs. The null hypothesis was that there would be no difference in the revision rates of cemented and cementless UKRs.

Materials and methods

This systematic review has been registered prospectively on PROSPERO, CRD42019134315 and follows the preferred reporting items for systematic reviews (PRISMA).

Inclusion and exclusion criteria

The inclusion criteria were studies in the English language that reported the ≥ 10-year outcomes of any primary medial UKR for osteoarthritis in adult patients. Studies included were from 2009 onwards to assess the outcomes of UKRs published within the last 10 years. The most commonly used cementless UKRs were introduced after cemented versions and first widely adopted from 2009 onwards [42]. Additionally using studies before this period would include a disproportionate number of older cemented UKR studies which would not be deemed as comparable to the more recent cementless UKR studies.

Exclusion criteria included registry studies given they tend not to subdivide implants according to fixation and whether the implant is medial/lateral and to prevent duplication of patients with existing studies in the literature [7, 48]. Additionally, registries can under-report revisions [57]. Additional exclusion criteria were case reports, abstracts, hybrid UKRs and any studies in which lateral UKRs formed more than 10% of the whole cohort given our study was focusing on medial UKR outcomes. Studies of all polyethelene tibial components, bicompartmental replacements and those looking at only certain subgroups of the population were excluded given these contribute potential unnecessary confounders. Details of the number of articles actually excluded from the study based on the inclusion and exclusion criteria specified are summarised in Fig. 1.

Fig. 1
figure 1

PRISMA flow chart

Full size image

Search strategy

Alongside an expert librarian, the databases Medline, Embase, Central were searched from their inceptions to 23/04/2019 and are summarised in the Appendix. Key words used in the search strategy included “knee arthroplasty” and “fixation” with all variations of these terms. In addition, reference lists of the included publications were also screened to identify any additional reports.

First, study duplicates were removed followed by a title and abstract screening based on eligibility criteria. All shortlisted papers had full texts extracted and were assessed. Where the same cohort was published more than once, the most recent publication using the full cohort was included. There was complete agreement between the independent authors (HRM, GSB) who screened the studies.

Outcomes of interest

The primary outcome of interest was revision. This was measured from; (1) revision rate and (2) 10-year survivals reported. Revision was defined as any removal/addition of any component to the knee joint as per the registries [7, 48, 62]. Secondary outcomes were (1) Revision indications and (2) PROMs.

The a priori analysis was to first compare fixation groups (cementless vs cemented), and then compare bearing-type (mobile vs fixed bearing) results within each fixation category.

Data collection and risk of bias

Two authors (HRM and GSB) independently extracted data from all included studies. Contact attempts were made for all authors to obtain missing information. In cases where a study reported results for both cemented and cementless fixation types, only UKR arms reporting long-term outcomes were included in the review as per the specified inclusion criteria.

All studies were assessed for risk of bias using the methodological index for evaluation of non-randomised studies (MINORs) as a percentage and an additional system based on the reporting of the primary outcome (A = clearly reported, B = not reported/unclear) and the number of cases (A > 100 cases, B 51–99 and C < 50) [10, 15, 44, 60]. Studies with a MINORs score over 80% were deemed at low risk of bias and those below 70% at high risk except those with three or more As in the primary outcomes [10, 15, 44, 60].

Data synthesis and analysis

The primary outcome, the revision rate, was calculated per 100 component years which is equivalent to the annual rate [% per annum (% pa)] as per the Australian Joint Registry [7] and previous reports [10, 25, 33, 44]. This involved dividing the total number of revisions by the total observed component years multiplied by 100 [50]. 95% confidence intervals (CI) were generated using the Clopper Pearson method [14]. Each revision indication rate was also calculated using the same methodology. Revisions and their indications per 100 component years were compared between groups using the chi-squared proportional test.

From the included studies, Campi et al. [11] reported the results for 1000 UKRs, but of these, 318 UKRs were also used in Mohammad et al.’s.[41] more recent study. The results of Campi et al’s. [11] unique 682 UKRs were obtained to prevent duplication of UKRs in the analysis. Therefore, in this systematic review, the number of UKRs reported for Campi et al. [11] is 682. This prevented overpowering the study and used the most up to date information for the cohort.

All statistical analysis was performed using Stata version 14 (STATA Corp, Texas, United States of America). p values of < 0.05 were deemed statistically significant.

Results

5835 articles were identified, which after duplicates and title/abstract screening were reduced to 39. Full text analyses deemed 25 articles eligible for inclusion (Fig. 1). Details of full text articles excluded are in Table 1. There were 21 studies [3, 4, 6, 12, 13, 16,17,18,19, 29, 31, 37, 38, 51, 55, 59, 61, 64, 65, 68, 69] reporting the long-term outcomes of cemented UKRs and 5 [11, 20, 34, 41, 59] reporting the outcomes of cementless UKRs. The majority of studies (15/25) were of the Oxford UKR (Table 2). All identified studies were observational studies with no long-term comparative studies. All studies scored low risk of bias except Aly et al. [4] (Table 3). The total number of UKRs in the cemented and cementless groups was 8790 and 1946.

Table 1 List of excluded full texts and reasons
Full size table
Table 2 Breakdown of UKR prostheses used in included studies
Full size table
Table 3 Risk of bias of cemented and cementless UKR studies
Full size table

Revisions by fixation type

24 out of 25 studies (n = 10,054) reported the number of revisions during the study period and the mean follow-up which allowed for quantitative analysis (Fig. 2). The only exception to this was Price et al. (n = 682) [55] which reported the median follow-up. Table 7 in the appendix summarises this in detail for each study.

Fig. 2
figure 2

Forest plot of revision rates per 100 component years of cemented and cementless UKR studies

Full size image

The overall revision rate for the cemented and cementless groups was calculated separately. In the cemented group (n = 8108), there were 456 revisions out of 62,637 component years resulting in a revision rate of 0.73% pa (CI 0.66–0.80) (Table 7). This equates to a 10-year survival of 92.7%. In the cementless group (n = 1946), there were 57 revisions out of 12,740 component years resulting in a revision rate of 0.45% pa (CI 0.34–0.58) (Table 7). This equates to a 10-year implant survival of 95.5%. The differences between cementless and cemented revision rates were significant (p < 0.001). The revision rates are plotted in Fig. 2.

There were 13 studies of the cemented Oxford UKR (n = 6326) and 2 studies of the cementless Oxford UKR (n = 1682). For the cemented Oxford, there were 381 revisions out of 49,384 component years giving a revision rate of 0.77% pa (CI 0.70–0.85). The cementless Oxford studies reported 37 revisions out of 9874 component years giving a revision rate of 0.37% pa (CI 0.26–0.52). The difference between the revision rates was significant (p < 0.001).

There were 10 studies of non-Oxford UKRs, of which 8 had cemented UKRs (n = 1782) and 3 had cementless UKRs (n = 264). For the cemented UKRs, there were 75 revisions out of 13,253 component years giving a revision rate of 0.57% pa (CI 0.45–0.71). For the cementless UKRs, there were 20 revisions from 2866 component years giving a revision rate of 0.69% pa (CI 0.43–1.10). There were no significant differences in the cemented and cementless non-Oxford UKR study’s revision rates (p = 0.41).

Revisions by bearing type

In the cemented group (n = 8108), there were 6478 mobile bearing UKRs [3, 4, 12, 16,17,18, 29, 31, 38, 51, 59, 65, 69] and 1071 fixed bearing UKRs [6, 19, 37, 61, 64, 68] clearly indicated. Chatellard et al. [13] (n = 559) had a mixture of mobile and fixed bearing UKRs and, hence, was not included in this analysis. There were 388 revisions out of 50,934 component years and 54 revisions out of 8813 component years for mobile and fixed bearing cemented UKRs, respectively. This resulted in revision rates of 0.76% pa (CI 0.69–0.84) and 0.61% pa (CI 0.46–0.80), respectively (Fig. 3, Table 7). The corresponding 10-year implant survival for mobile and fixed cemented UKRs were 92.4% and 93.9%. The differences in revision rates between mobile and fixed bearing cemented UKRs were not significant (p = 0.13).

Fig. 3
figure 3

Forest plot of revision rate per 100 component years by fixation and bearing type

Full size image

In the cementless group (n = 1946), there were 1760 mobile bearing UKRs [11, 41, 59] and 186 fixed bearing UKRs [20, 34]. There were 39 revisions out of 10,779 component years and 18 revisions out of 1961 component years. This resulted in revision rates of 0.36% pa (CI 0.26–0.50) and 0.92% pa (CI 0.54–1.45), respectively (Fig. 3, Table 7). The 10-year survival for mobile and fixed bearing cementless UKRs were 96.4% and 90.8% with this difference being significant (p = 0.001).

Implant survival reported

Of the 21 cemented studies, 19 studies reported the long-term survival. 15 studies reported the implant survivals at 10 years and ranged between 83.7 and 96.3% (Table 4). All 5 cementless studies reported long-term implant survivals. 3 studies reported the 10-year implant survival ranging between 96.6 and 97.5% (Table 4).

Table 4 Long-term implant survivals for the overall cohort from each of the included studies
Full size table

Indications for revision

All studies (23 studies, n = 9532; 7586 cemented and 1946 cementless) except Campi et al. [12] and Price et al. [55], reported the mechanisms of failure by fixation type and mean follow-up time (Table 8). The revision rates per 100 component years for aseptic loosening and disease progression were significantly lower (p = 0.02 and p = 0.009, respectively) in the cementless group compared to the cemented group (0.06 vs 0.13% pa and 0.10 vs 0.21% pa, respectively). The revision rate for polyethylene wear/impingement was significantly higher (p = 0.03) in the cementless group (0.05 vs 0.02% pa). No other revision indication was significantly different between cementless and cemented UKRs (Table 5).

Table 5 Indications for revision surgery in the cemented and cementless cohorts
Full size table

Patient-reported outcome measures

13/21 cemented studies and 4/5 cementless studies reported the long-term PROMS for the overall cohort studied. Studies reporting preoperative PROMs all showed an improvement at the ≥ 10-year scores. For the cemented UKRs the 10-year OKS reported ranged between 37 and 40 and for the cementless, it ranged from 38 to 41.7 (Table 6).

Table 6 Patient reported outcomes reported for the whole cohort at approximately 10 years
Full size table

Discussion

This is the first systematic review to the best of the author’s knowledge comparing the long-term outcomes of cemented and cementless UKRs. Overall cementless UKRs had a revision rate that was about one-third lower than cemented. This difference appears to be due primarily to the rates of revision for aseptic loosening more than halving.

Although historically cementless implants had a reputation of poor outcomes [10], this review suggests that they currently achieve similar if not better results than cemented implants. This review’s results are in agreement with a recent registry-based propensity-matched comparison of cemented and cementless Oxford UKRs [42], which found that the revision rate of the cementless was nearly a third less than the cemented and the revision rate for aseptic loosening more than halved. There are also concerns that cementless fixation is less forgiving than cemented and that only high-volume surgeons would benefit. However, another study has found that the merits of cementless are independent of surgeon volume [43]. Therefore, all surgeons should at least consider using cementless UKR implants.

The fact that cementless UKRs had significantly lower revision rates only for the Oxford UKRs could be because of the limited numbers in the non-Oxford implant group making it more prone to type 2 error. However despite this, the revision rates of the non-Oxford UKR fixation groups were essentially equivalent. The other possibility is the design of the Oxford UKR, which is ligament preserving with a mobile bearing resulting in predominantly compressive loads with minimal shear, and is therefore ideal for cementless fixation.

Randomised studies comparing UKR fixation showed a significant reduction of radiolucencies in cementless groups indicative of improved fixation [27, 52]. This probably explains why the rates of revision for loosening reduced. The decrease in revision rate for arthritis progression with cementless fixation is more difficult to explain. Possible explanations include cement fragments causing direct damage to the lateral compartment or cementing errors causing medial overstuffing resulting in lateral overload. The cementless Oxford UKR femoral component, compared to the cemented, has an additional anterior peg to improve fixation. To accommodate this more bone has to be removed anterior to the femoral component which may decrease the risk of the bone impinging on the bearing, which is known to increase the risk of disease progression [56]. Additionally, the overall mean follow-up weighted on each studies sample size differed between fixation groups (cemented = 7.7 years, cementless = 6.5 years). Arthritis progression tends to occur late, so the longer weighted mean follow-up in the cemented group will disproportionately increase its revision rate specifically for this indication.

No obvious differences in long-term PROMs were found between the cementless and cemented groups. However, both groups had better PROM scores than those commonly reported for TKR [46].

There are two fundamentally different design concepts for UKR; mobile and fixed bearings. The debate of which is better has been a contentious issue [49]. Theoretical advantages of a mobile bearing UKR include lower linear polyethylene wear, better long-term knee kinematics, and a more even load distribution at the implant–bone interface [35]. However, fixed bearings have the advantage of not dislocating. Our study showed that mobile bearing UKRs had significantly lower revision rates than fixed bearing UKRs in the cementless but not cemented groups. Other reviews and clinical studies, which were predominantly based on cemented components also found no differences in their outcomes [35, 45, 54].

The main limitation of this review is that all included studies were observational cohorts with no comparative control arm. Although there has been a formal comparison of the overall revision rates of cemented and cementless UKR from studies using the proportional chi-squared test, this must be interpreted with caution given this is an overall comparison between studies and not from a pooled comparison within studies. Therefore, it does not account for confounding factors, or selection bias related to the selection of patients included in these cohorts with different lengths of follow-up. There is considerable heterogeneity between studies (Fig. 2) where the revision rate can be seen to vary between studies, particularly for the cemented studies. Additionally, given the cementless studies are understandably from fewer centres, this may introduce a possible expertise bias. However it is encouraging that our results mirror those published from propensity-matched registry comparisons, which address these limitations [42]. Larger UKR numbers would cause the revision rate to fall, but this would not explain the lower revision rates in the cementless group as they had far fewer UKRs than the cemented group. Finally, the study was limited given most studies were of the Oxford UKR but this reflects the current literature and highlights the need for more long-term non-Oxford UKR studies of both fixation types.

Conclusions

Cementless fixation offers lower long-term revision rates than cemented, particularly in mobile bearing UKR, with a reduction in aseptic loosening rates suggesting improved fixation. All surgeons should, therefore, at least consider using cementless UKRs in their practice.