Part Three: Outcomes after joint replacement 2003 to 2017
Part Three of the 16th Annual Report provides outcome data in relation to hip, knee, ankle, shoulder and elbow replacements. It describes activity between 1 April 2003 and 31 December 2018.
There were 2,766,764 procedures entered into the NJR across all joint types, performed up to 31 December 2018. After removing procedures without linkage identifiers and those procedures where the linkage was not sufficiently clear to allow their use, there remained 2,332,798 primary cases and 66,248 linked revisions. This represents over a quarter of a million new cases being registered during the year.
There were 1,091,892 primary total hip replacements, 1,193,830 knee replacements, 5,587 ankle replacements, 37,916 shoulder replacements and 3,573 elbow replacements available for analysis, and these form the basis of the ensuing section concerning clinical outcomes, including revision rates, mortality data and other clinical outcomes where these are collected.
As previously, some figures in the latest year may not yet fully represent the final figures. There may be late data entry by units and further correction after the data quality audit and for this reason, readers should be wary of drawing conclusions about apparent sudden increases or decreases in a particular procedure in the latest year compared to previous years.
Hip replacement procedures
The number of hip replacement procedures recorded in the NJR continues to grow at a few percent per year continuing the pattern over the last decade since data entry became consistent at around 95% after the initial few years of “start-up”.
The steady gentle decline in cemented and uncemented hips since 2012 continues in favour of hybrid fixation. Few cemented hips have anything other than metal-on-polyethylene (MoP) or ceramic-on-polyethylene (CoP) bearings and the steady decline of MoP in favour of CoP continues. Likewise, those using hybrid hips appear to be increasingly favouring CoP. The consistent and dramatic decline (since 2011) in the use of ceramic-on-ceramic (CoC) bearings in favour of CoP bearings for uncemented hips continues. These combinations are used more in younger patients and this change to CoP is likely related to the excellent survivorship of this bearing surface combination as highlighted in previous years’ reports.
Metal-on-metal (MoM) bearings have declined to a trickle, and the brief burst of interest in ceramic-on-metal (CoM) bearings several years ago appears to have lapsed almost completely.
As with knee replacement, primary hip revision rates have declined steadily but progressively since about 2008, with this change being even more obvious for hips than for knees. In hips, this has often been attributed to the rapid decline in MoM usage and this has no doubt been a significant factor. However, the fact that we see a similar decrease in revision rates for knee replacements indicates that we need to look for alternative explanations for this improvement. It seems that the progressive decline in revision rates does coincide closely with the time at which the NJR began to provide personal revision rate feedback to surgeons. It is therefore likely that at least some element of the improvement in revision rates has come about due to the feedback of surgeons’ own audited results, and the actions that they have taken in response to this information.
The generally low revision rates for CoP and CoC bearing surfaces in primary hip replacement are quite striking features of the data relating to many of the sub-groups regardless of age and gender, but applies particularly to the younger age groups.
It is interesting to note that in cemented THR the CoP failures occur at a similar rate to MoP at 15 years despite the fact that up to 13-14 years the CoP combination fares slightly better. This observation is largely due to the small numbers of cases available for analysis at 15 years, meaning that less reliance can be placed upon the figures at that time point. This difference at 14-15 years may also relate in part to the fact that 14 years ago far fewer highly cross-linked polyethylene implants were being used, compared to the use over the last ten years; that change was a gradual one though and would, therefore, be unlikely to have produced any sudden change in failure rates. For uncemented THR the difference between MoP and CoP is similar. However, the failure rate of CoC splits the difference between MoP and CoP. Overall the variability of performance of different bearing surfaces within uncemented hips varies widely. Notably, CoP has slightly lower revision rates out to 13-14 years.
The pattern of differences in hybrid hip replacement also persists. With CoP bearings having lower revision rates when compared to other bearing types at 13-14 years, after which the numbers in these groups becomes small and therefore estimates are less reliable.
The large variability in revision rates with age is seen once again. Females under 60 years of age have higher revision rates than their male counterparts whereas those over the age of 60 have lower revision rates. Once MoM total hips and resurfacing hips are excluded however, the markedly higher revision rates in females under the age of 60 is not observed but revision rates remain lower in females over this age.
Knee replacement procedures
The analysis is now based on 1,193,830 primary knee replacements and there are 33,292 linked revisions from these primary operations. Amongst these primary numbers, there has been a slight increase in the proportion of unicompartmental knees, as has already been seen in the past five years. The continuing decline in hybrid, uncemented and patellofemoral knee operations means that these each represents really quite small numbers each year. There are some surgeons performing multiple partial replacements in the same knee in moderate numbers individually, but the overall numbers are currently insufficient to derive much useful information from analysis.
Overall revision rate for knees during the last 15 years appeared to alter from around 2008, such that 1-year revision rates peaked in 2009, 3-year rates peaked in 2011, 5-year rates peaked in 2013, 7-year rates peaked in 2015 and 10-year revision rates were highest in 2018. As discussed for hip replacement, this series of alterations and their timing implies some significant change in about 2008 which has had a knock-on effect on revision rates for procedures from that time onward. This is also seen to be the case for hips and although more dramatic in hips due to the additional effects of the decline in MoM hip bearings from a similar time, there is still clearly another factor at work because knees are affected as well as hips. Again, it is possible this could result from the NJR providing feedback to surgeons; a process which commenced in 2008.
The results show that posterior-stabilised fixed-bearing, posterior-stabilised mobile-bearing, unconstrained mobile-bearing and constrained condylar TKRs all seem to have slightly (but significantly) higher failure rates than unconstrained fixed-bearing cemented TKRs. The difference in failure rate between (uncemented/hybrid) posterior-stabilised and unconstrained TKRs is increasing year on year. This difference, which is also seen in data from other registries, has sometimes been attributed to the selective use of some of these more constrained implants for more complicated cases. This would be a plausible explanation for why constrained condylar implants might have a higher failure rate in primary cases, as they would not be used without good reason, as they are more complicated to use, more expensive and potentially have additional sources of post-operative complications. This explanation probably does not hold water in the case of posterior-stabilised or mobile-bearing TKRs however, as these are mostly selected by surgeons as a matter of choice (because they prefer them in principle) and they, therefore, use them routinely regardless of the patient’s specific characteristics.
It is interesting to note that the observed difference in revision rates between cemented unconstrained mobile-bearing TKR and cemented unconstrained fixed-bearing TKR is not seen for these bearing options in the uncemented/hybrid TKR groups. This lack of difference is being driven by the higher absolute failure rate of uncemented/hybrid unconstrained, fixed bearing. In contrast, the higher revision rate for fixed bearings in cemented posterior-stabilised TKRs becomes even more apparent in uncemented posterior-stabilised knees.
Considering the higher revision rate of primary unicompartmental knees that is seen across the 15 years of NJR data, it is of interest to see that up to 12 years the re-revision rate for unicompartmental knees that have been revised is lower in the NJR data compared to the re-revision of revised primary TKRs. This difference is small and the number at risk beyond seven years is low so the observation should be considered cautiously in light of the fact that the “first revision” of unicompartmental knees contains a mix of procedures ranging from simple bearing exchange in non-infected cases to full revision using a “revision” type of TKR.
The re-revision rate of revised primary patellofemoral joints is seen to be substantially lower in relation to the re-revision rate of primary total knees and of unicompartmental knees. Since first-revision of patellofemoral replacements is a less diverse procedure this lower re-revision rate is much more likely to be real than the more modest difference seen for unicompartmental knees. However, it is important to balance the survivorship of the revision against the likelihood of revising the primary.
Ankle replacement procedures
Ankle replacements have only been entered into the NJR since April 2010, and the numbers remain relatively small compared to hips, knees and shoulders. Nevertheless, the number of primary cases in the NJR rose by 843 during the year to reach 5,587, which now represents a very large cohort.
There have been 265 revision operations on these procedures, which include 37 conversions to arthrodesis. Unfortunately, the collection of data about arthrodesis and amputation as a “revision” outcome of ankle replacement is known to be incomplete in the NJR data. There may also have been a misunderstanding by some surgeons about whether those procedures are supposed to be registered as revision procedures by the completion of an A2 Minimum Dataset form and submission to the NJR. However, the mandatory reporting requirements and NJR definition of a revision are clear that the definition of revision does include any case where a component of an arthroplasty is either removed, modified or added at a subsequent procedure.
Since amputation is commonly performed by vascular surgeons in the UK, this may also have led to difficulties with the completeness of data since those surgeons are not within orthopaedic units and may not be familiar with the NJR and the mandatory reporting requirements. Analysis of data for failure rates and reasons for failure of ankle replacement therefore remains difficult and potentially inaccurate compared to some other joints. The overall revision rates for ankle replacement still need to be interpreted with caution.
It is clear that ankle replacement is being performed predominantly in male patients and that the overwhelming majority of those being registered involved uncemented implants. The numbers of ankle replacements being performed by each surgeon are remarkably similar to the numbers being performed by each unit where they are done, suggesting that surgeons have generally already selected one amongst their number in each unit to perform these procedures.
Notwithstanding the difficulties outlined above with respect to interpreting the data, it seems clear that the main reason for revision in these joints has been loosening of either the talar component, the tibial component or both. The rate of revision for infection must be interpreted with particular caution as that could be a failure mode which might particularly be expected to result in arthrodesis or amputation rather than revision surgery.
The comparative results of different implants and implant types are currently also difficult to interpret with confidence, not least because the most popular implant was voluntarily removed from the market by the manufacturer in 2014, and in the same year a quite different implant was introduced and immediately became the best-selling implant. This latter implant has a very short-term follow-up in the NJR despite the high numbers implanted.
Elbow replacement procedures
This section relates to radial head replacement, distal humeral hemiarthroplasty and total elbow replacement. As with ankle replacement, the numbers are relatively small compared with hips and knees, but the 3,573 cases still represent a substantial cohort of elbow replacements.
In contrast to some other joints, a large proportion of elbow replacements are performed for acute trauma, with this indication accounting for over 35% of the total. The trauma cases also differ significantly from the elective cases in terms of the type of implant used. In trauma, radial head replacement and distal humeral hemiarthroplasty make up over half the cases, whereas 90% of elective cases are total elbow replacement. Female patients make up 70.6% of cases.
The great majority of revision cases being entered into the NJR have been from primary cases performed before elbow replacements started to be included in the NJR in April 2012.
The number of surgeons entering primary elbow cases into the NJR has diminished slightly over recent years and the number entering revision cases were far fewer during 2018 than over the whole of the previous five years. This may reflect professional advice that elbow replacements are best concentrated in a small number of sites and surgeons so as to increase individual experience.
The majority of elective elbow replacements are still performed for inflammatory arthropathy.
The likelihood of having a revision elbow replacement is substantially higher during the first five years after elective replacements than after replacement for trauma (7.4% vs 3.0%). This could be for a number of reasons but it is important to note the very different spectrum of procedures being compared. At present, the numbers in the database do not allow for stratification and subset analysis to allow the reasons to be fully analysed by gender, age or individual procedure type.
Mortality after elbow replacement is seen to be 16.5% at five years and appears higher in trauma vs elective cases (19.8% vs. 15.2%) though the extent to which this is simply a manifestation of the mean ages is not yet clear.
Shoulder replacement procedures
Shoulder replacements began being registered in NJR in April 2012, and since that time 37,916 primary operations are available for analysis. Of these, 1,158 have undergone a revision operation.
Female patients accounted for 70.5% of shoulder replacements. There has been a quite dramatic change in the type of procedure being performed on the shoulder in recent years such that in 2018, 57% of all such operations were reverse polarity shoulders, which represents a 16% increase since 2015. Conventional shoulder replacements seem to be holding a fairly steady rate of implantation while humeral hemiarthroplasty is declining in numbers.
Relatively large numbers of surgeons appear to be performing shoulder replacements (722 surgeons in 395 units) considering the numbers being done overall. Consequently, the numbers performed by each surgeon remain relatively small compared to those performing hips and knees, although each surgeon still tends to perform a greater volume of procedures than surgeons performing ankle and elbow replacements.
A total of 91% of the shoulder replacements were performed for elective indications and 9% for acute trauma.
The changing spectrum of use of differing types of shoulder implant is a notable feature and clarification is still needed about many aspects of these changes. One of the underlying problems which were originally being addressed by reverse polarity shoulder designs was significant rotator cuff deficiency. The dramatic increase in the use of reverse polarity shoulders in recent years suggests that these devices are not now being used solely in patients with deficient rotator cuffs.
Revision rates are now available for the different types of shoulder implant and these demonstrate lower revision rates for total shoulders (stemmed, stemless and resurfacing) and for stemmed reverse polarity shoulders. The revision rates are higher for hemiarthroplasty (stemmed, stemless and resurfacing) and for stemless reverse polarity shoulders. There is evidence that fewer hemiarthroplasties are now being performed and this should be regarded as encouraging given the revision data.
As with unicompartmental knees, it is reasonable to think that the higher revision rate for some procedures such as hemiarthroplasty may be due to a number of complicated factors. There may be a perception that these operations would be simpler than more major revisions and there may be more willingness to undertake them. Similarly, revision is only one of the important endpoints and the issue of the actual symptomatic benefit is important. This aspect of outcomes may be more accurately reflected by the PROMs score and PROMs gain than by simple revision data.
Shoulder PROMs have therefore been introduced as an integral part of the NJR assessment of shoulders in particular as these have not previously been part of the National PROMs program. It is hoped that as this PROMs data increases in both breadth and duration it will help to demonstrate whether some of those implants which are “surviving” better may be doing so despite less good clinical outcomes or function, or whether these factors coincide to demonstrate the “best” procedures.
The PROMs data already demonstrate that considerable improvement is being achieved across the cohort of shoulder replacements, and it is anticipated that these data will be able to be stratified by implant and operation type in due course. This stratification will also be facilitated by the new implant database which is currently being introduced and which will allow for more granular comparisons when the numbers allow. The PROMs data do show that 12% of patients do not attain the minimum clinically significant improvement by six months post-operatively and all the major categories of shoulder replacement contain such patients. There are also seen to be 7% of elective patients who are worse after six months than pre-operatively, a matter that clearly needs investigation and explanation.
Young patients are seen to have high revision rates for shoulder replacement compared to those in similar aged hip and knee replacement patients. The revision rate is around 10% (10.2% males, 9.7% females) at four years in the under 55 year age group, compared to 4% at four years in TKR and 2.5% and 2% at four years for THR. This is important information in allowing properly informed consent for the patients, in particular for elective procedures.
These figures do nevertheless compare quite favourably to the 4-year revision rates for unicompartmental knees in the under 55 year age group (~7.5%).