The interaction of caseload and usage in determining outcomes of unicompartmental knee arthroplasty: A meta-analysis

Background: Outcomes following UKA are variable and influenced by surgical caseload (UKA/year) and usage (percentage of primary knee arthroplasty that are UKA), which relates to indications. This meta-analysis assesses the relative importance of these factors. Methods: MEDLINE (Ovid), Embase (Ovid) and the Web of Science (ISI) were searched for consecutive series of minimally invasive cemented Phase 3 Oxford medial UKA. The primary outcome measure was revision-rate/100 observed component years (%pa). Series were divided into groups according to caseload and usage. Results 46studies, including 12,520 knees, were identified. The annual revision-rate varied from 0%pa to 4.35%pa, mean 1.21%pa (95%CI 0.97-1.47). In series with mean follow-up of ten-years or more the revision-rate was 0.63%pa (95%CI 0.46-0.83), which equates to a ten-year survival of 94% (95%CI 92%-95%). Aseptic loosening, lateral arthritis, bearing dislocation, and unexplained pain were the predominant failure mechanisms with revision for patello-femoral problems and polyethylene wear exceedingly rare (<0.1%). Both increasing caseload (p=0.02) and usage (p<0.001) were associated with decreasing revision-rate. The lowest revision-rates were achieved with a caseload >24 UKA/year (0.88%pa, 95%CI 0.63-1.61) and usage >30% (0.69%pa, 95%CI 0.50-0.90). Usage was more important than caseload: with high-usage (≥20%) the revision-rate was low, whether the caseload was high (>12UKA/year) or low (≤12UKA/year), (0.94%pa (95%CI 0.69-1.23) and 0.85%pa (95%CI 0.65-1.08) respectively); whereas with low-usage (<20%) the revision-rate was high, whether the caseload was high or low (1.58%pa, 95%CI 0.57- 3.05 and 1.76%pa, 95%CI 1.21-2.41). Conclusion: To achieve optimum results with mobile-bearing UKA surgeons, whether high or low-caseload, should adhere to the recommended indications such that ≥20%, or ideally >30% of their knee replacements are UKA. If they do this then they can expect to achieve results similar to those of the long-term series, which all had high-usage (>20%) and an average ten-year survival of 94%