Between 1986 and 1991, 106 patients (127 knees) underwent uncemented knee arthroplasty for osteoarthritis. There were 106 total knee arthroplasties and 21 medial unicompartmental knee arthroplasties. The arthroplasties were evaluated for aseptic loosening during the year 2000. For total arthroplasty we used 77 porous-coated anatomic prostheses and 29 press-fit condylar prostheses. The mean bone mineral content of the proximal tibia, measured the day before surgery using dual-photon absorptiometry was 5.48 g/cm for the porous-coated anatomic prostheses which were revised for aseptic loosening (n = 9). This was significantly higher (p = 0.02) than the mean of 4.33 g/cm for those which were not revised. Values for the two revised press-fit condylar knees (4.78 and 4.93 g/cm) were above the mean value (4.23 g/cm) for those which were not revised. We found no statistically significant (p = 0.38) difference between the bone mineral content of the 12 revised and nine unrevised unicompartmental arthroplasties. Low trabecular bone quality, measured as the pre-operative bone mineral content of the proximal tibia, was not a predictor for later revision surgery following uncemented total knee or unicompartmental knee arthroplasty.