Introduction

Osteoarthritis affects nearly 27 million Americans [10] and is the leading cause of long-term disability [8, 13]. Total joint arthroplasty (TJA) is an effective alternative to medical management for the relief of pain in patients with osteoarthritis of the hip and knee [3, 19]. With an annual one million hip and knee arthroplasties performed in the United States [2], 400,000 in Germany [1], and 165,000 in the United Kingdom [16], TJA ranks among the most common high-cost procedures [21]. As a result of the success of these procedures, the British Medical Journal has termed TKA as the “joint of the decade” [14] and The Lancet has termed THA as the “operation of the century” [11].

However, TJA revisions are common. In the United States, for example, it has been reported that the revision burden is 17.5% after THA [9]. In Germany, as another example, there are an annual 160,000 primary THAs and 25,800 revisions [1], resulting in a burden of revision [9] of 14%. Revisions are an important problem both for society and for the individual patient. Several countries became aware of this problem and, as a consequence, have implemented national arthroplasty registries. The very first arthroplasty registry was started in Sweden in 1975 for registering knee arthroplasties and was followed by a hip arthroplasty registry, again in Sweden in 1979. After the implementation of these registries, the probability of revision was cut in half [6, 7, 12].

Encouraged by these experiences, several other countries started national joint registries such as Finland in 1980, Norway in 1987, Denmark in 1995, New Zealand in 1998, and Australia in 1999. After the disaster with the 3M Capital Hip system (St Paul, MN, USA), which was associated with a much higher than expected revision rate, England and Wales also implemented a national arthroplasty registry in 2003. All these national arthroplasty registries aim to report the survival of implants used for hip and/or knee arthroplasty to minimize the revision rate.

Given the high burden of revision in Germany, the national German Arthroplasty Registry was founded in 2011. When setting up the German registry, we sought to ensure high comparability to existing registries, including the definition of revision, which is the endpoint of all registries, for use in the German registry. However, when we looked at the definitions of revision in the reports of different national registries, it became evident that the definitions are quite different from one another.

Therefore we aimed to (1) compare the definition of revision in different reports; and to (2) apply different common clinical scenarios to these definitions to determine whether the different definitions might influence the likelihood that a particular procedure would be reported as a revision across the world’s registries.

Materials and Methods

Based on a Google search, the web site of the European Arthroplasty Register (EAR) (http://www.ear.efort.org/registers.aspx), and a Health Technology Assessment Report [4], we identified all available national joint arthroplasty registries in May 2012. Overall we could identify 23 registries, of which 13 had published reports that were available in English and were beyond the pilot phase. We downloaded the most current report from each of those registries in English to identify each registry’s definition of revision. If no such report was available on the web site of the register, we sent a written request.

A total of 13 reports could be obtained in English. Of these there were 11 reports covering knee and 11 covering hip arthroplasties (Table 1). Of the 13, a total of 11 provided definitions on two or more of our four scenarios, and so these 11 reports represented our study group (Table 2).

Table 1 List of the registry reports analyzed for this study
Full size table
Table 2 Definition of the endpoints in arthroplasty registers
Full size table

Two of us (TRL, FS) then separately applied four possible clinical scenarios to the definitions of revisions identified in the reports; discrepancies between our two observers were resolved by consensus. The four scenarios were: (A) wound revision without addition or removal of implant components such as hematoma evacuation; (B) exchange of the head and/or liner, e.g., for infection; (C) isolated secondary patella resurfacing; and (D) secondary patella resurfacing with routine (prophylactic) liner exchange. The rationale for scenario D is that in some countries with large arthroplasty registers, many patients do not undergo patella resurfacing during the index procedure, and it is common that during secondary patella resurfacing, a "prophylactic" liner exchange might be done, because the liner is the component that is most susceptible to wear. Of the 13 reports, nine, 11, eight, and nine provided definitions that answered questions A through D, respectively.

Results

Comparing Registries’ Definitions

All registries looked separately at the characteristics of (not revised) primary implantation, however only 11 of 13 registers reported on the characteristics of revisions. There were considerable differences across the reports. In 11 of 13 reports, the primary outcome was revision of the implant. In one registry the primary endpoint was “reintervention/revision” while another registry reported separately on “failure” and “reoperations”. In three registries, the definition of the endpoint was not provided, but in one registry a results list gave an indication for the definition of the endpoint (Table 2). If provided, there were considerable differences in the definition of the endpoint (Table 2).

Applying Clinical Scenarios

Wound revision without any addition or removal of implant components (scenario A) was considered a revision in three of nine reports that provided a clear definition on this question and two others did not provide enough information to allow this determination; exchange of head and/or liner, e.g., for infection (scenario B), was considered a revision in 11 of 11; isolated secondary patella resurfacing (scenario C) in six of eight; and secondary patella resurfacing with routine liner exchange (scenario D) was considered a revision in nine of nine reports (Table 3).

Table 3 Scenarios that are considered a revision, listed separately by register
Full size table

Discussion

The revision rates reported in national arthroplasty registries have a major impact on the care of patients and are important not only for surgeons, but also for implant manufacturers, healthcare decision-makers, healthcare insurance, and policymakers. Therefore, it is important that revision, which is the usual endpoint in arthroplasty registers, be universally defined so that the results of different registries are comparable. However, when we analyzed the definition of revision used in registries, it became apparent that revision is defined quite differently. Interestingly enough, there were even different definitions of revision used in registries within the same country (Swedish Hip versus Swedish Knee Registry; Norwegian Arthroplasty versus Norwegian Hip Fracture Registry, although the Norwegian Fracture Registry, because it is not an arthroplasty register, was not part of this analysis). This implies that some surgeries that are considered a revision in one registry are not considered a revision in another registry. For this reason, comparisons between different arthroplasty registries are difficult to conduct and pooling results from different registries [17] also needs to be performed with caution.

This study had a number of limitations. First, of the 23 registers that we could identify, we only obtained 13 reports in English for analysis. However, these 13 reports cover the largest and the longest established joint registries in the world such as England and Wales, most Scandinavian countries, Australia, and New Zealand to name a few. The registries not covered in our analysis are probably smaller or have not been established for a long time. Because it also appears likely that the publication of a report in English will increase its citation frequency, it could be argued that the reports not analyzed here are less influential. Moreover, it appears unlikely that the registries not analyzed here would have an identical definition of revision. Second, it is possible that we failed to find the definitions of the endpoints of interest in the registries’ reports or that the definitions are documented somewhere outside the report. However, we also searched the web site of registers in which we could not find the definitions without finding a clear definition that would be able to allow the classification of our predefined scenarios. Third, we acknowledge that some registries may have difficulties assessing every reoperation in every patient and that the definition of the endpoint is adapted to the national organization of the register. For example, in paper form-based registries, it appears comprehensible that only exchanges or additions of implant components are considered a revision, because the registration is based on the implant level and therefore the registry gets a form for every component that has been implanted. In these situations, registries might not be aware of other reoperations in which no implant component was added or exchanged. This situation has been acknowledged by the Swedish Knee Arthroplasty Registry, in which “it was noted early on that many surgeons did not report reoperations which they did not consider directly related to the prior knee arthroplasty” [22].

The definitions of what constitutes a revision procedure varied widely across the registries we surveyed. We are not aware of any other studies that have addressed this issue before. The sparse literature touching this topic deals with the collaboration of several Scandinavian registers [5] or a comparison of the Norwegian Knee Arthroplasty Register and a US Arthroplasty Register [18]. Given the different definitions of revisions, the question arises as to what definition is the “best” or “correct” one. To answer that question, it depends on whether a revision is viewed from the patient’s or the implant’s perspective: From the patient’s perspective, any additional surgery is a burden. Therefore, it is easy to understand that secondary resurfacing of the patella could be considered a revision. If this argument is followed further, however, other wound revisions such as lavage or hematoma evacuations should be considered revisions as well, even when the implants are not touched. However, these scenarios are considered a revision in only three of nine reports. Viewed from the implant’s perspective, the question arises if an implant should be counted as a failure if it remains in vivo while other components such as in secondary patella resurfacing are added. Counting such implants as a failure could dilute the effect of otherwise underperforming implants in comparative analysis, resulting in the delay of the identification of underperformers.

Because there is no definite answer as how to deal with this problem up to this time, registries have developed different strategies. For example, one registry has decided to publish additional tables in which exchange of liner (for infection) are not considered as revisions [22]. Another registry changed its reference for reporting: “In early reports we looked at survivorship in terms of revision and/or reoperation.[…] …, so the key message is that we have moved from revisions (and possible re-operations) to revisions alone. […] This allows us to exclude revision for infection from the analysis so we are able to comment on revision for aseptic loosening which may be more pertinent in terms of assessing implants rather than other environmental and surgical factors” [15]. In their 2013 report of the Swedish Knee Arthroplasty Register, the authors discuss their strict definition of revision. They mention that different types of soft tissue surgeries were never reported by some surgeons and therefore the register has decided to use a stricter definition “which surely had to do with the implant” [22]. They also mention that in up to 20% of all revisions for infection, the liner is exchanged, which results in these cases being counted as a revision. They address the additional problem of fixed liners (such as in all-polyethylene tibial components): if the surgeon chooses to not exchange the fixed liner, this will favor the implant in the registry report. If, on the other hand, the surgeon chose an exchange, this would “result in a reversed bias if the exchange of an inlay is not considered as being a revision” [22]. Therefore, it appears favorable that all reoperations could be captured by the registry, regardless of whether an implant component was touched. This would allow a definition of revision, which is detached from pragmatic reasons.

Applying some common clinical scenarios to the definitions of the worlds registers, we found large discrepancies in terms of whether each of those scenarios would be considered a revision among the registries we surveyed. The question arises if the different definitions used in the registries results in just a theoretical change of reported revision rate or if there is indeed a measurable change in the actual revision rate. In this respect, the authors of the most recent Swedish Knee Arthroplasty Report have performed separate analyses in which isolated exchanges of liners because of infection (scenario B of our study) were not counted as revisions [23]. They found that doing so does “affect the results and that the effect negatively affects the results of non-modular implants when compared to modular ones” [23]. The Italian registry does not consider isolated secondary patellar resurfacing to be a revision procedure (our scenario C); however, the authors of that registry have provided additional analyses in which they tallied those procedures as failures. In that reanalysis, the reported survival at 10 years decreased from 94.2% to 93.3% [20].

Because there are striking differences in the definition of revision in arthroplasty reports, we suggest that arthroplasty registries agree to standardize the definition of revisions. Once consent on this harmonization process has been achieved, we suggest that it be introduced step by step. To maintain comparability with previous reports of the same registry, it appears necessary that parts of the reports are prepared using the previously used definition of revision for the register, whereas other parts of the reports are prepared using the standardized definition of revision. Until or unless such consistency has been achieved, we suggest grouping surgical procedures into the following categories: (1) primary implantation; (2) exchange surgery (exchange or removal of implants); (3) secondary addition of components such as patella resurfacing or adding a second unicondylar component (regardless of routine liner exchange); and (4) reoperation without touching implant components (such as hematoma evacuation or lavage).