AIMS: While medial unicompartmental knee arthroplasty (UKA) is indicated for patients with full-thickness cartilage loss, it is occasionally used to treat those with partial-thickness loss. The aim of this study was to investigate the five-year outcomes in a consecutive series of UKAs used in patients with partial thickness cartilage loss in the medial compartment of the knee. PATIENTS AND METHODS: Between 2002 and 2014, 94 consecutive UKAs were undertaken in 90 patients with partial thickness cartilage loss and followed up independently for a mean of six years (1 to 13). These patients had partial thickness cartilage loss either on both femur and tibia (13 knees), or on either the femur or the tibia, with full thickness loss on the other surface of the joint (18 and 63 knees respectively). Using propensity score analysis, these patients were matched 1:2 based on age, gender and pre-operative Oxford Knee Score (OKS) with knees with full thickness loss on both the femur and tibia. The functional outcomes, implant survival and incidence of re-operations were assessed at one, two and five years post-operatively. A subgroup of 36 knees in 36 patients with partial thickness cartilage loss, who had pre-operative MRI scans, was assessed to identify whether there were any factors identified on MRI that predicted the outcome. RESULTS: Knees with partial thickness cartilage loss had significantly worse functional outcomes at one, two and five years post-operatively compared with those with full thickness loss. A quarter of knees with partial thickness loss had a fair or poor result and a fifth failed to achieve a clinically significant improvement in OKS from a baseline of four points or more; double that seen in knees with full thickness loss. Whilst there was no difference in implant survival between the groups, the rate of re-operation in knees with partial thickness loss was three times higher. Most of the re-operations (three-quarters), were arthroscopies for persistent pain. Compared with those achieving good or excellent outcomes, patients with partial thickness cartilage loss who achieved fair or poor outcomes were younger and had worse pre-operative functional scores. However, there were no other differences in the baseline demographics. MRI findings of full thickness cartilage loss, subchondral oedema, synovitis or effusion did not provide additional prognostic information. CONCLUSION: Medial UKA should be reserved for patients with full thickness cartilage loss on both the femur and tibia. Whilst some patients with partial thickness loss achieve a good result we cannot currently identify which these will be and in this situation MRI is unhelpful and misleading. Cite this article: Bone Joint J 2017;99-B:475-82.