Biventricular pacing has recently been proposed for treating patients with drug refractory heart failure and intraventricular conduction delay. The purpose is to restore ventricular relaxation and contraction sequences as homogeneously as possible. The aim of this study was to determine if some factors could predict the long-term clinical effectiveness of that new treatment. This study included 26 patients, aged 66 +/- 7 years, with drug refractory heart failure and wide QRS. Patients were implanted with a biventricular pacemaker. The left ventricle was paced through a coronary sinus tributary. New York Heart Association functional class, exercise tolerance, and left ventricular (LV) ejection fraction were collected at baseline and after pacemaker implantation. Patients were divided into 2 groups: group I = responders; group II = nonresponders. QRS duration and axis at baseline and during biventricular pacing, interventricular conduction time, and LV and right ventricular lead positions were compared between the 2 groups. Group I patients (n = 19) had a mean reduction of 1.3 in functional class and an increase in peak oxygen consumption rate by a mean of 50%. The only parameter that differed between the 2 groups was the QRS duration during biventricular pacing, with a significantly shorter value in group I than in group II (154 +/- 17 vs 177 +/- 26 ms; p = 0.016). Thus, a positive response to biventricular pacing is correlated with the quality of electrical resynchronization. The optimal positions of the right and LV leads would be those that could induce the greatest shortening of QRS duration.