Seventy-four primary total hip arthroplasties were performed in sixty-eight patients between August 1990 and September 1991. Clinical assessments were made with use of the Harris hip score and, specifically, the pain component of that score. The preoperative radiographs were digitally quantified for calculation of the so-called canal-to-calcar ratio and the so-called cortical index. The postoperative radiographs were evaluated for the percentage of the cross-sectional area of the femoral canal that was occupied by the prosthesis; subsidence of the prosthesis; and adaptive osseous changes, including hypertrophic cortical remodeling, osteolysis, formation of sclerotic radiolucent lines around the prosthesis, and formation of a pedestal at the tip of the prosthesis. The indication for the arthroplasty was osteoarthrosis in fifty hips (68 per cent), avascular necrosis in fourteen (19 per cent), congenital dysplasia in six (8 per cent), and another diagnosis in four (5 per cent). The average duration of follow-up was thirty-one months (range, eleven to forty-six months). The average Harris hip score (and standard deviation) was 75 +/- 16.8 points (range, 29 to 100 points), and the average score for the pain component was 37 +/- 7.5 points (range, 0 to 44 points). The average canal-to-calcar ratio of the hips was 0.44 (range, 0.32 to 0.74), and the average cortical index was 0.54 (range, 0.33 to 0.66). The average subsidence of the component was 0.6 centimeter (range, 0.0 to 2.3 centimeters). The average fill of the canal was 100 per cent proximally, 97 per cent at the middle of the stem, and 92 per cent distally as measured on the anteroposterior radiographs made immediately postoperatively and 100, 95, and 90 per cent, respectively, as measured on the lateral radiographs. A failure occurred in twenty-one hips (28 per cent) in twenty-one patients, with an average time to failure of 21 +/- 13 months (range, one to forty-four months). The Kaplan-Meier survival estimate (and standard error) for this population was 0.45 +/- 0.11 (confidence interval, 0.67 to 0.23) at forty-four months. The average subsidence of the components that failed was 0.7 centimeter (range, 0.1 to 2.3 centimeters). There was no significant relationship between failure of the component and the age or sex of the patient, the diagnosis, or the side of the operation. Postoperative severity of pain (p = 0.09) or subsidence (p = 0.08) alone did not reach significance for predicting outcome. The Harris hip score alone (p = 0.05), the Harris hip score in combination with subsidence of the femoral component (p = 0.01), and the pain component of the Harris hip score in combination with subsidence of the femoral component (p = 0.01) were all significant for predicting outcome. No other measured radiographic variable was predictive of failure. Despite optimization of the fit of the component within the femoral canal and the percentage of the cross-sectional area of the femoral canal occupied by the component, the clinical results indicated a high rate of failure. Thus, these criteria are not the only requisites for stabilization of these femoral components without cement. On the basis of these data, we have discontinued the use of these intraoperatively customized, non-porous, smooth femoral prosthesis.