Clinical, radiographic, biomechanical, and metallurgical data from fifty-eight patients with a fractured stem of the femoral component of a total hip replacement (thirty-seven Charnley, sixteen Müller, three Trapezoidal-28, and two Bechol prostheses) were analyzed, as well as the cases of twenty-seven control patients with matching clinical and radiographic data who had a Charnley prosthesis but no fracture of the stem. Radiographic data in the form of defined measurements were classified into discrete variables for statistical analysis. These radiographic variables, along with clinical variables, were rated by linear discriminant analysis and a fracture risk index (the sum of the rating scores or regressive coefficient) was derived for each patient. Using this index, two zones (of index values) were defined: one for the patients at risk for fracture and the other for those not at risk for fracture according to the discriminative scale. All patients in both the fracture group and the non-fracture group were separated successfully into one of these two zones, except for one patient with a metallurgical defect of the prosthesis and a stem in marked valgus orientation. Based on the values of the risk index. the zone for those at risk for fracture and the zone for those not at risk were each divided into three regions - questionable, marginal, and positive - to give more weight to the contributions of clinical, biomechanical, and material factors in the prediction of fractures of prosthetic stems. The mechanisms causing stem fracture that seemed to involve multiple variables of different origins were identified and an attempt was made to rank these groups of variables as causes of stem failures. A risk index developed in this way could be useful in predicting the possibility of fracture or loosening of a prosthetic stem after total hip replacement.