Inappropriate therapy remains an important limitation of implantable cardioverter defibrillators (ICD). PARAD+ was developed to increase the specificity conferred by the original PARAD detection algorithm in the detection of atrial fibrillation (AF). To compare the performances of the two different algorithms, we retrospectively analyzed all spontaneous and sustained episodes of AF and ventricular tachycardia (VT) documented by state-of-the-art ICDs programmed with PARAD or PARAD+ at the physicians' discretion. The results were stratified according to tachycardia rates <150 versus > or =150 beats/min. The study included 329 men and 48 women (64 +/- 10 years of age). PARAD was programmed in 263, and PARAD+ in 84 devices. During a mean follow-up of 11 +/- 3 months, 1,019 VT and 315 AF episodes were documented among 338 devices. For tachycardias with ventricular rates <150 beats/min, the sensitivity of PARAD versus PARAD+ was 96% versus 99% (NS), specificity 80% versus 93% (P < 0.002), positive predictive value (PPV) 94% versus 91% (NS), and negative predictive value (NPV) 86% versus 99% (P < 0.0001). In contrast, in the fast VT zone, the specificity and PPV of PARAD (95% versus 84% and 100% versus 96%) were higher than those of PARAD+ (NS, P < 0.001). Among 23 AF episodes treated in 16 patients, 3 episodes triggered an inappropriate shock in 3 patients, all in the PARAD population. PARAD+ significantly increased the ICD algorithm diagnostic specificity and NPV for AF in the slow VT zone without compromising patient safety.