Information governance and patient confidentiality

Data are collected via a secure web-based data entry application, then stored and processed in the NEC Software Solutions (NEC) data centre. NEC is ISO 27001 and ISO 9001 accredited and compliant with the NHS Data Security and Protection Toolkit. Data linkage to other datasets is approved by the Health Research Authority under Section 251 of the NHS Act 2006. Please visit https://www.hra.nhs.uk/about-us/committees-and-services/confidentiality-advisory-group.

Missing data

It is expected that neither the registry nor a local hospital’s system alone could be regarded as a definitive list of joint replacements, however the union of both registry and local hospital data can be considered the gold standard from which to calculate voluntary unprompted compliance at upload. This figure is important for healthcare providers as a measure of compliance with data entry processes but does not represent the final data completeness of records in the registry.

The effect of missing data on the statistical analysis of a dataset is well documented. Data which is systematically missing (Missing Not at Random) has the potential to induce bias i.e. to distort the truth. This is why compliance of reporting data to the registry by a specific surgeon or hospital is essential to the quality assurance process of surgeons and hospitals.

Analysis of data which are missing in either a random (Missing Completely At Random) fashion or random within known strata (Missing At Random), e.g. method of fixation, is known to yield unbiased results. We believe that a coordinated systematic agreement of individuals across the registry to under-report the failure of a specific implant is highly unlikely. Nevertheless, we believe if this did happen the issue would be identified and corrected by the NJR’s data quality audit process. The low revision rates of some replacements also make it difficult to predict which is likely to fail. Therefore, planning to omit selected primary joint replacements which are anticipated to fail within ten years following surgery would be unlikely to succeed. Increased centralisation of some revision joint replacement, by specialist revision surgeons, also means there is little motivation to omit revisions, which would largely have been primary cases of another surgeon or another hospital.

We believe that missing data within the registry can be considered missing completely at random. We propose that this missing data mechanism will ensure that the quality assurance process of implants and procedures entered into the registry is statistically unbiased.

Patient-level data linkage

Documentation of implant survivorship and mortality requires linkage of person-level identifiers in order to identify primary and revision procedures and mortality events for the same individual.

Starting with all NJR-sourced records, some were excluded because no suitable person-level identifier was found. Full details of the inclusion and exclusion criteria can be seen at the beginning of our analysis. Cases from Northern Ireland and Guernsey were also excluded because of unresolved issues around tracing mortality; and cases from the Isle of Man were also excluded due to the inability to audit them against local hospital data. Patients with longer follow-up may be less representative of the whole cohort of patients undergoing primary joint replacement than those patients with shorter follow-up, due to difficulties with data linkage and differential rates of reporting over time.

Linkage between primaries and any associated revisions (the ‘linked files’)

Implant survivorship is first described with respect to the lifetime of the primary joint. Where volumes allow, we also provide an overview of further revisions following a first revision procedure.

The unit of observation for all sets of survivorship analysis has been taken as the individual primary joint replacement. A patient with left and right replacements of a particular type, therefore, will have two entries, and an assumption is made that the survivorship of a replacement on one side is independent of the other. In practice, this would be difficult to validate, particularly given that some patients will have had primary replacements of other joints that were not recorded in the registry. Established risk factors, such as age, are recorded at the time of primary operation and will therefore be different for the two procedures unless the two operations are performed on the same date.

A revision is defined as any operation where one or more components are added to, removed from or modified in a joint replacement, or if a Debridement And Implant Retention (DAIR) with or without modular exchange is performed. Capturing DAIR with or without modular exchange commenced with the introduction of MDSv7 (June 2018). Prior to this, DAIR with modular exchange was included as a single-stage revision, but DAIR without modular exchange was not captured. Within the report each of these procedure types is included in the analyses as a revision episode. This is distinct from the analyses in the surgeon, hospital, and implant performance workstreams where DAIR without modular exchange is not currently included as a revision outcome. From MDSv8, DAIR without modular exchange is considered a reoperation procedure.

Analytical methods and terminology

The report uses a variety of statistical methods to reflect the diversity and range of performance within joint replacement. Analyses are tailored to ensure results are reported in units that can be easily interpreted. Here we define important concepts which underpin the analyses.

3.H.1. Overview of primary hip replacement surgery

In this report we address revision and mortality outcomes for all primary hip operations performed between 1 April 2003 and 31 December 2024, and submitted before 1 March 2025. Patients operated on at the commencement of the collection of data in the registry therefore had a potential 21.75 years of follow-up. This year, follow-up is reported at a maximum of 20 years in the tables, although beyond 15 years the numbers at risk are particularly low in some categories.

Figure 3.H1 (a) describes the data cleaning methodology applied to produce the total of 1,682,998 primary hip procedures included in the analyses presented in this report.

Figure 3.H1 (a)

Hip cohort flow diagram.

Over the lifetime of the registry, the 1,682,998 primary hip replacement procedures contributing to our revision analyses were carried out by a total of 4,342 unique consultant surgeons working across 496 hospitals. Over the last three years (1 January 2022 to 31 December 2024), 332,505 primary hip procedures (representing 19.8% of the current registry volume) were performed by 2,273 unique consultant surgeons working across 420 hospitals.

Looking at caseload over this three-year period, the median number of primary procedures per consultant surgeon was 69 (interquartile range (IQR) 3 to 225) and the median number of procedures per hospital was 689 (IQR 293 to 1,120). A proportion of surgeons will have commenced practice as a consultant during this period, some may have retired, and some surgeons may have periods of surgical inactivity within the time of coverage of the registry, therefore their apparent caseload would be lower.

The majority of primary hip procedures were carried out on females (females 59.9%; males 40.1%). The median age at primary operation was 69 (IQR 61 to 76) years. Osteoarthritis was given as a documented indication for surgery in 1,535,430 cases (91.2% of the cohort) and was the sole indication given in 1,479,386 (87.9%) primary hip replacements.

Table 3.H1 shows the breakdown of cases by the method of fixation and within each fixation sub-group, by bearing surfaces. Bearing surface combinations are reported as a separate group where there were 250 or more cases, unless there was only one type of bearing surface combination with a group size of fewer than 250. The most commonly used operation type over the life of the registry (2003 to present) remains as cemented metal-on-polyethylene (83.1% of all cemented primaries, 23.2% of all primaries). Dual mobility bearings are described either as dual mobility, to contrast to standard unipolar bearings, or where numbers allow, are categorised by the material of each part of the bearing surface (e.g. metal-on-polyethylene-on-metal (MoPoM) and ceramic-on-polyethylene-on-metal (CoPoM)). The numbers of other combinations of dual mobility (such as ceramic-on-polyethylene-on-ceramic (CoPoC)) were too small to include as separate groups this year.

Table 3.H1

Number and percentage of primary hip replacements by fixation and bearing.

Figure 3.H1 (b) and Figure 3.H1 (c) show the yearly number of primary total hip replacements performed for elective and acute trauma indications respectively. Elective procedures have been stratified by unipolar, resurfacing and dual mobility total hip replacements. Acute trauma procedures have been stratified by unipolar and dual mobility total hip replacements. Please note the difference in scale of the y-axis between each sub-plot.

Figure 3.H1 (b)

Frequency of primary hip replacements within elective cases stratified by procedure type, bars stacked by volume per consultant per year. Graphs by confirmed procedure type.

Figure 3.H1 (c)

Frequency of primary hip replacements within acute trauma cases stratified by procedure type, bars stacked by volume per consultant per year. Graphs by confirmed procedure type.

Each bar is further stratified by the volume of procedures that the consultant conducted in that year across both elective and acute trauma settings i.e. if a surgeon performed 25 elective unipolar THR procedures and 25 acute trauma unipolar procedures their annual total volume would be 50 procedures. Those 50 procedures would contribute to the black sub-division in both elective and acute trauma figures.

Figure 3.H1 (b) shows the annual rates of elective unipolar THR increasing, (with the exception of 2020 due to the COVID pandemic with rates partially recovered in 2021 and fully recovered by 2022), with the majority of additional procedures contributed by higher-volume surgeons i.e. those performing more than 49 hip procedures a year. In the acute trauma setting (Figure 3.H1 (c)) there was a rapid expansion of unipolar THRs recorded in the registry from 2011 until 2018, with a plateau in 2019 and then lower rates during the COVID pandemic, which have persisted but are partially compensated for by the volume of dual mobility THRs being performed for trauma.

Figure 3.H1 (b) also shows that after declining substantially in popularity, resurfacing has remained relatively stable over the past five years, with a slight increase in absolute numbers in 2021 and 2022. In 2024 over half of the resurfacing procedures were performed by consultants who used it in more than 25 cases per year.

Figure 3.H1 (b) and Figure 3.H1 (c) also illustrate the emerging use of dual mobility THR in the elective and acute trauma settings. Prior to 2013, dual mobility THR was relatively rare, but since 2013 its use has increased in both settings, other than in 2020 where COVID had an impact on case numbers, and it is now more common than hip resurfacing. Over half of dual mobility operations are performed by consultants who conduct 13 or more elective dual mobility hip replacements per year (seven or more for trauma cases), however, a greater proportion of dual mobility THRs are performed by lower-volume surgeons than other types of THR, in both the elective and acute trauma setting.

Figure 3.H1 (d) describes the funding status and organisation-type (based on organisation-type in 2024) of primary hip procedures collected by the NJR. The figure shows a steady increase in the number of THRs that were NHS-funded and performed in NHS hospitals from the beginning of the registry until 2014. After this time, this number plateaued up until 2019 and then reduced substantially due to the impact of COVID. The growth in the total number of THRs performed from 2014 to 2019 was largely driven by growth in the number of NHS-funded procedures being performed in independent hospitals. Although the total number of THRs performed since 2022 have recovered to exceed 2019 levels, the recovery of NHS-funded procedures being performed in NHS hospitals is only partial with an increase in the number of NHS-funded procedures performed in independent hospitals and independently-funded procedures performed in independent hospitals accounting for the overall volume recovery.

Figure 3.H1 (d)

Frequency of elective primary hip replacements by funding status and organisation type, per year.

Table 3.H2 shows the annual rates by fixation and bearing groups for each year for primary hip replacements. Hybrid fixation is the most common, accounting for 42.5% of all primary hip replacements undertaken in 2024. The percentage share of hybrid implants used has approximately tripled between 2008 and 2024, whilst the proportion of all hips that are cemented has more than halved, to 14.4% over the same period.

Table 3.H2

Percentage of primary hip replacements by fixation, bearing and year.

The percentage of uncemented implants used increased from 18% to 44% in the first ten years of the registry, but then steadily declined to 34.9% over the next eight years, before plateauing and then rising slightly in the latest figures (Figure 3.H2 (a)). Ceramic-on-polyethylene hybrid THR was the most common type in 2024, being used in 27.2% of cases, but ceramic-on-polyethylene uncemented THR is nearly as popular, accounting for 23.8% of cases.

Figure 3.H2 (a)

Primary hip type percentages by year of replacement.

Figure 3.H2 (b)

Primary hip type percentages by year of replacement, with dual mobility as a separate category.

Figures 3.H3 (a) to (d) illustrate the temporal changes in the bearing surface combinations used with the type of total hip replacement fixation. Groups that contain more than 500 procedures are plotted separately. Since 2012 there has been a steady increase in the use of ceramic-on-polyethylene bearings. The greatest variation in bearing use over time is noted in the uncemented fixation group.

Figure 3.H3 (a)

Cemented primary hip replacement bearing surface by year.

Figure 3.H3 (b)

Uncemented primary hip replacement bearing surface by year.

Figure 3.H3 (c)

Hybrid primary hip replacement bearing surface by year.

Figure 3.H3 (d)

Reverse hybrid primary hip replacement bearing surface by year.

Figure 3.H3 (e) illustrates the temporal changes in common head sizes, by method of fixation and bearing type in primary unipolar total hip replacement. In 2003, the vast majority of hip replacements utilised heads of 28mm or smaller, across all fixation methods. Since 2003, a progressive shift away from small (22.25mm or 26mm) heads in cemented hip replacements to larger head sizes (>28mm) with alternative fixation methods (uncemented or hybrid) has been observed.

Figure 3.H3 (e)

Trends in fixation, bearing and head size in primary unipolar total hip replacement by year.

In 2024, as in 2023, the three most common head sizes are 32mm (1st), 36mm (2nd) and 28mm (3rd), with 22.25mm and 26mm rarely being used. Only 12 cases of 26mm head usage were recorded for 2024. The use of ceramic-on-ceramic bearings across all head sizes, but most notably 36mm, has declined since 2011. This decline, conversely, corresponds with an increase in ceramic-on-polyethylene bearings with 32mm heads. The choice of bearing, head size and fixation method was much more heterogeneous in 2024 compared to 2003. The dominant choices in 2024 were 32mm and 36mm ceramic-on-polyethylene bearings.

Table 3.H3 provides a breakdown by fixation type and bearing surface, describing the age and sex profile of recipients of primary hip replacements. Patients receiving resurfacing and ceramic-on-ceramic bearings tended to be younger and those receiving metal-on-polyethylene-on-metal dual mobility bearings tended to be older than those in the other groups. Those receiving resurfacings were more likely to be younger males.

Table 3.H3

Age at primary hip replacement by fixation and bearing.

Table 3.H4 shows the American Society of Anesthesiologists (ASA) grade and indication for primary hip replacement by sex. A greater number of females than males undergo primary hip replacement and two-thirds of patients are ASA grade 2.

Table 3.H4

Primary hip replacement patient demographics.

Only a small number of patients with a grade greater than ASA 3 undergo a primary hip replacement. The majority of cases are performed for osteoarthritis. A total of 1,479,386 (87.9%) primary hip replacements have been recorded in the registry where the sole indication was osteoarthritis.

3.H.2. First revisions after primary hip surgery

A total of 50,946 first revisions of a hip replacement have been linked to a previous primary hip replacement recorded in the registry between 2003 and 2024. Figures 3.H4 (a) and (b) illustrate temporal changes in the overall revision estimates using Kaplan-Meier estimates; procedures have been grouped by the year of the primary operation. Figure 3.H4 (a) plots each Kaplan-Meier survival curve with a common origin, i.e. time zero is equal to the year of operation. This illustrates that revision estimates increased between 2003 and 2007/8 and then declined between 2007/8 and 2024.

Figure 3.H4 (a)

KM estimates of cumulative revision by year, in primary hip replacements.

Figure 3.H4 (b)

KM estimates of cumulative revision by year, in primary hip replacements plotted by year of primary.

Figure 3.H4 (b) shows the same curves plotted against calendar time, where the origin of each curve is the year of operation. In addition, we have highlighted the revision rate at 1, 3, 5, 7, 10, 13, 15, 17 and 20 years. Figure 3.H4 (b) separates each year, enabling changes in revision estimates over time to be clearly identified. If revision surgery and timing of revision surgery were static across time, it would be expected that all the revision curves would be the same shape and equally spaced; departures from this indicate a change in the number and timing of revision procedures. It is also very clear that the 3, 5, 7, 10, 13 and 15-year rate of revision increases for operations occurring between 2003 and 2007 and then reduces for operations occurring between 2008 and 2024. The early increases may be partly a result of under-reporting in the earlier years of the registry as this wasn’t mandatory at that time but is also contributed to by the usage of metal-on-metal bearings, which peaked in 2008 and then fell (see Table 3.H2).

A similar pattern, although smaller in effect, is also observed in knees. Knees were not affected by the high revision rates of metal-on-metal bearings, and thus the decreases observed since 2009 indicate a broader improvement in revision outcomes overall. It appears that this secular decline in revision estimate is still ongoing. This improvement suggests the adoption of evidence-based practice to which the NJR’s Clinician Feedback reporting has contributed. For example, for a primary hip replacement performed in 2014, the 10-year revision estimate is 2.7% (95% CI 2.6-2.8) which is below the current NICE recommended threshold of 5% at ten years (NICE, 2014). Prior to 2014, the revision threshold recommended by NICE was 10% at ten years.

Figure 3.H4 (c) illustrates the removal of all primary hips with a metal-on-metal bearing from Figure 3.H4 (b). The comparison between these charts illustrates the burden of revision which can be attributed to the revision of metal-on-metal bearings. We observe a secular decline in the rate of revision in the 3, 5, 7, 10 and 13-year revision estimates originating in 2008-2009 through to the present day which excludes the effect of metal-on-metal bearings.

Figure 3.H4 (c)

KM estimates of cumulative revision by year, in primary hip replacements (excluding metal-on-metal bearings) plotted by year of primary.

Table 3.H5 provides Kaplan-Meier estimates of the cumulative percentage probability of first revision for any cause, firstly for all cases combined and then by type of fixation and by bearing surface within each fixation group. The table shows updated estimates at 1, 3, 5, 10, 15 and 20 years from the primary operation together with 95% Confidence Intervals (95% CI). Estimates in blue italics indicate time points where 250 or fewer cases remained at risk, meaning that the estimates are less reliable and should be treated with some caution. Kaplan-Meier estimates are not shown at all when the numbers at risk fell below ten cases.

Table 3.H5

KM estimates of cumulative revision (95% CI) by fixation and bearing, in primary hip replacements. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Further revisions in the blue-italicised groups would be unlikely (due to such small numbers at risk) and, when they do occur, they may appear to have a disproportionate impact on the Kaplan-Meier estimate, i.e. the step upwards may seem disproportionately large. Furthermore, the upper 95% CI at these time points may be underestimated. Although a number of statistical methods have been proposed to deal with this, they typically give different values and as yet, there is no clear consensus for the large datasets presented here.

The revision estimate of metal-on-polyethylene-on-metal dual mobility bearings appears higher, up to five years across all fixation types, than that of most of the unipolar bearing combinations, except metal-on-metal and ceramic-on-metal. The ceramic-on-polyethylene-on-metal dual mobility bearings show lower revision estimates for cemented and uncemented THRs than the metal-on-polyethylene-on-metal combinations, but with overlapping confidence intervals. The currently relatively small numbers at risk in the dual mobility groups make it difficult to draw firm conclusions. The 1- and 3-year revision estimates for ceramic-on-ceramic resurfacing appear similar to those for metal-on-metal resurfacing which are generally higher than for other unipolar variants. The revision estimates at five and seven years appear lower, but the numbers at risk at all time points in the ceramic-on-ceramic resurfacing group are low, so this report should be interpreted with caution.

Figures 3.H5 to 3.H8 illustrate the differences between the various bearing surface sub-groups for cemented, uncemented, hybrid and reverse hybrid hips, respectively. Metal-on-metal bearings continue to perform worse than all other options regardless of fixation, apart from in cemented fixation where the results of the rarely used metal-on-metal combination are similar to metal-on-polyethylene-on-metal dual mobility. The revision estimates for ceramic-on-polyethylene bearings remain consistently low or equivalent to other well-performing alternatives across all fixation options out to 21 years and it is encouraging that these are becoming more widely used with time. Dual mobility bearings have higher early revision estimates than other options (not including metal-on-metal) for cemented and uncemented fixation. The revision estimates of uncemented metal-on-polyethylene-on-metal dual mobility bearings appear to rise markedly from nine years. There is also a small divergence in revision estimates for the metal-on-polyethylene-on-metal dual mobility bearings in hybrid THRs at a similar time point, although this is much less marked. After the first two years the revision estimates for uncemented and hybrid CoPoM dual-bearing hips appears to be following a similar trajectory to ceramic-on-polyethylene bearings. Given the relatively small numbers and the likely case mix selection, these patterns should continue to be monitored.

Figure 3.H5

KM estimates of cumulative revision in cemented primary hip replacements by bearing. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H6

KM estimates of cumulative revision in uncemented primary hip replacements by bearing. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H7

KM estimates of cumulative revision in hybrid primary hip replacements by bearing. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H8 (a) illustrates the revision estimates of metal-on-polyethylene and ceramic-on-polyethylene bearings used with reverse hybrid fixation in primary total hip replacement. Revision estimates are similar for the first eleven years, but after this there is a suggestion that outcomes are beginning to diverge with ceramic-on-polyethylene having slightly lower revision estimates. However, more data will be needed to ascertain if this trend represents a meaningful difference.

Figure 3.H8 (a)

KM estimates of cumulative revision in reverse hybrid primary hip replacements by bearing. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

In Figure 3.H8 (b) we present a comparison between metal-on-metal hip resurfacing and ceramic-on-ceramic hip resurfacing by sex. The numbers of ceramic-on-ceramic resurfacings are very small with very short follow-up and so should be interpreted with caution, but early trajectories between the two groups appear to be broadly similar.

Figure 3.H8 (b)

KM estimates of cumulative revision in resurfacing primary hip replacements by bearing and sex. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

In Figure 3.H9 (a), the whole cohort (including those with metal-on-metal bearings) has been sub-divided by age at primary operation and by sex. Across the whole group, there was an inverse relationship between the probability of revision and the age of the patient. A closer look at both sexes shows that the variation between the age groups was greater in females than in males; for example, females under 55 years had higher revision estimates than their male counterparts in the same age band, whereas females aged 80 years and older had a lower revision rate than their male counterparts.

Figure 3.H9 (a)

KM estimates of cumulative revision in all primary hip replacements by sex and age. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

In Figure 3.H9 (b), primary total hip replacements with metal-on-metal (or unconfirmed) bearing surfaces and resurfacings have been excluded. The revision estimates for the younger females are noticeably lower compared to the data in Figure 3.H9 (a) which includes metal-on-metal bearings; an age trend is seen in both sexes but estimates for females are lower than for males across the entire age spectrum. It is interesting to observe that age appears to have a greater effect on revision estimates in women than men with younger women having similar revision estimates to younger men, but older women having much lower revision estimates than their male counterparts.

Figure 3.H9 (b)

KM estimates of cumulative revision in all primary hip replacements by sex and age, excluding metal-on-metal hip replacement, unclassified replacements, and resurfacing. Blue italics in the numbers at risk table signify that 250 or fewer cases remained (more...)

Table 3.H6 further expands Table 3.H5 to show separate estimates for males and females within each of four age bands, <55, 55 to 64, 65 to 74 and ≥75 years. Estimates are shown at 1, 3, 5, 10, 15 and 20 years after the primary operation. These estimates refine results shown in earlier reports, but now with larger numbers of cases and therefore generally narrower confidence intervals. The relatively good results obtained with ceramic-on-ceramic and ceramic-on-polyethylene bearings in younger patients are striking. Resurfacing hip replacement continues to show high revision estimates in all groups, especially females. Even in males under 55 years of age, metal-on-metal resurfacing has twice the revision rate of some alternatives out to 15 years. Dual mobility age and sex sub-groups are too small at this stage to provide firm conclusions on relative revision estimates.

Table 3.H6

KM estimates of cumulative revision (95% CI) of primary hip replacements by sex, age group, fixation and bearing. Blue italics signify that 250 or fewer cases remained at risk at these time points.

3.H.3. Revisions after primary hip replacement: effect of head size for selected bearing surfaces / fixation sub-groups

This section looks at the effect of head size on the probability of revision following primary hip replacement. Fixation and bearing combinations with greater than 10,000 uses are included and head sizes with fewer than 500 implantations within each group are excluded.

This gave us 13 separate groups:

1. Metal-on-polyethylene cemented hip constructs n=389,581
2. Ceramic-on-polyethylene cemented hip constructs n=69,902
3. Metal-on-polyethylene uncemented hip constructs n=232,214
4. Metal-on-metal uncemented hip constructs n=25,829
5. Ceramic-on-polyethylene uncemented hip constructs n=209,062
6. Ceramic-on-ceramic uncemented hip constructs n=145,948
7. Metal-on-polyethylene hybrid hip constructs n=218,112
8. Ceramic-on-polyethylene hybrid hip constructs n=194,111
9. Ceramic-on-ceramic hybrid hip constructs n=27,818
10. Metal-on-polyethylene-on-metal hybrid hip constructs n=11,317
11. Metal-on-polyethylene reverse hybrid hip constructs n=27,185
12. Ceramic-on-polyethylene reverse hybrid hip constructs n=13,990
13. Metal-on-metal resurfacing n=42,474

Figures 3.H10 (a) to 3.H10 (m) show respective percentage cumulative probabilities of revision (Kaplan-Meier estimates) for various head sizes, for each of the groups with follow-up up to 20 years following the primary hip replacement.

In Figure 3.H10 (a), for cemented metal-on-polyethylene (MoP) hips, there was a statistically significant effect of head size (overall difference P=0.003 by logrank test) on revision estimates over the follow-up period. Overall, implants with head size 22.25mm had the worst revision estimates over the entire duration of follow-up, but implants with head size 36mm had marginally worse revision estimates in the first six years of follow-up. The numbers at risk after 12 years for patients who received 36mm heads are too small for meaningful comparison.

Figure 3.H10 (a)

KM estimates of cumulative revision of primary cemented MoP hip replacement by head size (mm). Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H10 (b) shows revision estimates for different head sizes for cemented ceramic-on-polyethylene (CoP) hips. There was a statistically significant effect of head size (overall P<0.001) with 36mm heads having the highest revision estimates, followed by 22.25mm heads. The lowest revision estimates were achieved with 28mm and 32mm heads.

Figure 3.H10 (b)

KM estimates of cumulative revision of primary cemented CoP hip replacement by head size (mm). Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H10 (c) shows revision estimates for uncemented metal-on-polyethylene (MoP) hips. There was a statistically significant effect of head size (overall P<0.001) with head sizes above 36mm having the highest revision estimates.

Figure 3.H10 (c)

KM estimates of cumulative revision of primary uncemented MoP hip replacement by head size (mm). Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H10 (d) shows revision estimates for uncemented metal-on-metal (MoM) hips, with a statistically significant difference between the head sizes overall (P<0.001) with the lowest revision estimates achieved with the smallest head sizes.

Figure 3.H10 (d)

KM estimates of cumulative revision of primary uncemented MoM hip replacement by head size (mm). Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

For uncemented ceramic-on-polyethylene (CoP) hips (Figure 3.H10 (e)), there was a statistically significant difference between the four head sizes shown (P<0.001) with 28mm and 40mm heads having higher revision estimates than 32mm and 36mm heads, although numbers at risk for patients who received 40mm heads after eight years are too small for meaningful comparison.

Figure 3.H10 (e)

KM estimates of cumulative revision of primary uncemented CoP hip replacement by head size (mm). Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H10 (f) shows revision estimates for uncemented ceramic-on-ceramic (CoC) hip replacements by head size. There are statistically significant differences between all five head sizes shown (P<0.001). In the short-term, the larger the head size, the lower the revision estimate of the construct, but revision estimates begin to rise in 44mm heads after five years.

Figure 3.H10 (f)

KM estimates of cumulative revision of primary uncemented CoC hip replacement by head size (mm). Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H10 (g) shows revision estimates for hybrid metal-on-polyethylene hip replacements by head size. There was a statistically significant difference between the six head sizes shown (P<0.001) with 22.25mm heads having higher revision estimates than the other heads. From 16 years the numbers at risk are generally low so apparent differences should be interpreted with caution.

Figure 3.H10 (g)

KM estimates of cumulative revision of primary hybrid MoP hip replacement by head size (mm). Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H10 (h) shows revision estimates for hybrid ceramic-on-polyethylene hip replacements by head size. Bearings with 28mm heads had higher revision estimates than those with 32mm and 36mm heads (P<0.001).

Figure 3.H10 (h)

KM estimates of cumulative revision of primary hybrid CoP hip replacement by head size (mm). Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H10 (i) shows revision estimates for hybrid ceramic-on-ceramic hip replacements by head size. Bearings with 36mm heads had a higher revision rate than 32mm and 28mm heads (P<0.001).

Figure 3.H10 (i)

KM estimates of cumulative revision of primary hybrid CoC hip replacement by head size (mm). Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H10 (j) shows revision estimates for hybrid metal-on-polyethylene-on-metal hip replacements by head size and outer diameter of the dual mobility polyethylene head. Those with 28mm heads combined with 49mm or 51mm polyethylene heads appear to fare better than other combinations, although comparison beyond six years is complicated by the small sample size.

Figure 3.H10 (j)

KM estimates of cumulative revision of primary hybrid MoPoM hip replacement by head size (mm) and outer diameter of the dual mobility polyethylene head (mm). Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at (more...)

Figure 3.H10 (k) shows revision estimates for reverse hybrid metal-on-polyethylene hip replacements by head size. There is no evidence that bearings with 28mm heads have a lower revision rate than those with 32mm heads or 36mm heads. (P=0.548).

Figure 3.H10 (k)

KM estimates of cumulative revision of primary reverse hybrid MoP hip replacement by head size (mm). Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H10 (l) shows revision estimates for reverse hybrid ceramic-on-polyethylene hip replacements by head size. There is some evidence of a difference in revision estimates between the head sizes, with 36mm heads having a higher revision rate in the first four years (P=0.018), after which the numbers at risk fall below 250.

Figure 3.H10 (l)

KM estimates of cumulative revision of primary reverse hybrid CoP hip replacement by head size (mm). Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H10 (m) shows revision estimates for resurfacing metal-on-metal hip replacements by head size. There is a strong trend to lower revision estimates with larger head sizes (P<0.001).

Figure 3.H10 (m)

KM estimates of cumulative revision of primary resurfacing MoM hip replacement by head size (mm). Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

3.H.4. Revisions after primary hip surgery for the main stem / cup brand combinations

In this section we present results for stem / cup brand combinations with more than 2,000 procedures for cemented, uncemented, hybrid and reverse hybrid hips or more than 1,000 procedures in the case of dual-mobility hips and resurfacings. Table 3.H7 (a) shows the Kaplan-Meier estimates of the cumulative percentage probability of revision of primary hip replacement (for any reason) for all the different stem / cup branding combinations that exist within the qualifying stem / cup combinations. Table 3.H7 (b) adds the head (and liner for modular acetabular components) branding information. Table 3.H8 (a) shows the results for all of the qualifying stem / cup combinations stratified by the bearing material combination. Table 3.H8 (b) adds the head (and liner for modular acetabular components) branding information. In Tables 3.H7 (b) and 3.H8 (b) the minimum threshold for inclusion was at least 500 procedures for unipolar brand combinations, and at least 250 procedures for dual-mobility procedures. The figures in blue italics are at time points where 250 or fewer cases remained at risk; no results are shown at all where the number had fallen below ten cases. No attempt has been made to adjust for other factors that may influence the chance of revision, so the figures are unadjusted cumulative probabilities of revision. Given that the sub-groups may differ in composition with respect to age and sex, the percentage of males and the median (IQR) of the ages are also shown in these tables.

Table 3.H7 (a)

KM estimates of cumulative revision (95% CI) of primary hip replacement by fixation, and stem / cup brand. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Table 3.H7 (b)

KM estimates of cumulative revision (95% CI) of primary hip replacement by fixation, and stem / head / cup brand (and liner in the case of modular acetabular components). Blue italics signify that 250 or fewer cases remained at risk at these time points. (more...)

Table 3.H8 (a) further divides the data by stratifying for bearing surface. This table shows the estimated cumulative percentage probability of revision for the resulting fixation / bearing sub-groups, provided there were more than 2,000 procedures for unipolar bearings, or more than 1,000 procedures for dual mobility bearings.

Table 3.H8 (a)

KM estimates of cumulative revision (95% CI) of primary hip replacement by fixation, stem / cup brand (and liner in the case of modular acetabular components) and bearing. Blue italics signify that 250 or fewer cases remained at risk at these time points. (more...)

Table 3.H8 (b) shows that there are ten cemented, ten uncemented, four hybrid and two reverse hybrid stem / head / cup (or liner / shell) / bearing combinations with revision estimates of less than 2% at ten years where more than 250 cases remain at risk at that time point. This is markedly lower than the current recommendation by NICE stating that implants with a revision rate of less than 5% at ten years should be selected for primary hip replacement for end-stage arthritis of the hip and the estimates required to achieve an Orthopaedic Data Evaluation Panel (ODEP) 10A* rating.

Table 3.H8 (b)

KM estimates of cumulative revision (95% CI) of primary hip replacement by fixation, stem / head / cup brand (and liner in the case of modular acetabular components) and bearing. Blue italics signify that 250 or fewer cases remained at risk at these time (more...)

3.H.5. Revisions for different indications after primary hip replacement

Overall, 50,946 (3.0%) of the 1,682,998 primary hip replacements had an associated first revision. The most common indications for revision were aseptic loosening (12,856), dislocation / subluxation (9,111), periprosthetic fracture (8,672), infection (8,380), adverse soft tissue reaction to particulate debris (6,826, a figure that is likely to be an underestimate due to changes in MDS collection, see later), and pain (5,357). Pain was not usually cited alone; in 3,648 out of the 5,357 instances (68.1%), it was cited together with one or more other indications. Associated PTIRs for these and the other indications are shown in Table 3.H9. Here, implant wear denotes wear of the polyethylene component, wear of the acetabular component or dissociation of the liner.

Table 3.H9

PTIR estimates of indications for hip revision (95% CI) by fixation and bearing.

The number of adverse reactions to particulate debris is likely to be underestimated because this was not requested as an indication for revision on the data collection forms in the earlier years of the registry, i.e. it was not included in MDSv1 and MDSv2. Some of these cases may have recorded the indication for revision as ‘other’ but this is not definitively known. Adoption of the later revision forms (MDSv3 onwards) was staggered over time and so a small number of revisions associated with a few primaries as late as 2011 still had revisions reported on MDSv1 and MDSv2 of the data collection forms. Restricting our analyses to primaries from 2008 onwards, as done in previous annual reports, ensures that >99% of revisions were recorded on later forms (MDSv3 onwards). It was noted that only 3,319 of the 6,826 instances (48.6%) of adverse reactions to particulate debris would thus be included, i.e. 3,507 of the earlier cases are therefore excluded from the analysis. Therefore, two sets of PTIRs are presented: one set for all primary hip replacements in the registry, which are likely to be underestimates of revisions for adverse reactions to particulate debris, and the other set for all primary hip replacements performed since the beginning of 2008, which has better ascertainment but does not include the cases with the longest follow-up.

Table 3.H9 reports revision by indication with further breakdowns by hip fixation and bearing. Metal-on-metal (irrespective of the type of fixation) and resurfacings seem to have the highest PTIRs for both aseptic loosening and pain, but ceramic-on-metal has similarly poor estimates. Metal-on-metal bearings have the highest incidence of adverse reaction to particulate debris. Although the numbers are relatively small in comparison to other groups, dual mobility bearings appear to have PTIRs for revision for dislocation / subluxation that are higher than or similar to alternative bearings and higher PTIRs for revision for periprosthetic fracture and infection. It is not yet known how much selection accounts for these observations.

In Table 3.H10, the PTIRs for each indication are shown separately for different time periods from the primary hip replacement, within the first year, and between 1 to <3, 3 to <5, 5 to <7, 7 to <10, 10 to <13, 13 to <15, 15 to <17, and ≥17 years after surgery (the maximum follow-up for any implant is now 21.75 years). Revision rates due to aseptic loosening are fairly constant until five years and then begin to steadily increase. Revision due to pain rises out to seven years and then declines. The revision rates due to subluxation / dislocation, infection and malalignment were all higher in the first year and then fell. In the case of periprosthetic fracture, the highest rates were seen in the first year, these then declined markedly before beginning to rise again at around five years. Revision for adverse reaction to particulate debris increased until 15 years before declining, whereas revision for lysis continued to rise over time.

Table 3.H10

PTIR estimates of indications for revision (95% CI) by years following primary hip replacement.

Figures 3.H11 (a) to 3.H11 (i) show how PTIRs of revision for aseptic loosening, pain, dislocation / subluxation, infection, lysis, adverse soft tissue reaction to particulate debris, periprosthetic fracture, and implant fracture changed with time. Only sub-groups with a total overall prosthesis-years at risk of more than 150,000 have been included. With time from the operation, PTIRs of revision for aseptic loosening (Figure 3.H11 (a)) tended to rise in cemented fixations and follow a fairly similar pattern in uncemented metal-on-polyethylene bearings. In uncemented metal-on-metal, they rose for the first seven years and then fell. In uncemented ceramic-on-polyethylene, ceramic-on-ceramic, hybrid ceramic-on-ceramic and resurfacings, the PTIRs were reasonably consistent over time. In hybrid metal-on-polyethylene and ceramic-on-polyethylene bearings, there were marked increases at later time points. For pain (Figure 3.H11 (b)), PTIRs were either fairly consistent or had a small initial peak followed by a decline to fairly constant rates for all bearings, apart from uncemented metal-on-metal and resurfacings where rates started high, rose to peaks at five years and then declined. Conversely, there was a high initial rate for dislocation / subluxation in all fixation / bearing groups which later fell but then began to rise in all groups from 13 years onwards apart from cemented metal-on-polyethylene, uncemented metal-on-metal, hybrid ceramic-on-ceramic and resurfacing (Figure 3.H11 (c)). Revision rates for infection were initially high and then fell in all groups apart from uncemented metal-on-metal primary total hip replacement and resurfacing (Figure 3.H11 (d)). The opposite was seen for lysis with increasing rates over time in all groups (Figure 3.H11 (e)).

Figure 3.H11 (a)

PTIR estimates of aseptic loosening by fixation and bearing.

Figure 3.H11 (b)

PTIR estimates of pain by fixation and bearing.

Figure 3.H11 (c)

PTIR estimates of dislocation / subluxation by fixation and bearing.

Figure 3.H11 (d)

PTIR estimates of infection by fixation and bearing.

Figure 3.H11 (e)

PTIR estimates of lysis by fixation and bearing.

Revision rates due to an adverse reaction to particulate debris increased with time, up to seven years in uncemented metal-on-metal primary total hip replacement and resurfacings (Figures 3.H11 (f) and (g)). Confidence intervals have not been shown here for simplicity but are wide in some groups.

Figure 3.H11 (f)

PTIR estimates of adverse soft tissue reaction by fixation and bearing.

Figure 3.H11 (g)

PTIR estimates of adverse soft tissue reaction by fixation and bearing, since 2008.

The revision rate for periprosthetic fracture (PPFx) reported by the NJR represent only those patients who have undergone a revision operation and not other types of surgery (e.g. Open Reduction and Internal Fixation). Revision for PPFx (Figure 3.H11 (h)) for uncemented THRs with all bearing combinations is substantially higher in the first year compared to cemented and hybrid THRs. The initial higher revision for PPFx in uncemented THRs then falls and then increases again with extended follow-up. Reverse hybrid constructs have a similarly high initial revision rate for PPFx compared to fully uncemented constructs, suggesting that post-operative fracture of the femur is a complication predominantly associated with uncemented stems. Hybrid fixation constructs have a lower revision rate for PPFx than for fully uncemented constructs, but greater than fully cemented constructs suggesting post-operative fracture of the acetabulum is a complication predominantly associated with uncemented acetabular cups and shells. Resurfacing THRs have a high revision rate for PPFx which is in excess of uncemented constructs. Currently, it is not clear how many PPFx occur in total or how many are fixed by other strategies, not including revision surgery.

Figure 3.H11 (h)

PTIR estimates of periprosthetic fracture by fixation and bearing.

Revision rates for implant fracture are generally low, the only exception is a small increase in implant fracture rates for ceramic-on-ceramic devices which is evident in both fully uncemented and hybrid constructs (Figure 3.H11 (i)).

Figure 3.H11 (i)

PTIR estimates of implant fracture by fixation and bearing.

3.H.6. Mortality after primary hip replacement surgery

In this section we describe the mortality of the cohort up to 21 years from primary hip replacement, according to sex and age group. Deaths recorded after 31 December 2024 were not included in the analysis. For simplicity, we have not taken into account whether the patient had a first (or further) joint revision after the primary operation when calculating the cumulative probability of death. While such surgery may have contributed to the overall mortality, the impact of this is not investigated in this report (see survival analysis methods note on page 36). Among the 1,682,998 primary hip replacements, there were 6,741 bilateral operations, with the left and right side operated on the same day; here the second of the two has been excluded, leaving 1,676257 primary hip replacements, of whom 409,111 of the recipients had died before the end of 2024.

Table 3.H11 shows Kaplan-Meier estimates of cumulative percentage mortality at 30 days, 90 days and at 1, 5, 10, 15 and 21 years from the time of the primary hip replacement, for all cases and by age and sex. Unsurprisingly, younger patients had a lower risk of death. These differences were apparent at 30 days, with approximately half the risk of death for a male patient under the age of 55 compared to one aged 65 to 69 years. These differences persisted to one year and then diverged further with approximately four times the risk of death in the older group at 21 years. For a similar age-group comparison, there was little initial difference for females, but by 21 years there was approximately three and half times the risk of death in the older group. It is worthy of note that for all cases in the registry, there is almost a 10% risk of death by five years, over 25% by ten years, over 40% by 15 years and over 60% by 21 years after primary hip replacement. The median age for undergoing a total hip replacement is 69 years, and for the 50% of patients over this age mortality rates are extremely high by 21 years ranging from 77.73% (95% CI 76.74-78.71) for women aged 70 to 74 years to 99.38% (95% CI 99.0-99.63) for men aged over 85 years.

Table 3.H11

KM estimates of cumulative mortality (95% CI) by age and sex, in primary hip replacement. Blue italics signify that 250 or fewer cases remained at risk at these time points.

3.H.7. Primary hip replacement for fractured neck of femur compared with other reasons for implantation

Total hip replacement is a treatment option for fractured neck of femur and in this section, we report on revision and mortality estimates for primary total hip replacements performed because of a fractured neck of femur compared to cases performed for other indications. A total of 65,143 (3.9%) of the primary total hip replacements were performed for a fractured neck of femur (NOF)^†.

Table 3.H12 shows that the proportion of primary hip replacements performed for an indication of a fractured neck of femur increased with time to a maximum of 7.6% in 2020 but has reduced since then, down to 3.7% in 2024 which is the lowest proportion since 2012. The proportion of THRs performed for fractured NOF in 2020 was artificially inflated by the dramatic decrease in elective THRs performed in 2020 due to the impact of COVID, prior to this the peak was 5.7%. The use of dual mobility bearings has become more popular in this group, peaking at 18.7% of cases in 2022, before reducing to 17.5% in 2024. The most striking feature is the marked drop in 2020 in the total annual number of THRs performed for a fractured NOF (4,361 compared to 5,684 in 2019). This is most likely due to the impact of the COVID pandemic possibly through a combination of fewer fractures occurring during lockdown and less or altered provision of care (with a possible shift from THR to hemiarthroplasty). This decrease has been sustained in 2021 through 2024 with 4,693 THRs performed for fractured NOF in 2021 and 4,347 in 2024, the lowest in the registry since 2015. There are usually late registrations of cases into the registry and thus the figures for 2024 may be revised upwards in next year’s report, but this observation may also be related to the publication of the HEALTH trial which demonstrated no difference in the risk of secondary procedures for patients receiving total hip replacement or hemiarthroplasty for a displaced hip fracture and a clinically unimportant improvement in function and quality of life for patients receiving a total hip replacement (Bhandari M, et al., 2019).

Table 3.H12

Number and percentage of fractured neck of femur in the registry by year.

Table 3.H13 compares the fractured NOF group with the remainder with respect to sex and age composition together and type of hip replacement received. A significantly larger percentage of the fractured NOF cases, compared with the remainder, were female (71.8% versus 59.2%: P<0.001, Chi-squared test).

Table 3.H13

Fractured neck of femur versus osteoarthritis only by sex, age and fixation.

The fractured NOF patients were significantly older (median age 73 years versus 70 years at operation). We found that cemented and hybrid hip replacements were used more commonly in fractured NOF cases than in hip replacements performed for osteoarthritis only, but cemented fixation was still used in under half of the patients. Figure 3.H12 (a) shows that the cumulative revision rate was higher in the fractured NOF cases group compared with the remainder (P<0.001, logrank test). The plotted cumulative revision lines diverge early in the first year and then remain approximately parallel out until about 13 years. This effect was not fully explained by differences in age and sex, as stratification by these variables left the result unchanged (P<0.001 using stratified logrank test: 14 sub-groups of age <55, 55 to 59, 60 to 64, 65 to 69, 70 to 74, 75 to 79, ≥80 for each sex). Figure 3.H12 (b) shows similar cumulative revision estimates for dual mobility compared to unipolar total hip replacement bearings in the hip fracture population out to ten years after which point the numbers fall below 250 in the dual mobility group. While the difference here is not significant, it is interesting that this is a different pattern seen to that for dual mobility bearings in cemented and uncemented fixation groups in elective total hip replacement where the early revision estimates appear higher in the dual mobility bearings.

Figure 3.H12 (a)

KM estimates of cumulative revision for fractured neck of femur and osteoarthritis only cases for primary hip replacements. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H12 (b)

KM estimates of cumulative revision by bearing type for fractured neck of femur cases in primary hip replacements. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H13 shows a markedly higher overall mortality in total hip replacements performed for hip fracture cases compared to cases implanted for osteoarthritis only (P<0.001, logrank test). As in the overall mortality section, the second of 6,741 simultaneous bilateral procedures were excluded. Sex and age differences did not fully explain the difference seen, as a stratified analysis still showed a difference (P<0.001).

Figure 3.H13

KM estimates of cumulative mortality for fractured neck of femur and osteoarthritis only in primary hip replacements. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

3.H.8. Overview of hip revisions

In this section we look at all hip revision procedures performed since the start of data collection by the NJR, 1 April 2003, up to 31 December 2024, for all patients with valid patient identifiers (i.e. whose data could therefore be linked).

In total, there were 157,412 revision procedures. These revisions were recorded on 133,301 hips in 125,118 patients. In addition to the 50,946 first revised primary hip replacements described in section 3.H.2 of this report, there were 95,715 additional revisions of a hip for which there is no associated primary hip replacement recorded in the registry. It is likely that the majority of the primaries associated with these revisions where the primary is not recorded in the registry would have been performed prior to the commencement of data capture by the NJR in 2003.The remaining 10,751 revision procedures were re-revisions, i.e. revision procedures subsequent to the first revision.

Revisions are classified as single-stage, stage one and stage two of two-stage revisions. Information on stage one and stage two revisions are entered into the registry separately, whereas in practice a stage two revision has to be linked to a preceding stage one revision. Debridement and Implant Retention (DAIR) with or without modular exchange are included as single-stage procedures. With the introduction of distinct indicators for the DAIR procedures with or without modular exchange in MDSv7 and introduction of a separate reoperations form in MDSv8, which now captures the DAIRs without modular exchange that do not meet the registry definition of a revision procedure as no implant is added, removed or modified, it may be possible to report these as distinct categories in future reports. Although not all patients who undergo a stage one of two revision will undergo a stage two of two revision, in some cases stage one revisions have been entered without a stage two, and vice versa, making identification of individual revision episodes difficult. We have attempted to do this later in this section.

The NJR asks surgeons and those responsible for healthcare delivery to ensure that when primary and revision joint replacement procedures of the hip, knee, ankle, elbow or shoulder are performed, that the relevant MDS form is completed and data entered into the registry. This is a requirement mandated by the NHS Standard Contract. For the purposes of the Annual Report, revision procedures include any addition, removal or modification of the implants and procedures such as debridement and implant retention with implant exchange, excision arthroplasty, amputation and conversion to arthrodesis. For data submitted on MDSv7 only, DAIRs without modular exchange are included as revision procedures. The completion of a revision MDS form is also mandatory for a procedure involving modification of a joint by adding another implant to another part of the joint. For the analyses of surgeon performance, hospital performance and implant performance, debridement and implant retention (DAIR) without implant exchange is currently excluded.

Table 3.H14 gives an overview of all hip replacement revision procedures carried out each year since April 2003. There were a maximum number of 13 documented revision procedures associated with a single hip, making up eleven revision episodes as two episodes consisted of a stage one of a two-stage procedure and a stage two of a two-stage procedure.

Table 3.H14

Number and percentage of hip revisions by procedure type and year.

The incidence of revision hip replacement peaked in 2012 and has declined since then, despite the increasing number of at-risk implants due to the increase in primary hip replacements and secular increases in the longevity of the lives of patients. In the COVID-impacted years of 2020 and 2021, the number of revision hip replacements performed were approximately half of the peak rate observed in 2012. The number of revisions performed in 2024 (7,253) remains 14% lower than the number performed in 2019 (8,274) prior to the impact of COVID.

Table 3.H15 (a) shows the stated indication for the revision hip replacement surgery. Please note that, as several indications can be stated, the indications are not mutually exclusive and therefore column percentages may add up to over 100%. Aseptic loosening was the most common indication for revision.

Table 3.H15 (a)

Number and percentage of hip revision by indication and procedure type.

Table 3.H15 (b) shows the stated indication for revision hip replacement surgery performed in the last five years (1,826 days). The most notable difference between all the data and that recorded in the last five years is pain as an indication for revision falling from 14% to 3% of single-stage revisions. There is also a higher proportion of cases revised for periprosthetic fracture in the last five years (20.5% compared to 13.3%) and a higher proportion of cases revised due to infection (12.6% compared to 6.4%). The ratio of stage two of two-stage, stage one of two-stage and single-stage revisions overall (1:1.03:13.9) is different compared to those performed in the last five years (1:1.44:16.9). Please note that higher percentage ratios do not equate to an absolute increase in revisions for a specific cause. Looking at the data for the last five years in comparison to data for the whole registry, the use of single-stage revision for infection in comparison to a two-staged revision approach has increased.

Table 3.H15 (b)

Number and percentage of hip revision by indication and procedure type in last five years.

3.H.9. Estimates of hip re-revision

In most instances (91.8% of 133,301 hips), the first revision procedure was a single-stage revision, however in the remaining 8.2% it was part of a two-stage procedure. For a given hip, survival following the first documented revision hip replacement procedure for those with a linked primary in the registry (n=50,946) has been analysed. This analysis is restricted to patients with a linked primary procedure so that there is confidence that the next observed procedure on the same joint is the first revision episode. If there is no linked primary record in the dataset, it cannot be determined if the first observed revision is the first revision or if it has been preceded by other revision episodes. The time from the first documented revision procedure (of any type) to the time at which a second revision episode was undertaken has been determined. For this purpose, an initial stage one followed by either a stage one or a stage two have been considered to be the same revision episode and these were disregarded, looking instead for the start of a second revision episode (the maximum number of distinct revision episodes was determined to be 11 for any hip).

In cases where a stage one of two procedure was followed by a stage two of two procedure within 365 days, we have treated this as a single distinct episode. This definition allows multiple stage one procedures to occur before a new revision episode is triggered. In situations where the first stage one procedure is not followed by a stage two procedure within a 365-day period, the next occurrence of a stage one procedure was considered as a new revision episode.

Kaplan-Meier estimates of the cumulative percentage probability of having a subsequent revision (re-revision) were calculated. There were 6,177 re-revisions and for 11,741 cases the patient died without having been re-revised. The censoring date for the remainder was the end of 2024.

Figure 3.H14 (a) plots Kaplan-Meier estimates of the cumulative probability of a subsequent revision between 1 and 21 years since the first revision operation.

Figure 3.H14 (a)

KM estimates of cumulative re-revision in linked primary hip replacements (shaded area indicates point-wise 95% CI). Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H14 (b) shows estimates of re-revision by type of primary hip replacement. Resurfacing has the lowest re-revision rate until approximately 17 years, after which the revision rate appears to accelerate and become worse than that associated with alternatives. However, after 17 years the numbers at risk are low and should therefore be interpreted with caution. Hybrid primary total hip replacements have the highest estimates of re-revision to alternatives up until approximately 16 years, after which the numbers at risk become small.

Figure 3.H14 (b)

KM estimates of cumulative re-revision by primary fixation in linked primary hip replacements. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H14 (c) shows the relationship between time to first revision and the risk of subsequent revision.

Figure 3.H14 (c)

KM estimates of cumulative re-revision by years to first revision, in linked primary hip replacements. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

The earlier the primary hip replacement is revised, the higher the risk of a second revision. There is a relationship between the indication for first revision and time to first revision; earlier in this report (section 3.H.5) we show, for example, that revisions for dislocation / subluxation, infection and malalignment were more prevalent in the early period after the primary hip replacement, and aseptic loosening and lysis were more prevalent causes later on.

For those with a documented primary hip replacement within the registry, Figures 3.H15 (a) to (e) show cumulative re-revision estimates following the first revision hip replacement, according to the main fixation used in the primary. Each sub-group, with the exception of reverse hybrid, has been further sub-divided according to the time interval from the primary hip replacement to the first revision, i.e. less than 1 year, 1 up to 3, 3 up to 5, 5 up to 7, 7 up to 10, and greater than or equal to ten years.

Figure 3.H15 (a)

KM estimates of cumulative re-revision in cemented primary hip replacement by years to first revision, in linked primary hip replacements. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points. (more...)

Figure 3.H15 (b)

KM estimates of cumulative re-revision in uncemented primary hip replacement by years to first revision, in linked primary hip replacements. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points. (more...)

Figure 3.H15 (c)

KM estimates of cumulative re-revision in hybrid primary hip replacement by years to first revision, in linked primary hip replacements. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.H15 (d)

KM estimates of cumulative re-revision in reverse hybrid primary hip replacement by years to first revision, in linked primary hip replacements. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points. (more...)

Figure 3.H15 (e)

KM estimates of cumulative re-revision in resurfacing primary hip replacement by years to first revision, in linked primary hip replacements. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points. (more...)

For reverse hybrid the overall numbers were too low for these sub-divisions and as such the maximum cut-off was greater than or equal to five years. For cemented, uncemented, hybrid, reverse hybrid and resurfacing hip replacements, there was a trend of higher observed re-revision estimates in those that had their first revision within one year, between one and three years, or three to five years of the initial primary hip replacement.

Table 3.H16 (a) shows the re-revision estimates of the 50,946 primary hip replacements in the registry that were revised, and of these, 6,177 were re-revised. Table 3.H16 (b) shows that primary hip replacements that fail within the first year after surgery have just over twice the chance of needing re-revision at each time point compared with primaries that last more than ten years.

Table 3.H16 (a)

KM estimates of cumulative re-revision (95% CI). Blue italics signify that 250 or fewer cases remained at risk at these time points.

Table 3.H16 (b)

KM estimates of cumulative re-revision (95% CI) by years since first revision. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Table 3.H16 (c) shows cumulative re-revision estimates at 1, 3, 5, 7, 10, 13, 15 and 18 years following the first revision for those with documented primary hip replacements within the registry, broken down by fixation types and bearing surfaces used in the primary hip replacement. The numbers are low for dual mobility hips and the duration of follow-up is short, but initial results show high failure estimates ranging from 6.5% to 22.9% at one year in dual mobility procedures.

Table 3.H16 (c)

KM estimates of cumulative re-revision (95% CI) by fixation and bearing used in primary hip replacement. Blue italics signify that 250 or fewer cases remained at risk at these time points.

The revision estimates for revisions following resurfacings were comparatively low, but Figure 3.H14 (b) shows that after 19 years the revision rate is becoming higher than those for alternatives.

3.H.10. Reasons for hip re-revision

Tables 3.H17 (a) and (b) show a breakdown of the stated indications for the first revision and for any second revision. Please note the indications are not mutually exclusive. Table 3.H17 (a) shows the indications for recorded revisions in the registry and Table 3.H17 (b) reports the indications for the first linked revision and the number and percentage of first linked revisions that were subsequently revised. In the final column in Table 3.H17 (b), we report the indications for all the second linked revisions e.g. 1,267 linked second revisions recorded aseptic loosening as an indication. It is interesting to note that both dislocation and infection are much more common indications for a second revision than for a first revision. This shows the increased risk of instability and infection following the first revision of a hip replacement compared to that of primary hip replacement.

Table 3.H17 (a)

Number of revisions by indication for all revisions.

Table 3.H17 (b)

Number of revisions by indication for first linked revision and second linked re-revision.

Tables 3.H18 (a) and (b) show that the numbers of revisions and the relative proportion of revisions with a linked primary in the registry increased with time. Approximately 60% of revisions performed in 2024 had a linked primary in the registry. This is likely to reflect improved data capture over time, improved linkability of records and the longevity of hip replacements with a proportion of primaries being revised being performed before data capture began or being outside the coverage of the registry.

Table 3.H18 (a)

Number of revisions by year.

Table 3.H18 (b)

Number of revisions by year, stage, and whether or not primary is in the registry.

3.H.11. 90-day mortality after hip revision

The overall cumulative percentage mortality at 90 days after hip revision was lower in the cases with a primary hip replacement recorded in the registry compared with the remainder (Kaplan-Meier estimates 1.81% (95% CI 1.69-1.93) versus 2.18% (95% CI 2.08-2.28)), which may reflect the fact that patients in this group were younger at the time of their first revision, median age of 71 (IQR 62 to 78) years compared to the group without primaries documented in the registry who had a median age of 74 (IQR 66 to 81) years. The percentage of males to females was similar in both groups (44.6% versus 42.6% respectively).

3.H.12. Conclusions

As in previous reports, our analysis of implants has been by revision of the construct, rather than revision of a single component, as the mechanisms of failure (such as wear, adverse reaction to particulate debris and dislocation) are interdependent between different parts of the construct. Revision analyses have also been stratified by age and sex. The new component database, introduced in the previous annual report, has provided increased granularity of implant data. This has primarily allowed more detailed reporting, such as the addition of the brand of the head and liner to the primary brand reporting tables. It has also allowed us to better resolve implant data where there was uncertainty previously, this is most noticeable in the increase in the proportion of total hip replacements using a dual mobility bearing in treating hip fractures but is also seen in newly identified cases in the overall data which provides longer-term outcome data for these bearings.

The highest revision estimates are among younger females and the lowest among older females. When data on metal-on-metal implants are excluded, younger females have similar revision estimates to younger males. Once again, it must be emphasised that implant survivorship is only one measure of success and cannot be used as an indication of patient satisfaction, relief of pain, improvement in function and the resulting greater participation in society. The data clearly show that constructs failing at different estimates is associated with the age and sex of the recipients.

Overall, the number of primary hip replacements recorded annually in the registry continues to increase, now with 1,682,998 eligible for analysis. The COVID pandemic had a marked impact on the provision of hip replacement with primary THR decreasing from 99,972 in 2019 to 57,551 in 2020, but procedure volumes have now recovered and surpassed previous years to 116,901 in 2024 (the highest annual number to date), and revision THR has fallen from 8,274 in 2019 to 5,225 in 2020 and partially recovered to 7,253 in 2024. Due to late data entry for 2024 the figures listed here will be revised upwards in subsequent reports, so the recovery will be greater than the current data suggests. The overall provision of primary hip replacement has recovered to above pre-pandemic levels, but a far greater percentage are now both funded and undertaken in the private sector, with overall NHS-provision still below pre-pandemic numbers.

It is interesting to examine the overall secular trends in provision of primary and revision hip replacements. Apart from the COVID-affected years of 2020 and 2021, the trend has been for an ever-increasing provision of primary hip replacement such that the volume of procedures now exceeds 100,000 cases per annum. The provision of, and presumably the requirement for, revision hip replacement increased markedly from 4,016 cases in 2005 to 10,510 in 2012 and then declined to 7,253 in 2024 (with lower numbers in COVID-affected years 2020 and 2021).

Looking at the relationship between year of primary and subsequent revision, between 2004 and 2007 the primaries undertaken each year were at higher risk of being revised than those undertaken the previous year, i.e. outcomes were getting steadily worse. This coincided exactly with the increased use of metal-on-metal stemmed hip replacements and hip resurfacings. This registry and other registries reported poor results with these types of prostheses. Their use then rapidly declined between 2007 and 2011 and the revision estimates for primaries performed over that period demonstrated a pronounced decline. Because of this disproportionate effect of metal-on-metal bearings on secular trends in revision, we reported the revision estimates over time excluding metal-on-metal. This showed that revision estimates have been decreasing since 2008/2009 for non-metal-on-metal bearing hip replacements. The reasons for this are likely to be multi-factorial, but surgeon performance reporting, which began at this time, is likely to be a contributing factor.

In addition, in the NJR Annual Report 2009, we commented that data suggested that ceramic-on-polyethylene bearings were associated with lower revision estimates. Between 2009 and 2024, the use of these bearings has increased approximately five-fold. In 2024, ceramic-on-polyethylene hybrid constructs were the most common type of hip replacement performed (27.2%), with the second commonest being ceramic-on-polyethylene uncemented hips which accounted for 23.8% of cases. The decline in revision estimates for primaries performed over this period has mirrored the increase in use of these bearings. This rate of decline in revisions by year of primary surgery has slowed over time, particularly since 2013, but is still evident.

The result of surgical practice changing in response to outcomes is that procedures now achieve remarkably low long-term revision estimates. The majority of patients undergoing THR are between 65 and 75 years old. It is striking that at 15 years the average revision rate for implants excluding MoM bearings is less than 5% i.e. the 10-year NICE benchmark for performance. Furthermore, revision estimates below 5% at 20 years have been achieved in three separate brand combinations (Table 3.H8 (b)), whilst MoM resurfacing has a revision rate of 15.53% (15.08-15.99) at the same time point.

When stratifying hip constructs by the brand of the stem / head / cup (or liner / shell) and bearing materials, we report revision estimates of less than 2% at ten years for ten cemented, ten uncemented, four hybrid and two reverse hybrid constructs, where more than 250 cases remain at risk with many more combinations on track to achieve these very low revision estimates. NICE currently recommend that when implants are selected for primary total hip replacement for end-stage arthritis, implants with an observed or predicted revision rate of less than 5% should be selected. The best rating issued by ODEP at ten years is 10A* which requires a revision rate of less than 5%. Given the large number of constructs achieving much lower revision estimates than these thresholds, it should be considered whether these thresholds should be revised to encourage the selection of implants that are associated with very low revision estimates for patients. We also present data here that show that it is very unusual for patients aged over 70 years to still be alive 20 years after their primary. Using existing implants and techniques, surgeons are thus capable of performing hip replacements that will last the entire life of nearly all patients above the median age of a patient undergoing hip replacement of 69 years.

This reinforces the argument that any new implants and techniques really need to focus on patients younger than 70 years of age and those undergoing revision surgery. Recent analysis of NJR data has shown strongly that revisions last significantly less long than primaries and that each subsequent revision lasts half as long as its predecessor (Deere et al 2022). Getting it right first time really is the solution.

The data demonstrating how widespread adoption of technology before long-term outcomes are available can be disastrous, continues to grow. The revision estimates with metal-on-metal resurfacing continue to increase over time, particularly in women, and the contrast with other implants is stark. For example, the revision estimates in women receiving metal-on-metal resurfacing are six-fold higher at 15 years than that achieved with some other commonly-used alternatives. This holds true even when stratified for age. Metal-on-metal stemmed and resurfacing implants continue to fail at higher than expected rates and their use is now extremely rare. The best-performing brand of resurfacing has a revision rate of 9.75% (95% CI 9.35-10.16) at 15 years. This contrasts with a revision rate of 2.57% (95% CI 2.01-3.29) for the best-performing cemented hip replacement, 1.82% (95% CI 1.18-2.81) for the best-performing uncemented hip replacement and 2.55% (95% CI 2.19-2.98) for the best performing hybrid hip replacement defined as the stem / head / cup (or liner / shell) and bearing material combination with more than 250 cases remaining at risk.

It is important that we monitor the performance of novel bearing designs of hip replacement closely. There is now sufficient data to report on ceramic-on-ceramic resurfacings. The numbers are low and follow-up is short and thus caution is required interpreting these early data, however revision estimates in young women appear to already be much higher than in young men. Patients undergoing these procedures need to be monitored very carefully. The use of dual mobility constructs continues to increase with over 33,000 of these now recorded in the registry. The early revision estimates with these appear to be slightly higher than alternatives, but 10-year revision estimates appear to be acceptable (3.09% (95% CI 2.60-3.67) for the commonest type (hybrid MoPoM)). Indications for usage should be carefully considered. It may be that higher early revision estimates are due to appropriate case mix selection, so it is important to closely monitor the emerging data on these implants. However, a higher early rate of revision compared to unipolar bearings was not observed in patients with a fractured neck of femur. This is an area which is developing and requires more in-depth analysis in the future.

Since the 12th NJR Annual Report in 2015, our data have been presented by age and sex, comparing combinations of fixation and bearing. This assists clinicians and patients in choosing classes of prostheses that are the most appropriate for particular patients. For example, in males aged 55 to 64 years, at 15 years post-surgery, hybrid and uncemented ceramic-on-polyethylene and ceramic-on-ceramic constructs, as well as cemented ceramic-on-polyethylene constructs have similarly low revision estimates of approximately 5%, while cemented metal-on-polyethylene constructs have revision estimates of 8.41% (95% CI 7.75-9.12) and uncemented metal-on-polyethylene bearings are 6.59% (95% CI 5.99-7.25). Metal-on-metal resurfacings in this group have a higher revision rate at 15 years of 8.92% (95% CI 8.40-9.48). Females aged 55 to 64 years have lower revision estimates than males for all fixation /bearing combinations at 15 years, except for those with metal-on-metal bearings such as resurfacings, where the revision estimates are markedly higher for females than males and also markedly higher than alternatives. For example, 15-year revision estimates with hybrid ceramic-on-polyethylene constructs in this group are 2.69% (95% CI 2.21-3.27) compared to metal-on-metal hip resurfacing of 21.63% (95% CI 20.41-22.91).

For patients over 75 years, all combinations except those with metal-on-metal bearings have good outcomes, with cemented and hybrid ceramic-on-polyethylene constructs possibly having the lowest revision estimates. The risk of revision at 20 years in this group is very small; males 6.09% (95% CI 5.47-6.79) and females 4.15% (95% CI 3.78-4.56). The 20-year mortality rate in males aged 75 to 79 years is 96.07% (95% CI 95.21-96.81) and in females aged 75 to 79 years is 92.22% (95% CI 91.53-92.88).

We have also examined outcomes of different head sizes (bearing diameters) with alternative fixation and bearing types and these results are interesting. With metal-on-polyethylene and ceramic-on-polyethylene, large head sizes appear to be associated with higher revision estimates particularly with 36mm heads used with cemented fixation and heads >36mm used with uncemented fixation. Ceramic-on-ceramic bearings have lower revision estimates with larger bearings when used with uncemented fixation in the short-term, but revision estimates begin to rise with the largest head sizes beyond six years. Higher revision estimates for 36mm compared to smaller heads are also seen in ceramic-on-ceramic hybrid fixations. This demonstrates the importance of examining the entire construct, not just the individual variables such as fixation, composition of bearing and head size.

With regard to specific branded stem / cup combinations, some of the best implant survivorships have still been found to be achieved by mix and match cemented hard-on-soft bearing constructs, although this practice remains contrary to both the MHRA and implant manufacturers’ guidelines for usage.

It is encouraging that the most commonly-used constructs by brand in cemented and hybrid fixation have good results. This does not hold true for uncemented fixation, but further breakdown by bearing type for commonly-used uncemented implants shows that results are acceptable if metal-on-metal bearings are excluded. It is important to note that there is variability in brand-level constructs with variation in revision outcomes according to factors such as the bearing combination used. It is therefore important to consider the construct when selecting implants for specific outcomes. We encourage all readers to view Table 3.H8 (b) for fine details of construct performance.

Risk of re-revision rate is strongly associated with time to first revision; as 19.35% (95% CI 18.55-20.19) of hips revised within a year of primary surgery are re-revised within ten years. In contrast, when the primary lasts at least ten years the re-revision rate is 8.99% (95% CI 8.10-9.97) at ten years after the first revision. Re-revision estimates up to ten years appear to be independent of the fixation and bearing of the primary hip replacement, except for resurfacing procedures which are initially associated with lower re-revision estimates, but this pattern appears to begin to wane between seven and ten years after the re-revision. At 15 years re-revision estimates are 18.03% (95% CI 16.68-19.47) for cemented primaries, 19.05% (95% CI 18.06-20.08) for uncemented primaries and 16.55% (95% CI 15.19-18.03) for resurfacings.

Overall, this latest report is good news for patients, clinicians, and the healthcare sector. Provision of hip replacement overall has recovered in volume and now surpasses pre-COVID levels, revision estimates continue to decline and clinicians increasingly use constructs with proven longevity. The detrimental effect of COVID on absolute provision has been short-lived, but profound. In 2020 there was a massive under-provision of primary hip replacement with over 42,000 fewer primary hip replacements performed than in 2019. In 2021, much of this decline in volume was reversed with only 10,000 fewer primary hip replacements than in 2019. In 2024 more primaries were performed than in 2019 (116,901 vs. 99,972), and numbers are in line with the long-term secular trend. It is noteworthy that NHS under-provision has been replaced with increased independent sector provision. The 2020/21 deficit of approximately 55,000 primary hip replacements has led to increases in waiting lists that will need comprehensive planning to resolve.

With the health service having to address an unprecedented backlog of joint replacement along with increasing pressure for cost-containment, the selection of clinically effective and value for money treatments with a good evidence-base will be increasingly important.

3.K.1. Overview of primary knee replacement surgery

In this section of the report, we address revision and mortality outcomes for primary knee operations performed and reported to the registry between 1 April 2003 and 31 December 2024. The very first procedures that were entered into the registry therefore had a potential 21.75 years of follow-up.

The outcomes of total and partial knee replacement procedures are discussed throughout this section, hereafter referred to as total (TKR), unicompartmental (UKR) and multicompartmental (MKR) knee replacement. Unicompartmental knee replacements include both unicondylar knee replacements and patellofemoral knee replacements. Multicompartmental knee replacement may include either a medial and lateral unicondylar knee replacement, or unicondylar knee replacement and patellofemoral knee replacement, or medial and lateral unicondylar knee replacement and patellofemoral knee replacement that are performed as part of the same procedure. Brief details of the type of orthopaedic surgery involved for each form of replacement can be found in the Summary of Sources section.

We note here that the NJR data collection process now distinguishes between medial and lateral unicondylar replacements, although this was not always the case in the past. This distinction is available for cases reported on the MDSv7 and MDSv8 forms but unicondylar cases reported on earlier versions of the MDS form do not make this distinction. We now use information from manufacturers entered into the new component database and the MDS form to determine whether the prostheses were used medially or laterally.

This year there have been a number of changes to how we have classified components. We have moved to using a new component classification system which has facilitated whole construct validation, laterality assignment within unicondylar knees, differentiation of the multi-compartmental knee replacements and refined constraint information. These enhancements will be visible throughout the report.

Figure 3.K1 (a) describes the data cleaning processes applied to produce the total of 1,805,000 primary knee procedures included in the analyses we present in this section and 1,788,187 ipsilateral procedures used for mortality analyses.

Figure 3.K1 (a)

Knee cohort flow diagram.

Over the lifetime of the registry, the 1,805,000 primary knee joint replacement procedures contributing to our revision analyses were carried out by a total of 3,829 unique consultant surgeons working across 485 hospitals.

Over the last three years (1 January 2022 to 31 December 2024), 357,524 primary knee procedures (representing 19.8% of primary knee replacements currently included in the registry) were performed by 1,942 consultant surgeons working across 401 hospitals. Looking at caseload over this three-year period, the median number of primary procedures by consultant surgeons was 135 (IQR 49 to 258) and the median number of procedures for hospitals was 770 (IQR 326 to 1,252). A proportion of surgeons will have commenced practice as a consultant during this period, some may have retired, and some surgeons may have periods of surgical inactivity within the timeline of coverage of the registry, therefore their apparent caseload would be lower.

Over this three-year period, there have been 302,497 primary TKRs performed by 1,916 consultant surgeons (median=116 cases per surgeon; IQR 46 to 208) in 401 separate hospitals (median=665 cases per hospital; IQR 275 to 1,069). In the same period, there have been 50,329 primary unicondylar knee procedures performed by 962 consultant surgeons (median=29 cases per surgeon; IQR 7 to 70) in 363 hospitals (median=91 cases per hospital; IQR 29 to 186).

The majority of primary knee replacements in the registry were carried out on females (females 56.1%; males 43.9%). The median age at primary operation was 70 years (IQR 63 to 76), see Table 3.K3 and commentary later for discussion of age at primary by type of knee replacement. Osteoarthritis was given as a documented indication for surgery in 1,760,724 procedures (97.6% of the cohort) and was the sole indication given in 1,738,995 (96.3%) primary knee procedures.

Table 3.K1 shows the breakdown of cases by type of knee replacement, the method of fixation, constraint and bearing used. Fixations of TKR are classified as “All cemented”, “All Uncemented”, “All Hybrid” (either a cemented tibia and uncemented femur or uncemented tibia and cemented femur) and “All pre-assembled/hinged/linked”. Bearings are now divided into “fixed” and “mobile” modular or pre-assembled or prefixed tibias and “MBT” which are fixed monobloc polyethylene tibias. The constraint has been reclassified across the registry; “unconstrained” devices include prostheses which retain both cruciate ligaments (BCR), cruciate substituting (no cam or post), posterior cruciate retaining / anterior cruciate sacrificing (CR). Posterior-stabilised knees include both posterior-stabilised knees which include cam or post, and knee prostheses that substitute both cruciate ligaments and contain a cam and post. Constrained condylar knees also known as Varus-valgus stabilised knees or CCK knees form a separate group. All pre-assembled, hinged or linked knees are now reported in isolation due to their limited and exceptional use in primary knee replacements. The constraint is based solely on the features of the tibial component. A breakdown within each method of fixation of the percentage of constraint and bearing types used is shown in a separate column. All cemented TKR is the most commonly performed type of knee replacement (82.8% of all primary knee replacements). A further 3.7% were either all uncemented or hybrid TKRs, and 0.4% were pre-assembled/hinged/linked. Most UKRs were unicondylar (10.3% of the total) with the remainder being patellofemoral (1.1%), multicompartmental (0.1%), or unconfirmed fixation, constraint and bearing (1.7%).

Table 3.K1

Number and percentage of primary knee replacements by fixation, constraint and bearing.

Nearly two-thirds of all operations (59.4%) were TKRs which were all cemented and unconstrained with a fixed bearing, followed by 18.4% which were all cemented and posterior-stabilised with a fixed bearing. Uncemented and hybrid prostheses are mostly unconstrained. While uncemented knees are almost equally likely to have a mobile or fixed bearing, hybrid knees are more likely to utilise a fixed bearing. Approximately two-thirds (71.7%) of cemented TKRs are unconstrained and have a fixed bearing. Medial unicondylar knee surgery has historically used a mobile bearing, but this has been changing in recent years (Table 3.K2) with a larger proportion using fixed bearings. Lateral unicondylar knee replacement make up less than 4% of all unicondylar knee replacements, with the majority comprising cemented fixed bearings. Some primary knee replacements could not be classified according to their bearing / constraint (approximately 1.7% of the total cohort).

Table 3.K2

Percentage of primary knee replacements by fixation, constraint, bearing and year.

Figures 3.K1 (b) to (d) show the yearly number of primary knee replacements performed for all indications. Procedures have been stratified by total knee, unicondylar and patellofemoral joint replacements. Please note the difference in scale of the y-axis between each plot.

Figure 3.K1 (b)

Frequency of primary total knee replacements within elective cases stratified by procedure type, bars stacked by volume per consultant per year.

Figure 3.K1 (c)

Frequency of primary unicondylar knee replacement within elective cases stratified by procedure type, bars stacked by volume per consultant per year.

Figure 3.K1 (d)

Frequency of primary patellofemoral knee replacements within elective cases stratified by procedure type, bars stacked by volume per consultant per year.

Each bar in the figure is further stratified by the volume of procedures that the consultant performed in that year within that joint replacement type i.e. if a surgeon performed 25 elective TKR procedures, 25 unicondylar knee replacements and 25 patellofemoral joint replacement procedures, their annual total volume would be 75 procedures. However, each 25 procedures are not aggregated and only contribute to the grey sub-division in each figure respectively.

Figure 3.K1 (b) shows that the volume of TKRs generally increased from when data collection started until 2020 when the impact of COVID was observed. From 2007 until 2020, the majority of TKR procedures were contributed by higher volume surgeons i.e. those performing 49 or more procedures annually. In 2020, the majority of procedures were performed by those performing 48 or fewer procedures annually before the previous pattern was restored in 2021, and since then (2022 and 2023) the majority of procedures have been performed by surgeons contributing 49 or more annually.

Figure 3.K1 (c) shows that the volume of unicondylar knee replacements increased rapidly from 2013 until the impact of COVID in 2020. In 2022 the volume of unicondylar knee replacements was greater than in 2019 and has continued to increase since. From 2014 until 2020, the majority of unicondylar knee replacement procedures were contributed by higher volume consultants i.e. those performing 25 or more procedures annually. In 2020, the majority of procedures were performed by those performing under 25 procedures annually, and then the previous pattern was restored in 2021. Only a small proportion of the procedures were contributed by consultants performing fewer than seven unicondylar knee replacements per year.

Figure 3.K1 (d) shows that the volume of patellofemoral knee replacements was fairly constant from 2008 until the impact of COVID in 2020 and since 2023 the volume has recovered to pre-COVID levels. Almost a third of procedures are carried out by surgeons with a yearly volume of four or less.

Figure 3.K1 (e) describes the funding status and organisation type (based on organisation type in 2024) of primary knee procedures collected by the NJR. The figure shows a steady increase in the number of knee replacements that were NHS-funded and performed in NHS hospitals from the beginning of the registry until 2012. After this time, the number plateaued until 2019 and then reduced substantially due to the impact of COVID. The growth in the total number of knee replacements performed from 2012 to 2019 was largely driven by growth in the number of NHS-funded procedures being performed in independent hospitals. Although the total number of knee replacement procedures performed in 2023 and 2024 has now exceeded the pre-COVID 2019 levels, where those procedures are performed has substantially changed. The number of NHS-funded procedures being performed in NHS hospitals has now recovered to the pre-COVID levels, however the increase in overall volume has been largely driven by increases in the number of NHS-funded procedures performed in independent hospitals and independently-funded procedures performed in independent hospitals. Where NHS-hosted and funded procedures used to account for approximately 60% of the total volume, they now represent less than 50%.

Figure 3.K1 (e)

Frequency of elective primary knee replacements by funding status and organisation type, per year.

Table 3.K2 shows the annual rates for the usage of the different types of primary knee replacements. Overall, more than 90% of all types of primary knee replacement utilised all cemented fixation, and since 2004 the share of all implant replacements of this type has increased by approximately six percentage points. The main decline in the type of primary knee replacements carried out has been in the use of all uncemented and hybrid TKRs over time (now only 2.4% of all knee replacements). Usage of each implant of this type has decreased proportionally to less than a quarter of those figures reported for 2004 (when they were 8.9% of all knee replacements).

Despite the trends in TKR, uncemented and hybrid unicondylar knee replacements are becoming increasingly popular, rising from less than 1% in 2010 to approximately 5% in 2021. Table 3.K2 also shows that in 2024 there was a record number of primary knee replacements (134,652), with the number of cemented unicondylar primary procedures rising above 10% of the yearly total for the first time.

Figure 3.K2 illustrates the temporal changes in fixation, highlighting the dominance of cemented TKR primaries.

Figure 3.K2

Fixation by year of procedure in primary knee replacement.

Table 3.K3 shows the age and sex distribution of patients undergoing primary knee replacement. The median age of a person receiving a cemented TKR was 70 years (IQR 64 to 76 years). Patients receiving cemented unicondylar prostheses were typically six years younger (median age 64 years; IQR 57 to 71) compared to all types of knee replacement, while those receiving uncemented / hybrid unicondylar prostheses were five years younger (median age 65 years; IQR 58 to 72). The patellofemoral group were typically 12 years younger (median age 58 years; IQR 50 to 66) compared to all types of knee replacement. Those receiving multicompartmental knee replacements were typically nine years younger (median age 61 years; IQR 54 to 68) compared to all types of knee replacement.

Table 3.K3

Age at primary knee replacement by fixation, constraint and bearing type.

Females who undergo a primary knee replacement are more likely to receive a TKR; they received 57.3%, 50.6% and 55.2% of cemented, uncemented and hybrid type procedures respectively. Conversely, cemented and uncemented / hybrid unicondylar surgery was performed on a higher proportion of males (54.3% and 55.5% respectively). Patellofemoral surgery was predominantly carried out on females (77% of patients) who are typically younger than a TKR or unicondylar patient, with a median age of 58 at time of operation.

Table 3.K4 shows the ASA grade, indication and age by sex for all primary knee replacements. ASA 2 is the most common ASA grade and only a small number of patients with a grade greater than ASA 3 undergo knee replacement. The majority of cases are performed with osteoarthritis as the sole indication; 1,738,995 (96.3%) of all 1,805,000 knee replacements.

Table 3.K4

Primary knee replacement patient demographics.

3.K.2. First revision after primary knee replacement surgery

In this section, estimates of cumulative revision in the tables are presented at 1, 3, 5, 10, 15 and 20 years. A total of 56,051 first revisions of a knee prosthesis have been linked to registry primary knee replacement surgery records of operations undertaken between 2003 and 2024. Figures 3.K3 (a) and (b) illustrate temporal changes in the overall revision estimates using Kaplan-Meier estimates; procedures have been grouped by the year of the primary operation.

Figure 3.K3 (a)

KM estimates of cumulative revision by year, in primary knee replacements.

Figure 3.K3 (b)

KM estimates of cumulative revision by year, in primary knee replacements plotted by year of primary.

Figure 3.K3 (a) plots each Kaplan-Meier curve with a common origin, i.e. time zero is equal to the year of operation.

Figure 3.K3 (b) shows the same Kaplan-Meier curves plotted against calendar time, where the origin of each curve is the year of operation. It separates each year enabling changes in revision estimates to be clearly identified. In addition, the revision estimates at 1, 3, 5, 7, 10, 13, 15, 17 and 20 years have been highlighted. If revision estimates and timing of revision estimates were static across time, it would be expected that all revision curves would be the same shape and equally spaced; a departure from this indicates a change in the number and timing of revision procedures. The cumulative probability of a knee joint being revised at three and five years increased for each operative year group between 2003 and 2008 and then reduced between 2009 and 2024. From the peak in 2008, the yearly survivorship curves are less divergent, i.e. a slowing in the observed trend.

Possible reasons for a peak in the probability of revision in the 2008 cohort out to ten years are firstly that the registry was not capturing the full range and number of operations taking place in hospitals in England and Wales until 2008, and secondly there could be bias in terms of the general overall health, risk of revision, and other key characteristics of the patients on record in the registry in the early years. Given that similar, more marked, patterns are observed in primary hip replacements and that the start of the reduction coincides with the timeline of when NJR Clinician Feedback and surgical performance analyses were introduced, it is likely that these patterns represent improved survival as a result of commencement of this feedback and the improved adoption of evidence-based practice.

Table 3.K5 shows Kaplan-Meier estimates of the cumulative percentage probability of first revision, for any cause, for the cohort of all primary knee replacements. This is broken down for TKR by knee fixation type (cemented, uncemented or hybrid) and sub-divided further within each fixation type by the constraint (unconstrained, posterior-stabilised, constrained condylar and pre-assembled, hinged or linked implants) and bearing mobility (fixed or mobile) and for UKR, by fixation type, laterality (medial or lateral) and bearing mobility (fixed or mobile). The table shows updated estimates at 1, 3, 5, 10, 15 and 20 years from the primary operation together with 95% Confidence Intervals (95% CI).

Table 3.K5

KM estimates of cumulative revision (95% CI) by fixation, constraint and bearing, in primary knee replacements. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Where groups have 250 or fewer cases remaining at risk, the figures are shown in blue italics. Further revisions in these groups would be highly unlikely, and when they do occur, they may appear to have a disproportionate impact on the Kaplan-Meier estimate, i.e. the step upwards may seem disproportionately large. Furthermore, the upper 95% CI at these time points may be underestimated. Although a number of statistical methods have been proposed to deal with this, they typically give different values and, as yet, there is no clear consensus for the large datasets presented here. Kaplan-Meier estimates are not shown at all when the numbers at risk fall below ten.

Figures 3.K4 (a) to 3.K4 (e) illustrate the differences in revision estimates between the types of knee replacement, fixation and constraint. It is worth noting the different vertical scales between the five figures. The results show the lowest revision estimates for cemented unconstrained fixed bearing TKRs and cemented TKRs with monobloc polyethylene tibias (Figure 3.K4 (a)). The revision estimates in cemented TKRs that are posterior-stabilised and those that have mobile bearings remain higher. The revision estimates of medial and lateral unicondylar replacements remain substantially higher than for TKRs, with lateral unicondylar knee replacements having a higher revision estimate than medial replacements (Figures 3.K4 (d) and 3.K4 (e)). Patellofemoral replacements have the highest revision estimates of all UKRs. Multicompartmental knee replacements have a higher revision estimate which is more than patellofemoral replacement up to seven years after surgery.

Figure 3.K4 (a)

KM estimates of cumulative revision in primary total cemented knee replacements by constraint and bearing. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.K4 (b)

KM estimates of cumulative revision in primary total uncemented knee replacements by constraint and bearing. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.K4 (c)

KM estimates of cumulative revision in primary total hybrid knee replacements by constraint and bearing. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.K4 (d)

KM estimates of cumulative revision in primary medial unicondylar, patellofemoral, and multicompartmental knee replacements by fixation, constraint and bearing. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk (more...)

Figure 3.K4 (e)

KM estimates of cumulative revision in primary lateral unicondylar, patellofemoral, and multicompartmental knee replacements by fixation, constraint and bearing. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk (more...)

It is important to note that the 20-year revision estimate of cemented, unconstrained knees with a fixed bearing is currently 5.28% in over one million patient procedures. Whilst the numbers at risk at 20 years is smaller, approximately 6,000 patients, the data is very encouraging.

Figure 3.K5 (a) shows that the chance of revision after primary TKR is far higher in younger patient cohorts and that males were slightly more likely, overall, to have a first revision compared to females of comparable grouped age, if they were under the age of 70 when they underwent primary surgery.

Figure 3.K5 (a)

KM estimates of cumulative revision in primary total knee replacements by sex and age. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.K5 (b) shows that the risk of revision of primary unicondylar knee replacement is substantially higher for younger patient cohorts, but that there are less marked differences in younger patients in the risk of revision according to sex. The risk of revision is approximately double in all age groups than it is for TKR. Please note the differences in the vertical axes between Figures 3.K5 (a) and (b).

Figure 3.K5 (b)

KM estimates of cumulative revision in primary unicondylar knee replacements by sex and age. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Table 3.K6 shows sex and age stratified Kaplan-Meier estimates of the cumulative percentage probability of first revision for any cause, firstly for all cases combined, then by knee fixation / constraint / bearing sub-divisions for TKR. Pre-assembled, hinged or linked knees are reported as a single group. Unicompartmental knee replacements are sub-divided into unicondylar, fixation, laterality, bearing type. Patellofemoral replacements are reported as a single group. Multicompartmental knees are stratified by the combination of compartments that they replace. Estimates are shown, along with 95% CIs, for males and females within each of four age bands, <55, 55 to 64, 65 to 74 and ≥75 years for revision estimates at 1, 3, 5, 10, 15 and 20 years after the primary operation.

Table 3.K6

KM estimates of cumulative revision (95% CI) by sex, age, fixation, constraint and bearing, in primary knee replacements. Blue italics signify that 250 or fewer cases remained at risk at these time points.

UKRs have worse revision estimates compared to TKRs, with the chance of revision at each estimated time point being approximately double or more than that of a TKR (Table 3.K5). The revision estimate for cemented (medial or lateral) unicondylar knee replacements is approximately three times higher than the observed estimate for cemented TKR at ten years and 3.6 times higher at 20 years. The revision estimate for uncemented (medial or lateral) unicondylar knee replacements is 2.2 times higher than for cemented TKR at ten years and three times higher at 15 years. The revision estimate for patellofemoral replacement is 5.5 times higher than for cemented TKR at ten years and 5.9 times higher at 20 years although, we advise that this is viewed with a degree of caution since the number of patellofemoral replacements at risk at 20 years is small. Multicompartmental knee replacements have relatively small numbers, and at ten years the risk of revision is 4.7 times higher than for cemented TKR, 1.6 times higher than for cemented unicondylar knee replacements and 2.1 times higher than for uncemented unicondylar knee replacements. The early (less than 10-year) revision estimates of multicompartmental knee replacements are elevated compared to patellofemoral replacements.

First revision of an implant is slightly less likely in females than in males overall for cemented TKR but, broadly, a patient from a younger age group is more likely to be revised irrespective of sex, with the youngest group having the worst predicted outcome in terms of the risk of subsequent revision (Table 3.K6). Conversely, female patients are more likely to have a unicondylar implant revised in the longer-term compared to their male, age-equivalent counterparts. For patellofemoral implants, males are generally more likely to undergo revision than their age-matched female counterparts.

The numbers of multicompartmental knee replacements are small in the age and sex-stratified analyses, but overall, the risk of revision is markedly higher than all modular and unlinked TKR and higher than a patellofemoral replacement in the first ten years after surgery.

3.K.3. Revisions after primary knee replacement surgery by main brands for TKR and UKR

As in previous reports, only brands that have been used in a primary TKR in 1,000 or more operations have been included (Tables 3.K7 (a) and Table 3.K8). Table 3.K7 (b) shows a breakdown of the brands included in Table 3.K7 (a) according to whether the patella was resurfaced or not at the time of the primary procedure. In Table 3.K9 (a) brands are displayed with a breakdown according to fixation, constraint and bearing mobility where there are more than 1,000 operations for TKR and UKR procedures and more than 100 operations for multicompartmental procedures. Table 3.K9 (b) provides an additional breakdown for the TKRs displayed in Table 3.K9 (a) according to whether the patella was resurfaced at the time of primary procedure or not. The reduction in reporting threshold from 2,500 procedures, as in reports prior to 2024, was used to ensure visibility of the broader range of constructs that can be identified within the new component classification.

Table 3.K7 (a)

KM estimates of cumulative revision (95% CI) by total knee replacement brands. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Table 3.K7 (b)

KM estimates of cumulative revision (95% CI) in total knee replacement brands by whether a patella component was recorded. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Table 3.K8

KM estimates of cumulative revision (95% CI) by unicompartmental knee replacement brands. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Table 3.K9 (a)

KM estimates of cumulative revision (95% CI) by fixation, constraint and brand. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Further breakdowns by component are available from other sources such as ODEP, but they may use external data sources which are not generalisable to our registry population. The figures in blue italics are at time points where 250 or fewer primary knee replacements remained at risk. No results are shown where the number had fallen below ten cases. We have made no attempt to adjust for other factors that may influence the chance of revision, so the figures are unadjusted probabilities. Given that the sub-groups may differ in composition with respect to age and sex, the percentage of males and the median (IQR) of the ages are also shown in these tables.

Tables 3.K7 (a) and (b) and Table 3.K8 show the Kaplan-Meier estimates of the cumulative percentage probability of first revision, for any indication, of a primary TKR (Tables 3.K7 (a) and (b)) and primary UKR (Table 3.K8) by implant brand.

Table 3.K9 (a) shows Kaplan-Meier estimates of the cumulative percentage probability of first revision of a primary TKR or primary UKR by implant brand and bearing / constraint type for those brands / bearing types which were implanted on at least 1,000 occasions for TKR and UKR and 100 occasions for multicompartmental procedures. Patient summaries of age and sex by brand are also given. There are many brands achieving less than 3% revision at ten years used in hundreds of thousands of patients. It is important to note Table 3.K9 (b) shows the variation by brand, constraint, bearing mobility and whether or not the patella had been resurfaced. It is very clear that the performance of some specific brands depends on constraint, mobility and whether or not the patella had been resurfaced, whereas for other brands it does not.

Table 3.K9 (b)

KM estimates of cumulative revision (95% CI) by fixation, constraint, brand and whether a patella component was recorded. Blue italics signify that 250 or fewer cases remained at risk at these time points.

3.K.4. Revisions for different indications after primary knee replacement

Table 3.K10 shows the revision incidence rates for each indication recorded on data collection forms for knee revision surgery, for all cases and then sub-divided by fixation type and whether the primary procedure was a TKR or a UKR.

Table 3.K10

PTIR estimates of indications for revision (95% CI) by fixation, constraint, bearing type and whether a patella component was recorded.

For all knee replacements, the highest Prosthesis Time Incidence Rates (PTIRs) for the five most common indications for revision were for, in descending order: aseptic loosening / lysis, infection, progressive arthritis, instability, and pain. For cemented TKR, the highest PTIRs, also in descending order were aseptic loosening / lysis, infection, instability, pain and ‘other’ indication. Revision incidences for TKRs which were uncemented were lower than cemented TKR for infection, similar for periprosthetic fracture but higher for all other recorded indications.

For cemented unicondylar knee replacements (medial or lateral), the highest three incidence rates for indications for revising the implant were for: progressive arthritis, aseptic loosening / lysis and pain, respectively. For uncemented / hybrid unicondylar knee replacements (medial or lateral) the highest rates were for: progressive arthritis, dislocation / subluxation, and implant wear. The incidence of revision for pain, aseptic loosening / lysis, implant wear and progressive arthritis were lower for uncemented / hybrid fixation than for cemented but the incidence was higher for dislocation / subluxation and periprosthetic fracture. For patellofemoral replacements, the top three indications for revision were: progressive arthritis, pain, and ‘other’ indication. For multicompartmental knee replacements, the highest incidence for revision was for progressive arthritis, pain and aseptic loosening / lysis.

In Table 3.K11, the PTIRs for each indication are shown separately for different time periods from the primary knee replacement, within the first year from primary operation, and between 1 to <3, 3 to <5, 5 to <7, 7 to <10, 10 to <13, 13 to <15, 15 to <17 and ≥17 years after surgery (the maximum follow-up for any implant is now 21.75 years). Most of the PTIRs for a particular indication do vary, especially for infection, aseptic loosening / lysis, pain and progressive arthritis for different time intervals after surgery. Infection is most likely to be the reason that a joint is revised in the first year but after seven years or more, is comparatively less likely than some of the other reasons. Conversely, revision between one and three years after surgery is more likely for aseptic loosening / lysis and pain, with incidence rates dropping off for pain later but rising again for aseptic loosening / lysis. Aseptic loosening / lysis and implant wear PTIRs continue to remain relatively higher than other indicated reasons for revision for implants surviving for longer periods after surgery.

Table 3.K11

PTIR estimates of indications for revision (95% CI) by years following primary knee replacement.

3.K.5. Mortality after primary knee replacement surgery

In this section we describe the mortality of the cohort up to 21 years from primary operation, according to sex and age group. Deaths recorded after 31 December 2024 have not been included in the analysis. For simplicity, we have not considered whether the patient had a first (or further) joint revision after the primary operation when calculating the cumulative probability of death (see survival analysis methods note on page 36). Of the 1,805,000 records of a primary knee replacement, 29,811 unknown knee type records were excluded and there were 16,813 bilateral operations in which the patient had both knees replaced on the same day; here the second of the two has been excluded, leaving 1,558,302 TKR procedures (of whom 393,766 had died before the end of 2024) and 200,666 UKR or multicompartmental procedures (of whom 24,206 died before the end of 2024).

Note: These cases were not censored when further revision surgery was undertaken. While such surgery may have contributed to the overall mortality, the impact of this is not investigated in this report. Furthermore, exclusions for unknown knee type and same-day bilateral operations were not mutually exclusive; there was an overlap of 592 cases of unknown knee types with same day bilateral procedures.

Tables 3.K12 (a) and (b), show Kaplan-Meier estimates of cumulative percentage mortality at 30 days, 90 days and at 1, 5, 10, 15 and 21 years following a TKR, UKR or MKR, for all cases and by age and sex. Fewer males than females have had a primary knee replacement and, proportionally, more females than males undergo surgery above the age of 75. Males, particularly in the older age groups, had a higher cumulative percentage probability of dying in the short or longer-term after their primary knee replacement operation than females in the equivalent age group. The mortality estimates are lower in males and females following UKR than TKR, but these figures do not adjust for selection and hence do not account for residual confounding (Hunt et al., 2018). Similarly, the overall mortality estimates in multicompartmental knee replacements are lower than unicondylar knee replacements and patellofemoral replacements. As these analyses are unadjusted these differences are likely to be explained by the number of younger patients that receive this type of joint replacement.

Table 3.K12 (a)

KM estimates of cumulative mortality (95% CI) by age and sex, in primary TKR. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Table 3.K12 (b)

KM estimates of cumulative mortality (95% CI) by age and sex, in primary unicompartmental replacements. Blue italics signify that 250 or fewer cases remained at risk at these time points.

It is important for surgeons to consider the longevity of knee replacements when operating on patients. A knee replacement with the lowest 20-year failure rate will likely be the only knee replacement that an 80-year-old, male or female, patient will require.

3.K.6. Overview of knee revisions

In this section we look at all recorded knee revision procedures performed since the registry began collecting data on 1 April 2003 up to the end of December 2024, for all patients with valid patient identifiers (i.e. whose data could be linked).

In total there were 111,772 revision procedures (Table 3K.13). These revisions were recorded on 91,901 individual knees in 87,066 patients. In addition to the 56,051 revised primaries described previously, there were 42,695 additional revisions of a knee for which there is no associated primary operation recorded in the registry. The remaining 13,026 revision procedures were re-revisions, i.e. revision procedures subsequent to the first revision.

Table 3.K13

Number and percentage of revisions by procedure type and year.

We have classified revisions as single-stage, stage one of two-stage, or stage two of two-stage revisions. Information on stage one and stage two of two-stage revisions is entered into the registry separately. Debridement and Implant Retention (DAIR) with or without modular exchange are included as single-stage procedures. With the introduction of distinct indicators for the DAIR with or without modular exchange procedures in MDSv7 and the introduction of a separate reoperations form in MDSv8, which captures the DAIR without modular exchange procedures, it may be possible to report these as distinct categories in future reports. Not all patients who undergo stage one of a two-stage revision will undergo a stage two of two-stage revision. In some cases, stage one revisions have been entered without stage two, and vice versa, making identification of entire patient revision episodes difficult. We have attempted to address this later in this section.

The NJR asks surgeons and those responsible for healthcare delivery to ensure that when primary and revision joint replacement procedures of the hip, knee, ankle, elbow or shoulder are performed, that the relevant MDS form is completed and data entered into the registry. This is a requirement mandated by the NHS Standard Contract. For the purposes of the Annual Report, revision procedures include any addition, removal or modification of the implants and procedures such as debridement and implant retention with implant exchange, excision arthroplasty, amputation and conversion to arthrodesis. For data submitted on MDSv7 only, DAIRs without modular exchange are included as revision procedures. The completion of a revision MDS form is also mandatory for a procedure involving modification of a joint by adding another implant to another part of the joint. For the analyses of surgeon performance, hospital performance and implant performance, debridement and implant retention (DAIR) without implant exchange is currently excluded.

Table 3.K13 gives an overview of all knee revision procedures carried out each year since April 2003. There were a maximum of 15 documented revision procedures associated with any individual knee, making up 14 revision episodes as one episode consisted of a stage one and a stage two of a two-stage procedure. The increase in the number of operations over time, until 2020 when the volume of procedures was impacted by COVID, is likely to reflect the increasing number of at-risk implants prevailing in the dataset.

Table 3.K14 (a) shows the stated indications for the revision knee surgery. As more than one reason can be selected, i.e. indications are not mutually exclusive, column percentages do not add up to 100%. Aseptic loosening / lysis is the most common indication for revision, accounting for over 36% of single-stage revision operations, while instability, wear, pain and other indications account for between 10% and 20% each.

Table 3.K14 (a)

Number and percentage of knee revision by indication and procedure type.

Of the two-stage revision operations, infection is the main indication recorded in approximately 80% of either stage one or stage two procedures. Table 3.K14 (b) presents these results, restricted to the last five years of data, and shows that aseptic loosening / lysis represents a lower proportion of revisions in this period but progressive arthritis, infection and periprosthetic fracture represent a greater proportion of cases revised.

Table 3.K14 (b)

Number and percentage of knee revision by indication and procedure type in the last five years.

3.K.7. Estimates of knee re-revision

In most instances (87.3%), the first revision procedure was a single-stage revision, in the remaining 12.7% it was part of a two-stage procedure. For a given knee, the implant survival following the first documented revision procedure linked to a primary in the registry (n=56,051) has been analysed. This analysis is restricted to patients with a linked primary procedure so that there is confidence that the next observed procedure on the same joint is the first revision episode. If there is no linked primary record in the dataset, it cannot be determined if the first observed revision is the first revision or has been preceded by other revision episodes. The time from the first documented revision procedure (of any type) to the time at which a second revision procedure was undertaken has been determined. For this purpose, an initial stage one followed by either a stage one or a stage two of a two-stage procedure have been considered to be the same revision episode and so these were disregarded, looking instead for the start of a second revision episode.

The maximum number of distinct revision episodes for any knee was determined to be 14. In cases where a stage one of two procedure was followed by a stage two of two procedure within 365 days, we have treated this as a single distinct episode. This definition allows multiple stage one procedures to occur before a new revision episode is triggered. In situations where the first stage one procedure is not followed by a stage two procedure within a 365-day period, the next occurrence of a stage one procedure was considered as a new revision episode.

Kaplan-Meier estimates of the cumulative percentage probability of having a subsequent revision (re-revision) were calculated. There were 6,554 re-revisions and for 10,451 cases the patient died without having been re-revised. The censoring date for the remainder was the end of 2024.

Figure 3.K6 (a) plots Kaplan-Meier estimates of the cumulative probability of a re-revision in linked revised primary knee replacements grouped by type of primary implant between 1 and 20.94 years since the primary operation.

Figure 3.K6 (a)

KM estimates of cumulative re-revision, in linked primary knee replacements (shaded area indicates point-wise 95% CI). Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.K6 (b) shows estimates of re-revision by type of primary knee replacement. Revised patellofemoral knee replacements have the lowest risk of re-revision, although after 12 years the numbers at risk fall to 250 or fewer and should be interpreted with caution. Revised cemented unicondylar knee replacements have the next lowest risk of re-revision until 16 years, after which the numbers at risk become too small for an accurate interpretation.

Figure 3.K6 (b)

KM estimates of cumulative re-revision by primary fixation, in linked primary knee replacements. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Revised uncemented / hybrid unicondylar knee replacements appear to have a higher risk of re-revision than their cemented counterparts and are equivalent to the estimates seen for revised cemented TKRs until eight years, after which the numbers in the revised uncemented / hybrid unicondylar group become small.

It is important to note that interpretating the risk of revision in isolation is challenging, it must be considered in light of the revision risk of the primary procedure. For example, the low re-revision estimate of patellofemoral replacements must be considered with due deference to the high primary revision estimate of this specific type of knee replacement.

Figure 3.K6 (c) shows the relationship between time to first revision and risk of subsequent revision. The earlier the primary knee replacement is revised, the higher the risk of second revision. For example, if a primary knee replacement is revised within the first year of the primary replacement being performed, there is an 9.1% (95% CI 8.5-9.8) re-revision estimate at one year following the first revision, rising to 19.8% (95% CI 18.9-20.7) by five years; if a primary knee replacement is not revised until between five and seven years after the primary procedure, the re-revision estimate is 2.7% (95% CI 2.3-3.1) at one year following the first revision, rising to 8.3% (95% CI 7.6-9.1) by five years. It is worth noting here that these results are unadjusted and are potentially confounded by indication.

Figure 3.K6 (c)

KM estimates of cumulative re-revision by years to first revision, in linked primary knee replacements. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

For those with documented primary knee replacements within the registry, Figures 3.K7 (a) to (f) show cumulative re-revision estimates following the first revision, according to the main type of primary knee replacement. We have further sub-divided each sub-group according to the time interval from the primary to the first revision, i.e. less than 1 year, 1 to <3, 3 to <5, 5 to <7, 7 to <10, and greater than or equal to 10 years, where sample size permitted. 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. However, for hybrid TKRs, the worst re-revision estimates were experienced by those who had their first revision within three to five years of the initial primary knee replacement. However, the numbers at risk were small in each of the hybrid stratifications and therefore we advise that the results should be interpreted with caution.

Figure 3.K7 (a)

KM estimates of cumulative re-revision in primary cemented TKRs by years to first revision. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.K7 (b)

KM estimates of cumulative re-revision in primary uncemented TKRs by years to first revision. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.K7 (c)

KM estimates of cumulative re-revision in primary hybrid TKRs by years to first revision. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.K7 (d)

KM estimates of cumulative re-revision in primary patellofemoral knee replacements by years to first revision. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.K7 (e)

KM estimates of cumulative re-revision in primary cemented unicondylar knee replacements by years to first revision. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.K7 (f)

KM estimates of cumulative re-revision in primary uncemented / hybrid unicondylar knee replacements by years to first revision. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Table 3.K15 (a) shows the re-revision rate of the 56,051 revised primary knee replacements (54,419 (97.1%) with known knee type at primary procedure) recorded in the registry. Of these, 6,554 were re-revised.

Table 3.K15 (a)

KM estimates of cumulative re-revision (95% CI). Blue italics signify that 250 or fewer cases remained at risk at these time points.

Table 3.K15 (b) shows that primary knee replacements that are revised within the first year after surgery have approximately two to four times the chance of needing re-revision at each time point compared with primaries that last more than five years.

Table 3.K15 (b)

KM estimates of cumulative re-revision (95% CI) by years since first revision. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Table 3.K15 (c) shows cumulative re-revision estimates at 1, 3, 5, 10, 15 and 18 years (data not presented for 20 years due to low numbers) following the first revision for those with documented primary knee replacements within the registry, broken down by type of knee replacement, constraint, mobility and whether a patellar component was recorded. Overall, the worst re-revision estimates were demonstrated in those where the initial primary had been a cemented TKR, uncemented TKR or an uncemented unicondylar knee replacement although the confidence intervals broadly overlap after five years in the cemented TKR group and earlier in the other groups. However, it should be noted that interpreting re-revision estimates is challenging and due consideration should be given to the primary revision estimate.

Table 3.K15 (c)

KM estimates of cumulative re-revision (95% CI) by fixation and constraint and whether a patella component was recorded. Blue italics signify that 250 or fewer cases remained at risk at these time points.

3.K.8. Reasons for knee re-revision

Tables 3.K16 (a) and (b) show a breakdown of the stated indications for the first revision and for any second revision. Please note the indications are not mutually exclusive. Table 3.K16 (a) shows the indications for all knee revisions recorded in the registry and Table 3.K16 (b) reports the indications for the first linked revision and the number and percentage of first recorded revisions that were subsequently re-revised. The final column reports the indications for all the second linked revisions. It is interesting to note that infection, dislocation / subluxation, instability and stiffness are more common indications for second revision than for a first revision. This reflects the factors that infection, surgical complexity and soft tissue elements also contribute to the outcome of revision knee replacement.

Table 3.K16 (a)

Number of revisions by indication for all revisions.

Table 3.K16 (b)

Number of revisions by indication for first linked revision and second linked re-revision.

Tables 3.K17 (a) and (b) show that the number of revisions and the relative proportion of revisions with an associated primary in the registry increased with time. The number of revisions peaked in 2019 (5,768) before the impact of COVID. The number of revisions has only partly recovered to 5,239 in 2024. More than 82% of those revisions performed in 2024 had a linked primary in the registry. We propose that this is likely to reflect improved data capture over time, improved linkability of records and the longevity of knee replacements, with a proportion of those primaries being revised having been performed before registry data capture began or are outside the coverage of the registry.

Table 3.K17 (a)

Number of revisions by year.

Table 3.K17 (b)

Number of revisions by year, stage, and whether or not primary is recorded in the registry.

3.K.9. 90-day mortality after knee revision

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 0.90% (95% CI 0.83-0.98) versus 1.15% (95% CI 1.04-1.26)), which may reflect the fact that this patient group was younger at the time of their first revision, with a median age of 69 (IQR 61 to 76) years, compared to the group without primaries documented in the registry who had a median age of 73 (IQR 65 to 79) years. The percentage of males was similar in both groups (45.1% versus 46.6% respectively).

3.K.10. Conclusions

There are now over 1.8 million primary knee replacements recorded in the registry with a maximum follow-up of 21.75 years, making this the largest dataset of its kind in the world. Of these, 96.3% of the procedures were performed for osteoarthritis as the only indication. In 2024, 134,652 primary knee replacements procedures were recorded in the registry, the largest number entered into the registry in a single year. The introduction of the new component classification and database has allowed us to provide additional categories to this year’s report. Examples include the separation of medial and lateral unicompartmental knee replacements, capture of multicompartmental primary knee replacements and increased granularity to define the implant constructs used in knee replacement as reflected in the brand revision outcome tables. To allow transparent and inclusive reporting, we have reduced the threshold for inclusion of brands in these tables and adopted a lower threshold for lower-volume procedures to enable comparison of outcomes.

Within the registry, approximately 86.5% of recorded procedures are TKRs, 10.3% medial or lateral unicondylar knee replacements and 1.1% patellofemoral replacements. The annual proportion of unicondylar knee replacements has risen since 2013, reaching approximately 10% for the first time in 2017 and rising to 14.7% in 2024, with cemented unicondylar above 10% of all primaries for the first time. The popularity of uncemented unicondylar replacements has risen relatively rapidly. These made up less than 1% of knee replacements in 2010 and accounted for 5.2% in 2021, although this appears to be reducing in the last couple of years. Uncemented or hybrid unicondylar primaries now represent a quarter of all the primary unicondylar knee replacements performed. This increase in the proportion of primary knee procedures that are unicondylar knee replacements is supported by guidance from NICE published in 2020 and Quality Standards published by NICE in 2022 (NICE,2020; NICE,2022).

Cemented, unconstrained (cruciate retaining), fixed bearing TKR remains by far the most common type of knee replacement with excellent results at 20 years after surgery with an expected revision estimate of 5.28%, followed by cemented, posterior-stabilised, fixed bearing TKR, which only have a marginally higher revision estimate at 20 years of 6.23%. Patients who received unicondylar or patellofemoral knee replacement were typically younger than those receiving a TKR. Both TKR and patellofemoral replacement are more likely to be performed on females, whereas unicondylar knee replacement is more likely to be performed on males.

TKRs with a monobloc polyethylene tibia consistently show some of the lowest unadjusted revision estimates, although the numbers at risk at 20 years are small, so must be interpreted with caution. Cemented TKRs that are unconstrained with a fixed bearing, as well as being the most common type of TKR, consistently show low revision estimates in comparison to alternatives; unadjusted revision estimates are approximately one percentage point lower in comparison to cemented unconstrained TKRs with a mobile bearing and cemented TKRs that are posterior-stabilised, with either a fixed or mobile bearing at 15 years.

Age and sex are associated with the risk of revision surgery. Younger patients and males are more likely to undergo revision, and it has previously been felt that this may explain the higher revision estimates observed in UKR. We present results divided by sex and age group and these show the risk of revision of a cemented unicondylar knee replacement is approximately twice as high in males and 2.4 times higher in females at ten years than a cemented TKR. Uncemented mobile unicondylar knee replacements show that revision estimates are lower than for cemented mobile unicondylar replacements, but remain higher than those for cemented TKR. The risk of revision of a patellofemoral replacement is at least 2.9 times higher in both males and females than a cemented TKR across all age groups at ten years and the results of multicompartmental knee replacements show similarly high revision estimates. The emerging niche use of multicompartmental knee replacements and their associated high revision estimates is concerning.

The most common causes of revision across all primary knee replacements were aseptic loosening / lysis, infection and progressive arthritis. For uncemented TKRs, the incidence of revision for infection was lower than for cemented TKRs but higher for nearly all other indications. Progression of osteoarthritis elsewhere in the knee is also the fourth most common indication for revision knee replacement. The risk of revision for progressive arthritis, aseptic loosening / lysis and pain were all higher for UKRs than TKRs, but the risk of revision for infection was lower. For cemented unicondylar knee replacements, the highest risk of revision was for progressive arthritis, aseptic loosening / lysis and pain. For uncemented unicondylar knee replacements, the second most common indication was dislocation / subluxation ahead of implant wear as the third most common indication. The incidence of revision for indications such as pain and aseptic loosening / lysis was lower for uncemented unicondylar knee replacements than for cemented, but higher for dislocation / subluxation and periprosthetic fractures.

Infection accounts for the majority of the two-stage revision procedures performed. Approximately 9% of revisions for infection that have been recorded in the registry to date have been single-stage procedures. At this time, the single-stage group includes DAIR procedures, so this indicates low usage and take-up of single-stage revision in the treatment of knee prosthetic joint infection. The soft tissue envelope makes single-stage knee revision surgery potentially more challenging than that in the hip, which may explain the differences in utilisation of a single-stage approach.

The risk of re-revision of a UKR following a revision procedure is higher than the risk of revision of a primary TKR across all types of knee replacement. However, there is diversity in the revision estimates for UKRs that can be influenced by implant choice and surgeon procedure volume. The risk of re-revision of a revised patellofemoral replacement is slightly lower than the other types of knee, with the rest being broadly similar. The risk of re-revision is higher for those revised after a shorter period of time following the primary and is associated with the specific indication for revision. This suggests that not all of the processes that lead to revision are the same and that some have greater impact than others with consequences beyond the initial revision.

Knee replacement remains a safe procedure with low rates of peri-operative mortality. The rates of mortality are higher for males than those for females. The average age of a patient undergoing a TKR is approximately 70 years; approximately 56% of males and 45% of females in the 70 to 74 age bracket will have died within 15 years of their knee replacement. This means that for the average patient undergoing a knee replacement, their knee replacement should last them for the rest of their life, without the need for revision surgery.

3.A.1. Overview of primary ankle replacement surgery

In this section of the report, we look at revision and mortality for all primary ankle operations performed between 1 January 2010 and 31 December 2024, and submitted before 1 March 2025. There were, after data cleaning, 11,321 primary ankle operations available for analysis on 10,695 patients. A total of 626 patients had bilateral operations (11 had both sides operated on the same day), which can be seen in the patient flow diagram in Figure 3.A1.

Figure 3.A1

Ankle cohort flow diagram.

The median age at primary surgery was 69 years (IQR 62 to 75 years), with an overall range of 17 to 97 years. More procedures were performed in men (60.6%) than in women.

All ankle replacement brands recorded in the registry are uncemented implants, but cement can be used occasionally by surgeons in circumstances such as poor bone stock or low-demand patients. Of the 11,321 primary procedures, a total of 10,926 (96.5%) procedures were implanted without cement being listed in the component data. Cement was listed in 395 (3.5%) primary procedures. Of all total ankle replacement (TAR) procedures, 241 (2.1%) were classified as unconfirmed. Figure 3.A2 illustrates the temporal changes in fixation of primary ankle replacements.

Figure 3.A2

Fixation by year of primary ankle replacement.

Figure 3.A3 describes the funding status and organisation type (based on organisation type in 2024) of ankle replacement procedures collected by the NJR. Prior to 2020 (the start of COVID) we saw an increase in the absolute number of ankle replacements being provided, which in part was facilitated by an expansion of NHS-funded procedures in both the NHS and the independent sector. As of 2024 the recovery of ankle replacements, due to an expansion of provision within NHS hospitals, as well as a substantial increase in the number of independently-funded procedures, has markedly overtaken the pre-2020 data.

Figure 3.A3

Frequency of elective primary ankle replacements by funding status and organisation type, per year.

Figure 3.A4 and Figure 3.A5 show the yearly number of primary ankle replacements performed for all indications and ankle replacements stratified by fixed and mobile bearings, please note the difference in scale of the y-axis between each plot. Each bar in the figure is further stratified by the volume of procedures that the surgeon conducted in that year, and when procedures are stratified by fixed and mobile bearings the volume of procedures is calculated separately. For example, if a surgeon performed 25 primary ankle replacement procedures of one prosthetic type, their procedures would have contributed to the grey sub-division in Figure 3.A4. If those procedures consisted of 12 fixed bearings and 13 mobile bearings, those procedures would be represented by green and purple bars respectively in Figure 3.A5. In total, 59 consultants split their practice between mobile and fixed prostheses within the same year.

Figure 3.A4

Frequency of primary ankle replacements, bars stacked by volume per consultant per year.

Figure 3.A5

Frequency of primary ankle replacements stratified by fixed and mobile bearings, bars stacked by volume per consultant per year. Graphs by confirmed procedure type.

Figure 3.A4 shows the volume of primary ankle replacements recorded in the registry increasing since 2015 (except for a large drop in 2020 due to the impact of COVID). Whilst the number of primary ankle replacements performed by higher-volume ankle surgeons (i.e. surgeons who perform 13 or more TAR procedures annually) has increased substantially, the number performed by low-volume surgeons (i.e. surgeons who perform six or fewer TAR procedures annually) has also increased. Figure 3.A5 illustrates that the expansion of TAR procedures has largely been of a fixed bearing design and that the use of mobile bearing has steadily been decreasing, although with some recovery since COVID.

Table 3.A1 shows the number of annually reported cases over the 14-year observation period. In 2019 there were just over one thousand ankle replacement procedures being performed annually. The COVID pandemic resulted in a reduction of procedures with approximately half the volume of procedures being conducted in 2020 compared to 2019. The volume of procedures has again started to increase, and in 2024 the number of procedures was considerably higher than in 2019.

Table 3.A1

Descriptive statistics of ankle procedures performed by consultant and hospital by year of surgery.

A total of 357 consultants carried out the 11,321 reported primary procedures over the 14-year period. The annual median number of procedures per consultant was two in 2010 and four in 2024, with little change in the median number performed since 2018. We can see that 8.0% of consultants performed 20 or more primary ankle replacements in 2024, with a further 16.6% performing between 10 and 19 of these. Of the 308 hospitals who submitted data to the registry, 20 (6.5%) had carried out 20 or more procedures in any single year since the start of data collection. The percentage of hospitals submitting 20 or more ankle primary operations each year does not exceed 8.6%. The number of hospitals submitting more than 20 primary ankle procedures per year has changed from three in 2010 to 16 in 2024 and the median number of primary replacements per unit has also changed from two to four respectively across the same time period.

Table 3.A2 shows the number of replacements by implant brand and year of primary operation. The most frequently used brand is the fixed bearing, Infinity, which represented 66.1% of primary ankle replacements performed in 2024. The use of this brand has risen considerably since its introduction in 2014.

Table 3.A2

Number and percentage of primary ankle replacements by ankle brand.

The NJR identifies when components within joint replacements come from different brands and/or manufacturers (termed as mix and match). There are no examples of mix and match between manufacturers for ankle replacements. The Infinity and Inbone implants, both manufactured by the same company, were designed to be interchangeable with a matched articulating surface. This combination represented 8.3% of primary ankle replacements in 2024. The most common brand in 2024 was the Infinity tibial component representing 74.2% of all ankle replacements, with the tibial implant matched to the Infinity talar component in 66.1% of cases (Infinity), and the Inbone talar component in 8.1% of cases (Infinity[Tibial]Inbone[Talar]). The Hintegra was implanted in 7.7% of procedures, and Vantage in 6.3% of procedures. It was not possible to identify the type of constructs implanted in 16 procedures.

3.A.2. Revisions after primary ankle replacement surgery

A total of 615 out of the 11,321 primary procedures had a linkable A2 MDS form completed to indicate a revision before the end of 2024. The first revisions shown here include 78 conversions to arthrodesis, 400 single-stage procedures, 105 two-stage procedures, 32 DAIRs, 23 with modular exchange and nine without. No other procedures, such as amputations, have been recorded. Given the low number reported for conversion to arthrodesis, we believe that these small numbers are likely to be a reflection of under-reporting (Jennison et al., 2023).

Figure 3.A6 and Table 3.A3 show the overall estimated cumulative percentage probability of (first) revision over time. Table 3.A3 and Figure 3.A7 show the same results stratified by sex and age at primary. Younger people, and particularly younger women, were more likely to experience a revision.

Figure 3.A6

KM estimates of cumulative revision of primary ankle replacement (shaded area indicates point-wise 95% CI).

Table 3.A3

KM estimates of cumulative revision (95% CI) of primary ankle replacement, by sex and age. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Figure 3.A7

KM estimates of cumulative revision of primary ankle replacement, by sex and age. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Figure 3.A8 shows estimated cumulative revision curves plotted against calendar time, where the origin of each curve is the year of operation. For example, the yellow line shows cumulative revision over time for primary ankle operations conducted in 2016. In addition, we have highlighted the revision estimates at 1, 3, 5, 7, and 10 years, using red dashed lines with different markers. For example, the red dashed line with ‘x’ markers, representing revision estimate at 1 year, has a first marker in 2011 (representing revision at 1 year after 2010). If revision surgery and timing of revision surgery were static across time, it would be expected that all the revision curves would be the same shape and equally spaced; departures from this indicate a change in the number and timing of revision procedures. The 3, 5, 7, and 10-year rates of revision initially increase over time, and then reduce for operations occurring after 2014. The early increases in revision estimates may partly be a result of under-reporting in the earlier years of the registry. The later decreases in revision estimates can largely be attributed to the voluntary withdrawal of the Mobility implant and the introduction of the Infinity implant during 2014.

Figure 3.A8

KM estimates of cumulative revision by year, in primary ankle replacements plotted by year of primary.

Table 3.A4 and Figure 3.A9 show the estimated cumulative percentage probability of (first) revision by implant brand with at least 250 uses. Estimates are not reported when there are fewer than ten primary procedures at risk of revision for the considered time period. At one year post-operation, estimates of revision were heterogeneous between brands, varying from 0.20% (95% CI 0.03-1.38) to 1.50% (95% CI 0.63-3.57). Larger variations between brands were observed for later post-operative periods, with estimates varying from 0.41% (95% CI 0.06-2.90) to 8.53% (95% CI 6.78-10.72) at five years post-operation. At ten years post-operation, the 95% confidence intervals are large, overlapping each other and making interpretation difficult.

Table 3.A4

KM estimates of cumulative revision (95% CI) of primary ankle replacement, by brand. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Figure 3.A9

KM estimates of cumulative revision of primary ankle replacement, by brand. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Figure 3.A10 shows the estimated cumulative percentage probability of (first) revision by implant brand, stratified by sex with at least 250 uses overall. The large relative differences between the lowest and highest estimates seem to be related to the implant’s brand and are unlikely to be entirely due to patient sex.

Figure 3.A10

KM estimates of cumulative revision of primary ankle replacement, by brand for males and females. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Table 3.A5 shows the indications for revision of ankle replacements, with aseptic loosening and infection as the most commonly cited indications.

Table 3.A5

Indications for the first revisions following primary ankle replacement. Note: These are not mutually exclusive.

Of the revisions for infection, 35 (20.6%) were recorded as having a high suspicion of infection (e.g. pus or confirmed micro), whilst 120 (70.6%) had a low suspicion of infection (awaiting microscopic histology). From MDSv8, there is no distinction between high and low suspicion infection. Out of the 271 revisions for aseptic loosening, 41.7% were performed because of loosening of both the tibial and talar components and 35.0% of patients revised for an indication of lysis had lysis of both tibial and talar components.

Of the 31 revisions for implant fracture, 27 (87.1%) were performed for a fractured meniscal insert and fewer than four were performed to treat implant fracture of both tibial and talar components.

The NJR asks surgeons and those responsible for healthcare delivery to ensure that when primary and revision joint replacement procedures of the hip, knee, ankle, elbow or shoulder are performed, that the relevant MDS form is completed and data entered into the registry. This is a requirement mandated by the NHS Standard Contract. For the purposes of the Annual Report, revision procedures include any addition, removal or modification of the implants and procedures such as debridement and implant retention with implant exchange, excision arthroplasty, amputation and conversion to arthrodesis. For data submitted on MDSv7 only, DAIRs without modular exchange are included as revision procedures. The completion of a revision MDS form is also mandatory for a procedure involving modification of a joint by adding another implant to another part of the joint. For the analyses of surgeon performance, hospital performance and implant performance, debridement and implant retention (DAIR) without implant exchange is currently excluded.

3.A.3. Mortality after primary ankle replacement surgery

In this analysis, the second of each of the 11 (same day) bilateral procedures were excluded. Among the remaining 11,310, a total of 1,341 patients had died before the end of 2024, 456 of these were female and 885 were male.

Figure 3.A11, Table 3.A6, and Figure 3.A12 show the estimated cumulative percentage probability of death at different times after surgery, by sex and age at primary. Male patients and patients of older age were more likely to have died.

Figure 3.A11

KM estimates of cumulative mortality after primary ankle replacement (shaded area indicates point-wise 95% CI).

Table 3.A6

KM estimates of cumulative mortality (95% CI) after primary ankle replacement, by sex and age group. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Figure 3.A12

KM estimates of cumulative mortality after primary ankle replacement by sex and age group. Blue italics signify that 250 or fewer cases remained at risk at these time points.

3.A.4. Conclusions

Compared to hip, knee, and shoulder replacements included in the NJR Annual Report, primary ankle replacement is a low-volume procedure, and linked first revisions consequently amount to quite small numbers. A recent study by Jennison et al. (2023) suggests that up to one-third of revisions are not reported to the NJR, and in particular there is significant under-reporting of revision to arthrodesis procedures, or revision to amputation, making outcome analysis difficult.

The fixed bearing Infinity implant is the market leader and survivorship data monitoring its use for ten years remains encouraging, but as with all implants continued data capture is essential.

Although there has been a trend towards an increasing volume of ankle replacement procedures, the mean number per hospital had only risen from 3.9 to 6.3 per year between 2010 to 2019, with an expected decline in numbers due to the impact of COVID, that has now recovered to be 7.3 per year.

In 2024 only 20.5% of hospitals conducting ankle replacements performed more than ten primary operations per year and just 8.6% of hospitals performed more than 20 primary procedures per year. The British Orthopaedic Foot & Ankle Society (BOFAS) and NHS Getting It Right First Time (GIRFT) encourage surgeons to pool resources and create networks, where practicable, to ensure the sharing of best practice and the achievement of the highest standards of care and outcome quality for patients.

The cumulative percentage probability of 90-day mortality following primary ankle surgery is very low (0.15% (95% CI 0.09-0.24)) and the cumulative percentage of revision at ten years following a primary ankle replacement is 9.54% (95% CI 8.75-10.39). Although this is likely to be a modest underestimate given the findings of the recent study by Jennison et al. (2023). Substantial heterogeneity in the rates of revision was observed between the implant brands used in primary ankle replacement surgery. It is likely that any data missing is ‘Missing At Random’ in relation to brands and therefore the heterogeneity observed is robust. Our data quality audit programme now routinely captures missing ankle procedures and so these missing data effects will reduce over time.

3.E.1. Overview of primary elbow replacement surgery

In this section we detail primary elbow replacements performed between 1 January 2012 and 31 December 2024 and entered into the registry before 1 March 2025. Data on linked first revision episodes and linked mortality data are presented. Primary elbow replacement in this section refers to total elbow replacement (with or without radial head replacement), distal humeral hemiarthroplasty, lateral resurfacing and radial head replacement. We conducted an extended review of the component labels reported on the primary elbow (E1) MDS form. Our analysis has been able to identify inconsistencies between the type of procedure reported on the MDS form and the component label data uploaded to the registry. Procedures where the reported type of surgery did not match the components listed on the MDS form are classified as unconfirmed.

A total of 11,381 primary replacements were available for analysis for a total of 11,037 patients (Figure 3.E1). Of these patients, 344 had documented elbow replacements on both left and right sides, and in 22 patients these were both performed on the same day (bilateral).

Figure 3.E1

Elbow cohort flow diagram.

The majority of replacements were performed on women (66.6%) and the median age at the time of primary operation was 64 years (IQR 52 to 74), with an overall range of 14 to 99 years.

Table 3.E1 shows that the annual number of primary elbow replacements entered into the registry has increased since 2012. While the increase in the early years is in part due to improvement in data capture, the consistent increase observed year-after-year from 2015 to 2019 mostly reflects an increase in the volume of procedures, improved reporting of radial head replacement and inclusion of distal humeral hemiarthroplasties, or a combination of these factors. COVID reduced the number of procedures being conducted annually between 2020 and 2023. Procedure volume is now similar to pre-COVID levels.

Table 3.E1

Number of primary elbow replacements by year and percentage of each type of procedure.

Table 3.E1 provides a breakdown by the stated type of replacement. Of all procedures, including the unconfirmed, 46.4% were classified as a total elbow replacement. A total of 547 (4.8%) primary elbow replacements had an unconfirmed status.

Table 3.E2 details the type of primary operation in each year and we show that 6,577 (57.8%) elbow replacements were carried out for acute trauma indications. These have been separated from the remaining 4,804 cases performed for elective indications in the rest of this section. Two-thirds (67.2%) of the elbow procedures performed for trauma were confirmed radial head replacements.

Table 3.E2

Types of primary elbow replacements used in acute trauma and elective cases, by year and type of primary procedure.

Figure 3.E2 describes the funding status and organisation type (based on organisation type in 2023) of elective elbow replacement procedures collected by the NJR. Prior to 2020 (start of COVID) we can see a steady number of elective elbow replacements being provided, mostly by NHS providers with NHS funding. Since 2020 we can see that the recovery of elbow replacements follows a similar pattern to pre-2020, with mostly NHS-funded NHS provision.

Figure 3.E2

Frequency of elective primary elbow replacements by funding status and organisation type, per year.

Figure 3.E3 and Figure 3.E4 show the yearly number of primary elbow replacements performed for elective and acute trauma indications respectively. Elective and acute trauma procedures have been stratified by total elbow replacements (with or without a radial head replacement), radial head replacements and distal humeral hemiarthroplasty, please note the difference in scale of the y-axis between each sub-plot. Each bar in the figure is further stratified by the volume of procedures that the surgeon conducted in that year across both elective and acute trauma settings i.e. if a surgeon performed 12 elective primary total elbow replacement procedures and 12 acute trauma primary total elbow replacement procedures their annual total volume would be 24 procedures. Those 24 procedures would contribute to the dark purple sub-division in both elective and acute trauma figures shown here.

Figure 3.E3

Frequency of primary elbow replacements within elective cases stratified by procedure type, bars stacked by volume per consultant per year. Graphs by confirmed procedure type.

Figure 3.E4

Frequency of primary elbow replacements within acute trauma cases stratified by procedure type, bars stacked by volume per consultant per year. Graphs by confirmed procedure type.

Figure 3.E3 shows that the volume of elective primary total elbow replacements peaked in 2017 before falling slightly in 2018 and 2019 before the impact of COVID in 2020. The number of surgeons performing one or two procedures annually was falling prior to COVID but remained steady since. Elective radial head replacements are increasingly being recorded in the registry, however the majority of consultants perform fewer than five procedures annually.

The volume of elective distal humeral hemiarthroplasty has recovered to above pre-pandemic levels but numbers remain very low. Figure 3.E4 shows the volume of primary total elbow replacements for acute trauma cases staying relatively constant over the last five years. In the last three years, there has been an increasing proportion of primary total elbow replacements performed by higher volume elbow surgeons i.e. those performing more than 13 procedures a year. Radial head replacements for acute trauma peaked in 2019 before falling back due to COVID in 2020, figures have recovered since COVID but the number reported in 2023 was low. The proportion of consultants performing three or more procedures per year was increasing prior to 2020, indicating a degree of specialisation among a minority of consultants. The number of distal humeral hemiarthroplasties for trauma has continued to increase since recording started in 2018, with a greater volume now being performed by lower volume consultants. Together with total elbow replacements there was a 42.6% increase in procedures performed for trauma in 2024 compared to 2023. The reason for this sudden increase is unknown but is unlikely to reflect a true increase in the incidence of elbow trauma, and may point to a change in practice.

Table 3.E3 describes the indications for the primary operation separately by type of primary elbow replacement. Primary operations with an unconfirmed procedure type are excluded from this table.

Table 3.E3

Indications for main confirmed types of primary elbow replacements, by year and type of primary operation.

Please note that the indications for primary elbow replacement are not mutually exclusive since more than one indication could have been provided. Only one indication for surgery, as defined in Table 3.E3, was given for all 6,287 acute trauma cases with a confirmed type of primary procedure. In 204 (4.5%) of the 4,545 elective cases with a confirmed type of primary, more than one indication was given.

Over the last three years (from 2022 to 2024), 3,151 primary elbow replacements were entered into the registry, of which 1,052 had confirmed components consistent with a total elbow replacement (with or without radial head replacement).

Table 3.E4 (a) and Table 3.E4 (b) show the number of all types of elbow replacement by year and the NJR geographical region over this time period, together with the number of hospitals and consultants. A list of hospitals within each NJR region is provided in the downloads section of reports.njrcentre.org.uk and further information can be found on https://surgeonprofile.njrcentre.org.uk.

Table 3.E4

Number of hospitals and consultant surgeons (cons.) providing primary elbow replacements during each year from the last three years, by region.

The median number of elbow replacements per hospital has changed very little over the last three years and remains around three per annum with up to five replacements per hospital in the North East & Yorkshire region and as low as two replacements per hospital in the East of England region in 2024. These figures are subject to change, as some hospitals may not have submitted all of their procedure data for 2024 by the time of our data analysis.

Table 3.E5 lists the brands used in elbow replacement by confirmed procedure type, with sub-division by acute trauma and elective cases.

Table 3.E5

Brands used in primary elbow replacement by confirmed procedure type.

The top five constructs (Coonrad Morrey[Hum:Ulna], Discovery[Hum:Ulna], Nexel[Hum:Ulna], Latitude[Hum]Latitude EV Stem[Ulna], Latitude EV Stem[Hum:Ulna]) account for nearly 90% of total elbow replacements performed. All total elbow replacements with radial head replacement were performed using the Latitude family of implants. One implant (RHS[Rad]) accounts for 89.7% of the bipolar radial head replacements and two implants (Anatomic[Rad] and Evolve Proline[Rad]) account for 84.8% of the monopolar radial head replacements. All (100%) lateral resurfacing procedures have been performed using the LRE[LHR:LRR] brand. The Latitude system was used for all distal humerus hemiarthroplasty procedures, with the Latitude[DHH] accounting for 57.1%.

3.E.2. Revisions after primary elbow replacement surgery

The NJR asks surgeons and those responsible for healthcare delivery to ensure that when primary and revision joint replacement procedures of the hip, knee, ankle, elbow or shoulder are performed, that the relevant MDS form is completed and data entered into the registry. This is a requirement mandated by the NHS Standard Contract. For the purposes of the Annual Report, revision procedures include any addition, removal or modification of the implants and procedures such as debridement and implant retention with or without implant exchange, excision arthroplasty, amputation and conversion to arthrodesis. For data submitted on MDSv7 only, DAIRs without modular exchange are included as revision procedures. The completion of a revision MDS form is also mandatory for a procedure involving modification of a joint by adding another implant to another part of the joint. For the analyses of surgeon performance, hospital performance and implant performance, debridement and implant retention without implant exchange is currently excluded.

We found that a total of 520 elbow primaries in the registry (158 acute trauma cases and 362 elective) had linked revision procedures recorded up to the end of 2024, including 29 excision procedures, 248 single-stage revisions, 19 DAIRs (14 with modular exchange and five without modular exchange) and 120 stage one of a two-stage procedure.

Table 3.E6 shows Kaplan-Meier estimates of the cumulative percentage probability of revision up to 12 years after the primary operation, together with 95% confidence intervals for all cases and for acute trauma and elective cases separately.

Table 3.E6

E6 KM estimates of cumulative revision (95% CI) by primary elbow procedures for acute trauma and elective cases. Blue italics signify that 250 or fewer cases remained at risk at these time points.

For the sub-group of total elbow replacement, shown in Figure 3.E5, we found that the survival of total replacements was comparable for trauma and elective indications up to one year. From one year post-operation onwards, the revision estimates were higher for the elective total elbow replacements, but the data for acute trauma are less certain for longer follow-up periods due to the low numbers in the registry and because the confidence intervals of the estimates in both groups overlap. There are insufficient data to compare lateral resurfacing, distal humeral hemiarthroplasty and the other unconfirmed types of primary procedure between elective and trauma indications.

Figure 3.E5

KM estimates of cumulative revision of primary total elbow replacement (with or without a radial head replacement) by acute trauma and elective cases. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these (more...)

Figure 3.E6 shows Kaplan-Meier estimates of the cumulative percentage probability of revision by acute trauma and elective cases in radial head replacements. Revision of radial head replacement may be under-reported as they are frequently revised to an excision arthroplasty.

Figure 3.E6

KM estimates of cumulative revision of primary radial head replacement by acute trauma and elective cases. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Table 3.E7 presents data for primary total elbow replacement (with and without radial head replacement), radial head replacement, and distal humeral hemiarthroplasty stratified by acute trauma and elective indications, sex, and three age groups. Whilst numbers are currently small for many of the groups, we hope that this provides useful information for surgeons and for their patients when they are deciding whether or not to have a joint replacement.

Table 3.E7

KM estimates of cumulative revision (95% CI) for primary elbow replacement for acute trauma and elective indications by procedure type, sex, and age group. Blue italics in the numbers at risk table signify that 100 or fewer cases remained at risk at these (more...)

Figure 3.E7 shows cumulative estimates of revision within the acute trauma cases. These differences remain uncertain and should be treated with caution as the number of procedures and the number of revisions within these groups remain low.

Figure 3.E7

KM estimates of cumulative revision of total elbow replacements and distal humeral hemiarthroplasty within the acute trauma cases. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Table 3.E8 shows the cumulative probability of revision for brands used in at least 100 primary elbow replacements with a confirmed procedure type. For total elbow replacement, the cumulative revision estimates varied between brands from 0.9% to 2.4% in the first post-operative year. At five years post-operation, the estimates still varied between brands from 5.2% to 7.7%. However, we note that as numbers are small, this may simply be due to chance. For radial head replacement, the cumulative revision estimates varied between brands from 0.7% to 2.2% in the first post-operative year.

Table 3.E8

KM estimates of cumulative revision (95% CI) for all primary elbow procedures by implant brand. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Figure 3.E8 shows the rate of revision by implant brand within the elective cases. Brand comparisons will become more reliable as the size of the elbow cohort increases over time, and allow further stratification by patient characteristics, acute/elective status and indication for primary surgery.

Figure 3.E8

KM estimates of cumulative revision of total elbow replacements by implant brand within the elective cases. Elbow replacements with fewer than 150 procedures are excluded. Blue italics in the numbers at risk table signify that 250 or fewer cases remained (more...)

Table 3.E9 gives a breakdown of the indications for the first data-linked revision procedure. The most common indications for revision remain aseptic loosening, lysis and infection. The indications for revision were not mutually exclusive; in 150 of the 520 first revisions more than one indication was stated. A few cases (n=135) had gone on to have further revision procedures. The numbers are too small for any further analysis nor to draw any reliable conclusions.

Table 3.E9

Indications for first data-linked revision after any primary elbow replacement. Acute trauma and elective cases are shown separately, for total elbow replacement, lateral resurfacing, distal humeral hemiarthroplasty, and radial head replacement.

3.E.3. Mortality after primary elbow replacement surgery

For this analysis, the second procedure of a pair of bilateral operations performed on the same day were excluded (Figure 3.E1). Among the remaining 11,359 procedures, 1,636 of the recipients had died by the end of December 2024.

Table 3.E10 and Figure 3.E9 show the overall cumulative percentage probability of mortality shown separately for acute trauma and elective cases.

Table 3.E10

KM estimates of cumulative mortality (95% CI) by time from primary elbow replacement, for acute trauma and elective cases. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Figure 3.E9

KM estimates of cumulative mortality of total elbow replacement and radial head replacement for acute trauma and elective cases. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

The mortality estimate at five years after primary total elbow replacement for trauma is 111.0% higher than the estimate in elective total elbow joint surgery, with a five-year mortality estimate of 30.6% for trauma indications. These differences are likely to be due to demographic differences in patient characteristics and indications for undergoing surgery.

3.E.4. Conclusions

The annual number of primary elbow replacement procedures entered into the registry has increased since 2012. Between 2020 and 2023 procedure volumes were profoundly affected by COVID and annual numbers have recovered to pre-COVID levels during 2024. The registry now has the largest dataset of elbow replacements globally.

The type of procedure reported is determined from two sources of information. The first is the procedure type recorded on the MDS data collection form by the surgeon at the time of the procedure. The second source is the set of component labels attached to the MDS form and recorded at the time of upload of the record. When there is a mismatch between these two sources, i.e. the components entered do not match the procedure type recorded or in the case where there are no component data at all in the data entry record, the procedure type is reported as unconfirmed. Work is ongoing to reconcile these unconfirmed procedures and reduce their ‘unconfirmed’ status, and the data show significant reduction since 2012. This will enhance the comprehensiveness and utility of the data moving forward, whilst the audit of procedures recorded has led to an improvement in the completeness of data available for analysis.

Distal humeral hemiarthroplasty was not included in the MDS until June 2018. Despite this, it appears to be increasing overall, while total numbers remain low. Most distal humeral hemiarthroplasty and radial head replacement procedures are performed for acute trauma and trauma sequelae as expected. Early results suggest that revision estimates up to three years are higher for distal humeral hemiarthroplasty than total elbow replacement or radial head replacement for acute trauma patients, but should be treated with caution due to low numbers.

The distribution of indications for elective total elbow replacement has been consistent over the last three years of data entry with inflammatory arthropathy accounting for 49.4% of cases. In 2024 there were 389 confirmed elective and acute trauma primary total elbow replacements (including 15 with radial head replacements) performed in 98 hospitals by 120 consultants. The volume of procedures does not show large variation, however the number of hospitals performing elbow replacements has increased from 96 in 2022 to 98 in 2024. It has been the intention of the NHSE GIRFT programme to centralise total elbow replacement surgery across fewer specialist centres. It should be noted that the median numbers of primary procedures per hospital and per surgeon have not changed significantly from 2022 to those reported in 2024.

The Kaplan-Meier estimate of cumulative revision of total elbow replacement at five years was 4.61% (95% CI 3.38-6.28) for trauma patients and 7.74% (95% CI 6.80-8.80) for elective cases. Minor disparities in the estimates of revision were observed between implant brands. Brand comparisons will become more evident and reliable as the size of the elbow cohort increases over time. We note that the main indications for revision were infection and aseptic loosening and this is observed for both acute trauma and elective cases.

The five-year mortality estimate for elbow replacement in all cases is 11.63% (95% CI 10.95-12.35) with little difference between trauma and elective surgery, mostly because of the large number of radial head replacements in the trauma group. When considering only total elbow replacement without radial head replacement, the five-year mortality estimate for trauma cases is double that for elective indications.

3.S.1. Overview of primary shoulder replacement surgery

Shoulder replacements have been recorded in the registry since 2012. In this section we provide an overview of the (data linked) primary shoulder replacements performed up to 31 December 2024 and also document the first revision surgery and mortality, when these events had occurred following a primary shoulder replacement.

In 2018 and 2019 a rigorous review of the shoulder data was undertaken by the NJR due to the rapid expansion of shoulder implant types available. As a consequence of this review, new classifications and component attributes are now used within the report to define the primary groupings throughout the whole of this section. The report uses whole construct validation, ensuring all relevant elements required to build a construct are present in a procedure. We have cross-checked the implanted construct with the indicated procedure at the time of the surgery and positively confirmed the implanted construct matches the reported procedure. This has led to the definition of unconfirmed constructs where there are either insufficient implants listed to make up a complete construct, or the implants used do not match the indicated procedure. A total of 7,473 (8.6%) procedures are unconfirmed; although the volume is expected to improve in future reports, with the completion of a national shoulder audit and the development of more rigorous checks.

We define a stemmed humeral component as a humeral component in which any part enters the humeral diaphysis, while a stemless humeral component is defined as being completely confined to the metaphysis with no part entering the diaphysis.

There has been an increasing use of reverse total shoulder replacements for acute trauma and osteoarthritis indications over the lifetime of the registry. This should be considered when comparing shoulder types and brands.

A total of 86,882 primary shoulder replacements were available for our analysis in a total of 78,603 patients. Of these patients, 8,279 had documented replacements on both left and right sides, 49 of which were bilateral simultaneous operations (left and right on the same day). See Figure 3.S1 for a detailed description of patients included in this section.

Figure 3.S1

Shoulder cohort flow diagram.

Table 3.S1 illustrates the number of shoulder replacements and how they have changed across time. There was a steady increase in the number of primary shoulder replacements year-on-year prior to the COVID pandemic. Since 2020 the annual number of shoulder replacements has increased again and has now substantially surpassed the level recorded in 2019. Table 3.S1 also illustrates relative proportions of proximal humeral hemiarthroplasty (HHA), conventional total shoulder replacement (TSR) and reverse polarity total shoulder replacement (RTSR). There was an increasing preference for reverse polarity total shoulder replacement year-on-year until 2019 and since then it continues to increase but less rapidly.

Table 3.S1

Number and percentage of primary shoulder replacements (elective or acute trauma), by year and type of shoulder replacement.

Procedures with insufficient prostheses elements to build a valid construct, or a construct that disagrees with the procedure indicated at the time of surgery are identified in the tables as ‘Unconfirmed’. It is noted that entering all the elements of reverse polarity total shoulder replacements appears to have been particularly challenging and so it is urged that those completing the data entry forms and entering data should pay particular attention to these procedures.

Figure 3.S2 describes the funding status and organisation type (based on organisation type in 2024) of elective primary shoulder replacement procedures collected by the NJR. Prior to 2020 (which marked the start of COVID) we can see an increase in the absolute number of joint replacements being provided, which in part was being facilitated by an expansion of NHS-funded procedures in both the NHS and the independent sector. Since 2020 we can see that the recovery of shoulder replacement has been due to an expansion of provision within the independent sector (both NHS- and independently-funded). Notably, there has been a substantial increase in the number of independently-funded procedures compared to pre-2020 data. In 2024, NHS-funded procedures in NHS hospitals have surpassed their pre-COVID level.

Figure 3.S2

Frequency of elective primary shoulder replacements by funding status and organisation type, per year.

Figure 3.S3 and Figure 3.S4 show the yearly number of primary shoulder replacements performed for elective and acute trauma indications respectively. Elective and acute trauma procedures have been stratified by procedure type. (Please note the difference in scale of the y-axis between each sub-plot.) Each bar is further stratified by the volume of procedures that the surgeon conducted in that year across both elective and acute trauma settings i.e. if a surgeon performed 24 elective primary stemmed humeral hemiarthroplasty procedures and 24 acute stemmed humeral hemiarthroplasty procedures their annual total volume would be 48 procedures. Those 48 procedures would contribute to the grey sub-division in both elective and acute trauma figures.

Figure 3.S3

Frequency of primary shoulder replacements within elective patients stratified by procedure type, bars stacked by volume per consultant per year. Graphs by confirmed procedure type.

Figure 3.S4

Frequency of primary shoulder replacements within acute trauma patients stratified by procedure type, bars stacked by volume per consultant per year. Graphs by confirmed procedure type.

Figure 3.S3 shows a complex pattern of increasing and decreasing treatment preferences for elective indications. Resurfacing humeral hemiarthroplasty and resurfacing total shoulder replacements have declined since the start of data collection, while stemless total shoulder replacements have steadily increased, and the volume of stemmed reverse polarity total shoulder replacement has increased substantially. There has been a decrease in the use of stemmed humeral hemiarthroplasty and stemmed total shoulder replacements, whilst the growth in stemless total shoulder replacements and stemmed reverse polarity total shoulder replacements appears to be occurring in higher-volume shoulder surgeons.

Figure 3.S4 shows the popularity of stemmed humeral hemiarthroplasty for acute trauma indications has reduced in recent years, while the popularity of stemmed reverse polarity total shoulder replacements has been steadily increasing. Stemmed reverse polarity total shoulder replacements are increasingly conducted by higher-volume surgeons.

Figure 3.S5 and Figure 3.S6 illustrate the age and sex differences between the different types and sub-types of shoulder replacements for elective indications and acute indications respectively, using a modified ‘Box and Whiskers’ plot. The whiskers represent the 2.5 and 97.5 centile of the distribution. The figures also show the frequency of procedures by sex and procedure type. Women tend to be older than men at the time of primary operation and those receiving reverse polarity total shoulder replacements tend to be older than those receiving proximal humeral hemiarthroplasty or conventional total shoulder replacements. These differences are consistent over the lifetime of recording of shoulder replacements in the registry. The majority of procedures recorded within the registry are reverse polarity total shoulder replacements, and the majority of unconfirmed procedures consist of reverse polarity total shoulder replacements.

Figure 3.S5

Age (Box and Whiskers*) and frequency of primary shoulder replacements by sex and type of shoulder replacement for elective indications.

Figure 3.S6

Age (Box and Whiskers*) and frequency of primary shoulder replacements by sex and type of shoulder replacement for acute trauma indications.

Table 3.S2 displays similar information to Figure 3.S5 and Figure 3.S6, with results separated by acute trauma and elective procedures.

Table 3.S2

Demographic characteristics of patients undergoing primary shoulder replacements, by acute or elective indications and type of shoulder replacement.

Table 3.S3 illustrates the number of primary shoulder replacements and the number of hospitals and consultants conducting shoulder replacements within the registry. The table also illustrates the median and interquartile range of the number of replacements performed within each hospital or by each consultant. This is displayed overall, aggregated by the last five years of data, and by year of data collection.

Table 3.S3

Numbers of hospitals and consultant surgeons providing primary shoulder replacements and median and interquartile range of procedures performed by hospital and consultant, by year, last five years and overall.

The results illustrate that the median, and interquartile range, number of procedures performed by hospitals and consultants has remained static for the last few years until 2019 and the subsequent impact of COVID. There are currently 19 (9 to 33) procedures per hospital and 14 (6.5 to 26) procedures per consultant, which is above pre-COVID levels.

Table 3.S4 illustrates the number and percentage of primary shoulder procedures by the type and sub-type of shoulder replacement for both acute trauma and elective procedures. The indication for surgery in elective procedures is also illustrated. The majority of proximal humeral hemiarthroplasty and conventional total shoulder replacement procedures recorded in the registry are for an indication of osteoarthritis. Reverse polarity total shoulder replacements are the most common procedures performed for cuff tear arthropathy, trauma sequelae, other inflammatory arthropathy, avascular necrosis, other causes and cuff tear without arthropathy. It is important to note that the indications for shoulder surgery recorded in the registry are not mutually exclusive; 82.2% of procedures list a single indication for the cause of surgery with the remainder recording more than one indication.

Table 3.S4

Number and percentage of primary shoulder replacements by indication and type of shoulder replacement.

Table 3.S5 (a) to Table 3.S5 (h) illustrate the shoulder construct used by sub-type of the primary shoulder replacement for overall procedures and by acute and elective sub-divisions. Implants are only listed if they have been used on ten or more occasions overall, or five occasions within the last year, respectively. Results illustrate the frequency of all implanted constructs across all years of data collection within the registry i.e. between 2012 and 2024. The frequency of shoulder constructs within the last year of the data collection is also illustrated to indicate contemporary practice. Constructs and prostheses elements are suffixed ‘[ ]’ to indicate the implants that make up the construct. In the cases of ‘within manufacturer and brand’ construct, this suffix is placed after the brand name; whereas within ‘mix and match’ constructs, the suffix is placed immediately after the brand of the implanted element. While the detail in reporting of constructs has become more granular, the complexity has necessarily increased to reflect the diversity of implanted elements and will facilitate improved implant scrutiny. Given the rapid evolution and heterogeneity of shoulder prostheses, it is expected that the classification system will evolve year-on-year with the introduction of new types of prostheses and the combinations in which these are used by surgeons.

Table 3.S5 (a)

Number of resurfacing proximal humeral hemiarthroplasty replacements between 2012 and 2024 and within the last year by brand construct.

Table 3.S5 (b)

Number of stemless proximal humeral hemiarthroplasty replacements between 2012 and 2024 and within the last year by brand construct.

Table 3.S5 (c)

Number of stemmed proximal humeral hemiarthroplasty replacements between 2012 and 2024 and within the last year by brand construct.

Table 3.S5 (d)

Number of resurfacing total shoulder replacements between 2012 and 2024 and within the last year by brand construct.

Table 3.S5 (e)

Number of stemless conventional total shoulder replacements between 2012 and 2024 and within the last year by brand construct.

Table 3.S5 (f)

Number of stemmed conventional total shoulder replacements between 2012 and 2024 and within the last year by brand construct.

Table 3.S5 (g)

Number of stemless reverse polarity total shoulder replacements between 2012 and 2024 and within the last year by brand construct.

Table 3.S5 (h)

Number of stemmed reverse polarity total shoulder replacements between 2012 and 2024 and within the last year by brand construct.

3.S.2. Revisions after primary shoulder replacement surgery

We present results in this section as percentage cumulative revision of primary shoulder replacements. Results are estimated using the 1 - Kaplan-Meier method; 95% CIs are shown within tables and when the number at risk falls below 250, estimates are shown in blue italics to indicate that caution is required in interpreting the results. Data are presented up to 12 years, which is the last full year of data collection within the registry. Figures also include an ‘at-risk table’ which presents the number of individuals at risk of revision at the time indicated.

Figure 3.S7 and Table 3.S6 illustrate the cumulative revision of primary shoulder procedures performed overall (shown in Table 3.S6 only) and by acute trauma and elective procedures. Our results indicate that the risk of revision is comparable for the first three years following surgery, at which point it starts to diverge, with lower revision rates for acute trauma than for elective patients. This is not related to mortality because patients are censored from the analysis at the date of death.

Figure 3.S7

KM estimates of cumulative revision for primary shoulder replacement by acute trauma and elective cases. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Table 3.S6

KM estimates of cumulative revision (95% CI) for primary shoulder replacement for all cases, acute trauma and elective cases. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Table 3.S7 further breaks down the cumulative revision of primary shoulder procedures for elective patients, by sex and age group. Results indicate that females have a lower risk of revision in the long-term compared to males, and that younger patients have an increased risk of revision compared to older patients.

Table 3.S7

KM estimates of cumulative revision (95% CI) for primary shoulder replacement for elective cases by sex and age group. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Table 3.S8 and Figure 3.S8 report cumulative revision of primary shoulder procedures, for elective patients, by type (Table 3.S8 only) and sub-type of shoulder construct.

Table 3.S8

KM estimates of cumulative revision (95% CI) for primary shoulder replacement for elective cases by shoulder type. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Figure 3.S8

KM estimates of cumulative revision for primary elective shoulder replacement by type of shoulder replacement. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Proximal humeral hemiarthroplasties undergo revision at a higher rate than either conventional total shoulder replacements or reverse polarity total shoulder replacements. The extent to which proximal humeral hemiarthroplasty procedures are seen as ‘revisable’ procedures compared to total shoulder replacements should be considered when interpreting the results. Furthermore, while Table 3.S8 and Figure 3.S8 suggest a stemmed proximal humeral hemiarthroplasty might be the better choice over a stemless or resurfacing humeral hemiarthroplasty, the latter group are more straightforward to revise than a stemmed implant and so caution is again needed when interpreting these sub-group results.

The cumulative risk of revision of stemless reverse polarity total shoulder replacements is higher compared to stemmed versions. This needs careful interpretation as the number of stemless reverse polarity replacements is low, however, it is worth noting that some stemless reverse polarity brands have been withdrawn from the market. The performance of stemmed conventional total shoulder replacement, compared to stemmed reverse polarity shoulder replacements, is of particular interest. Reverse polarity total shoulder replacements tend to have an initially higher revision rate (predominantly for infection and instability) which then plateaus, whereas the conventional total shoulder replacements increase more slowly but at a constant rate (rotator cuff failure being the commonest indication for revision) and therefore exceed the cumulative risk of revision of reverse polarity total replacements and overall is 3.5% higher at 12 years. The extent to which the different indications for surgery are confounding results is not clear and therefore results should be interpreted with caution.

Table 3.S9 and Figure 3.S9 report the cumulative revision of primary shoulder procedures, for acute trauma patients, by type (Table 3.S9 only) and sub-type of shoulder construct. Proximal humeral hemiarthroplasties undergo revision at a higher rate than reverse polarity total shoulder replacements.

Table 3.S9

KM estimates of cumulative revision (95% CI) for primary shoulder replacement for acute trauma cases by shoulder type. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Figure 3.S9

KM estimates of cumulative revision for primary shoulder replacement for acute trauma cases by shoulder type. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

The extent to which proximal humeral hemiarthroplasty procedures are seen as ‘revisable’ procedures compared to reverse total shoulder replacements and the demographic characteristics of the patients should all be considered when interpreting the results.

Table 3.S10 reports cumulative revision of primary shoulder procedures for elective patients by shoulder construct. All constructs that have been used on more than 250 occasions are reported. Where the construct is solely built from within the same product line, the elements used to build the construct are suffixed in [ ] following the brand. Where the construct is built from different product lines, the elements used to build the construct from each brand are indicated in [ ] immediately after the brand. The description of constructs is necessarily complex, this reflects the extensive modularity of modern shoulder prostheses. All results should be viewed in the context of observational data and due consideration given to the volume of unconfirmed prostheses.

Table 3.S10

KM estimates of cumulative revision (95% CI) for primary shoulder replacement for elective cases by brand construct in constructs with greater than 250 implantations. Blue italics signify that 250 or fewer cases remained at risk at these time points. (more...)

Table 3.S11 and Table 3.S12 describe the prosthesis time incidence rate (PTIR) per 100 years of follow-up for the reported indication for revision in acute trauma patients receiving a primary shoulder replacement. Table 3.S11 reports indications for all patients across the life of the registry i.e. between 2012 and 2024, this was achieved by aggregating indications for revision across the different minimum data sets. Table 3.S12 reports data for patients whose information was entered following the introduction of MDSv7.

Table 3.S11

PTIR estimates of indications for shoulder revision (95% CI) for acute trauma by type of shoulder replacement between 2012 and 2024.

Table 3.S12

PTIR estimates of indications for shoulder revision (95% CI) for acute trauma by type of shoulder replacement using reports since MDSv7.

Cuff insufficiency is the leading indication for revision for those who receive a proximal humeral hemiarthroplasty for acute trauma, whereas instability or dislocation, or infection are the leading causes in reverse polarity total shoulder replacements when performed for acute trauma, see Table 3.S11. The low number of primary replacements and even lower frequency of revisions for patients whose data were entered using the most recent minimum data sets makes results difficult to interpret. It is important to note that the indications for revision are not mutually exclusive and 10.3%, 75.7%, and 11.1% recorded none, one and two indications for revision respectively.

Table 3.S13 and Table 3.S14 describe the prosthesis time incidence rate (PTIR) per 100 years of follow-up for the reported indication for revision in elective patients receiving a primary shoulder replacement by type and sub-type of shoulder replacement.

Table 3.S13

PTIR estimates of indications for shoulder revision (95% CI) for elective procedures by type of shoulder replacement between 2012 and 2024.

Table 3.S14

PTIR estimates of indications for shoulder revision (95% CI) for elective procedures by type of shoulder replacement using reports since MDSv7.

Table 3.S13 reports indications for all patients across the life of the recording of shoulders in the registry i.e. between 2012 and 2024. This was achieved by aggregating indications for revision across the different minimum data sets. Table 3.S14 reports data for patients whose information was entered following the introduction of MDSv7.

For elective primary replacements, cuff insufficiency is the leading indication for revision for those who receive a proximal humeral hemiarthroplasty or conventional total shoulder replacement, whereas instability or dislocation is the leading cause for revision in reverse polarity total shoulder replacements, see Table 3.S13. It is important to note the indications for revision are not mutually exclusive and 15.3%, 66.7%, and 14.6% recorded none, one and two indications for revision, respectively.

The NJR asks surgeons and those responsible for healthcare delivery to ensure that when primary and revision joint replacement procedures of the hip, knee, ankle, elbow or shoulder are performed, that the relevant MDS form is completed and data entered into the registry. This is a requirement mandated by the NHS Standard Contract. For the purposes of the Annual Report, revision procedures include any addition, removal or modification of the implants and procedures such as debridement and implant retention with implant exchange, excision arthroplasty, amputation and conversion to arthrodesis. For data submitted on MDSv7 only, DAIRs without modular exchange are included as revision procedures. The completion of a revision MDS form is also mandatory for a procedure involving modification of a joint by adding another implant to another part of the joint. For the analyses of surgeon performance, hospital performance and implant performance, debridement and implant retention (DAIR) without implant exchange is currently excluded.

3.S.3. Patient Reported Outcome Measures (PROMs) Oxford Shoulder Scores associated with primary shoulder replacement surgery

The Oxford Shoulder Score (OSS) is a validated patient reported outcome measure for use in shoulder surgery. It consists of 12 pain and function items which address problems that the patient may have encountered with their shoulder over the preceding four weeks (Dawson et al., 1996). The score is coded from zero to four (from ‘worst’ to ‘best’) and then summed in line with updated OSS recommendations (Dawson et al., 2009). The final total score ranges from zero to 48, with 48 representing the ‘best’ outcome and zero the ‘worst’. Where up to two items were missing, the average of the remaining items can be substituted for the missing values (Dawson et al., 2009). If more than two items are missing, the results have to be disregarded.

Table 3.S15 provides a detailed description of the number of patients reporting an OSS questionnaire pre-operatively, 6 months, 3 years and 5 years following surgery for patients undergoing primary shoulder replacement for acute trauma or elective indications. The responses are further divided by how close to the time point of interest the questionnaire was collected and its completeness. The results are expressed absolutely (N) and as a percentage (%) of ‘Eligible’ participants and those who ‘Responded’ to the PROMs questionnaires. Eligibility is defined as being alive at the time point of interest and also having sufficient follow-up time following primary surgery.

Table 3.S15

Number and percentage of patients who completed an Oxford Shoulder Score (OSS) by acute trauma and elective indications, by the collection window of interest at different time points.

How close the response was to the time point of interest is categorised by defining ‘windows of interest’. The pre-operative window of interest is 90 days prior to the primary surgery, until the day of the primary operation. The 6-month data collection window of interest ranges from 5 months to 8 months, i.e. spanning a 3-month window of interest. The 3- and 5-year data collections had windows of interest ranging from 1 month prior to 3 and 5 years respectively, to 6 months after i.e. spanning a 7-month window of interest.

Ensuring data is collected pre-operatively by hospitals is very important. To assess the efficacy of a surgical technique or implantable construct, understanding where the patient started is critical to understand how the patient is likely to respond to surgery. Collecting pre-operative PROMs post-operatively is likely to induce recall bias and for this reason the end of the pre-operative window was strictly defined as ‘the day of surgery’. Table 3.S15 clearly illustrates only a small minority of eligible patients complete an OSS questionnaire prior to surgery and within the window of interest.

Given the low compliance in pre-operative score collection by hospitals delivering shoulder replacement surgery, the potential for bias in interpreting results is clear. Collection and compliance with reporting at 6 months, 3 and 5 years is substantially better than pre-operative rates, but the response rate of all eligible participants is still less than 50% in all instances. For the last decade, the British Elbow and Shoulder Society (BESS) have strongly advocated and promoted to national bodies the collection of shoulder PROMs as essential in the assessment of patient outcomes and surveillance after shoulder replacement surgery. As part of the NJR Minimum Data Set, the collection of shoulder PROMs for shoulder replacement surgery is mandated by the NHS Standard Contract, and so the ongoing low pre-operative compliance with PROMs data collection by NHS trusts remains particularly concerning.

Table 3.S16 provides a detailed description of the number of patients reporting complete OSS within the window of interest pre-operatively and at 6 months, 3 years and 5 years by the year of surgery for patients undergoing primary shoulder replacement for acute trauma or elective indications. The denominator used to calculate percentages is the number of patients alive at the time point of interest. The data illustrate that collection and submission of pre-operative PROMs by hospitals is consistently poor, with less than 30% of elective patients having their PROMs data submitted. Compliance with 6-month PROMs reporting appeared to be improving in the years prior to the COVID pandemic, but has reduced again since.

Table 3.S16

Number and percentage of patients who, cross-sectionally, completed Oxford Shoulder Score (OSS) by acute trauma, elective and by year of primary operation, within the collection window of interest, with valid measurements at the time points of interest. (more...)

Table 3.S17 describes the number and percentage of paired measurements available for longitudinal analyses for all patients undergoing primary shoulder replacement for acute trauma or elective indications. The denominator used to calculate percentages is the number of pre-operative measurements. The numerator is the number of responses within the window of interest, see Table 3.S15, with no more than two items missing responses. The proportion of patients available for a paired longitudinal analysis at any time point is low, and the proportion of patients with serial measurements at any time point is even lower. While the proportion of patients with both pre-operative and 6-month OSS has been higher since 2017 (excluding 2020 due to COVID), this still only represented 13.6% of all eligible elective primary replacements in 2023.

Table 3.S17

Number and percentage of patients who completed longitudinal Oxford Shoulder Score (OSS) by acute trauma, elective and by year of primary operation, within the collection window of interest, with valid measurements at the time points of interest.

Figure 3.S10 reports the cumulative revision rate for elective patients undergoing primary shoulder replacements who completed pre-operative and 6-month PROMs assessments within the specified window of interest. Results indicate a different cumulative revision rate for patients who are included in the PROMs cohort versus those who are not. This difference suggests the group of patients responding to the PROMs questionnaires is different to the patients who are not responding, and so is not representative of the larger population. This highlights the risk of using incomplete datasets to make inferences for the larger cohort, and so the data from this PROMs cohort need to be interpreted cautiously despite their relatively large size. If anything, it indicates that the PROMs cohort is likely to be a more ‘satisfied’ group of patients as their revision rates are lower than the non-PROMs cohort.

Figure 3.S10

KM estimates of cumulative revision for primary elective shoulder replacements for patients with and without valid PROMs. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Figure 3.S11 illustrates the distribution of pre-operative OSS and change in OSS between the pre-operative and the 6-month assessment. Results are displayed for patients with elective indications for primary shoulder replacement only. It also illustrates the association between pre-operative OSS and the change in OSS. While pre-operative and change in OSS are approximately normally distributed, this hides the ceiling effect within the assessment of the change score. This makes the interpretation of change in OSS particularly challenging and highlights the necessity of ascertaining a pre-operative PROMs when assessing the efficacy of any intervention associated with a primary shoulder replacement. In the absence of specialist methods which account for floor and ceiling effects, a simple analysis of change scores is reported to be the most appropriate (Glymour et al., 2005). At six months following surgery, 5.3% of patients reported a score worse than they did pre-operatively. This figure is reduced compared to previous years due to the more refined inclusion/exclusion criteria of the PROMs cohort as defined previously.

Figure 3.S11

Distribution and scatter of pre-operative Oxford Shoulder Score (OSS) and the change in OSS (post-pre) score for those receiving elective shoulder replacements for valid measurements within the collection window of interest.

Table 3.S18 presents descriptive statistics, mean and standard deviation, median and interquartile range, by year of primary shoulder replacements for acute trauma or elective indications. Results are presented only for those with measurements pre-operatively and at six months, within the window of interest and with no more than two items missing. The number of patients with valid OSS that receive primary shoulder replacements is relatively low, however the results appear to be broadly concordant with those receiving primary shoulder replacement for elective indications. The change in OSS has tended to improve across the years of data collection, but the significance of this is very unclear given the potential for bias due to the lack of a representative sample.

Table 3.S18

Descriptive statistics of the pre-operative, 6-month and the change in Oxford Shoulder Score (OSS) by acute trauma, elective and by year of primary operation, within the collection window of interest, with valid measurements pre-operatively and 6 months (more...)

Table 3.S19 presents descriptive statistics, mean and standard deviation, median and interquartile range, by type and sub-type of primary shoulder replacements for acute trauma or elective indications. Results are presented only for those with measurements pre-operatively and at six months, within the window of interest and with no more than two items missing. The number of patients receiving a primary shoulder replacement for acute trauma indications is small.

Table 3.S19

Descriptive statistics of the pre-operative, 6-month and the change in Oxford Shoulder Score (OSS) by acute trauma, elective and by shoulder type, within the collection window of interest, with valid measurements pre-operatively and 6 months post-operatively. (more...)

Table 3.S19 clearly illustrates the change and improvement between pre-operative and 6-month assessment of OSS. While positive, the change of score is substantially less for patients receiving a proximal humeral hemiarthroplasty compared to either a conventional total or reverse polarity total shoulder replacement. The change in OSS between conventional total shoulder replacement versus reverse polarity total shoulder replacement is broadly similar.

3.S.4. Mortality after primary shoulder replacement surgery

The following section describes the mortality profile for patients receiving primary shoulder replacements. Where patients received same-day bilateral procedures (N=49), see Figure 3.S1, they were excluded from the analysis to avoid double counting. This results in 86,833 patient procedures being included in the analysis, with 15,643 observed deaths.

Figure 3.S12 and Table 3.S20 describe the mortality of patients receiving a primary shoulder replacement up to 12 years following the primary procedure for all patients (Table 3.S20 only) and patients undergoing surgery for acute trauma and elective indications separately. Data is shown at 30 and 90 days following the primary procedure and then at 1, 3, 5, 7, 10 and 12 years. Table 3.S20 indicates the importance of separating the data for patients receiving a primary shoulder replacement for acute trauma, from the data for those with elective indications, due to the differences in the frailty of the patient population despite their similar age profile, see Table 3.S2.

Figure 3.S12

KM estimates of cumulative mortality by acute trauma and elective indications for patients undergoing primary shoulder replacement. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

Table 3.S20

KM estimates of cumulative mortality (95% CI) by acute trauma and elective indications for patients undergoing primary shoulder replacement. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Table 3.S21 and Figure 3.S13 describe the mortality of patients receiving a primary shoulder replacement up to 12 years following the primary procedure, by sex and age group of the patients undergoing surgery for elective indications only. Data are shown at 30 and 90 days following the index procedure in Table 3.S21 and then at 1, 3, 5, 7, 10, and 12 years. When mortality is divided by age (see Figure 3.S13), it is clear that older males have higher mortality than females, this pattern first becomes evident after the age of 65.

Table 3.S21

KM estimates of cumulative mortality (95% CI) for primary shoulder replacement for elective cases by sex and age group. Blue italics signify that 250 or fewer cases remained at risk at these time points.

Figure 3.S13

KM estimates of cumulative mortality for primary elective shoulder replacement by age group and sex. Blue italics in the numbers at risk table signify that 250 or fewer cases remained at risk at these time points.

3.S.5. Conclusions

In this year’s report, we provide extensive insight into the use and performance of shoulder constructs in primary shoulder replacements and also give a detailed description of revision rates by the indication for surgery. A detailed description of the longitudinal PROMs data collection is also provided for both elective and trauma patients.

The pattern of use of primary shoulder replacements has continued to be documented. In recent years, we have extensively revised shoulder implant data processing and, building on the recent internal and external validation, we now report at the level of the construct. This more detailed level of reporting has led to new insights, but has also highlighted some inconsistencies within the data recorded that have led to unconfirmed procedures being reported. The volume of unconfirmed proximal humeral hemiarthroplasty is consistently low, and the volume of unconfirmed conventional total shoulder replacements has fallen since their being recorded on the registry. However, the volume of unconfirmed reverse polarity total shoulder replacements is consistently high, which is of concern and our recent national shoulder audit aims to reduce the number of unconfirmed cases. This lack of completeness hampers one of the core functions of the registry, which is to provide a comprehensive record of all implanted prostheses.

There are now 86,882 shoulder replacements eligible for analysis, after the application of our data cleaning processes. Patterns of use and the completeness of data are becoming clearer and revision rates out to 12 years can be analysed. PROMs data continue to be collected so that patient outcomes in terms of pain and function can also be assessed alongside revision rates. It has previously been identified that some patients who have worse post-operative PROMs scores, i.e. a poor outcome, are not captured by the metric of revision surgery.

Confirmed reverse polarity total shoulder replacements made up 70.4% of all shoulder replacements in 2024 and the patterns of use observed in previous reports continue. This high level of use across indications points to a growing confidence in this implant and a sustained change of practice in the NJR’s operational areas, despite limited high-level outcome evidence. As a consequence, proximal humeral hemiarthroplasties, and to some extent conventional total shoulder replacements, are now declining in numbers.

Revision rates this year have not altered much from the pattern observed last year. Revision rates in patients under the age of 55 years continue to be high and are now 10.6% and 8.9% in males and females respectively at five years, and 16.9% and 13.9% at ten years. These informative revision figures provided by the registry could usefully be addressed in clinical discussions with younger patients wishing to undergo shoulder replacement surgery. The maturity of the shoulder registry data for revision rates now provides valuable medium-term and long-term information for patients and clinicians.

At present, reverse polarity total shoulder replacement demonstrates the lowest revision rates at 12 years. However, it is worth highlighting that these procedures have a higher early revision rate compared to stemmed conventional total shoulder replacements, until approximately three years following surgery. After three years the revision rate of stemmed reverse polarity shoulder replacements falls below stemmed conventional total shoulder replacements. The observed non-proportionality between conventional and reverse bearings combined with the differing indications between the two procedures does not necessarily mean that reverse polarity shoulder replacements should be favoured over conventional total shoulder replacement, particularly for indications that would normally indicate the latter.

Revision of all three sub-types of humeral hemiarthroplasty is higher than for both conventional and reverse polarity total shoulder replacement. While it may be argued that the higher revision rate for humeral hemiarthroplasty is mediated by the ease of the revision procedure, the PROMs data evidenced in this report do not support this. The change in PROMs score between the pre-operative and 6-month assessment following surgery suggests less improvement in the group of patients that receive a humeral hemiarthroplasty as their primary operation, as compared to those that receive total shoulder replacements.

More in-depth analysis which accounts for case-mix should be conducted as, while the age and sex distribution is similar, the distribution of indications for which patients undergo proximal humeral hemiarthroplasty is different to that of either conventional total shoulder replacement or reverse polarity shoulder replacement, with a much higher proportion of patients indicating avascular necrosis. There has also been an increasing use of reverse total shoulder replacements for acute trauma and osteoarthritis indications over the lifetime of the registry. An in-depth analysis accounting for the many indications collected by the registry and other clinically relevant factors may help surgeons select different treatment modalities for different patients and indications.

We have presented a detailed description of PROMs data with reference to not only those who have responded, but the entire cohort of patients receiving a primary shoulder replacement. The pre-operative scores are administered and collected by hospitals and our analysis demonstrates that hospital compliance remains poor. The post-operative shoulder PROMs are administered directly to patients on the NJR’s behalf by their authorised contractor, NEC Software Solutions. The completeness of measures cross-sectionally and importantly from a longitudinal perspective and how this has changed across the years has been described. A pre-operative and 6- month matched elective cohort of 7,358 patients is now available for analysis, but the representative nature of these data compared to the whole cohort is not clear. It illustrates, for those who completed the PROMs, that shoulder replacement surgery results in substantial improvement in both pain and function for patients. However, it is less clear how those who do not complete the PROMs fare.

The largest benefit gains by elective patients can be observed in those patients receiving a conventional total shoulder replacement, followed closely by those receiving a reverse polarity shoulder replacement, which is thereafter followed by those receiving a proximal humeral hemiarthroplasty.

Overall, in this report we have shown that the volume of shoulder replacement surgery in the registry continues to grow rapidly and now presents an opportunity for outcomes to be assessed by revision rates and by PROMs, although careful consideration of the latter is needed in respect to its generalisability. Importantly, our approach of whole construct validation using new classifications and component attributes will lead to more meaningful analysis and provision of more useful information for patients, surgeons and other interested stakeholders.