We studied primary total knee replacements (TKRs), reported to the Norwegian Arthroplasty Register, operated on between 1994 and 2000. A Cox multiple regression model was used to evaluate differences in survival among the prosthesis brands, their types of fixation, and whether or not the patella was resurfaced. In Norway in 1999, the incidence of knee prosthesis operations was 35 per 100,000 inhabitants. Cement was used as fixation in 87% of the knees, 10% were hybrid and 2% uncemented implants. Bicompartmental (not resurfaced patella) prostheses were used in 65% of the knees. With all revisions as endpoint, no statistically significant differences in the 5-year survival were found among the cemented tricompartmental prostheses brands: AGC 97% (n 279), Duracon 99% (n 101), Genesis I 95% (n 654), Kinemax 98% (n 213) and Tricon 96% (n 454). The bicompartmental LCS prostheses had a 5-year survival of 97% (n 476). The type of meniscal bearing in LCS knees had no effect on survival. Survival with revision for all causes as endpoint showed no differences among types of fixation, or bi- or tricompartmental prostheses. Pain alone was the commonest reason for revision of cemented bicompartmental prostheses. The risk of revision because of pain was 5.7 times higher (p < 0.001) in cemented bicompartmental prostheses than cemented tricompartmental ones, but the revisions mainly involved insertion of a patellar component. In tricompartmental prostheses the risk of revision because of infection was 2.5 times higher than in bicompartmental ones (p = 0.03). Young age (< 60) and the sequelae after a fracture increased the risk of revision. The 5-year survival of the 6 most used cemented tricompartmental knee prostheses brands varied between 95% and 99%, but the differences were not statistically significant. There were more revisions because of pain in bicompartmental than in tricompartmental knees. In tricompartmental knees, however, there were more revisions because of an infection. The relatively few patients with uncemented and hybrid implants showed no improvements in results compared to cemented knee prostheses.