Executive Summary

Shoulder replacements have been collected by the NJR since 2012 and there are now 99,184 primary shoulder replacements available for analysis. Due to the complexity of some of the constructs used, some cases could not initially be classified with certainty since the description of the operation entered by the surgeon did not always coincide accurately with the components entered. Extensive additional auditing of the data over the last few years has allowed many of these uncertainties to be resolved and as a result these ‘unconfirmed’ entries have been greatly reduced. Nevertheless, there remain 8.1% of shoulder replacements which are ‘unconfirmed’ and these are predominantly reverse polarity shoulder replacements.

A substantial change in practice away from conventional total shoulder replacement (TSR) and proximal humeral hemiarthroplasty (HHA) in favour of reverse polarity total shoulder replacement (RTSR) has been observed over the last decade accompanied by substantial change in the indications for which RTSR is used by many surgeons. This change is seen to continue in these latest data (Table 3.S1) with stemless RTSR and stemmed RTSR both increasing in number. Stemless TSR is also seen to have increased albeit by much smaller numbers, All other types of HHA and TSR can be seen to have progressively decreased in usage (Figure 3.S3).

Confirmed RTSR constituted 74.4% of all shoulder replacements in 2025 and this high level of use points to growing confidence in this type of procedure as well as substantial change in the indications accepted by surgeons. These changes in practice have the capacity to influence the relative outcomes of different classes and brands of shoulder replacement and these analyses should be considered with those potential confounders in mind.

While this changing practice may affect the analysis by brand and class, the analysis of sub-groups may still offer valuable and interesting results. For example, the revision rates for shoulder replacements in patients under the age of 55 remain high. The cumulative revision rate (CRR) is 10.5% in males and 9.2% in females at five years, rising to 16.8% in males and 14.6% in females at ten years (Table 3.S7). This is more than twice the figure for hips in the same age-group and appears to reverse the modest difference between males and females. This provides important information that patients require if they are to provide properly informed consent for their operation.

As the shoulder database grows, its reliability and value has increased, and we are now able to provide very useful medium and longer-term information on survival by brand and by types of shoulder replacement. For elective indications, stemmed RTSR is seen to have the lowest overall revision rate at 13 years but up to three years post-op the revision rate is higher than for stemmed conventional TSR (Figure 3.S8). The extent to which such variations may be affected by the changes in usage and indications is not yet clear. Revision rates are provided for individual constructs in Table 3.S10. These provide overall revision rates with those constructs, and considerable caution should be exercised in interpreting these results as the indications and demographics may vary widely. For most constructs the numbers are such that further stratification by indication and demographics would currently result in very wide confidence intervals.

Revision rate is of course only one outcome measure, and it is useful, where available, to relate this to other outcome metrics such as Patient Reported Outcome Measures (PROMs). At present the proportion of cases where we have valid PROMs for shoulder cases remains disappointingly low. The demographics of the cases where valid pre-and post-operative PROMs are available are not entirely congruent with those where the PROMs are not available (Figure 3S.10); nevertheless, some useful information is to be found. For all sub-types of humeral hemiarthroplasty, the revision rates are higher than for conventional or reverse total shoulder replacement. It has been suggested that this is due, at least in part, to the operation to revise a humeral hemiarthroplasty being easier to perform. The PROMs results provide some evidence that this is not the reason for the higher revision rate because the hemiarthroplasties show less improvement in their PROMs following operation than total shoulder replacements. Further work is needed to define the utility of humeral hemiarthroplasty as the indications for those operations remain very different from those in either TSR or RTSR, although the age and sex distribution is similar. 

Analysis of the PROMs data for shoulders describes in detail not only the data where patients have responded with valid PROMs, but also the trends in completing the forms and the degree to which the data may not be representative of those who failed to return valid PROMs. With those caveats the data do show that for those completing the PROMs there is substantial improvement in both pain and function following shoulder replacement surgery.

The results show that the greatest improvement has been seen for those patients having conventional total shoulder replacement surgery with reverse shoulder replacement having nearly as good improvement. At present it remains to be seen whether, when much more comprehensive PROMs results are available, both revision rates and PROMs results support the use of certain articular brands and types of shoulder replacement or whether the two outcome measures are at odds with one another.

Substantial disruption to the provision of shoulder replacement surgery occurred with COVID and it has taken some time to recover. By 2023 more shoulder replacements were being done than before COVID and by 2024 the numbers were in line with the pre-COVID growth expectations. Prior to COVID the numbers were rising in both NHS and independent hospitals. Since COVID the majority of the increase in volume has been in the independent hospitals and this increase is proportionally higher for independently-funded cases than for NHS-funded cases in those independent units (Figure 3.S2).

Elbow replacements differ from other joints in the registry in many ways and some of these render comparisons of results with other joints somewhat unhelpful. The previously noted preponderance of elbow replacements for acute trauma in our data continues to be seen and this, along with the small number of cases and the different indications for different varieties of elbow implants (i.e. total elbow, radial head and humeral hemiarthroplasty) make it difficult to draw some general conclusions.

In contrast to other joints collected by the NJR, elbow replacements seem to be performed annually in similar numbers overall with little evidence of the steady and significant growth in numbers seen for hips, knees and shoulders. Of approximately 1,200 cases per year about half are radial head replacements, 10% are distal humeral hemiarthroplasties and nearly 36% are total elbow replacements. Since 2024, of all cases, 60% were performed for acute trauma and of those two-thirds were radial head replacements. This emphasises the relative difference in usage for the different classes of implant and consequently the analysis has been largely divided into those performed for acute trauma and those for various elective indications.

The overwhelming majority of elbow replacements are performed in NHS hospitals and while this to some extent reflects the fact that almost all acute trauma surgery is performed in NHS hospitals, in the case of elbow replacement this largely also applies to elective elbow replacement.

Several notable trends are seen in the data. There has been a steady and continued growth in the number of radial head replacements performed for trauma; at the same time there has been a slight fall in the number of radial head replacements for elective indications. A slight increase in humeral hemiarthroplasty and total elbow replacement for acute trauma has occurred while for elective cases there has been a small drop for each of these (Figures 3.E3 and 3.E4).

Most surgeons who perform elbow replacements can be seen to do quite small numbers with very few doing more than one per month. This might seem to suggest elbow replacement would be a good target for centralised service provision. The large preponderance of acute trauma cases may make this a less desirable arrangement for many patients, and perhaps for surgeons, but it is regarded by some to be inefficient to have 200 hospitals performing these operations when they do only around 1,200 per year between them. Given that in other joints with more data, higher volume surgeons have tended to produce better outcomes it is likely that pressure for such centralisation will continue. There is already quite substantial variation in the number of cases performed per surgeon around the country (Tables 3.E4 (a) and (b)).

In many of the tables (e.g. Table 3.E1 and 3.E2) some cases are referred to as ‘unconfirmed’. This happens when there is a discrepancy between the description of the type of operation entered by the surgeon, and the combination of implant details entered on the registry. These numbers have fallen substantially in recent years due to the comprehensive audit undertaken and hopefully this improvement will continue in coming years.

Table 3.E6 shows revision estimates according to type of procedure. For many of these the numbers available are currently still insufficient to derive statistically robust survival estimates and these numbers are therefore represented in blue. It is notable that, performed for acute trauma, radial head replacement is seen to have a cumulative revision rate of around 2% at ten years, whereas performed for elective indications the cumulative revision rate (CRR) is approaching 6% at seven years. In contrast total elbow replacement for acute trauma has a CRR approaching 6% at only seven years, and for elective total elbow replacement the CRR is nearly 10% at seven years. Clearly these procedures are performed on quite different populations of patients.

Some of the tables show data from just the preceding three years. These are the years for which the data are the most complete so although they may give limited insight for survival analysis they do give useful information about demographics and the indications for operation. These indications appear to have remained static over that period as do the numbers of cases performed per consultant.

Comparison of different brands’ survival estimates shows some variation between brands, but for most this is relatively minor. At present the numbers mean that there are fairly wide confidence intervals, so we will need to await increased numbers and increased years of follow-up before much useful detail will hopefully emerge. Elbow replacement around the world is relatively rare, however, so this collection of data in the NJR represents probably the largest such collection in one health system, and consequently our best opportunity of obtaining useful outcomes data in the near future.

While ankle replacements are performed in very small numbers compared to hip, knee and shoulder replacements, there is nevertheless considerable increase in their use as demonstrated by the virtually linear growth year on year. This has seen numbers grow from 600 per annum to almost 1,600 per annum and while some of the early growth was due to lower compliance with entering cases at that time and some recent growth may be catch-up following COVID, the number of replacements being performed clearly suggests a growing need (Figure 3.A3). 

The small numbers do necessarily mean that the number of linked revisions will be small, and this is compounded by the finding by Jennison et al (2023) that up to a third of revision ankle joints may not be captured in the registry. In particular it seems likely that those ankle replacements revised to arthrodesis or amputation may be missed by the standard data collection processes. As ankles have only been collected in the registry for 15 years, a large proportion of the revisions will have been on cases where the primary operation occurred before the NJR started collecting ankle procedures.

The Infinity fixed-bearing implant is the market leader and established this position almost immediately when it came on the market.  It is therefore gratifying to see that the data show that it now has one of the lowest cumulative revision rates out to ten years (Figure 3.A9). 

It is notable that the various ankle components used in a single procedure are almost always from the same manufacturer, and the only use of a mixture of ankle brands recorded in the NJR occurs with Infinity and Inbone implants.  This could give rise to some confusion, but the use of these together is legitimate as they have the same manufacturer and the same articular surface design. The difference is in the nature of the bone resection required on the talar side, and although these two talar designs are presented separately they are in many ways identical. Over half of all ankle implants now being used are Infinity tibia and talus, but Infinity tibia with Inbone talus accounted for 7.9% of procedures in 2025. 

In contrast, Infinity Adaptis is considered a different implant by the NJR, having been introduced much more recently and having an element of additive technology in its manufacture despite it having similar bearing surfaces.

Numbers of cases in the registry are now sufficient to give quite narrow confidence intervals to the revision estimates, and at ten years this is now 9.5% overall. Although this is probably a modest underestimate bearing in mind the findings of Jennison et al (2023) the improvement in overall revision rate year-on-year demonstrated in Figure 3.A8 (Appendix 6) shows the seven-year cumulative revision rate (CRR) falling to approximately half its value only five years ago. The CRR for ankles is still very much higher than that seen with total hip, total knee or unicondylar knee replacements.

A greater proportion of ankle replacements are now being performed by surgeons who do more than 12 per annum. Nevertheless, there are still around 200 ankle replacements being performed each year by surgeons who do four or fewer per year (Figure 3.A4). 

During 2024, the USA’s Food and Drug Administration (FDA) issued device safety information about the Hintermann H3 ankle replacement related to a higher than expected device failure rate. Subsequent investigation by the FDA has revealed higher than expected rates of revision in several national registries and in June 2026 the FDA issued a notice that the Hintermann H3 ankle replacement has a higher than expected rate of revision and also a higher than expected rate of implant fracture. The FDA has recommended that surgeons seek alternative treatments to the Hintermann H3 for their patients.

Jennison, T., Ukoumunne, O., Lamb, S., Sharpe, I., & Goldberg, A. J. (2023). How long do ankle arthroplasties last? Bone Joint J, 105(3), 301-306.

There are now over 2.2 million knee procedures recorded in the registry, with over 1.95 million consented and validated primary knee joint replacement procedures available for analysis, representing over 15 million ‘years at risk’. There has been an annual increase in knee procedures post-COVID to a record annual high of 143,972 primary knees in 2025 (Table 3.K2). However, NHS-funded procedures performed in NHS facilities have plateaued and the main increase in volume has been in private facilities with increases in privately-funded and NHS-funded procedures.  NHS-hosted and funded procedures historically accounted for over 60% of the total knee volume but they now represent less than 50%. This continues to pose questions regarding in-house NHS capacity, facilities, staffing, and productivity.

Over the last three years, consultant surgeons have performed a median of 153 knee procedures representing an increase from 135 in the last report. The median number of knee procedures for hospitals was 879. Consultants completed a median of 130 total knee replacement (TKR) cases and 35 unicondylar cases in the past three years. Hospitals performing unicondylar replacements completed a median of 110 cases in the three years.

The majority of primary TKR cases continue to be done by surgeons performing over 49 cases per year. The majority of unicondylar knee replacement procedures were contributed by high volume consultants, i.e. those performing 25 or more procedures annually.

Across all registry knee replacement data, 56% of all patients are female and the typical patient is 70 years of age. Patients receiving unicondylar prostheses were typically six years younger (64), patients receiving patellofemoral prostheses were twelve years younger (58) and patients receiving multicompartmental prostheses were typically nine years younger (61). These differences in cohort age should be considered when reviewing revision rates, as revision rates predictably increase with younger age for all types of knee procedure. 

The most common knee procedure is a TKR which is cemented, unconstrained and fixed bearing, representing 59.7% of all knee replacements. By fixation type, over 90% of all primary knee replacements in the registry are all-cemented. There has been a decline in uncemented and hybrid TKR since the start of the registry, but the percentage of uncemented procedures has begun to increase again in the last five years. 

Unicondylar knees have now reached their highest proportion to date representing 15.8% of all knees. The best performing unicondylar brands now show revision rates of 1.85% (Persona) at five years and 4.60% (medial Physica ZUK) at ten years (Table 3.K9 (b)) and are fixed bearing designs. Cemented unicondylar primary procedures have risen to 12% of all primaries in 2025 and the largest growth in any knee procedure last year was in cemented unicondylar replacement. Last year 71% of unicompartmental procedures were fixed bearing (Table 3.K2).

The most common TKR configuration remains a cemented unconstrained (cruciate retaining) fixed bearing implant. This configuration has a revision rate of 5.22% at 20 years, which should inform our predictions of TKR survivorship in our discussions with patients. 

This year’s report again shows very detailed analysis of TKR brands by constraint, mobility and patella resurfacing status (Table 3.K9 (b)). It is very clear that the performance of some specific brand constructs is affected by these variables, whereas for others it is not. This makes it important for surgeons to be aware of the performance of the specific brand construct they use. 

A total of 5,732 known first revision knee procedures were recorded in 2025. In total there are 119,239 revision knee procedures (Table 3.K13) recorded in the registry of which 14,168 revision were re-revisions. Over 61,113 first-time revision procedures are now linked with a known primary on the registry, which is 62.4% of the total number of revision knees on record. In 2025, 83.5% of first revisions performed were linked to a primary on the registry. It is encouraging that the cumulative probability of a TKR being revised at three, five, seven and ten years has continued to reduce between 2009 and 2024 (Figure 3.K3 (b)). 

Multicompartmental knee replacements, where there is simultaneous replacement of two or more compartments of the knee with unicondylar devices, have the highest revision rates followed by patellofemoral, lateral unicondylar, medial unicondylar and then TKR (Figures 3.K4 (d) and 3.K4 (e)). However, revised patellofemoral knee replacements have the lowest risk of re-revision, followed by revised cemented unicondylar replacement (Figure 3.K6 (b)). Surgeons therefore need to weigh up the revision rate of the first procedure and the re-revision rate in considering the overall ‘arthroplasty life’ for the patient, particularly in younger patients. Revised uncemented unicondylar replacements have a higher risk of re-revision than cemented unicondylar replacements and are equivalent to estimates seen for revised TKR until eight years. 

In most instances (87.8%), the first revision procedure was a single-stage revision; in the remaining 12.2% it was part of a two-stage procedure. For the specific indication of infection, single-stage revision has now been performed twice as frequently as stage one of two stage revision in the past five years.

The chance of re-revision after first revision is approximately 15% at ten years which is five times higher than current revision rates for primary TKR. Revisions of patellofemoral and cemented unicondylar procedures have the lowest chance of re-revision. 

Early revision of a primary TKR is associated with a higher failure rate of that revision. A revision within one year of primary knee replacement has a re-revision rate of almost 10% within one year and approximately 20% by five years. This highlights the importance of getting the first procedure right. For cemented TKRs, uncemented TKRs, unicondylar and patellofemoral knee replacements, those who had their first revision within one year of the initial primary knee replacement experienced the worst re-revision estimates.

The overall cumulative percentage probability of mortality at 90 days after knee revision was lower in the cases with their primaries documented in the registry compared with the remainder (Kaplan-Meier estimates 0.94%).

The NJR has collected re-operation procedures via the RO form since 2023. This covers any unplanned return-to-theatre events that do not involve revision or exchange of joint replacement components. Reported activity has increased since the year of introduction, which currently is likely to just reflect increased awareness of the form and the need to complete it. Forms for 1,292 re-operations were received for knee procedures in 2025: 377 were for peri-prosthetic fracture, 173 were washout and implant-retention procedures (without modular exchange) for suspected or confirmed infection, and 578 were manipulation under anaesthesia (MUA). Knee surgeons need to be aware of when these forms need to be completed and as the number of these increase, more useful analysis will be possible.

For knee replacement, 471,831 linked procedures had pre- and post-operative PROMs. Primary bicondylar knee replacement showed improvement in Oxford Knee Score (OKS) from 18.5 to 35.5 (mean gain of 16.5) and EQ5D improvement from 0.393 to 0.738. Unicondylar knee replacement showed improvement in OKS from 20.7 to 38.5 (mean gain of 17.2) and EQ5D improvement from 0.451 to 0.790. Patellofemoral knee replacement showed improvement in OKS from 20.3 to 36.3 (mean gain of 16.0). Revision knee replacement showed improvement in OKS from 15.8 to 29.0 (mean gain of 13.2). EQ-VAS scores improved in all procedures with post-operative mean scores of 73.9 for TKR, 76.8 for unicondylar replacements, 74.8 for patellofemoral and 66.5 for revisions. All knee procedures showed substantial improvement in PROMs. Patient satisfaction was also high, with 56.4% of patients undergoing primary bicondylar knee replacement, and 61.8% undergoing unicondylar knee replacement reporting the highest satisfaction category. Unicondylar knee replacements therefore achieved the highest post-operative PROMs and the highest proportion of highly satisfied patients. Primary bicondylar knee replacements started with lower pre-operative PROMS than other primary procedures. Please read our ‘PROMs Report 2026’ for a detailed analysis, available in Downloads.

Robotic usage recorded in the NJR is growing and predominantly driven by usage for knee replacements. A total of 11,641 robotic knee procedures were performed in 2025 compared with only 3,762 robotic hip procedures. The robotic knee market has representation from more providers than hips, with six systems having data recorded. The most commonly used system was Mako with 7,869 recorded procedures in 2025, followed by CORI with 1,590. CORI usage grew the most compared with 2023, approximately doubling in the two years. Velys had its first recorded usage in 2024. While these numbers are dwarfed by the number of knee procedures done without a robot, usage is growing and appears to be the dominant technique in selected centres.

Primary TKR and unicondylar numbers have continued to rise post-COVID with the highest volumes in the registry reported in 2025. The predicted large increase in revision TKR burden has not yet been shown in reported revision numbers. The analyses in this report provide surgeons with evidence to support many TKR implant options with revision rates of under 2% at ten years and under 4% at 15 years, and medial unicondylar implant options with revision rates under 2% at five years and under 5% at ten years.

Primary hip replacement using unipolar implants entered into the registry reached 100,000 cases per annum in 2023 and has increased since so that in 2025 it reached a per annum record of 100,000 cases in 2023, and has increased to 113,472 in 2025. So the overall number of primary hip replacements entered into the registry was another record high this year, for the fourth year running. While the overwhelming majority of these were unipolar implants, the number of dual mobility hips has grown steadily and has almost doubled since 2021. Some cases receiving a dual mobility hip are of a more complex nature than the ‘average hip’ but this does not seem to be the same for all dual mobility cases. For this reason, the generally higher revision rate of dual mobility hip replacements demands ongoing vigilance and study. Resurfacing hip procedures are now done in very small numbers in comparison to standard unipolar replacements (Figure 3.H1 (b)) although the numbers have increased somewhat since COVID.

Over the last three years the median number of hip replacement operations per consultant was 75 (IQR 4 to 234) and for hospitals the median number was 745 (IQR 340 to 1,175). Although the number of resurfacing hips is small, more than half of these are performed by surgeons who do more than 25 such cases per year. 

There has been a modest but steady decline in unipolar hip replacements performed for acute trauma since 2022. Although some hips receive dual mobility implants for acute trauma, over the same four years there does not seem to have been a significant growth in these suggesting that overall total hip replacement (THR) for acute trauma has become less popular (Figure 3.H1 (c)). This would be in keeping with the results of the HEALTH trial (2019) which failed to show a difference in risk of secondary procedures in total hip replacement compared to hemiarthroplasty for fractured neck of femur.

The growth in overall numbers of hip replacements to 20% higher than prior to COVID can be seen to have occurred due to independently-funded cases and NHS-funded cases performed in independent sector hospitals. Although NHS hospitals have recovered to pre-COVID levels of hip replacement since the pandemic, the number of primary hip replacements performed in NHS hospitals has not increased since 2014 (Figure 3.H1 (d)).

Cemented hip replacement continues to decline in number while uncemented hips, following a period of decline, have increased over the last five years, becoming the most common fixation method once again in 2025 (Figure 3.H2 (a)). There has been a slight fall in hybrid hip numbers after a steady increase in each year since 2008. In the past, the overall revision rate for uncemented hips was higher than for hybrid hips, which were in turn higher than for cemented hips. This is still seen in some of the data, for example in Table 3.H5 which shows the data broken down by bearing surface as well as fixation method. We know from the data in the ten-year rolling window used to compare surgeons and units that the data for the last ten years now shows very little difference in the ten-year revision rates between cemented, hybrid and uncemented hips (PTIR data). Variability in the proportion of cases in these fixation groups at different times over the last 22 years may therefore be a contributory factor in some of the overall revision rate differences depicted in Table 3.H5.

For all varieties of hip fixation, the selected bearing surface has shown a gradual change over the years away from metal-on-polyethylene towards ceramic-on-polyethylene. This is seen to a greater extent in uncemented and hybrid hip replacements than in cemented and reverse hybrid. The dramatic fluctuations over time seen for uncemented hip bearing choice, which were to some extent related to the rise and fall of metal-on-metal bearings, seem now to have settled to a steady increase in ceramic-on-polyethylene (Figure 3.H3 (a) to (d)). These variations over time with bearing surfaces differing relatively in the four fixation groups mean that comparisons in overall revision rates between the sub-groups need careful interpretation. The different fixation groups have been stratified by implant femoral head diameter as well as by bearing surface material in Figure 3.H3 (e) in order to minimise the effect of one of the additional variables.

There are now 55,104 first revision operations which can be linked to a primary hip replacement which has been entered into the registry. These provide an excellent body of evidence to study the changes of revision rates over time. Figures 3.H4 (a) to (c) show the overall revision rates changing with time plotted by year of implantation and because metal-on-metal (MoM) bearings were known to have a significant impact, the results are also plotted with these excluded. These show that there has been a steady improvement in revision rates year-on-year and at all time points, perhaps most clearly seen for the ten-year and 13-year data. With MoM hips excluded (Figure 3.H4 (c)) the ten-year revision rates have improved from 3.4% to 2.5% since ten years ago. The 13-year revision rates have fallen from 5% to 3.6% in the last nine years.

It seems likely that this steady improvement in revision rates has occurred because clinicians are receiving direct feedback about their own revision rates and that they have modified and adapted their practice as a result, to the benefit of patients.

It remains uncertain why revision rates were apparently increasing prior to 2008 even with MoM hip bearings excluded from the data, but this may be related to poorer data quality and completeness in the first five years of the registry.

Table 3.H3 shows a breakdown of hip replacements by fixation and bearing surface, demonstrating large differences in age and sex between some of the groups. As we know that age and sex are important factors determining revision rates, it may be that some of the differences shown in the table have been affected by these demographic differences. For example, ceramic-on-ceramic bearings and resurfacing have been used in young patients. Resurfacing of hips is a procedure used much more frequently in males, which is perhaps unsurprising since the revision rates were widely reported to be much higher in female patients over a decade ago.

Over 80% of patients receiving a primary hip replacement are seen to be in ASA grade 1 or 2, regardless of sex, and very small numbers have an ASA grade higher than 3 (Table 3.H4). This indicates that hip replacements are largely being performed on people who are otherwise in reasonable health.

The substantial variations in revision rates according to age and sex have been recognised for a long time. Figure 3.H9 (a) shows these differences are not diminishing with time. However, when the metal-on-metal cases are removed (Figure 3.H9 (b)) the higher revision rates in younger females disappear, and females of all ages are seen to have lower revision rates than males.

Analysis of the revision estimates for all major stem and cup combinations is shown in Tables 3.H7 (a) and (b), the latter also showing the femoral head brand. Tables 3.H8 (a) and (b) show similar information including bearing surface materials. As these data are not adjusted for age or sex it is important to check the additional data presented to ascertain whether a particular implant combination is used in a different patient group to the norm.

A notable finding is that there are 23 combinations of stem/cup/head/bearing/surface which show revision estimates of less than 2% at ten years. These include choices from each of the fixation groups. Currently the NICE guidance about an ‘acceptable’ hip implant has stood at a 5% ten-year revision rate since 2014. Similarly, the newly introduced ODEP 10A* rating for either a hip stem or a cup is based on a 4.2% cumulative revision estimate at ten years. It seems clear that there are ample choices available to a surgeon which have a good chance of achieving a much lower revision rate than has been set by either of these standards. It is therefore important that surgeons scrutinise carefully the revision estimates of the precise combinations of implant they wish to use from these data in the report. Then they should be able to offer their patients the best chance of avoiding a further operation.

It is well known that some indications for hip replacement revision occur earlier than others after the primary operation and detailed information about this is presented in Table 3.H10. Derived from over 55,000 first revision cases of primary hips in the registry, the table illustrates the timeframe for indications occurring early such as dislocation, infection and periprosthetic fracture, and those occurring later such as aseptic loosening and bone lysis.

There are 165,602 hip revisions recorded in the registry and while 55,104 of these are first revisions that can be linked to a primary, there are 98,800 revision procedures which cannot be linked to a primary. The overall number of revision operations increased from 2003-2012, but then declined steadily. There has been a slight increase in the last five years but the extent to which this may be related to cases delayed by COVID is not clear (Table 3.H14).

It is interesting that more than half of over 19,000 revisions for infection had this done as a single-stage operation. Overall, the largest indication for hip revision in the whole registry is aseptic loosening (39.8%) which falls to 31% of all revisions done in the last five years, while the proportion performed for infection has risen from 16.2% to 22.9% in the same time periods. It is important to note that the changes in proportions of indications for revision operations does not necessarily reflect an absolute increase or decrease in revision for any individual indication for revision.

The proportion of eligible procedures contributing linked PROMs for hips (i.e. a 6-month post-operative PROMs form following a pre-operative form for the same patient) increased from 45.9% in 2009 to 78.5% in 2012. Completion remained broadly stable through to 2019, before rapidly declining to just 18.1% in 2024.

Patients reported substantial improvements in hip pain and function following surgery across all procedure types. For primary stemmed total hip replacement, the mean Oxford Hip Score (OHS) increased from 17.2 before surgery to 39.6 six months after surgery (mean gain of 21.8) and EQ5D improvement from 0.324 to 0.792. Hip resurfacing demonstrated similarly favourable outcomes, with the mean OHS increasing from 21.8 to 41.9 (mean gain of 19.6) and EQ5D improvement from 0.453 to 0.857. Patients undergoing revision hip replacement also experienced clinically important improvements, with mean OHS increasing from 19.9 to 33.9 (mean gain of 13.2) and EQ5D improvement from 0.367 to 0.674. Although revision procedures achieved lower post-operative scores than primary procedures, all procedure groups demonstrated substantial improvements in patient-reported pain and function.

Patient-rated health scores (EQ-VAS) also improved with post-operative mean scores of 76.7 for primary stemmed hip replacement, 80.7 for resurfacing and 70.3 for revision hip replacement. Patient satisfaction was also high, with 59.3% of patients undergoing primary stemmed hip replacement and 58.1% undergoing hip resurfacing reporting the highest satisfaction category, compared with 42.0% of patients undergoing revision hip replacement.

While with some of these PROMs the highest six-month levels were achieved in resurfacing hip procedures, the primary stemmed THR group started with lower values than the resurfacing group and the gain demonstrated by the stemmed primary THR group was slightly higher for all three PROMs analysed. Please read our ‘PROMs Report 2026’ for a detailed analysis, available in Downloads.

Robotic surgery has been adopted in fewer cases for hip replacement than for knee replacement, and adoption has been much slower in the NJR catchment than in some other countries. Nevertheless, there has been significant growth in robot-assisted THR in the last few years, and NICE has conducted an Early Value Assessment on the use of robots in orthopaedic surgery.

There has been a 32% increase in robot-assisted THR from 2,854 in 2023 to 3,762 in 2025 and these numbers are very unevenly distributed, demonstrating that some centres are using a robot as a principal mode of performing THR. The most used robot in THR is the Mako but the CORI has also shown significant growth in its use over three years.

Since 2023 the NJR has been collecting information about clinically important returns to theatre which do not meet our definition of a ‘revision’. This is likely to prove very important in the analysis of hip replacement outcomes both in a general sense and regarding the relative outcome of different hip brands. In particular, the collection of data on hip reoperations where a periprosthetic fracture has been treated by internal fixation is expected to confirm that a large proportion of cases with periprosthetic fracture have been ‘missed’ in the revision data. It is also anticipated that once these ‘non-revision’ failures are being captured comprehensively, it will be clear to what extent periprosthetic fractures of some varieties of hip implant may be being preferentially treated by internal fixation.

Bhandari M et al.; Total Hip Arthroplasty or Hemiarthroplasty for Hip Fracture. Value Health. N Engl J Med 2019; 381:2199-2208.
 

Mr Andrew Porteous - Chair, Editorial Committee
Mr Tim Wilton - NJR Medical Director

The NJR continues to engage with surgeons and to explain our methodology and key messages from analysis of the data, supporting surgeons in making wise implant choices, and monitoring and improving their results.  We do this by attendance at many orthopaedic specialist society meetings and the British Orthopaedic Association (BOA) Annual Congress, as well as via our annual report, website and direct communication with surgeons. 

The number of cases submitted to the registry has grown and is now at higher annual levels than achieved prior to the COVID pandemic. The increase in procedures varies slightly between joints and whether they were performed within the NHS or independent sector, but hip and knee replacement procedures have increased in each of the past four years to their highest levels since the start of the registry. Approximately 1.8 million hip replacements and 1.95 million knee replacements are now available for analysis. Further detail is available in the joint-specific chapters of the report. Small changes in data numbers can occur each year due to ‘data lag’, e.g. when a hospital uploads data after the reporting cut-off these are then captured in the next data analysis period. 

This year’s report has seen a reduction in the length of each chapter to focus on more clinically relevant data and analysis. The registry continues to do the same analyses as previous years but some tables and graphs can now be found in ‘Appendix 6’. Readers are also encouraged to visit the NJR Centre website, where additional data about the NJR, which may not appear in the annual report, are available. 

Enhanced analysis by indication (either trauma or elective) now provides more clinically relevant information on shoulder and elbow implants.  Dished, medial stabilised or medial pivot designs have become more popular in knee replacement, and some manufacturers have re-classified some of their knee implants in terms of their ‘cruciate stability’. This can mean that the results reported for the brand sub-types may be altered compared to previous years but should be more accurate and reliable in future as a result of the re-classification. Ongoing refinement of the implant classification system allows more granular reporting and analysis, especially for those implants where there are multiple variants within a single brand. These improvements will continue to make it increasingly difficult for a poorly performing implant or construct to be camouflaged or hidden within a brand that generally has good performance. Similarly with more granular detail in the analysis we will be able to identify a smaller variant within a brand where the variant performs much better than the brand as a whole.

Revision is a clearly defined and important endpoint for analysis. Collecting data on all re-operations, and qualitative data in the form of Patient Reported Outcome Measures (PROMs) is an important focus for the NJR as these metrics are important to both surgeons and patients. 

National PROMs data for hip and knee replacement surgery is not collected by the NJR. Historically the NJR have received retrospective datasets of PROMs from NHS England's Data Access Request Service (DARS), enabling PROMs analyses to be incorporated into previous Annual Reports. Shoulder PROMs are collected by the NJR across our geographical areas of operation. Following a prolonged interruption to data access, the NJR has now received an updated national PROMs dataset although the dataset was received too late in the Annual Report production cycle to undertake the full range of analyses previously presented by the NJR. The NJR intends to produce the full set of PROMs analyses, including implant-level reporting, in future Annual Reports where data availability and production timelines allow. A detailed analysis of the data quality, collection rates and outcome of PROMs data for hip and knee replacements is reported in a new ‘PROMs Report 2026’ available in Downloads. 

National PROMs data quality shows completion rates of 50% for hips and 52.7% for knees in 2012, that has fallen to 15% and 16% respectively in 2024. Revision hip and knee procedures showed worse PROMs completion rates. It is important to realise that although the proportion of valid pre- and post-operative PROMs is disappointing, this is to some extent a reflection of how the data have been collected. Where the pre-operative PROMs were completed but the operation was delayed by more than three months those data were not deemed valid. Similarly, where the gap between operation and post-operative PROMs collection exceeded the allowance, those data have been deemed invalid. Obviously, major changes to the system for PROMs collection will be required to improve this, and we are working closely with NHSE to find a better way forward. There were, however, still 435,138 linked hip replacements and 471,831 linked knee replacements with pre- and post-operative data for analysis. For all categories of hip and knee replacement procedures, these show substantial improvement in joint specific PROMs scores and health-related quality of life.  

Since 2023, the NJR has broadened its surveillance capability through the collection of re-operation procedures, capturing clinically important return-to-theatre events that do not involve addition, removal or exchange of joint replacement components. This is collected on the ‘Reoperations other than revision’ (RO) form. Submissions using this form have grown substantially from 1,885 in 2023 to 4,076 in 2025. Two-thirds of these forms relate to hips and the majority of the remainder relate to knees. Peri-prosthetic fracture is the most common indication for re-operation overall and 74% of these occur in hips. This reflects the increasing clinical importance of fragility fracture, trauma around implants, and complex reconstruction in an ageing joint replacement population. Manipulation under anaesthesia (MUA) is the commonest re-operation recorded for knees. Future work for the NJR’s Data Quality Audit programme will be the benchmarking of provider completeness of re-operation reporting, improving consistency of procedure coding, validating capture of non-theatre-based procedures and supporting hospitals to strengthen local submission processes.

Robot-assisted primary joint replacement continues to show year-on-year growth, with a 68% increase in robotically-assisted procedures over the last three years. Usage is predominantly in hip and knee replacement and volumes remain small compared with non-robotic procedures. In 2025, 3,762 robotic hip procedures and 11,641 robotic knee procedures were recorded.  In April 2025, the National Institute for Health and Care Excellence (NICE) published Early Value Assessment guidance on robot-assisted orthopaedic surgery which concluded that robotic systems show clear gains in surgical precision, but that current evidence does not yet demonstrate a clear improvement in revision rates or patient-reported outcomes compared with conventional surgery. 

The NJR has been involved with the manufacturers in agreeing a minimum data set that could be included in registry data to allow meaningful comparison of the outcomes both of robotic and conventional surgery and between the robots themselves. Critically, this involves collection of the software and hardware versions rather than simply whether a robot is used or not. 

The NJR is identified as a key partner in collecting the longitudinal data needed to evaluate clinical effectiveness, safety and value. Our data suggest that robotic surgery is moving from early adoption toward mainstream use in selected centres. The NJR’s role in supporting registry-based surveillance will be critical in determining whether greater surgical precision translates into measurable long-term patient benefit or whether there is evidence of increased or novel complications related to robot use.

The NJR continues to see an increasing number of high-quality scientific studies using NJR data or performed in collaboration with the NJR. These publications are listed in the research section of the annual report with some papers each year being selected for more in-depth coverage in the report or podium presentation during the NJR session at the BOA.

We urge surgeons to examine the updated revision rates shown in this report and scrutinise the tables to make sure they are viewing results of the specific variants they use, as this will supplement other sources of information such as ODEP ratings. It should be remembered that ODEP ratings are voluntary and usually only updated every two to three years and therefore may not capture all implant constructs or the most recent data. Multiple hip and total knee implant options now have 10-year revision rates that are less than half the ODEP 10A* benchmark rate (although the ODEP target rate does refer to the upper confidence limit of the cohort).

The NJR dataset continues to expand rapidly and the size of datasets for hip and knee replacement now allow very detailed analysis and granularity. Increasing numbers of shoulders, elbows and ankles have provided the opportunity for more detailed analysis in this report and going forward. New information regarding PROMs, robots, re-operations and analysis by elective or trauma indication should continue to increase the clinical relevance of the annual report for surgeons and allow them to make better informed decisions for the ultimate benefit or their patients.

Acknowledgements

The NJR continues to work collaboratively with our many stakeholders; the most important of these being our surgeons and patients. We would like to thank the surgeons for completing the data input forms and the patients for allowing us to use their data.

The NJR operational collaboration is a huge team effort. This year saw the retirement of Elaine Young as NJR Director of Operations, after 19 years of service to the NJR, and we are hugely appreciative of her valuable contribution. 

Many thanks also to the following without which the NJR could not function:

All members of the NJR Board and members of the NJR committees:
Executive 
Data Quality 
Editorial
Implant Scrutiny 
Medical Advisory
Regional Clinical Coordinators 
Research
Surgical Performance

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
Government of Jersey
Independent Healthcare Providers Network 
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)
Computer Assisted Orthopaedic Surgery (CAOS)
British Orthopaedic Directors Society (BODS)
Orthopaedic Trauma Society (OTS)
British Orthopaedic Oncology Society (BOOS)
British Orthopaedic Trainee Association (BOTA)

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; and Deirdra Taylor, Associate Director of Communication and Stakeholder Engagement for getting the final report into shape.