Does Knee Prosthesis Survivorship Improve When Implant Designs Change? Findings from the Australian Orthopaedic Association National Joint Replacement Registry

Abstract

Background

TKA generally has excellent long-term survivorship. When a new knee system supersedes a previous model, increased survivorship, improved functional performance, or both may be expected, because key areas of design modification are often targeted to address wear, stability, and the patellofemoral articulation. However, not all design changes are beneficial, and to our knowledge, knee arthroplasty has not been systematically evaluated in the context of design changes that occur during the development of new knee arthroplasty systems.

Questions/purposes

Using the Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR) we performed multiple old-to-new comparisons of frequently used contemporary knee implants to ask: (1) does overall prosthesis survivorship free from revision increase when a new knee prosthesis system is introduced to replace a prior prosthesis system? (2) Has survivorship free from revision improved for the revision indications of wear, instability, and patellofemoral articulation issues, where development efforts have been concentrated?

Methods

Data from the AOANJRR from September 1999 to December 2017 were used to compare the survivorship of prostheses free from revision at a maximum of 17 years in procedures where a new design model was introduced to replace a prior knee system from the same manufacturer. Only prosthesis systems used in a minimum of 2000 primary TKA procedures for osteoarthritis that had a minimum of 5 years of follow-up were included. Varus-valgus constrained and hinge TKA designs were excluded. Cruciate-retaining, posterior-stabilized, and medial pivot-design knees were considered separately. The new and old prosthesis systems were paired for analysis. Survivorship was calculated with Kaplan Meier estimates and comparisons were performed using the Cox proportional hazards method. Subanalyses according to the three main revision indications were performed, and where possible, analyses were performed based on polyethylene types (highly cross-linked polyethylene and ultra-high-molecular-weight polyethylene), combined and separated. Revision was defined as a reoperation of a previous knee arthroplasty in which one or more of the components was removed, replaced, or added. There were 323,955 TKA procedures and 11 new prosthesis system designs that were introduced to replace an earlier knee system from the same manufacturer. Of these prosthesis system pairs, six were cruciate-retaining prostheses, four were posterior-stabilized designs, and one was a medial pivot design.

Results

Six of the 11 knee system pairs showed improved survivorship with the new design, three were no different, and in two, the newer prosthesis systems had a higher rate of revision than the old one did. When revision for wear was analyzed, five prosthesis systems showed improvement, five were no different, and one had a higher rate of revision than the previous system did. There was no improvement in the rate of revision for instability; seven new prosthesis systems showed no difference from the previous system and four new prosthesis systems had a higher rate of revision than the previous system did. A subanalysis of revision for patellofemoral complications showed improvement in two comparisons, no difference in six, and a higher revision rate in two; one could not be calculated because of an insufficient number of revisions for this reason.

Conclusions

It is difficult to predict whether a new system will demonstrate better survival than a previous one, and widespread uptake of a new design before a benefit is shown in robust clinical studies is unwise. Similarly, adoption of a new system for which there is no difference in survivorship from a previous model may be premature because a new device may have associated unknown and unintended consequences. Healthcare policy makers and therapeutic device regulators should similarly be guided by results and seek out peer-reviewed evidence before accepting change to established practice. Surgeons must be aware that implant changes may not translate into better survivorship and must seek compelling evidence of improvement in survival and/or function before changing systems.

Level of Evidence

Level III, therapeutic study.

Introduction

Each year, new knee replacement models are made available for use by surgeons. In 2017 alone, 75 new combinations of prostheses were recorded by the Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR) [4]. Although the frequency of the use of many of these combinations is small [4], in some instances, entire new knee systems are added to substitute for a previous older design. Manufacturers, surgeons, and patients should all rightfully expect better survivorship when a new system supersedes a previous model [15]. Additionally, survivorship studies with responsible step-wise introduction of new technology outlined by Malchau [20] are required. There is, however, little available information as to whether the introduction of new prosthesis systems is beneficial [1, 15, 17]. One registry study [1] and two meta-analyses [16, 18] examining changes to two different prosthesis system designs have studied survivorship, each showing no improvement with the newer designs.

Previous studies comparing old and new prosthesis systems are limited to two successive designs from the same company [8, 13, 14, 17, 21, 22, 24-26, 31, 37]. New prostheses are typically developed to address deficiencies or problems of previous designs. The key areas that designers generally aim to improve are stability and kinematics, wear, and patellofemoral complications [8, 14, 17, 24-26, 31, 37]. These issues, along with infection, are among the more common indications for TKA revision [4, 11, 30, 34]. It is reasonable to conclude that if these problem areas are rectified, then survivorship should improve. However, this has not been the focus of these prior comparisons. Measuring improvement after design change is challenging, even in studies examining these specific aspects, because not only is the resultant change often difficult to measure, but differences between one prosthesis system and another may also be difficult to show. Studies investigating design changes related to stability have examined the femoral component radius of curvature or a high flex design using knee outcome scores or functional questionnaires [8, 18, 25]. Researchers studying design change related to wear have used retrieval analyses, contact patterns, and radiostereometric analyses [14, 22, 24]. Patellofemoral studies of design improvements have investigated the presence of anterior knee pain, patellofemoral crepitus, or mal-tracking, and the frequency of patellar clunk or use of lateral release [17, 21, 31, 37]. Most of these studies adopted a matched-pair design [17, 21, 22, 25, 31, 37].

There is evidence that the expected benefit is not always achieved with design modifications. Some design comparisons have revealed no difference between the original design and the newer one, particularly when knee outcome scores are studied [6, 17, 25]. In other instances, even simple changes have produced poorer outcomes [3, 12, 23, 29, 32]. For example, the keel of the tibial component for both the Nexgen (ZimmerBiomet, Warsaw, IN, USA) and Genesis II (Smith & Nephew, Memphis, TN, USA) knee systems was shortened, and this change was associated with increased failure because of loosening [3, 32]. Many of the direct comparison studies claimed to be limited by small sample sizes or short follow-up periods; as a consequence, these studies have recommended that further comparisons should be made to determine whether progress is truly evident [8, 17, 22, 25, 28].

Using the AOANJRR, we performed multiple old-to-new comparisons of frequently used contemporary knee implants to ask: (1) does overall prosthesis survivorship free from revision increase when a new knee prosthesis system is introduced to replace a prior prosthesis system? (2) Has survivorship free from revision improved for the revision indications of wear, instability, and patellofemoral articulation issues, where development efforts have been concentrated?

Patients and Methods

Data from the AOANJRR from September 1999 to December 2017 were used to assess the performance of knee prosthesis systems. The AOANJRR began collecting data on September 1, 1999, achieving complete national implementation by mid-2002. Since then, it has collected data on more than 98.8% of THAs and TKAs performed in Australia [4]. These data are validated against patient-level data provided externally by all Australian state and territory health departments. A sequential, multilevel matching process is used to identify any missing data, which are subsequently retrieved by contacting the relevant hospital. Each month, in conjunction with internal validation and data quality checks, all primary procedures are linked to any subsequent revision involving the same patient, joint, and side. Data are also matched biannually with the Australian Government’s National Death Index to obtain information on the date of death. Linking revision and death to the primary procedure enables revision rates to be determined.

The study included all new knee prosthesis systems used in primary TKAs for osteoarthritis that were introduced to replace a knee prosthesis system from the same manufacturer. Knee prosthesis systems were limited to those used in a minimum of 2000 TKA procedures for at least 5 years. Varus/valgus constrained and hinged prostheses were excluded from the analysis because these are usually reserved for difficult or unusual clinical situations. Because the rates of revision are known to differ according to the type of constraint used, prosthesis systems using cruciate-retaining (CR), posterior-stabilized (PS), and medial-pivot design (MPD) knees were considered separately [4, 35]. The new and old prosthesis systems were paired for each analysis. Overall, 323,955 TKAs using 11 pairs of old and new prosthesis system designs that had the same manufacturer were included in the study. Of these, six were CR prostheses, four were PS, and one was an MPD.

Seven comparisons had the initial and new prosthesis knee systems still in use, and one comparison no longer had either prosthesis system in use, while one older and one newer system had been discontinued in two comparisons. The patients had a mean age of 69 years (standard deviation 9.2 years), and most were women (57%) (Table 1).

Table 1.

Age and gender of patients undergoing primary TKA by prosthesis system, manufacturer, and model (primary diagnosis of osteoarthritis)

During the study period, highly cross-linked polyethylene (XLPE) was introduced, and its proportional use varied between older and newer prostheses. Analyses were performed with polyethylene types combined and separated, where possible, to reduce the potential influence of the polyethylene type on outcomes. Six prostheses used only ultra-high-molecular-weight polyethylene (UHMWPE), including two pairs in the comparisons (AGC CR/Maxim CR [Zimmer-Biomet, Warsaw, IN, USA] and the Advance MPD/Evolution MPD [Microport Orthopedics, Arlington, TN, USA]). Two prostheses used XLPE exclusively (Attune CR and PS [DePuy, Warsaw, IN, USA]). The proportion of resurfaced patellae ranged from 36% for the Duracon CR knee (Stryker, Kalamazoo, MI, USA) to 93% for the Attune PS (DePuy, Warsaw, IN, USA) (Table 2).

Table 2.

Polyethylene type and patella component use for primary TKA by prosthesis model (primary diagnosis of osteoarthritis)

Revision was defined as reoperation of a previous knee arthroplasty in which one or more of the components was removed, replaced, or added. Revision diagnoses recorded by the surgeon at the time of revision were used for further analyses and, where possible, fitted to three categories. These categories were wear and associated consequences, which included revision for wear of the tibial insert; tibial wear, lysis, or loosening after 2 years (early loosening is more likely a failure of initial fixation); instability, which included instability, prosthesis dislocation, and bearing dislocation; and patellofemoral causes, which included patellofemoral pain, patella erosion, patella mal-tracking, patella wear, and patella component breakage. The latter comparisons were further performed for resurfaced and un-resurfaced patellae.

Statistical Analysis

Kaplan-Meier estimates of survivorship were used to report the time to revision of knee prosthesis systems used in primary TKA procedures, with censoring at the time of death or closure of the dataset at the end of December 2017. The follow-up time period varied from the minimum inclusion time of 5 years to a maximum of 18 years. The unadjusted cumulative percent revision with 95% confidence intervals was calculated using unadjusted point-wise Greenwood estimates until the number at risk for each group reached 40. Age- and sex-adjusted hazard ratios calculated from Cox proportional hazard models were used to compare the rate of revision between the groups. A minimum of four revisions in each group were required to calculate HRs to avoid uninformative and imprecise estimates. The assumption of proportional hazards was checked analytically for each model. If the interaction between the predictor and the log of time was statistically significant in the standard Cox model, then a time-varying model was estimated. Timepoints were selected based on the greatest change in hazard, weighted by a function of events. Timepoints were iteratively chosen until the assumption of proportionality was met, and HRs were calculated for each selected time period. For the current study, if no time period was specified, the HR was calculated for the entire follow-up period. All tests were two-tailed at 5% levels of significance. The statistical analysis was performed using SAS software version 9.4 (SAS Institute Inc., Cary, NC, USA).

Results

All-cause Revision

Six knee prosthesis system pairs showed improved survivorship with the new prosthesis design: the Duracon (n = 19,493) and Triathlon CR (n = 70,145), Scorpio CR (n = 19,323) and Triathlon CR (n = 70,145) (after 6 months), Scorpio PS (n = 11,202) and Triathlon PS (n = 10,464) (initially lower survivorship until 6 months and higher survivorship between 9 months and 5 years), PFC Sigma PS (n = 15,843) and Attune PS (n = 3961), Genesis II PS (n = 34,319) and Legion PS (n = 15,916), and Advance MPD (n = 2282) and Evolution MPD (n = 4810) (Table 3).

Table 3.

CPR and HRs of primary TKA by prosthesis model for all data, XLPE only, and UHMWPE only (primary diagnosis of osteoarthritis, all-cause revision)

Three knee prosthesis system pairs had no difference in survivorship. These were the PFC Sigma CR (n = 33,770) and Attune CR (n = 8729), Genesis II CR (n = 31,971) and Legion CR (n = 7047), and Maxim CR (n = 2317) and Vanguard CR (n = 24,285). Lower survivorship was evident in two knee prosthesis systems pairs: the Genesis II PS (n = 34,319) and Journey (n = 3068) (with variation during the time periods), as well as the AGC CR (n = 5019) and Maxim CR (n = 2317) (Table 3).

Where possible, the polyethylene type was stratified. Of the six prosthesis pairs that showed an improvement, four could be assessed with both XLPE and UHMWPE. The Scorpio CR-to-Triathlon CR comparison revealed improved survivorship with both polyethylene types, the Genesis II PS (n = 8149)-to-Legion PS (n = 8828) comparison showed improvement only with XLPE, and the Genesis II CR (n = 28,511)-to-Legion CR (n = 3628) pair’s survivorship only improved when UHMWPE was assessed. The Scorpio PS-to-Triathlon PS survivorship had variable improvement during the time period. The Genesis II CR-to-Legion CR pair showed no difference in the overall assessment and with XLPE but showed an improvement with UHMWPE (Table 3).

Revisions for Wear, Instability, and Patellofemoral Indications

When a subanalysis of revision for wear and its consequences was performed, five comparisons showed improved survivorship, five prosthesis systems showed no difference, and the Genesis II PS (n = 34,319)-to-Journey (n = 3068) comparison showed a lower survivorship with the newer design than with the old one. When stratification by polyethylene was possible, only the Scorpio PS-to-Triathlon PS comparison showed improvement in survivorship with both polyethylene types. The Scorpio CR (n = 14,672)-to-Triathlon CR (n = 10,139), Genesis II PS (n = 26,158)-to-Legion PS (n = 7084), and Genesis II CR (n = 28,511)-to-Legion CR (n = 3628) pairs showed increased survival with UHMWPE (Table 4).

Table 4.

CPR and HRs of primary TKA by prosthesis model for all data, XLPE only, and UHMWPE only (primary diagnosis of osteoarthritis, revision for wear)

When revision for the indication of instability was further analyzed, no system comparisons showed improvement, seven showed no difference, and four comparisons showed decreased survivorship with the newer prosthesis system. Of the four new prosthesis systems with decreased survivorship for instability, two were CR (Scorpio CR [n = 19,323]-to-Triathlon CR [n = 70,145] and AGC CR [n = 5019]-to-Maxim CR [n = 2317]) and two were PS designs (Genesis II PS [n = 34,319]-to-Journey [n =3068] and Genesis II PS [n = 34,319]-to-Legion PS [n = 15,916]) (Table 5).

Table 5.

CPR and HRs of primary TKA by prosthesis model (primary diagnosis of osteoarthritis, revision for instability)

A subanalysis of revision for patellofemoral indications showed improvement in two comparisons (Scorpio CR [n = 19,323]-to Triathlon CR [n = 70,145] and PFC Sigma CR [n = 33,770]-to-Attune CR [n = 8729]), no difference in six, and a lower survivorship result in two (Genesis II CR [n = 31,971]-to-Legion CR [n = 7047] only until 1.5 years; Genesis II PS [n = 34,319]-to-Journey [n = 3068] for the entire period). One pair could not be calculated because of an insufficient number of revisions for this reason. When stratified by patella component use, of the two overall comparisons that showed improved survival, survival for the Scorpio CR (n = 9642)-to-Triathlon CR (n = 34,022) comparison only improved when the patella was not resurfaced, while the PFC Sigma CR (n = 16,800)-to-Attune CR (n = 3176) comparison showed no difference without resurfacing. The data for resurfaced patellae with this pair could not be calculated because of the low number of revisions. Both comparisons that had lower overall survival for patellofemoral causes only showed this difference when the patella was not resurfaced. These were the Genesis II CR (n = 18,207)-to-Legion CR (n = 2688) for the first 1.5 years and the Genesis II PS (n = 11,698)-to-Journey (n = 1868) (Table 6).

Table 6.

CPR and HRs of primary TKA by prosthesis model for all data, patella resurfaced, and not resurfaced (primary diagnosis of osteoarthritis, revision for patella reasons)

Discussion

Increased survivorship and improved functional performance are expected when a new knee system supersedes a previous model. Although new knee systems often have design rationales to address wear, stability, and the patellofemoral articulation, not all design changes result in improved survivorship. We found that when a new prosthesis system replaced a previous model, survivorship was improved in only six of 11 comparisons. It is difficult to predict whether a new system will outperform a previous one. It is also unwise to participate in widespread adoption of a new design before a benefit is shown in robust clinical studies. Similarly, adoption of a new system for which there is no difference in survivorship from a previous model may be irresponsible because there may be unknown and unintended consequences of the new device. One of the newer prostheses in this analysis (Journey, Smith & Nephew, Memphis, TN, USA) has been identified as an underperforming implant and was removed from the market.

Limitations

Our study has certain limitations. This study used implant survival as the outcome measure. Although revision surgery is a recognized and widely accepted method of measuring outcome, improvements in prosthesis performance in areas such as satisfaction and function may occur without being captured by a survivorship method. Revision rates provide a distinct and unequivocal measure, whereas other measures such as ROM, knee scores, or pain scores are more subjective. Revision may underestimate the number of failures because some patients with failed knee prostheses may decide against, or may be too unwell to undergo, revision surgery. Revision only captures more-severe complications and may underestimate the total number of patients with a particular problem such as patella pain, where the impact may not be enough to warrant revision surgery. Using revision as the endpoint may also be problematic from a statistical standpoint alone because knee arthroplasty survivorship is generally very good, and it may be difficult to show further improvement. Although registry analyses may have some limitations, we feel this method is perhaps simpler and more easily understood than other methods of measuring the effects of design change, which also have shortcomings. Randomized controlled trials or even case-controlled studies are complex to design and complete, large sample sizes are required, and there is the added difficulty of choosing responsive and appropriate outcome measures. We also acknowledge that when deciding which prosthesis to use, despite no measured improvement in survivorship, surgeons might choose a particular design that may lead to improved patient satisfaction.

During the study period, XLPE was introduced and was disproportionately used with newer prostheses. There are prosthesis-specific differences in survivorship with the use of XLPE [9]. Although there may be bias in the overall and wear-related comparisons in favor of this bearing surface, we attempted to control for this factor by stratifying by polyethylene type. We similarly stratified by patella component use because prosthesis-specific variations in survival may affect the revision rates [4]. Stratification decreases the numbers available for each comparison, which may make detecting a difference more difficult. Although stratification has some limitations, the complexities of differences in use can be managed with this method.

Although we controlled for prosthesis stability, we did not further subdivide by the use of cement fixation or mobile bearings. Similarly, other factors were not considered, such as hospital caseload volume, surgeon experience, and the use of assistive technology such as computer navigation, image-derived instrumentation, or robotic assistance for bone preparation. Although these factors may influence prosthesis survival, we expected that further subdivision would not only make the analysis more complex, but also render the groups too small for meaningful comparisons to be made. Although all of these attributes may influence survivorship, they were considered to be distributed uniformly and therefore had a minimal effect on the results.

Although a “learning curve” may be associated with the introduction of a new prosthesis, we expected that the potential influence of this learning curve would be diminished by similarities in contemporary knee arthroplasty designs and instrumentation. The learning effect would also be present in the “older” prosthesis group as new surgeons commenced their individual joint replacement practice during the study period.

A final limitation is that the revision causes are recorded by the surgeon at the time of the revision procedure, and there may be variation in surgeon interpretation, not only between surgeons but also over time. An example is the increased recognition of variations in instability (such as mid-flexion instability) during the study period [36]. This may partially explain the lack of improvement in revision for instability, but because the time periods of prosthesis pair use overlap, the effect of this factor may be diminished without discounting the findings of this study.

All-cause Revision

We found improved survivorship when a new prosthesis system replaced a previous model in only six of 11 comparisons. In two comparisons, the newer prosthesis had a higher cumulative percent revision than the old one did. This finding is contrary to what would be anticipated from design changes intended to improve survivorship [15]. However, the results are not completely unexpected, because previous authors have found that the introduction of new technologies or prosthetic designs does not necessarily lead to improvement [1, 23, 27]. Our findings are in contrast to the results of numerous other studies comparing old and new versions of a single prosthesis system that show either encouraging early results or at least a non-inferior outcome for the new design [1, 13, 17, 18, 21, 22, 31, 37]. Many of these studies were performed with acknowledged manufacturer sponsorship or are design surgeon series, which may account for the different outcomes observed in an unselected larger registry study [10, 16, 21, 22, 25, 26, 37].

The introduction of XLPE during the study period made comparisons of old and new knee designs more complex, because knee prostheses using XLPE have a rate of revision of 3.9% at 10 years compared with 5.7% for UHMWPE [4]. Of the six possible comparisons using XLPE, three showed improvement (with one for only a portion of the time period), while three showed no difference. The two comparisons with decreased survivorship both used only UHMWPE, and five of nine comparisons using UHMWPE showed increased survivorship (with one for only a portion of the time period).

Revisions for Wear, Instability, and Patellofemoral Indications

In five of the 11 prosthesis system pairs, there were fewer revisions for wear and its consequences with the newer design than with the old design. This number remained at five when only UHMWPE was assessed, while one comparison showed improvement when XLPE alone was analyzed. Improvement in prosthesis wear performance may be due to enhancements of the tibiofemoral articulation from a design or mechanical perspective, as well as the use of XLPE [7, 24]. A similar improvement in wear performance during this time period has been noted by others [11, 30, 33, 34].

Despite implant design changes intended to enhance stability, there was no improvement in revisions for instability, and there was decreased survivorship for this revision indication in four new prostheses. This lack of improvement may be due to the relative infrequency of revisions for these reasons. One implant design (Journey, Smith & Nephew, Memphis, TN, USA) was substantially different from its predecessor, with altered kinematics that led to increased instability-related revisions. Our finding of poorer survival for instability using newer prostheses contrasts with the results of large studies from Norway and the United States, which found that the rate of revision for instability has remained stable or decreased during this time period [11, 33]. Across all TKAs recorded in the AOANJRR, there is an increasing rate of revision for instability over time, which may be due to improved recognition or acknowledgement of this failure mechanism [5].

There was increased survivorship for revision for patellofemoral indications in only two of the 11 comparisons (which both involved CR prostheses), despite new designs that have specifically aimed to improve the patellofemoral articulation [8, 18, 21, 26, 37]. After stratification by patella component use, this improvement was only seen in one pair when the patellar remained unresurfaced, and in no instance when patellar resurfacing was performed. There was an increased rate of revision for patellofemoral reasons in two of the newer designs, and this finding was only evident when the patella remained unresurfaced. There were fewer revisions with the use of patella components, and in seven instances, revisions were so infrequent that calculating an HR was not possible. Although there is little evidence to suggest widespread improvement in the patellofemoral articulation, these data support the finding of a reduced rate of revision when a patella component is inserted [19]. The lack of difference may be because of the low frequency of revisions.

Conclusions

Although many products have a limited market life expectancy characterized by a cycle of introduction, growth, maturity, and decline, it is important that when producing new prosthesis systems, manufacturers continue to strive for improvements, closely monitor the results of these new systems, and not introduce new designs simply for marketing reasons [2]. Responsible introduction pathways for new technologies and devices should be followed. Surgeons should refrain from adopting a new design system until the results of robust clinical studies are known and must be aware of the unintended consequences of design changes. Healthcare policy makers and therapeutic device regulators should be guided by results and seek out peer-reviewed evidence before accepting change to established practice. The adoption of registry-nested clinical trials for new devices is a potential solution to this problem.