Part Three: Outcomes after joint replacement 2003 to 2013
Part Three of the NJR 11th Annual Report provides information regarding the survivorship of implants as well as mortality in relation to hip and knee replacement operations carried out in England, Wales and Northern Ireland between 1 April 2003 and 31 December 2013. This is the first time the NJR has reported survivorship information in excess of ten years. Short term information is provided for ankle replacements and early data for shoulder replacements is currently being investigated and will be published separately online in the Autumn. Elbow data will be reviewed when numbers are sufficient for meaningful analysis.
For the purposes of this analysis in excess of 1.6million records were available but a number of records were excluded. The final analysis comprises 620,400 primary hip replacements, 676,082 primary knee replacement and 2,004 ankle replacements.
HIP REPLACEMENT PROCEDURES
The 620,400 primary hip replacements were carried out by a total of 2,906 consultants within 460 hospitals. The predominant diagnosis was osteoarthritis in 93%, 40.3% were males and the median age was 69 years (IQR 61-76). The maximum follow-up of these patients was 10.75 years. The data is presented in terms of fixation and bearing with further analysis to take age and gender into consideration and finally the survivorship of common brands that were used.
In 2003 cemented hip replacement was used in 60.5% of cases but this reduced year on year to 31.9% in 2009, then increased slightly to 33.2% in 2013. Over the same time period there was a corresponding increase in the use of uncemented implants from 16.8% in 2003 to 42.5% in 2013. Over the same period, hybrid hip replacement increased from 12.3% to 20.2% and resurfacing reached a peak of 10.8% in 2006 but then declined to just 1.1% in 2013. There were important differences in the choice of bearing surface and mode of fixation. For example, with cemented hip replacement, a metal-on-polyethylene bearing was used in 88.2% of cases whereas when the replacement was uncemented there was greater diversity of bearing selection: metal-on-polyethylene was used in 37.1%, ceramic-onceramic in 33.8%, ceramic-on-polyethylene in 14.4% and a metal-on-metal articulation in 11.9%.
There was a dramatic rise and then fall in the use of a metal-on-metal bearing with an uncemented construct. For example in 2003, metal-on-metal bearings were used in 7.5% of cases and this peaked in 2007 to 31.1% and then fell away rapidly from 2009 onwards when the adverse effects of metal-on-metal bearings became clear. In 2013 their use had fallen to just 0.1% of cases. In relation to implant survivorship, cemented fixation without adjustment by bearing, age and gender appeared to be the most successful when there was a 3.20% (95% confidence interval 3.03-3.39) cumulative percentage probability of revision at 10 years. This compared to 7.68% (7.34-8.03) with an uncemented, hybrid 3.95% (3.60-4.34), reverse hybrid 4.77% (3.07-7.37) and resurfacing 13.01% (12.40-3.65).
However, as mentioned previously, the metal-on-metal articulation was used much more frequently with uncemented fixation and if one compares the most ‘successful’ bearing combination ceramic-on-polyethylene, the differences in fixation become less marked. For example, when cement was used the ten-year revision probability with ceramic-on-polyethylene was 2.09% compared to 3.73% with uncemented ceramic-on-polyethylene, 2.19% with hybrid ceramic-on-polyethylene. These figures were not adjusted for age and gender.
A wider range of revision probability was seen when hip replacements were broken down by fixation, bearing, age and gender. The data demonstrated low revision rates with ceramic-on-polyethylene bearings and high revision rates with young patients (less than 55 years at the time of surgery). For example, for male patients with all fixation groups and all bearings, the ten-year revision probability was 9.12% in patients less than 55 years compared to 3.45% in patients greater than 75 years. When cemented fixation and a ceramic-on-polyethylene bearing was used in patients less than 55 years, in male patients the ten-year revision estimate was lower at 3.68% and in male patients over 75, was 1.49%.
When uncemented fixation and a metal-on-metal bearing was used in male patients less than 55 years, the revision estimate was at 21.82% and females of the same age group 28.39%. In terms of brand analysis, considering the combination of stem and cup, it could be seen that many combinations had a relatively low revision probability at ten years. For example, the ten-year revision probability in all patients when a cemented Charnley stem was used with a cemented Charnley cup was 2.87% (2.43–3.39) and when an Exeter V40 was used with a Contemporary socket the revision risk at ten years was 3.15% (2.62–3.79). It was interesting to note that one of the lowest revision risks was that of an Exeter V40 with an Elite Plus cemented cup with a rate of just 1.36%. This represents a construct where the stem is made by one manufacturer and the socket another, a so-called mix and match system.
In relation to resurfacing, it could be seen that different brands had different revision probabilities at ten years. The revision probability of the ASR was 30.36% (27.49–33.45) compared to the BHR of 9.04% (8.37–9.76). As reported last year the importance of the choice of bearing material could be seen within one particular brand combination, the Corail stem and Pinnacle socket, both made by the same manufacturer. At seven years when a metal-on-metal articulation was used with this construct the revision risk was 8.82% compared to just 1.84% when the ceramic-on-polyethylene bearing was used.
The different causes for revision were explored in more detail which identified some intuitive trends; namely a decreasing risk of dislocation with a PTIR of 2.33 per 1000 patient-years in the first year, reducing to 0.66 between years one and three. In comparison the PTIR for adverse reaction to particulate debris was 7.2 with an uncemented metal-on-metal articulation compared to just 0.14 with ceramic-on-polyethylene. The overall PTIR of adverse reaction to particulate debris was 0.10 within the first year increasing to 3.27 between years five and seven.
In-depth study on the effect of short-term mortality
The effect of short-term mortality after hip replacement was studied in detail and published in The Lancet in November 2013. The 90-day mortality halved from 0.56% to 0.29% over the eight year period examined. A multivariable analysis failed to demonstrate any significant effect of the type of replacement on short term mortality. Variables that were associated with decreased mortality were spinal anaesthetic, posterior approach, the use of mechanical thromboprophylaxis and the use of chemical thromboprophylaxis.
KNEE REPLACEMENT PROCEDURES
The 676,082 primary knee replacements were carried out by a total of 2,813 surgeons within 450 hospitals. The diagnosis was osteoarthritis in 96% of cases, 57% of patients were female and the median age for primary surgery was 70 years. The follow-up period for these patients extended to 10.75 years.
The joints replaced and patients were analysed according to the type of replacement (total, unicondylar etc), fixation method, constraint and then by brand. In comparison to hip replacement procedures there has been less variation in practice over the last ten years. In 2003, 81.5% of procedures were total knee replacements using a cemented method of fixation and this increased to 88% by 2013. Over the same time period, uncemented replacements fell from 6.7% to 2.4% and the use of hybrid fixation from 2.8% to 0.4% of all primary surgeries. In spite of the observation of higher revision rates amongst partial knee replacements (unicondylar and patello-femoral) the proportional use of these has remained at about 8% for unicondylar and 1% for patella-femoral replacement.
The most common type of implant and method of fixation used was a cemented, unconstrained (post cruciate retaining) fixed bearing type total knee replacement which increased in use from 66% in 2003 to 71% in 2013. Where a unicondylar knee was implanted there was a trend away from the use of a mobile bearing; from 81.3% of all unicondylar replacements in 2003 to 60% in 2013. The ten-year risk of revision for all knee types surviving to that point in time was 4.47% (4.34-4.60). For cemented, unconstrained fixed bearing total knee replacement the ten-year revision risk was 3.06% (2.93-3.19), for cemented unconstrained mobile 4.26% (3.63-5.06), cemented posterior stabilised fixed 3.63% (3.39-3.88) and cemented posterior stabilised mobile bearing 4.74% (3.66-6.15 (fewer than 250 cases remain at risk at the time shown)). Generally the risk of revision is low at ten years (less than 5%) with relatively small differences between the types of bearing, although some fixation/constraint methods do fare better than others (e.g. amongst cemented implants, unconstrained implants have significantly lower risk of revision after ten years).
Of those unicondylar implants surviving for ten years, the chance of revision was nearly three times higher than the equivalent group of total knee replacements surviving with little difference between fixed and mobile varieties (11.78% for fixed unicondylar and 12.95% for mobile unicondylar). The age of the patient at time of primary surgery had a profound effect on the revision risk. For example, for female patients aged less than 55 years, the ten year revision risk was 11.07% (10.06-12.16) compared to 2.0% (1.82-2.19) for female patients aged 75 or more. When a cemented, unconstrained fixed bearing knee was used the comparable revision risks were 6.79% in female patients under 55 years compared to 1.58% in patients over 75 years. When a unicondylar knee was used in a female, the corresponding risk of first revision for those implants surviving up to ten years was 18.98% and 9.69%.
For total knee replacement, analysis by brand revealed similar findings to total hip replacement in that many different brands demonstrated revision risk at ten years, for those surviving to ten years, between 3% and 4% with overlapping confidence intervals indicating similar performances in terms of risk of implant revision. There were some exceptions to this trend for total knee implant brands, notably the Rotaglide + mobile knee had a ten-year revision risk of 7.14% compared with, for example, the PFC which had a risk of implant revision for those surviving to ten years of 2.66%. For unicondylar knees, the risk of implant revision tends to be higher than total knee brands in general and more variable.
Amongst unicondylar implant brands with ten years follow-up, the ten-year risk of revision was highest for a Preservation implant at 20.14% compared to an Oxford partial knee implant with a 12.42% chance of being revised after ten years. The risk of revision of different brand types at varying time points after primary surgery are presented in further detail according to surgical method of fixation and constraint choice opted for. The different causes for revision were studied as well as the temporal changes in revision risk by reason for revision. As expected, revision for infection was more common in the first year of surgery and then fell in subsequent years but the opposite was true for pain. Not all types of replacement had a similar risk of revision by reason. For example, when the reason for revision was infection, the PTIR for cemented knees was 1.13 revisions per 1,000 patient-years (1.08-1.17) compared to 0.7 (0.6-0.81) for a unicondylar type of replacement. However, for pain, the corresponding rates were higher for unicondylar knees at 3.84 (3.60-4.10) compared to 0.76 (0.73-0.80) for cemented total knee replacements.
In-depth study on the effect of short-term mortality
The effect of short-term mortality after knee replacement was studied in detail and published online in The Lancet in July 2014. The 45-day mortality almost halved from 0.37% in 2003 to 0.20% over the eight year period examined. A multivariable analysis confirmed mortality rates for unicondylar knee were lower than total knees, which is perhaps to be expected as the former is a less invasive procedure. This advantage, however, would need to be weighed up with the higher rates of revision that are consistently reported for unicondylar than for total knee replacement.
ANKLE REPLACEMENT PROCEDURES
A total of 2,004 primary ankle replacements were recorded on the NJR between April 2010 and December 2013. These were carried out by a total of 184 surgeons in 196 hospitals. The median age of the patients was 68 years (IQR 61-74) and 57% were male.
The Mobility was the most commonly used brand of replacement which decreased in market share from 62.0% in 2010 to 35.6% in 2013. The Zenith increased its share from 19.0% to 24.6% over the same period. Nearly all replacements were uncemented. There were 29 revisions recorded which may well be a significant under-reporting. At three years the revision risk was 2.78% (1.89-4.10) although these early estimates should be viewed with caution.
A NOTE ON PATIENT REPORTED OUTCOME MEASURES (PROMs)
Last year, the NJR reported PROMs for the first time. This year, no data from the NHS England PROMs programme was available to the NJR from the NHS Health and Social Care Information Centre.