The use of bipolar endocardial electrogram characteristics to guide radiofrequency (RF) current catheter ablation of accessory pathways (APs) has been advocated by several investigators. However, the influences of a varying anatomy of the AP and the atrioventricular groove, of different ablative approaches, and of RF current pulses preceding the final pulse have not been adequately addressed. Local bipolar endocardial electrograms were retrospectively analyzed in a uniform cohort of 62 consecutive patients with a single manifest AP located on the left free wall; in all patients, the AP had been ablated by a uniform approach with a single catheter advanced retrogradely toward the mitral annulus. Electrogram parameters assessed were the presence or absence of a presumed AP potential, the atrial-to-ventricular (A/V) amplitude ratio, the A-V interval, and the onset of delta wave to local ventricular activation (delta-V) interval. The AP location was classified on fluoroscopy as anterior, lateral, or posterior. Catheter stability was verified by comparing pre- and post-RF amplitudes of local atrial potentials. The ablation site was ventricular in 52 patients (group A) and atrial in 10 (group B). In group A, 26 APs (50%) required a single RF current pulse for ablation. These APs showed no anatomic predilection and no statistically significant differences in electrogram parameters from 24 APs that were ablated only after a median of three pulses had failed, suggestive of a wider ventricular insertion of the latter APs. A lower A/V ratio and a higher incidence of transient AP block found in the remaining 2 group A patients, who had anteriorly located APs requiring > 10 failed pulses, suggested an adverse anatomy of the A-V groove in that region. A stepwise multivariate logistic regression analysis revealed that the simultaneous presence of (1) a presumed AP potential, (2) an A/V ratio > or = 0.10, (3) an A-V interval < or = 40 milliseconds, and (4) a delta-V interval < or = 0 milliseconds was associated with a specificity of 94% and a positive predictive accuracy of 87% for an RF pulse to be successfully applied to the ventricular insertion to the AP. Compared with APs of group A, APs of group B were distinguished by unsuccessful ventricular pulses associated with a delta-V interval > 10 milliseconds in the presence of an A/V ratio > 0.33 (specificity of 97% and positive predictive accuracy of 82%), which is suggestive of a more epicardial ventricular insertion of these APs. The effect of anatomic variations of the AP and the A-V groove is reflected in the bipolar endocardial electrogram and needs to be considered in the approach to AP ablation. The stepwise inclusion of the four electrogram criteria introduced in this study may improve the efficacy of RF catheter ablation of a manifest left free-wall AP at its ventricular insertion. Whenever mapping cannot improve on a delta-V interval > 10 milliseconds despite apparently close contact with the mitral annulus ("good" A/V ratio), attempts at ablation are likely to be successful at the atrial aspect of the mitral annulus.