Background

Knee arthroplasty is a highly successful procedure and, for most patients, is definitive surgery expected to last a lifetime [1]. However, particularly for younger patients, a previously successful knee arthroplasty may deteriorate over time (for example, due to wear of the components) and return to the attention of the surgeon. For other patients, primary knee arthroplasty may have failed to treat the original symptoms. Previous studies have suggested that around 13% of patients are dissatisfied with their outcome following knee arthroplasty [2] and up to 20% of patients have chronic pain [3]. Whilst many of these patients improve with support, those who do not may look to explore revision surgery.

Revision knee arthroplasty (rKA) can be defined as further surgery to an existing knee arthroplasty where a component is added, replaced or modified or the joint is debrided and irrigated [4]. For some patients, there is an absolute indication for rKA, and alternative treatment options are reserved for those unfit (or unwilling) to undergo surgery. This group can include a variety of diagnoses, but urgent indications (such as prosthetic joint infection [PJI] and certain types of fracture) provide unambiguous examples [5]. Elective, aseptic rKA is more common (> 80% cases) [4, 6] and the decision of whether (or when) to undergo rKA follows a shared decision-making process between a patient and their surgeon after discussion of the risks, benefits and alternative treatment options [7]. The goals of surgery in these cases are often similar to primary knee arthroplasty: to reduce pain, improve quality of life and minimise the risk of future complications.

For patients considering elective, aseptic rKA, it follows that full participation in a shared decision-making process requires clear information (supported by high-quality evidence) on the expected outcome should they choose surgery, do nothing or select another type of care [8]. However, the evidence to support these discussions is limited, and has not previously been addressed with a systematic review. As such, the aim of this systematic review was to summarise the evidence for the clinical effectiveness of rKA compared to non-operative treatment for the management of patients with aseptic, non-urgent causes for failed knee arthroplasty.

Methods

Patient and public involvement (PPI)

This study is supported by the SORE (Surgery Or REstraint for elective, aseptic revision) knee arthroplasty PPI group and a recent James Lind Alliance Priority Setting Partnership [9].

Ethics

Research ethics committee (REC) approval was not required for this review.

Registration and reporting

The study was prospectively registered with PROSPERO (CRD42020196922) and is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 Statement [10]. The completed PRISMA checklist is provided as a Supplementary file.

Search strategy

Our search strategy (Appendix 1) was designed with an experienced information specialist. MEDLINE, Embase, AMED and PsychINFO were searched from inception to 1st December 2020. There was no restriction on language of publication. Reference lists of included studies were examined to identify further relevant publications.

Types of study

Randomised and non-randomised studies of patients with a failed rKA treated with elective, aseptic rKA or non-operative management were eligible for inclusion. Randomised studies of any size were eligible for inclusion, whilst non-randomised studies with fewer than 100 patients were excluded due to feasibility.

Population, Interventions, Comparisons and Outcomes (PICO)

The PICO framework for this study is illustrated in Fig. 1 and described below.

Fig. 1
figure 1

A diagram to illustrate the study population, interventions and comparisons, types of study and patient-relevant outcomes

Full size image

Population

Patients aged 18 years or older with a failed KA were eligible for inclusion. A failed KA was defined when patients were explicitly stated to be candidates for rKA. We anticipated that this definition may have failed to identify some studies reporting on suitable patients receiving non-operative management. However, we considered it important to be able differentiate this patient group from the larger population with a poor outcome after KA, where revision surgery is often not discussed or offered. We did not consider patient-report outcome measures (PROMs) to be suitable to define failure, since no clear threshold has been defined and current evidence suggests this is likely to vary widely between patients and surgeons [11].

Interventions and comparators

Revision knee arthroplasty (rKA) was defined as any procedure following primary knee arthroplasty where a component of an arthroplasty was removed, modified, or added [4]. This included isolated exchange of a polyethylene insert, secondary patella resurfacing after total knee arthroplasty, arthroplasty of a further compartment of the knee after partial knee arthroplasty and re-revision surgery. Studies with any procedures for malignancy were excluded. We included studies with up to 30% of procedures for urgent indications (infection or fracture). This threshold was chosen to maximise inclusion of the available literature, without compromising the population of interest. Approximately, 20% of all rKA are performed for ‘urgent’ indications [4, 6]. Non-operative management was defined as any intervention to the joint arthroplasty other than revision arthroplasty (including no treatment).

Outcomes

The time-points of interest (unless otherwise stated) were defined as: immediate (“in-hospital” or up to 30 days), early (up to 1 year); medium-term (1–5 years); and longer-term (over 5 years). Studies were required to report on one or more of the following outcomes:

Implant survivorship

The primary outcome of interest was all-cause re-revision surgery (which included both elective, aseptic and non-elective, aseptic reasons for re-revision). Studies were required to report implant survivorship using the Kaplan-Meier method. The time-points of interest were 1, 5, 10 and 15 years. Studies reporting implant survivorship at other time points were rounded down to the nearest of these milestones. A further analysis was performed based on the calculation of person-time incidence rates (PTIRs).

Patient-reported outcome measures (PROMs)

‘Joint-specific’ PROMs were defined as instruments addressing one of the following domains: pain, function, combined pain and function, joint-related health status, or patient activity. These instruments were required to be supported by a validation study in a rKA population and to have at least ‘potential for recommendation’ as defined by the COnsensus-based Standards for the selection of health status Measurement INstruments (COSMIN) initiative [12]. The Knee Injury and Osteoarthritis Outcome Score (KOOS) [13], Lower Extremity Activity Scale (LEAS) [14], Oxford Knee Score (OKS) [15, 16], and Western Ontario and McMaster Universities Arthritis Index (WOMAC) [17] instruments met these criteria. A clinically meaningful change following elective, aseptic rKA has only been defined for the OKS (where the MICgroup = 9.5 points) [16, 18]. For health-related quality of life (QoL) and anxiety or depression we did not require instruments to have been validated specifically for elective, aseptic rKA.

Acquired comorbidity (including mortality)

Acquired comorbidity following rKA was recorded for immediate and early follow-up. The following adverse events were recorded: death, allogeneic blood transfusion, cardiac complications, central nervous system complications, deep vein thrombosis, pulmonary embolism, genitourinary complications, renal complications, respiratory complications, post-operative infection (such as deep surgical site infection or sepsis) and wound dehiscence. We also recorded the incidence of ‘any complication’ where reported as such in a study. This system was chosen based on prior knowledge of World Health Organisation (WHO) International Classification of Disease (ICD) codes, which it was anticipated that many studies would use [19].

Hospital admission impact

Hospital admission impact was evaluated according to length of stay, requirement for high-dependency or intensive care, and hospital re-admission.

Data extraction and management

All citations were imported to the web application Rayyan [20]. De-duplication and abstract screening was performed by two review authors (SS and JX/LF). The full-text of each study potentially meeting inclusion criteria was screened by two reviewers (SS and AS/EH/RB/TH). Disagreements were resolved through discussion. A standardised data collection form was created using the Research Electronic Data Capture (REDCap) data management platform and piloted to ensure consistency and ease of use [21]. Data were extracted on study design, dates of study, number of sites and location, and study setting. Participant enrolment and withdrawals were recorded, together with demographic information (age, gender, comorbidities and revision diagnosis). The funding source and notable declarations of interest for trial authors were recorded. Data were extracted from figures at the discretion of the lead author.

Data analysis

Meta-analysis was performed for implant survivorship at 1, 5, 10 and 15 years following assessment of clinical and methodological homogeneity. The included studies were required to report survivorship using Kaplan-Meier estimates, under the assumption that these estimates approximated risk. The Stata package metan was used for analysis. A random effects model was used to account for variability among the included studies (for example, due to different characteristics of the patient groups). Statistical heterogeneity was assessed by visual inspection of the forest plot for obvious differences in results between the studies, and by using the I2 and chi2 statistical tests. Where studies did not report a 95% confidence interval around the Kaplan-Meier estimate, simple imputation was performed to impute the mean standard error calculated from the other studies reporting at that time point. A sensitivity analysis was performed to examine the effect of excluding studies with imputed data. Since not all studies reported Kaplan-Meier estimates, an additional analysis was performed for studies that provided data where person time incidence rates (PTIRs) could be calculated. The denominator for rate was calculated by multiplying the number of patients with the mean follow-up. The numerator was calculated by totalling the number of first re-revisions over the study follow-up. The PTIR was then expressed as the number of re-revisions per 100 patient years at risk (which corresponds with current NJR methodology) [22]. Secondary outcome measures (patient reported outcome measures, acquired comorbidity, and hospital admission impact) were evaluated using narrative synthesis with results organised into tables.

Quality assessment

Two authors (SS and EH/RB) independently assessed study quality according to the checklist proposed by Wylde et al. [23], which was designed for studies on joint arthroplasty. The tool evaluates bias due to patient selection (two items), missing data (one item) and confounding (one item). Each item is rated either ‘adequate’ or ‘inadequate’ and reported individually, rather than as a summary score. An adequate rating is given to (i) recruitment of consecutive patients, (ii) recruitment of patients from multiple centres, (iii) follow-up of more than 80% of patients and (iv) use of a multivariable model.

Missing data

We did not contact investigators or study sponsors to obtain missing outcome data.

Software

Statistical analyses were performed using Stata (StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC.)

Results

After deduplication, the titles and abstracts were screened for 4297 articles. 149 full-text articles were assessed for eligibility. The PRISMA flow diagram is provided in Fig. 2. No randomised or non-randomised studies were identified that reported on patient-relevant outcomes following elective, aseptic rKA compared to another form of care. No studies reported on patient-relevant outcomes following non-operative management for failed KA. Forty non-randomised, uncontrolled studies (434,434 rKA) [24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63] reported on patient-relevant outcomes following elective, aseptic rKA and were included in this review (Table 1).

Fig. 2
figure 2

PRISMA flow diagram

Full size image
Table 1 Overview of included studies
Full size table

Outcome measures

Implant survivorship

Fifteen studies [24, 33, 36,37,38, 41, 49, 50, 52, 53, 55, 56, 61,62,63] reported all-cause implant survivorship for 7227 rKA (Fig. 3). Seven studies (5524 rKA) reported survivorship at 1 year, 13 studies (5,754 rKA) at 5 years, nine studies (2188 rKA) at 10 years and two studies (452 rKA) at 15 years. Pooled analysis of data found all-cause implant survivorship of 95.5% (95% CI 93.2–97.7%) at 1 year, 90.8% (95% CI 87.6–94.0%) at 5 years, 87.4% (95% CI 81.7-93.1%) at 10-years, and 83.2% (95% CI 76.7–89.7%) at 15 years. These estimates changed little when studies that did not report confidence intervals for survivorship estimates were excluded (Appendix 2 Figure 1). Eighteen studies (3205 rKA) [24, 25, 28, 31,32,33, 40, 41, 45, 46, 49, 50, 53, 58, 59, 61,62,63] provided data from which person-time incidence rates could be calculated. These are provided as a further sensitivity analysis in Appendix 2 Table 1.

Fig. 3
figure 3

Forest plot of estimates for reported implant survivorship following elective, aseptic rKA

Full size image

Patient-reported outcome measures (PROMs)

Twelve studies [25, 34, 38, 41, 42, 47, 48, 51, 57, 59, 62, 63] reported on the outcome of 2382 rKA with one or more returned PROM questionnaires (Appendix 3 Table 1). The instruments used to report joint function were: KOOS (2 studies), OKS (5 studies), and WOMAC (5 studies). The instruments used to report QoL were EQ-5D (2 studies), SF-12 (2 studies) and SF-36 (3 studies). None of the included studies measured anxiety or depression using dedicated instruments, though these domains were assessed within some of PROMs listed above. Ten of the 12 studies (83.3%) reporting on joint-function and 6 of the 7 studies (85.7%) reporting on QoL provided both pre-operative and post-operative summary statistics. Each of these studies reported improvement in joint-function and QoL following elective, aseptic rKA. Indeed, the two studies that reported mean change in score using the OKS [48, 59], both found that improvement in joint function exceeded the MICgroup estimate of 9.5 points at all post-operative timepoints.

Acquired comorbidity

Mortality

Four studies (353,064 rKA) reported mortality rates after rKA [26, 27, 35, 54] (Table 2). Three studies reported on immediate-term mortality [27, 35, 54] with estimates ranging from 0.16 to 0.30%. Yao et al [26] reported an early (1 year) mortality rate of 1–2% for indications other than fracture and infection from a single tertiary centre in the USA between 1985 and 2015.

Table 2 Studies reporting on mortality after rKA
Full size table

Blood transfusion

Four studies [27, 30, 35, 39] (53,618 rKA) reported on the need for blood transfusion following rKA (Appendix 4 Table 1). All studies were based in the USA and the rate of blood transfusion ranged from 8.4% [39] to 18.4% [35]. Nichols et al. [39] analysed the Marketscan administrative claims dataset and reported a rate of allogeneic blood transfusion of 7.9% during the index hospitalisation, with a further 0.5% requiring autologous blood transfusion. Dai et al. [35] reported a transfusion rate of 18.4% during the index hospitalisation from 5187 patients within the US Nationwide Inpatient Sample (NIS). Burnett et al [30] analysed the Humana Inc. administrative claims database where 11.9% of patients required blood transfusion within 3 days of rKA between 2007 and 2015. Most transfusions (92.0%) were with allogeneic packed red blood cells and they found a 72% reduction in requirement for blood transfusion from 2007 (15.9% rKA) to 2015 (4.5% rKA). Boddapati et al [27] analysed data from the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) where they found a transfusion rate of 11.9% within 30 days of aseptic rKA.

Complications

Seven studies [27, 29, 35, 39, 43, 54, 60] (408,050 rKA) reported on complications after rKA (Table 3). Three studies reported immediate complications [27, 35, 54] and four studies reported early complications [29, 39, 43, 60]. The rate of any complication at 90 days ranged from 9.1 [43] to 37.2% [39]. The reported rate of surgical site infection ranged from 15.6 [60] to 24.1% [39] in the two studies reporting early complications. Studies reporting immediate complications all reported lower rates of post-operative infection (< 1%). The specific complication of wound dehiscence was identified in 0.3% [54] to 1.7% [39] rKA. Medical complications included: deep vein thrombosis (0.2% [35] to 1.7% [29] rKA), pulmonary embolism (0.1% [39] to 0.6% [29] rKA), cardiac complications (0.3% [39] to 0.9% [54] rKA) and central nervous system complications (0.1% [27, 54] rKA).

Table 3 Early complications after rKA
Full size table

Hospital admission impact

Length of stay (LOS)

Eight studies [27, 39, 41,42,43,44, 54, 59] (372575 rKA) reported on LOS after rKA (Table 4). Among the studies based in the USA mean LOS ranged from 3.4 days [27] to 5.6 days [39]. Bin Abd Razak et al [59] reported a mean LOS of 7.7 days at a single tertiary centre in Singapore. Whilst Kim et al [41] reported a mean LOS of 16 days following rKA in the Republic of Korea from a single surgeon series.

Table 4 Studies reporting on length of stay (LOS) after rKA
Full size table

High-dependency care

None of the included studies provided information on high-dependency care utilisation after rKA.

Hospital re-admission

Three studies (36,200 rKA), all from the United States, reported on hospital re-admission after rKA [27, 39, 44] (Table 5). Boddapati et al [27] analysed data from 10584 aseptic rKA within ACS-NSQIP between 2005 and 2015 where they identified a readmission rate of 6% at 30 days. Nichols et al [39] reported a 23% re-admission rate at 90 days based on data from 25,354 rKA registered with the Truven MarketScan database in North America from 2009 to 2013. Schairer et al. [44] reported a 13% re-admission rate at 90 days using a hospital administrative claims database of 262 rKA from 2005 to 2011.

Table 5 Studies reporting on hospital re-admission after rKA
Full size table

Quality assessment

Among the 40 studies included, 21 studies (53%) recruited consecutive patients, 15 studies (38%) were multicentre, 31 studies (78%) had adequate patient follow-up and 20 studies (50%) included a multivariable regression model (Appendix 5 Table 1).

Discussion

This study has summarised patient-relevant outcomes (PROs) following elective, aseptic revision knee arthroplasty (rKA). The quality of the included studies was low, comprising uncontrolled observational series. We did not find any studies comparing PROs following revision surgery to non-operative management or no treatment at all. We have addressed the question: “How long is an elective, aseptic rKA expected to last?”. We found rKA survivorship ~ 96% at 1 year, ~91% at 5 years, ~87% at 10 years and ~ 83% at 15 years. All studies reporting on joint function and quality-of-life showed large improvements at early timepoints following rKA. We also reported the rate of complications following elective, aseptic rKA. These estimates may be useful to support the process of informed consent. The risk of death in the immediate post-operative period was low, with reported rates of 0.16% to 0.30%. Only one study reported mortality at 1 year, with a rate of 2%. The rate of any complication was highly variable (from 9.1 to 37.2% at 90 days following surgery). This is likely to reflect the heterogeneity of both patients undergoing elective, aseptic rKA and the procedures themselves. Post-operative infection (which is a set of administrative codes incorporating both systemic sepsis and surgical site infections) was one of the most common complications. There were large differences between studies reporting rates at immediate timepoints (< 1% “in-hospital” or within 30 days) compared to those reporting at early timepoints (15.6% to 24.1% within 90 days). Cardiac, central nervous system, genitourinary, renal and respiratory complications were all rare (~ 1% or less at 90 days). The rate of deep vein thrombosis ranged from 0.2 to 1.7%, while the rate of pulmonary embolism ranged from 0.1 to 0.6% at 90 days. With respect to the hospital admission, the mean length of stay (LOS) in the USA ranged from 3.4 days to 5.6 days. The two studies included from Singapore and Korea both reported longer mean LOS. We have not explored the reasons for this. The rate of re-admission to hospital ranged from 13 to 23% within the first 90 days. Patients undergoing elective, aseptic rKA were at high risk for blood transfusion, with rates of 8.4 to 18.4% reported.

A number of relevant studies have been published since the literature search for this review was completed. Deere et al. [64] reported on implant survivorship following first and multiple rKA procedures using data from the National Joint Registry (NJR) for England, Wales, Northern Ireland, the Isle of Man and the States of Guernsey. They reported Kaplan Meier survivorship estimates for first rKA procedures of 96.4% at 1 year, 87.4% at 5 years and 82.9% at 10 years. The reported re-revision rates at 5- and 10 years were higher than in the present study, which may be due to the inclusion of ‘urgent’ rKA procedures. They found that male gender and younger age were risk factors for multiple revisions. A recent study from our group reported on mortality and complication rates following 30,826 elective rKA procedures recorded in Hospital Episode Statistics (HES) in the UK [5]. This found a 90-day mortality rate of 0.44%, which is comparable to the estimates reported in this review, and similar to primary KA (0.46%). Of note, the early mortality rate following infected rKA appears to be greater (2.04% at 90 days) [5]. A further study from our group reported on patient-reported outcome measures following elective, aseptic rKA in 10,727 patients from the NHS PROMs dataset [11]. This found that two-thirds of patients experienced a meaningful improvement in joint function after rTKA, 69.4% were satisfied with the procedure and 74.1% felt that surgery was a success [11]. However, the rate of early patient-reported complications was very high (46.0% at 6 months)—which is much higher than reported in administrative datasets, as confirmed by the current review—and this finding requires further exploration.

A major strength of this study is that we have reported domains of outcomes following surgery that patients themselves have identified to be important [65]. Whilst the quality of the included studies was low, we predicted this with the design of our review. Due to the preponderance of small, low-quality studies reporting on elective, aseptic rKA, one inclusion criterion we enforced (based on feasibility) was to exclude studies with fewer than 100 participants. This has resulted in bias towards larger studies (such as those reporting data from joint registries and other routine healthcare datasets). On the one hand, these studies have enabled us to capture data on rare outcomes (such as mortality and a range of different complications). However, the limitations of administrative data coding and the restricted perspective of these datasets must also be understood. For example, whilst many studies reported the diagnosis of a complication, this was not always paired with information on the treatment that the patient subsequently went on to receive. Re-operations not classified as re-revisions were poorly reported and so were not summarised. We recognise that elective, aseptic rKA is an ‘umbrella term’ with heterogeneity in patients, indications for surgery, severity of the disease, and types of procedure. In the future, estimates for clinical outcomes should be tailored to these different groups. To aid future systematic reviews and meta-analyses, studies reporting on rKA would benefit from consensus on how causes of failure should be categorised. In the meantime, use of a hierarchical system may be beneficial [66].

Conclusion

Higher-quality evidence is needed to support patients with the decision of whether to undergo elective, aseptic rKA. This should include studies comparing operative and non-operative management. Implant survivorship following elective, aseptic rKA was ~ 96% at 1 year, ~ 91% at 5 years and ~ 87% at 10 years, with most studies identifying large improvements in pain and joint function. Early complications were common after elective, aseptic rKA and the rates summarised here can be shared with patients during informed consent.