The Norwegian Arthroplasty Register recorded 24,408 primary total hip replacements from 1987 to 1993; 2907 of them (13%) were performed with uncemented femoral components. We have compared the results of eight different designs, each used in more than 100 patients. Survivorship of the components was estimated by the Kaplan-Meier method using revision for aseptic loosening of the femoral component as the end-point. At 4.5 years, the estimated probability of revision for aseptic loosening for all implants was 4.5%, for the Bio-Fit stem 18.6% (n = 210) and for the Femora stem 13.6% (n = 173). The PM-Prosthesis and the Harris/Galante stem prostheses needed revision in 5.6% and 3.6%, respectively. The clockwise threaded stem of the Femora implant needed revision in 20% of right hips, but in only 4% of left hips. The short-term results of the four best uncemented femoral components (Corail, LMT, Profile and Zweimuller) were similar to those for cemented stems, with revision for loosening in less than 1% at 4.5 years. The importance of the control of innovative designs and the registration of early results is discussed.