Data from the Framingham Heart Study are used to derive equations for long-term predicted probabilities for death and a variety of cardiovascular endpoints. An accelerated failure method is employed, first the standard Weibull model and then a useful extension. The extension relaxes the assumption of proportional hazards implied by the standard Weibull model. Models differ markedly in form for the various endpoints, but in every case the varying scale model provided a significantly better fit. The resulting differences in predicted probability may be important in planning community health projects or clinical trials and in carrying out cost-benefit analyses.