Outcome and reproducibility of data concerning the Oxford unicompartmental knee arthroplasty

BACKGROUND AND PURPOSE: The reproducibility of results and potential confounders in sample-based studies is important to consider in the assessment of studies. Comprehensive arthroplasty registers could serve as a reference dataset for comparative analyses. We analyzed an implant that is frequently used worldwide, the Oxford unicompartmental knee replacement, in order to identify potential confounders inherent in the datasets and to evaluate the outcome achieved with this implant. METHODS: We performed a structured literature review of the data published on the revision rate of the Oxford medial unicompartmental arthroplasty. Both clinical follow-up studies and worldwide registry data were included. Confidence intervals were calculated to determine the statistical significance of differences. RESULTS: A substantial proportion of the published data (52-68% depending on the method of calculation) is derived from studies involving participation of the institution that developed the implant. The results published by this group show a statistically significant deviation from the reference datasets from registers or independent studies. Data from the developing hospital show mean revision rates that are 4 times lower than those based on worldwide register data, and 3 times lower than the ones quoted in independent studies. On average, the data published in independent studies are reproducible in registry data. INTERPRETATION: A conventional meta-analysis of clinical studies is substantially affected by the influence of the developing hospital, and is therefore subject to bias. For assessment of the outcome of implants, registry data are superior and, in terms of reference data for the detection of potential bias factors in the literature, could make an essential contribution to meta-analyses.