Prognostic variables in late follow-up of aortic valve replacement using the proportional hazard model: a study on patients using the Medtronic-Hall cardiac prosthesis.

An alternative approach to the study of the follow-up of patients with heart prostheses is the use of the reliability theory (hazard function) and proportional hazard model (Cox's model). In a population of 480 patients who underwent AVR in the period from June 1977 to January 1983, with a mean follow-up time of 2.8 years, 16 preoperative variables were considered. From this pool of variables, six entered the regression model in a time-independent mode. These were age at operation, sex, preoperative NYHA classification, presence of AI, presence of endocarditis and presence of atrial fibrillation on ECG, none of which entered the model in the time-related mode. Another multifactorial approach, using a stepwise regression analysis to examine primary predictive factors that independently correlate with survival, while simultaneously accounting for the other previous variables, showed that the variables with additive prognostic value were age at operation, presence of AI and presence of endocarditis. Based on this model, a forecast five-year survival rate ranging from 88 to 14 per cent was found at the end of the fifth year. For the most favourable and the worst combinations of these prognostic variables, a patient-specific forecast five-year survival rate was drawn up. Our results were compared, using univariate and multivariate methods, with the results found in the literature, and the implications of this comparison were discussed.