Unicompartmental knee arthroplasty (UKA) is known to have a higher risk of revision than tricompartmental arthroplasty (TKA), while UKA implants are generally less expensive than TKA implants. We estimated the costs of implants and hospital stay of both procedures and related the cost difference at primary operation to the difference in number of revisions to be expected. We compared 15,437 primary TKAs and 10,624 primary medial or lateral UKAs. The operations were all done on patients with arthrosis during 1985-1995. By matching patients in the Swedish Patient Administration System with the Swedish National Knee Arthroplasty Register, the groups could be compared regarding the length of the hospital stay. The cumulative revision rate (CRR) and the relative risk of revision were calculated with survival statistics, as well as the risk of a second revision and the risk of infection. The weighted mean cost of the commonest implants in each group was used as an estimate of the implant cost. We found that the TKA patients were, on average, 2 years older at operation and had a lower CRR than the UKA patients-i.e., 10-year CRR of 12% and 16%, respectively. After adjusting for age, gender and year of operation, UKA patients were found to have a 2-day shorter hospital stay and fewer serious complications than TKA patients. The mean estimated cost of a unicompartmental implant was 57% of that of a tricompartmental implant. We conclude, that by using UKA instead of TKA in appropriate patients, money can be saved, even after taking into account the increased number of revisions to be expected.