Professor Mike Reed, Chair of the NJR Editorial Committee
Mr Tim Wilton, NJR Medical Director
There has been considerable work going on in the NJR over the last year relating to developments of various kinds that may be of interest to the reader. Some of this development work was deferred due to the pandemic and has now been re-introduced and/or completed. The executive summary therefore deals with the impact of some of these changes separately from the summary of findings relating to individual joints.
It has been increasingly unclear for some years whether the use of revision as the main metric for outcome analysis can lead to distortion of some of the overall outcomes attributed to surgeons, units and implants. This is more so when dealing with some joints than others. This may more greatly impact the results when assessing the outcome of surgeons and units than the results relating to implants given the smaller numbers involved with surgeons and units. In both cases, some additional secondary measure is desirable and we have used the national PROMs results as one such secondary measure. Unfortunately, this year the usual provision of the PROMs data by NHS Digital was not possible due to changes in the coding within NHS Digital’s systems, and we are actively seeking alternatives to enable the future provision of this important additional metric. Shoulder PROMs have not been part of the national PROMs programme and are collected by the NJR. We are actively pursuing ways in which similar PROMs data can be collected across all joint replacements to fulfil this requirement.
The treatment of periprosthetic fractures varies considerably across the country with some surgeons and units preferring internal fixation and some preferring revision of the implant where this is feasible. In the past, internally fixed periprosthetic fractures have not been analysed by the NJR and this has led to an underestimation of the failures of this procedure. The new Minimum Data Set (MDSv8) seeks to address this problem by specifically collecting this information. In addition, other forms of intervention on replaced joints are also now being collected to include most of those interventions that fall short of the strict definition of a ‘revision’. We expect this to provide a richer assessment of the overall outcomes of replaced joints in due course. MDSv8 also introduces the collection of more detailed information about revision operations that should allow more appropriate stratification of the outcomes of revision operations according to the complexity of the procedure.
Camouflage of poor results of an implant due to variants being analysed within an entire family (brand) has also been identified as a more significant issue during the last two years. As a result, the larger families of knees, where numbers allow, can be broken down and analysed in sub-groups within each brand. This is being performed routinely for such variants as cruciate-retaining and posterior-stabilised knees (Table 3.K9 (a)) and within those variants according to whether the patella is resurfaced or not (Table 3.K9 (b)). However, where there are sufficient numbers of cases, and particularly where there has been a concern raised about an implant, the analysis has now been more detailed and granular. It is our intention to widen the scope of these more detailed analyses and screening tools are being developed to guide when and how this should be done.
Hip implants have already been mostly separated into the pertinent sub-brand variants for the purposes of the comparative analysis, but it is clearly important that this is done similarly across all joints where relevant variants exist.
The data storage system for the NJR has been re-developed in a major way over the past two years and a number of key functions are now cloud-based. This change has allowed the various systems that previously co-existed but functioned in varying ways, to be re-designed so that they will allow interrogation and report production in the same way and using the same techniques.
There have been changes to UK legislation that make the collection of detailed information about many implantable devices mandatory. This opportunity to collect data about hip hemiarthroplasty used for the treatment of hip fractures has been taken up by the NJR and these will now form part of our routine data collection.
Analysis of the results of procedures at unit and surgeon level has been changed in the last year. Surgeons performing knee replacement are now assessed separately for their outcomes on total knee replacement, unicondylar replacement and patellofemoral replacement, although they are still being provided with their overall outcomes of all three types of operation for continuity and retrospective comparison. This has changed the status of a number of units and surgeons so that a few who were previously at outlier status have found they no longer are, while for some the opposite has occurred.
During the last two years there has been a great deal of work done with the International Society of Arthroplasty Registries (ISAR). They have agreed to adopt the classification system for hip and knee implants which had been developed jointly by the German Arthroplasty Registry (EPRD) and the NJR. This classification system will form the basis of their International Prosthesis Library (IPL). This decision has the capacity to allow all registries to analyse the results of implants in the same way, confident in the knowledge that they are discussing exactly the same variant of a device. Given the variations in sales around the world, and the fact that different variants are used in different parts of the globe and for differing indications, the ability to describe the implant construct with greater accuracy will be crucial when sharing outcome data. Work has now begun jointly with ISAR to develop a similar classification system for shoulder implants.
Commentary on findings
In this annual report there are excellent summaries and commentary in each of the joint replacement sections and we would encourage specialists to read their area of interest in full. We have summarised our key learning points and thoughts.
Hip replacement
This year's annual report is based on almost 1.5 million primary hip replacements performed by over 4,000 surgeons in almost 500 units.
We are now at 20 years since the NJR’s data collection commenced and we are reporting a maximum of 19.75 years of follow-up, although the size of some of the groups at longer follow-up is modest.
In the last three years, during and in the aftermath of COVID, the median number of procedures performed by a consultant over a three-year period was 59 (approximately 20 per annum) with a median number of procedures per unit of 492 (approximately 164 per annum). This represents a drop since pre-COVID times when surgeons were performing a median of 64 (approximately 21 per annum) over three years (see NJR Annual Report 2020).
In terms of bearing surface combinations ceramic-on-polyethylene (CoP) is now dominating in both hybrid and uncemented fixations (see Table 3.H2). Metal-on-polythene still dominates in cemented fixations, although fully cemented fixation is now used in less than 20% of all cases. Ceramic-on-ceramic bearings are now infrequently used. However, across the whole life of the registry approximately 30% of hip primaries have been cemented, 37% uncemented, and 25% are hybrid hip replacements.
Resurfacing is continuing at low levels (around 700 per year) and ceramic-on-ceramic hip resurfacing is now shown in Table 3.H1 for the first time, although this is performed in very small numbers.
There is a significant and consistent rise in the numbers of dual mobility hip replacements being performed, with the maximum surgeon volume being over 100 cases per year. This is somewhat surprising given the large number of very successful combinations of unipolar total hip replacements that are demonstrated in the registry with very long follow-up. One might specifically question the underlying reasons for this increase with the relatively low level of revision for dislocation in unipolar bearings that is shown in the registry.
It is worth studying Figure 3.H1 (d). This shows the location and funding of joint replacements over the last 20 years. It demonstrates that NHS-funded operations in NHS facilities peaked in 2014. They stayed level until COVID but have now dropped back to lower than 2007 levels. The independent sector provision has increased hugely over this period, particularly in the last couple of years of COVID recovery and there are now more hip replacements performed in the independent sector than in the NHS. Despite the cost-of-living crisis the number of hip replacements paid for privately has almost doubled since 2019. In terms of overall numbers of hip replacements, 2022 was similar to 2019.
Figure 3.H3 (e) demonstrates an increasing trend towards the 32mm and 36mm CoP bearings in both uncemented and hybrid fixations.
In trauma, the absolute number of total hip replacements performed for hip fractures is lower than recent years pre-COVID levels. It is noted that dual mobility operations for trauma are increasing (Table 3.H12).
Figure 3.H6 looks at revision of uncemented primary hip replacements by bearing. It is worth noting that failure rates of metal-on-polyethylene-on-metal (MoPoM) dual mobility bearings in this group are very high in the first couple of years, exceeding early failure in other implants including resurfacing and metal-on-metal (MoM) hip replacement. Revision rates in the early years are also high in the MoPoM dual mobility hip replacements in hybrid fixations (Figure 3.H7). In both groups the ceramic-on-polyethylene-on-metal (CoPoM) dual mobility bearing appears to be performing better than its metal counterpart, although the numbers are small.
Given the numbers of procedures now being performed, it is reassuring to see (in Figure 3.H10 (h)) the estimates of revision of primary hybrid CoP hip replacements for both the 32mm, and to a lesser extent, the 36mm bearing, have excellent survival. Revision rates in hybrid CoP are less than 2.5% at ten years.
Table 3.H8 details the success rates by brand and bearing surface. There are some outstanding leaders here with a significant number of combinations having a revision rate of less than 2.5% at 15 years. This calls into question whether the current National Institute for Health and Care Excellence (NICE) benchmark of a failure rate of less than 5% at ten years (NICE 2014) remains an appropriate contemporary standard. There is good cause to revisit this benchmark.
Of note, the best performing resurfacing brands have a revision rate of around 10% at 15 years. Analysis of the NJR data demonstrates that for every 100 MoM hip-resurfacing procedures it is estimated that there would be 7.8 excess revisions by ten years (Hunt et al., 2018). In 2022 there were approximately 700 resurfacing procedures, and this would approximately equate to an additional 55 revision procedures in the first ten years.
In Table 3.H9 we detail the causes for revision by fixation and bearing type. It is worth noting the higher failure rate in MoPoM dual mobility hips compared to its unipolar counterpart. Infection is responsible for almost 2.5 revisions per 1,000 prosthesis-years for these implants, with an all-cause revision rate currently running at nine revisions per 1,000 prosthesis-years. The figure for alternative treatment options, namely unipolar total hip replacement, is around three; although of course the groups of patients may not be directly comparable. The risk of infection, dislocation and periprosthetic fracture all appear higher in dual mobility implants compared to unipolar total hip replacement patients.
Knee replacement
We now have over 1.5 million primary knee joint replacement procedures within the registry performed by 3,613 consultant surgeons in 479 units.
Over the last three years contributing consultant surgeons have performed a median of 89 knee procedures (approximately 30 per annum), and each unit around 492.5 knee procedures (approximately 164 per annum). Knee replacements remain more common in females (56%), with a median age of 70. Over 97% of the cohort are documented as having osteoarthritis.
Cemented total knee replacements make up around 84% of primary knee replacements. Unicondylar knee replacements constitute around 10%.
Close to 60% of total knee replacements are cemented and unconstrained (cruciate-retaining) with a fixed bearing. A much smaller proportion, around 20%, are uncemented or posterior-stabilised, which do not appear to have the same results as cemented and unconstrained. Table 3.K6 demonstrates this across all age groups at ten years and beyond. With very long follow-up some of the groups are too small for results to be conclusive.
Figure 3.K1 (c) clearly shows that primary unicondylar knee replacement is on the rise. Although these did see a drop during 2020 and 2021 because of COVID, there has been a further increase with more unicondylar knee replacements being performed than ever before in the registry. The recovery of unicondylar knee replacements has been much better than total knee replacements post-COVID, and one could speculate this is because of the relatively greater ability for these joint replacements to be performed as day cases. Most unicondylar knee replacement procedures are now performed by surgeons performing more than 25 cases per year.
In contrast Figure 3.K1 (d) shows patellofemoral knee replacement is becoming less popular. There are now fewer than 1,000 cases being performed per year.
As noted with hip replacement, we can see that since 2012 most of the growth in NHS-funded knee replacement procedures has been in the independent sector. There are now fewer NHS-funded knee replacements performed in the NHS than there were in 2007.
Table 3.K2 demonstrates the continued decline in use of both uncemented and hybrid total knee replacement. These now represent 2.1% and 0.2% of our primary procedures respectively.
Females who undergo knee replacement are more likely to receive a total knee replacement than men who are relatively (but not absolutely) more likely to receive unicondylar knee replacement.
There are now multiple brands of knee replacement implant that are performing extremely well with very large numbers being tracked. Surgeons can choose from a wide number of brands with failure rates of less than 4% at 15 years. It is also worth noting that there is some separation in results at 19 years among the big brands, so please refer to Table 3.K9 (b) for further details. Broadly, across the brands, there is a higher revision rate for non-patella resurfaced cemented constrained total knee replacements. This appears to be mainly driven by pain but is very brand-specific (Tables 3.K9 (b) and 3.K10). Figure 3.K4 (a) shows exceptional survival for monobloc polyethylene tibias, although they appear to have been performed in a slightly older age group, this demands more exploration.
As has been noticed previously, Figure 3.K3 (b) demonstrates that although revision rates increased in the early years of the registry, they have been consistently reducing since 2008. This is consistent with, but less obvious than, the effect in hip replacement that was also influenced by MoM implants. One key point to make around knee replacement is that despite a tsunami of knee revision being predicted in the literature the number of knee revisions has remained remarkably stable, even declining in recent years, although this particular effect is likely to be related to COVID. Figure 3.K5 (a) demonstrates the chance of revision after primary knee replacement is higher in younger patients, and in males.
If unicondylar knee replacements are revised they do not behave like a primary total knee replacement in terms of longer-term survival and some differences are being seen between the commonly used implants. It appears that if a revision of a unicondylar knee replacement is required then the risk of re-revision is higher in uncemented and hybrid components, than it is in unicondylar implants that were initially cemented. Overall, it can be seen in Table 3.K5 that there is still a very significant difference in the revision rate for cemented unicondylar (medial or lateral) knee replacements which is 3.1 times higher than for cemented total knee replacements at ten years, and 3.5 times higher at 15 years. Even the best performing unicondylar knees (cemented fixed or uncemented mobile) have over double the revision rate of the popular unconstrained fixed cemented total knee replacement at ten years.
Patients will often ask how many times a knee replacement can be revised. In practice, there are six patients that have had ten or more revision procedures out of 1.5 million patients with primary procedures. Whether these revisions have ultimately resulted in a good outcome is not known from these data.
Mortality after primary knee replacement surgery is explored in Table 3.K12 (a). This shows some groups, particularly men over 85, are at relatively higher risk with mortality being almost 2% at 90 days, a factor which should be discussed with patients as part of a shared decision-making process for whether to undergo elective knee replacement at this age.
Ankle replacement
In this report we have a maximum follow-up of 12 years for ankle replacements. This cohort represents over 8,000 procedures.
Compared to pre-pandemic rates there has been a reduction in NHS-funded ankle replacements, and an increase in privately-funded cases. Reassuringly, it can be seen in Figure 3.A4 that most ankle replacements are now being conducted by surgeons performing more than seven ankle replacements per year, with large numbers being performed by surgeons performing more than 13. However, around a third of ankle procedures are being performed by surgeons who implant less than seven cases a year. In 2022, only seven units of 161 were performing more than 20 ankle replacement procedures per year. The British Orthopaedic Foot and Ankle Society (BOFAS) has recommended the use of networks and the pooling of resources to encourage specialist units to perform ankle replacement at higher volumes.
The overall headline revision rate is approximately 10% at 12 years. This is very implant specific however, with noticeable differences between implants (Figure 3.A8). It is also clear from Figure 3.A7 that younger patients and female patients are more likely to have a revision. From Table 3.A5, it can be seen that although aseptic loosening remains the most common reason for revision, infection comes a close second. Overall, there is a growth in fixed bearing ankle replacements and a distinct decline in mobile bearings (Figure 3.A5).
The Infinity implant was introduced in 2014 as part of a large multi-centre post-market surveillance study following the discontinuation of the Mobility implant which was the market leader up until that time. The gamble of moving to a fixed bearing implant is thus far supported by outcome data, as both the Infinity and the related prosthesis Inbone appear to have revision rates of less than 5% at seven years. Clearly this is still relatively short follow-up, with the uncertainty of small numbers and ongoing monitoring is essential.
There remains significant concern that we are not capturing arthrodeses or amputations following ankle replacements and thus the failure rate is probably higher than reported. We are hopeful this problem will be addressed by our data quality audits and the introduction of the forthcoming 'reoperation' data entry form.
Elbow replacement
In this report we present data for the first ten years after elbow replacement. This refers to total elbow replacement (with or without radial head replacement), lateral resurfacing and radial head replacement, and since 2018, distal humeral hemiarthroplasty which amounts to over 8,000 procedures. The majority are performed on women (67%). Roughly half of the implants required cement. There has been an increase, apart from during the COVID years, in data entry of elbow replacements. This is likely to be due in part to an increase in volume of procedures, improved reporting of radial head replacement, and inclusion of distal humeral hemiarthroplasties. Around half the cases were performed for trauma but over half of these were radial head replacements. Figure 3.E4 details the increasing proportion of primary total elbow replacements that are performed by higher volume surgeons (those performing more than 13 procedures a year). Figure 3.E3 shows that there still has not been a consistent recovery in practice since COVID.
Table 3.E4 (a) and (b) show the median number of elbow replacements per unit remains around three. Fewer units and surgeons are performing cases however. Some regions do appear to be performing significantly more replacements in elbow replacing units. This is likely to be the result of centralisation of services as part of the Getting It Right First Time (GIRFT) agenda.
It is clear from Figure 3.E5 that for primary total elbow replacement the revision rate for trauma is roughly half than that for an elective indication. This may well describe the frailty of these patients, higher mortality, and their suitability for revision.
Figure 3.E7 details survival rate of distal humeral hemiarthroplasty versus total elbow replacement with acute trauma as the indication. Numbers are small, particularly for the distal humeral hemiarthroplasty, but at the moment they certainly do not appear to be outperforming total elbow replacements.
There is a relative absence of long-term data for elbow replacement with only very small numbers at ten years. Table 3.E8 shows that at five years the linked total elbow replacements brands all have relatively similar survival of around 6 or 7%. These are small numbers in most brands.
The distribution of indications for elective elbow replacement has been consistent over the last five years of data entry with inflammatory arthropathy accounting for 32% of cases.
Although the five-year mortality rate after elbow replacement is consistent between trauma and elective surgery, when radial head replacement is taken out of the data the five-year mortality rate for trauma cases is almost double that of elective indications.
Shoulder replacement
Shoulder replacements have been recorded in the registry since 2012, so we present up to ten years of data. New classifications are now used for analysis. We now have almost 64,000 shoulder replacements under review.
Since the inception of data collection by the NJR, there has been a marked increase in stemmed reverse total shoulder replacements for trauma. Figure 3.S9 appears to demonstrate low revision rates of these stemmed reverse total shoulder replacements performed for trauma with revision at ten years being less than 3%. Reverse polarity shoulder replacements now dominate in trauma, and in elective practice they dominate for the cuff tear arthropathy indication.
Overall, from Table 3.S7 it is clear that men, particularly younger men, have higher failure rates.
Most humeral hemiarthroplasties and total shoulder replacements continue to be performed for osteoarthritis. Elective primary shoulder replacement for trauma appears to have a lower revision rate than when it is performed for elective indications although this may simply be due to the frailty of the patients and therefore revisions perhaps being avoided. This is not because the patients are dying before revision however, as this is accounted for in the data.
Shoulder replacement is the only area where the NJR collects PROMs. PROMs responses appear to be relatively poor and Figure 3.S10 demonstrates that those filling in PROMs questionnaires have a slightly different revision outcome to those that do not complete PROMs. Interestingly at ten years the revision rates of the groups appear similar.
PROMs results are explored in Table 3.S19. In elective practice the PROMs scores for humeral hemiarthroplasties appear lower than those for patients having a reverse total shoulder replacement or a standard total shoulder replacement, although the patients and the indications may differ. We do not have enough data to make comparisons in the trauma group. In elective practice less than 10% of patients have completed a pre-op and a 6-month post-op score (Table 3.S17). Figure 3.S11 clearly demonstrates the reduced chances of a patient gaining improvement if they have a higher pre-op shoulder score. Patients with a pre-op Oxford shoulder score over 40 appear to be more likely to get worse post-operatively.
Figure 3.S8 shows excellent long-term results with large numbers of stemmed reverse polarity total shoulder replacements. The key indication for these appears to have been rotator cuff replacement.
In elective practice, in Table 3.S8, the performance of stemmed conventional total shoulder replacements compared to stemmed reverse polarity shoulder replacements does differ, and at ten years the stemmed reverse polarity shoulder replacement appears to have the edge although it must be appreciated that the indications for both these replacement types are different.
Concluding acknowledgements
The NJR continues to work collaboratively with our many stakeholders; the most important, of course, are the patients we serve, and whom we would like to thank for allowing us to use their data.
The NJR operational collaboration is a huge team effort. Elaine Young, NJR Director of Operations, has demonstrated the great versatility of her leadership and her team.
Many thanks also to the following without which the NJR could not function:
All members of the NJR Steering Committee
Members of the NJR sub-committees:
Executive
Data Quality
Editorial
Implant Scrutiny
Medical Advisory
Regional Clinical Coordinators
Research
Surgical Performance
Members of the Data Access Review Group
Members of the NJR Patient Network
Other organisations:
Medicines and Healthcare products Regulatory Agency (MHRA)
Care Quality Commission (CQC)
NHS England (NHSE)
Welsh Government
Northern Ireland Executive
Isle of Man Department of Health
States of Guernsey
Independent Healthcare Providers Network Services
Getting It Right First Time (GIRFT)
British Orthopaedic Association (BOA)
British Hip Society (BHS)
British Association for Surgery of the Knee (BASK)
British Elbow and Shoulder Society (BESS)
British Orthopaedic Foot and Ankle Society (BOFAS)
European Orthopaedic Research Society (EORS)
Healthcare Quality Improvement Partnership (HQIP)
Confidentiality Advisory Group (CAG)
Association of British HealthTech Industries (ABHI)
We are most grateful to our NJR delivery contractors for their very valuable input into the NJR Annual Report and their many other functions. NEC Software Solutions, University of Bristol and University of Oxford teams help us refine and improve each year.
We offer our personal thanks to Vicky McCormack, Report Project Manager, NEC; Deirdra Taylor, Associate Director of Communication and Stakeholder Engagement and Oscar Espinoza, Communication and Design for the NJR, for getting the final report into shape.
A note on Patient Reported Outcome Measures (PROMs)
In the last NJR Annual Report, we published an exploration of the level of completeness and quality of data from the national PROMs programme and a proposal on how we might report implant-level PROMs in the report in future.
Over the course of the year, we have further consulted with stakeholders, including orthopaedic surgeons and representatives of the implant manufacturing industry, and have received broad support for the inclusion of implant level PROMs using the tables we had proposed last year.
PROMs data for hip and knee replacement surgery is not routinely collected by the NJR, but is a separate programme managed by NHS Digital (now part of NHS England (NHSE)). The NJR accesses the cumulative national PROMs data retrospectively annually through an application to NHSE’s Data Access Request Service.
Unfortunately, due to circumstances beyond our control, we have not been able to secure access to these datasets this year. NHSE report that “In 2021 significant changes were made to the processing of Hospital Episode Statistics (HES) data and its associated data fields which are used to link the PROMs-HES data. Redevelopment of an updated linkage process between these data are still outstanding with no definitive date for completion at this present time. This has unfortunately resulted in a pause in the current publication reporting series for PROMs at this time.”
NHSE are currently working to identify solutions and once this has been resolved we hope to be able to readdress the reporting of PROMs in the next NJR Annual Report. We are disappointed that we are unable to proceed with this important work to consider these outcomes in respect of implant performance.
Shoulder PROMs collection is overseen directly by the NJR within our geographical areas of operation and so is unaffected by these issues. Please see the shoulder section of the report for more information about shoulder PROMs.
https://www.nice.org.uk/guidance/ta304/chapter/5-Implementation
Hunt LP, Whitehouse MR, Beswick A, Porter ML, Howard P, Blom AW; Implications of Introducing New Technology: Comparative Survivorship Modelling of Metal-on-Metal Hip Replacements and Contemporary Alternatives in the National Joint Registry. J Bone Joint Surg Am. 2018 Feb 7;100(3):189-196.