During testing of implantable defibrillators, ability to sense ventricular fibrillation is assessed by observing electrograms and the emitted ECG interpretation channel during induced ventricular fibrillation. We hypothesized that ventricular electrogram amplitude in sinus rhythm could be used to predict the ventricular electrogram amplitude in ventricular fibrillation and serve as a first approximation of the "safety margin" for sensing ventricular fibrillation. We compared the peak-to-peak epicardial ventricular electrogram during sinus rhythm and ventricular fibrillation in 12 patients undergoing defibrillator implantation. The ventricular electrogram was recorded with an integrated bipolar lead and filtered at 10-50 Hz. Ventricular fibrillation was induced by alternating current and the ventricular electrogram measured from cessation of alternating current to the first countershock. The mean ventricular electrogram amplitude in sinus rhythm was 15.3 +/- 5.4 mV (range 7.1-25.5) and in 37 episodes of ventricular fibrillation was 8.3 +/- 3.6 mV (range 2.1-16.3). There was a significant relationship between the mean ventricular electrogram amplitude in sinus rhythm and in ventricular fibrillation (R = 0.7, P less than 0.001). There was wide variation among individuals in the decrease in the mean ventricular electrogram amplitude during ventricular fibrillation, with the ratio of mean ventricular electrogram in sinus rhythm to mean ventricular electrogram in ventricular fibrillation ranging from 0.29 to 1.05 (mean 0.55 +/- 0.20). This suggests that up to a fourfold decrease may be expected in the mean ventricular electrogram amplitude during ventricular fibrillation.(ABSTRACT TRUNCATED AT 250 WORDS)