Time-dependent variability of electrocardiogram (ECG) in patients with Brugada syndrome could affect the interpretation of provocative testing. The aim of this study was to characterize ECG changes during and after flecainide infusion. We studied 59 consecutive patients. The ECG was continuously analyzed during the first 30 minutes of provocative testing, and a single ECG was recorded 60 minutes later. We analyzed CYP2D6 and CYP3A5 variants affecting flecainide metabolism and performed blinded measurements at lead II. At baseline, ECG patterns were classified as follows: type II in 31 patients (53%), type III in 15 (25%), and normal ECG in 13 (22%). Because of induction of type I ECG, the percentage of responders progressively increased with longer recording time periods (6.8% in 10 minutes vs 11.9% in 20-30 minutes vs 18.6% in 90 minutes; P < .01). Four patients displayed a late response, which was evidenced 90 minutes after the initiation of provocative testing. QRS width differentially increased between responders and nonresponders (P < .01), with a maximum QRS width of 110 ms during the first 30 minutes being effective for identifying possible late responders (sensitivity 100%; specificity 85.6%; positive predictive value 88%; negative predictive value 100%). The incidence of CYP2D6 variants was lower in late responders than in early or delayed responders (0% vs 75% vs 100%; P = .04), while a homogeneous distribution of CYP3A5*3/*3 was observed in our population. Response to flecainide exhibits time-dependent variability of ECG patterns and intervals. Longer periods of ECG recording increase the recognition probability of type I ECG.