Of 285 total hip arthroplasties (260 patients) performed for primary osteoarthritis during a six-year period, 135 were resurfaced using a Tharies prosthesis (total hip articular replacement with internal eccentric shells) and 150 were treated with the Trapezoidal-28 total hip replacement. From each of these two groups 100 hips (ninety-one patients in the Tharies group and eighty-six in the Trapezoidal-28 group) that had been followed for two to seven years were evaluated at the time of follow-up in accordance with a predetermined protocol. The patients were younger in the Tharies than in the Trapezoidal-28 group (average ages, fifty-eight and sixty-six years), included more men (sixty compared with thirty-five), and were more active postoperatively. The average follow-up was forty-seven months for the total joint-replacement group and thirty-eight months for the surface replacement group. At follow-up the ratings for pain, walking, and function according to the University of California at Los Angeles 10-point scale and the clinical results were identical in the two groups. Heterotopic ossification (Brooker grade III or IV) developed after thirteen Trapezoidal-28 and twenty-two Tharies arthroplasties. Radiographs made at six and twelve months and at final follow-up showed that the incidence of radiolucencies about the acetabular component was higher in the resurfacing group: fifty-seven with complete radiolucent lines after an average follow-up of thirty-eight months compared with thirty-six with complete lines after an average follow-up of forty-seven months. There were three failures in the joint-replacement group: a hematogenous staphylococcal deep infection that required a Girdlestone procedure, a femoral stem fracture that required revision, and loosening of an acetabular component for which revision was performed. There was also one dislocation, successfully treated by closed reduction. Similarly, in the resurfacing group there were three failures: two loose acetabular components, revised successfully, and one loose femoral component that necessitated total joint arthroplasty. Multivariate stepwise regression analysis showed that the factors that affected the final extent and width of the acetabular radiolucencies adversely after resurfacing were: any radiolucent lines that were visible at six months, a high level of physical activity after arthroplasty, and a thin superior cement mantle.(ABSTRACT TRUNCATED AT 400 WORDS)