The purpose of the current study was to examine the long-term survival of ankle replacements, to examine between-country differences in ankle revision surgery, and to compare temporal trends in revision rates between countries. Across the four joint registries used, we observed between-country variation in survival rates in both the short-term (2–10 years) and in the long-term (> 10 years).
Few studies have examined and compared ankle replacement survival using data from national joint registries [9, 10]. We go beyond these studies to examine primary ankle replacement survival for a longer follow-up period (up to 20 years) and using the most complete and up to date data. This allowed us to examine temporal trends using robust country-level data. Our estimates of ankle replacement survival are similar to those previously reported by Bartel and Roukis [10]; 0.94 vs. 0.94 at 2 years, 0.87 vs. 0.86 at 5 years, and 0.81 vs. 0.77 at 10 years [10]. Using more current data, we were able to examine survival rates in the long-term at 15- and 19-years follow-up. Compared to the most recent data on primary ankle replacement survival in the long-term, our estimates of primary ankle replacement survival were comparable. For instance, a single study which examined long-term ankle replacement survival in Sweden, survival was reported as 0.63 (CI 0.58–0.67) at 15-years and 0.58 (CI 0.52–0.65) at 20 years [11].
We observed between-country differences in primary ankle replacement survival. The available data did not permit any adjustment for age and gender differences in the population receiving implants. Another possible explanation is that there are different thresholds (e.g. surgical requirements) for revision in different countries although there are, to our knowledge, no national guidelines governing indications for revision – future work in the field should include the development standardised indications for revision. Further, the failure mechanisms which lead to ankle revision are highly contested. There are several factors outside of the ankle replacement which are likely to influence rates of ankle revision and primary ankle replacement survival. For instance, improved patient selection (e.g. age at intervention) [23,24,25], type of implant used, frequency of primary replacement and, surgical caseload and skill [26, 27] affect revision and survival rates. This suggests that risk factors for ankle revision following primary ankle replacement may be, to some extent, modifiable: that is, if countries with high revision rates adopt the practices of countries with low revision rates, the full benefits of a primary ankle replacement may be gained without the consequence of high rates of revision/poor replacement survival. Specialist centres for the management of ankle replacements may facilitate improved survival rates.
Temporal changes in disease indications, such as a decline in severe destructive rheumatoid arthritis (RA) [28], as well as within country demographic shift for example operating on older people are also likely to influence implant survival rates.
We also considered differences in the indications for revision, which dependent on the registry, fell into one of the following six categories: 1) fracture / dislocation, 2) pain, 3) instability / reduced mobility, 4) prosthesis issues, 5) pathology and 6) ‘other’ (see Supplementary 1). We were unable to undertake a robust analysis of differences between registries in the proportions with these indications. We suggest that there is an international coordinated effort to harmonise the coding of these indications. A quick analysis suggested that ankle pain, prosthesis issues (e.g. loosening, defective polyethylene) and malalignment/fracture were among the most common reasons for revision which would need further exploration with higher quality data.
The use of different ankle prostheses will also give rise to different survival rates. For instance, in the current analysis implants from Australia and New Zealand demonstrated greater levels of survival compared to those from Norway and Sweden. These lower levels of survival in Scandinavia, to some extent, may be explained by the more long-term use of the early Scandinavian Total Ankle Replacement (STAR) design, with higher rates of prosthesis loosening reported for the first-generation LINK® STAR prothesis compared to the second-generation prosthesis [29].
There are other limitations to this study which require careful consideration. For instance, we assessed ankle replacement survival at a population-level as patient-level statistics were not available. Subsequently, we analysed the number of replacements rather than the number of patients; we were unable to report on whether one patient had multiple revisions. Whilst this method agrees with previous studies and has been shown to have little effect on the accuracy of survival estimates [11], we acknowledge that this may limit the generalizability of our findings. In addition, there is a degree of uncertainty regarding the reporting of the registry data. For instance, there may be under-reporting of either primary surgery and/or revision to these registers. It is unclear whether the counts of primaries and/or revisions reported here are matched within individuals (i.e. revisions may be reported in people who did not have their primary surgery entered and vice-versa). A high level of data completeness for the capture of primary ankle replacements has, however, been previously reported for the period investigated in the current study [30].
There are challenges to using ‘revision’ as an endpoint due to varying between-registry definitions. Whilst the included registries were similar in their definitions of the primary endpoint [10], slight variations were apparent between registries. For instance, the Norwegian registry counts all re-operations, including debridements, as revisions whilst in other registries ‘debridement’ is not specified as a revision. Such variations could give rise to differences in the estimates of annual incidence, particular if ‘revisions’ include re-operations; the challenges of registry terminology have been reported previously [22, 31]. More so, variations in the definitions of disease indications will also influence ankle replacement survival rates. Future work should aim to harmonise registry definitions of both replacement, revision and disease indications which could be achieved through consensus study of the international datasets.
We did not assess ankle replacement survival and rates of revision by model of implant used. One of the main surgical factors which has been associated with ankle replacement survival rates is the model/type of implant used. There is evidence, using data from joint registries, to suggest that more modern ankle replacements have better rates of survival at 5 (0.81 vs. 0.88) and 10-years (0.69 vs. 0.84) follow-up compared to older prosthesis designs [11]. The speed at which new implants are introduced and the time required for surgical training and education will influence the need for revision and subsequently, survival of primary ankle replacements. Specifically, any benefits of more modern implants may not be observed for several years after introduction, following a period of surgical learning and national adoption. There was evidence of this in the current study, specifically one of the oldest joint registries, Sweden, has only recently started to show a decline in the annual incidence of total ankle revisions. In the current study, we did not request data on the number of revisions by implant type due to the low counts of total replacements; we were concerned that small numbers and confounding by surgeon and year of operation would prevent meaningful analysis of replacement survival by implants. In addition, we were unable to examine generational differences in survival rates because we did not have data on the date of first acquisition of survival data. Subsequently, we were unable to examine secular trends between registries as we were unable to compare calendar years. Lastly, we did not undertake formal significant testing to compare the curves between the 4-time series, as the results from such an exercise are complex to interpret and would have added little to the interpretation of the results from simply comparing the shape of each country’s curve. Future work aims to compare secular trends across registries during which both time-dependant and implant-dependant factors are less likely to affect survival rates.
In addition to surgical factors, the between-country differences in rates of primary ankle replacement will vary by population size and demographic structure. There is variation in the age structure of these populations with New Zealand for example having the lowest proportion aged over 65 [32]. Thus, even without formal age adjustment the between-country differences do not appear to be explained by age. There are limitations to using KM estimates to examine joint replacement survival. For instance, it is assumed that the survival probabilities are the same for patients who entered the registry at study inception compared to patients who entered more recently [33]. This assumption may not hold true due to the continuing improvement and safety of ankle implants with the introduction of new replacements over time.