Several authors have reported good success rates using the Oxford Unicompartmental knee replacement system [14, 16]. It has been suggested that results are comparable to that of Total Knee Arthroplasty (TKA) .
In our small and retrospective study, 4 of the 48 Oxford unicompartmental knee replacements had been revised within the 4.5 year follow up period and our outcomes in the surviving knees were disappointing compared with other studies [3, 4, 7, 14, 16–19], with 7.5% of our patients achieving 'poor' results according to the 'Oxford Knee Scoring' system. Having said this although we were disappointed with our average Oxford core of 33, the average Oxford score following Total knee replacement has been quoted as 34.82 at two years in a recent large study .
Our results are similar to those reported by Van Isaker et al, who demonstrated functional results to be poor in 10% of their followed up knees , and Cottenie et al  in which 6% had poor and 4% fair functional ratings. Both of these studies used the 'Hospital for Special Surgery' score, not the Oxford functional rating system that we used.
In our study four UKAs required revision: two were revised for pain secondary to progressive lateral tibiofemoral compartment degenerative change, one was revised after avascular necrosis developed within the lateral femoral condyle, and one was revised because of persitent and unexplained medial pain, in all cases symptoms resolved with conversion to TKA.
We found little correlation between component mal-positioning and poor oxford scores. This is in keeping with very recent work by the Oxford group who concluded that because of the spherical femoral component, the Oxford UKR is tolerant to femoral mal-alignment of 10° and tibial mal-alignment of 5° .
We feel medial knee pain is problematic in this prosthesis. There are several possible aetiologies for medial discomfort including: impingement; medial overhang of the tibial component; cementing errors; aseptic loosening of femur or tibia; soft tissue irritation (MCL, Pes Anserinus); and neuroma formation. Unfortunately there are a group of patients that get unexplained medial pain which is not attributable to any of these factors. Of those with unexplained pain occasionally these will often settle after 1-2y, however it is our experience that an unacceptable number (22/40) persist beyond this time. Our study included only patients of > 3y post op and therefore those 'early settlers' are excluded automatically.
Patients reporting medial knee pain had poorer Oxford scores (Figure 4). 91.6% (22/24) of those with medial pain had scores of 37 or less, as far as we are aware this close correlation has not been previously reported. It is noteworthy that we found a relatively high incidence of medial knee pain despite the fact that phase III Unicompartmental replacements were used.
Although excessive medial overhang of the tibial component (more than 2 mm) was seen in 4/40 knees this did not seem to correlate with poor Oxford scores or medial knee discomfort. This is in keeping with the most recent results reported by Murray et al . They reported that medial overhang of < 3 mm and did not worsen Oxford scores when compared with an overhang of > 3 mm which did have a negative impact on the scores, they did not report an association with medial joint discomfort or pain. It should be noted that in Figure 8 the Radiograph is rotated so the overhang visible is likely to be mostly posteromedial, which could be less problematic than direct or anteromedial overhang. This may have some bearing on the lack of correlation between overhang and medial pain as some reported overhangs could have been the less significant 'posteromedial' type. This, however, still does not help in our understanding of why medial pain occurs in high numbers of patients (in our study) following Oxford unicompartmental knee replacement.
A large proportion of our patients experienced medial knee pain (more than half). We believe that this medial discomfort does correlate with poorer results, as none of those with scores > 37 complained of the symptom and all those with scores below that did. However it is not the single most important determinant of poor functional results as several patients (18/22 complaining of medial pain) had outcomes which were 'moderate' to 'good'. Is it possible that the presence of medial knee pain is irrelevant to the outcome of these knees? Certainly we do not believe this to be the case as we have found that medial joint discomfort was a common reason for patient dissatisfaction with the Oxford UKA, with one patient requiring revision to TKR (With successful outcome).
There are suggestions that patients with lesser degrees of osteoarthritis preoperatively do not achieve such good results with arthoplasty as those with greater wear. Within our small sample we did not find this to be the case, and furthermore, we did not note a correlation between severity of preoperative osteoarthritis and presence of post op persistent medial discomfort.
There are limitations to our study including being a retrospective review of a small cohort. Due to the fact that we excluded all patients with significant patellofemoral arthritis, we performed very few UKAs (48) when compared with TKAs (around 740) during the period studied and this may, of course, have a significant bearing on our results. It has been suggested that as the Oxford unicompartmental arthroplasty is a demanding procedure that the outcomes are better in units where the operation is being performed frequently [18, 23–25]. When the cause for revision of Knee replacement was studied from the New Zealand Joint registry data, it was noted that the early revision rate for the Oxford unicompartmental knee was 2.9 times greater than that for Total knee replacement. However, higher-use surgeons (i.e. those performing one/month or more) had a revision rate comparable to TKA. Those performing > 12 per year had a revision rate of 0.99%, those performing 8-11 per year had revision rates of 4%, those performing 2-7 per year 6.4% and those performing 1 per y had an 8% revision rate .
We used standard post operative Xrays to score alignment of prostheses, rather than 'screened' radiographs, and we accept this may affect the calculation of the radiographic scores.