When national registers compare different implants, they tend to use raw, unmatched data. Based on raw data, the revision rate of UKA is about three times higher than that of TKA. Raw data relating the death rate after UKA and TKA is also contained in the registers and shows that the death rate five years after a total knee is about twice as high as after a partial knee (NJR, 2010). This difference in death rate is clearly not a simple manifestation of the different types of operation and demonstrates how invalid a comparison using raw data is. Furthermore, if a comparison based on raw data is invalid for death, then it will also be invalid for revision. The reason for the dramatic difference in the death rate is that UKA tend to be implanted in younger and more active patients than total knees. An identical explanation must at least in part explain the difference in the revision rate because young and active patients tend to have a high revision rate.
Liddle et al. (2015) compared adverse events in matched UKA and TKA. Data was obtained not just from the NJR for England and Wales, but also the Hospital Episodes Statistics (HES) database and the Office for National Statistics (ONS). Patients were matched, using a propensity score analysis, on 20 variables including preoperative score, patient demographics, co-morbidities and deprivation indices. 25,334 UKA were matched against 75,996 TKA. Almost all the UKA with data were matched to TKAs. All patients who were treated with UKA could alternatively have been treated with a TKA so the conclusions are generalisable.
Based on the matched cohort, it was found that there were many advantages of UKA compared to TKA. For example, the length of stay was 1.38 (95% CI 1.33-1.43) days shorter with a UKA. The re-admission rate within the first year 0.65 (95% CI 0.58 – 0.72), intra-operative complications 0.73 (95% CI 0.58 – 0.91), and transfusions 0.25 (95% CI 0.17 – 0.37) were all less. Complications also occurred significantly less frequently, for example the incidence of thromboembolism with the UKA was 0.49 (95% CI 0.39 – 0.62) that of TKA, infection was 0.5 (95% CI 0.38 – 0.66), stroke was 0.37 (95% CI 0.16 – 0.86) and myocardial infarct was 0.53 (95% CI 0.30 – 0.90).
The mortality following UKA was also significantly lower (Fig. 10.5). For example, during the first 30 days, the hazard ratio was 0.23 (95% CI 0.11 – 0. 50 p<0.001) and during the first 90 days it was 0.46 (95% CI 0.31 – 0.69 p<0.001). The difference in mortality was not just seen in the short term. The survival curves progressively separated for about four years and thereafter remained parallel until the study stopped at eight years suggesting that the effect of surgery on mortality lasted for four years (Fig. 10.5). At eight years, the mortality following UKA was 0.87 that of TKA (95% CI 0.80 – 0.94 p<0.001). Overall there was an appreciable difference in death rate. If 62 patients (95% CI 43 – 116) were treated with a UKA rather than a TKA, over the eight year period one life would be saved. Furthermore, if within the National Health Service, the proportion of knee replacements that were UKA increased from about 7% to 20%, then about 160 deaths per year would be saved.
Figure 10.5 Post-operative mortality: for matched UKA and TKA (a) for the first year and (b) for eight years.
In the matched comparison, it was found that the revision rate and reoperation rate were still higher after UKA than TKA. At eight years, the revision rate was 2.12 (95% CI 1.99 – 2.26) times higher and the overall reoperation rate was 1.38 (95% CI 1.31 – 1.44) times higher with UKA. However, to put the adverse outcomes in perspective, it was concluded that “if 100 patients receiving TKA received UKA instead, the result would be around one less death and three more reoperations in the first four years after surgery” (Liddle et al, 2014).
It has been suggested that this analysis is biased (Knee Roundup, 2015). The data is, however, freely available and, if the analysis were to be repeated, similar results would be obtained. Futhermore, there is another similar study done under the auspices of the NJR assessing 45-day mortality that found results virtually identical to ours (Hunt et al, 2014).
Liddle et al. in a separate matched study, compared the patient reported outcome scores (PROMS) of UKA and TKA (Liddle et al, 2015). 3,519 UKA were matched with 10,557 TKA. The main outcome measure was the Oxford Knee Score (OKS). Excellent matching was achieved with the preoperative knee scores (UKA being 21.8 (SD 7.6) and TKA being 21.7 (SD 7.7)). At six months following UKA, the OKS was significantly better (p<0.0001) with the UKA being 38 and the TKA 36. Although this difference in Oxford Score is relatively small, many more patients achieved an Excellent OKS (> 41) with UKA rather than TKA (Odds ratio 1.59 95% CI 1.47 – 1.73, p<0.001). EQ-5DTM (EuroQuol, Rotterdam, The Netherlands) was also collected and a significantly better overall score was achieved with UKA rather than TKA (p<0.001). Furthermore, the four subscales relating to mobility, pain, function and self care were significantly better but, in the subscale of anxiety, there was no significant difference. The level of patient satisfaction was also assessed and more patients achieved Excellent satisfaction with UKA compared to TKA (Odds ratio 1.3 times CI 1.3 (SD 1.2 – 1.4, p<0.0001)) .
As far as we are aware, there are no other matched comparisons of UKA and TKA based on PROMS from national registers. Unmatched comparisons based on data from the New Zealand Joint Registry showed a significantly better OKS with UKA compared to TKA (UKA 39, TKA 37 p<0.0001) (Rothwell et al, 2010). Similarly, an unmatched comparison by Baker et al. (2012) from the NJR showed significantly better OKS compared to TKA (35.5 (95% CI 34.5 – 36.4) to 34.0 (95% CI 33.0 – 34.2)) although, in this study, no significant difference was found in the improvement in OKS between UKA and TKA when adjusted for other factors.