We compared two cementless femoral components, the 'isoelastic' Butel stem and the more rigid PCA design, in a randomised, prospective study of 56 patients with a mean follow-up of 4 years (2 to 5). Patients were matched in 28 pairs, and one of each pair was treated with each femoral component. The isoelastic stem gave fewer signs of stress shielding radiologically, but both the Harris hip score and the visual analogue pain scale showed significantly inferior results at the two-year and three-year follow-up. The overall failure rates for the femoral components were 43% for the Butel and 11% for the PCA. These results contrast with those of earlier experimental and clinical studies, in which isoelastic properties appeared to be advantageous. This review emphasises the importance of controlled studies before an altered or new design of prosthesis is released on to the market for general use.