Sixty-three patients took part in a follow-up study ten to seventeen years after valve replacement for aortic stenosis. Data obtained were used to predict cardiac deaths (n = 14) occurring in the following three-year period. The degree of residual left ventricular (LV) hypertrophy correlated inversely with indices for LV systolic and diastolic performance (radionuclide cardiography). A logistic regression (LR) model had a positive predictive value for cardiac deaths of 100% (10/10) with 8% (4/53) false negatives. Evaluating LV performance indices exclusively, LR analysis showed that only peak filling rate had independent predictive value. Subnormal and normal LV ejection fractions were associated with equal three-year cardiac death rates (29%). Depressed LV function was related to residual hypertrophy. Impaired diastolic performance was the prime predictor of cardiac deaths.