We investigated whether increasing or decreasing beat-to-beat variability of repolarization (BVR) would change drug-induced proarrhythmic outcome accordingly. Increased variability of repolarization has been suggested as a prelude to proarrhythmic circumstances in experimental and clinical situations. The non-cardiovascular, I(Kr)-blocking drug sertindole was administered to anesthetized dogs with chronic atrioventricular block. Three interventions were used to prevent or suppress sertindole-induced torsades de pointes (TdP). Supratherapeutic doses of sertindole (1.0 mg/kg intravenously) induced TdP in 10 of 13 dogs whereas 0.2 mg/kg induced no TdP, despite increases in QT intervals by both doses. The BVR, quantified as short-term variability (STV) from Poincaré plots, was the only parameter that predicted TdP outcome (1.0 mg/kg sertindole: 2.3 +/- 0.7 ms to 5.1 +/- 2.1 ms, p < 0.05; 0.2 mg/kg sertindole: 2.3 +/- 0.8 ms to 3.2 +/- 1.1 ms, p= NS). 1) KCl, intravenous, reduced the incidence of sertindole-induced TdP from 6 of 7 to 1 of 7 dogs (p<0.05) and prevented sertindole-related increase of STV: 3.0 +/- 1.1 ms vs. 4.5 +/- 1.3 ms (p < 0.05); 2) levcromakalim (I(K,ATP) activator) reduced sertindole-induced TdP and decreased STV from 4.9 +/- 2.1 ms to 2.6 +/- 0.9 ms (p < 0.05); 3) steady-state ventricular pacing (60 beats/min) abolished sertindole-induced TdP and decreased STV from 4.9 +/- 1.5 to 3.2 +/- 1.0 (p < 0.05). Torsades de pointes reappeared upon return to non-paced idioventricular rhythm. None of the 3 interventions reduced the sertindole-induced prolonged QT interval. Proarrhythmic intervention is related to an increase in BVR, whereas antiarrhythmic treatment is associated with a decrease in BVR. The BVR is superior to QT interval prolongation in the prediction and prevention of drug-induced TdP in this experimental model.