This study examines whether the current clinical practice of using a 5 mV minimum amplitude during normal sinus rhythm (NSR) ensures adequate detection during subsequent episodes of ventricular fibrillation (VF) at the time of the implantable cardioverter-defibrillator (ICD) threshold testing. Risk of nondetection occurs with ICDs when a substantial portion of the individual cardiac events on an electrogram goes undetected. Detection risk was assessed by 2 methods: percentage of missed cardiac events (incidence of signal dropout), and mean electrogram amplitude. During ICD implantation and testing in 27 patients utilizing 41 lead positions, 135 episodes of VF were induced and analyzed. On 64 occasions, the countershock was not successful in achieving cardioversion, and the continuing electrical activity was analyzed as a separate group of postshock waveforms. Thresholds of 1 and 2 mV were applied to each individual cardiac depolarization in a VF episode. Significant risk of nondetection was assumed when > or = 10% of individual events displayed dropout. Underdetection by signal dropout occurred in 11 of 135 preshock arrhythmia signals (8.1%) from 3 patients at a 2 mV threshold, and in 6 of 135 signals (4.4%) at a 1 mV threshold. A mean NSR amplitude > or = 5 mV was associated with significantly lower risk of nondetection during subsequent VF episodes at both 1 and 2 mV thresholds (largest p < 0.001). Similar results were observed in analysis of postshock arrhythmia signals. Further examination of signal dropout and linear regression criteria suggest that in order to eliminate the possibility of nondetection at a 1 mV threshold, minimum NSR amplitudes of 8.5 and 10.0 mV, respectively, are required.