Eighty-seven valgus osteotomies of the tibia were performed in seventy-three patients for osteoarthrosis of the medial compartment of the knee; the median follow-up was ten years (range, three to fourteen years). The data were subjected to univariate and multivariate statistical analysis and to survivorship analysis. For these calculations, the end-point of failure was defined as an arthroplasty of the knee, and additional calculations were performed with the end-point defined as the performance of an arthroplasty or moderate or severe pain in patients who had declined an arthroplasty. None of the many risk factors that were evaluated could be found to be associated with the duration of survival, except for relative weight and angular correction. The median loss of correction after the osteotomy was 1 degree. If, at one year after the operation, the valgus angulation was 8 degrees or more, or if the patient's weight was 1.32 times the ideal weight or less, the probability of survival five years thereafter was at least 90 per cent and the probability ten years thereafter was at least 65 per cent. However, when valgus angulation at one year was less than 8 degrees in a patient whose weight was more than 1.32 times the ideal weight, the rate of survival decreased to 38 per cent five years thereafter and to 19 per cent ten years thereafter. There is a considerable risk of failure of a proximal tibial osteotomy if the alignment is not overcorrected to at least 8 degrees of valgus angulation and if the patient is substantially overweight.