C resynchronization therapy (CRT) by atriobiventricular pacing has been proposed to treat severe heart failure in patients with intraventricular conduction delay resulting in ventricular mechanical discoordination. The clinical efficacy of CRT was recently demonstrated in a single-blind, randomized, crossover trial, namely the MUltisite STimulation In Cardiomyopathies (MUSTIC) study.1 The purpose of this substudy was to assess the effect of CRT on heart rate variability (HRV) and to analyze its correlation with clinical efficacy. • • • All patients provided written informed consent before enrollment. All had severe heart failure due to idiopathic or ischemic left ventricular systolic dysfunction, left ventricular ejection fraction 35%, and end-diastolic diameter 60 mm. All patients were in sinus rhythm with a QRS interval 150 ms and no classic indication for pacemaker implantation.2 Patients had been in New York Heart Association (NYHA) class III for 1 month before inclusion while receiving individually optimized treatment, including diuretics and angiotensin-converting enzyme inhibitors at the maximum tolerated dose. Exclusion criteria have been already reported.1 The study included a 6-month randomized crossover phase during which atriobiventricular (active) pacing was compared with inhibited (inactive) ventricular pacing at the basic rate of 40 beats/min, each for a period of 3 months in random order (Figure 1). Implantation was performed after a 1-month observation period to verify the stability of heart failure. Pacemakers were programmed as inactive after implantation. Patients were randomly assigned to experimental groups within the following 2 weeks after verification of pacemaker proper performance. Treatment order randomization followed a block design with stratification according to center. The single-blind, crossover phase was followed by a complementary 6-month period during which the pacing system was programmed according to patients’ preferences. All leads were implanted transvenously. The left ventricular lead was placed in a tributary of the coronary sinus according to a previously described method.3 Specially designed electrodes were used. The preferred target site was the lateral wall. The pacemakers were triple-output devices using standard dual-chamber technology, with built-in adapters to synchronize pacing in both ventricles. Pacemakers were programmed as either active or inactive at randomization. The baseline pacing rate was set at 40 beats/min and the upper rate limit at 85% of the highest predictable heart rate according to patients’ age and gender. Each patient underwent Doppler echocardiography to determine their optimal atrioventricular delay (electrical delay between atrial and ventricular excitation) during atriobiventricular pacing. No modification of medical treatment except diuretic dose adjustment was permitted between enrollment and the end of the crossover phase of the study. Twenty-four-hour ambulatory electrocardiograms with a 2-channel recorder (ELA Medical, le PlessisRobinson, France) were obtained in all subjects during normal activity at baseline, at randomization, at the end of each of the 2 crossover periods, and at the end of each longitudinal follow-up period (9 and 12 months). Tapes were replayed through a Elatec analyzer (ELA Medical). HRV analysis used specifically validated software.4 Ambulatory electrocardiographic recordings that lasted 16 hours and whose analyzable proportion was 90% were excluded to avoid bias linked to HRV circadian variations. After initial arrhythmia analysis and editing, the remaining normal-to-normal RR intervals in suitable recordings were measured and HRV time-domain analyses were performed according to published guidelines.5 For the purposes of this study, 4 different time-domain HRV indexes were measured: (1) the SD of the normal-to-normal (NN) interval over the recording duration (SDNN); (2) the SD of the 5-minute mean NN interval over the entire recording (SDANN [milliseconds]). Because SDANN values are obtained from successive 5-minute periods, they reflect heart rate changes caused by 5-minute cycles. They reflect long-term variations but provide no short-term variability data; (3) the square root of the mean squared successive differences between adjacent RR intervals over the entire recording (rMSSD [milliseconds]). This correlates more with the parasympathetic (vagal) activity; (4) proportion of successive interval differences 50 ms as computed over the entire recording (pNN50 [percent]). This also reflects parasympathetic activity. From the Departement de Cardiologie et Maladies Vasculaires, Centre Cardio-Pneumologique, Hopital Pontchaillou-CHU, Rennes, France. Dr. Daubert’s address is: Departement de Cardiologie et Maladies Vasculaires, Centre Cardio-Pneumologique, Hopital PontchaillouCHU, 35033 Rennes Cedex, France. E-mail: jean-claude. daubert@chu-rennes.fr. Manuscript received August 23, 2002; revised manuscript received and accepted January 23, 2003.