Studying atrial fibrillation: what can we learn from the AFTherapy study?

The AFTherapy study 1 represents an ambitious project with the intention to settle a number of questions about pacemaker programming to prevent the start of atrial fibrillation (AF). This controversial but also promising and insufficiently understood non-pharmacological therapy of AF raises a number of questions. The first question concerns the ideal pacing rate. An individualized pacing rate of 10 bpm faster than the average intrinsic heart rate (taken from pacemaker memory 2 ), a pacing rate 20% faster than the intrinsic rate, 3 and a rate 10–19% faster than the average intrinsic sinus rate (in atrial single-chamber pacing 4 ) have been reported to successfully prevent AF. In contrast, the PA 3 trials showed no difference between pacing at a (non-individualized) rate of 30 and 70 bpm. 5 The second question concerns the use of rate-adaptive pacing. Bellocci et al . 6 reported that in patients with chronotropic sinus node incompetence, pacing in the rate-adaptive mode was superior in suppressing AF than the DDD mode. However, this has never been reproduced. Finally, an AF-suppressive effect of pacing has been reported to require atrial stimulation during sinus rhythm for >80 or 90% of the time. 7 , 8 This is best achieved by overdrive pacing algorithms that pace the heart dynamically just above the intrinsic rate. Similarly, Murgatroyd et al . 9 found an antiarrhythmic effect of a pacing algorithm that prevented short–long cycles after atrial premature beats (APBs). These findings suggest that overdrive algorithms ‘on top’ of conventional pacing may optimize results of pacing to prevent AF. This is the background and rationale of the AFTherapy study and explains the complex design of the study using several randomization periods with different pacing rates (40, 70, 85 bpm), pacing modes (DDD vs. DDDR), and the use of dedicated overdrive and APB response algorithms ( Figure 1 in Camm et al . 1 ). As in most studies on rhythm control therapies in AF in the last decade, the final results of AFTherapy are disappointing: different pacing rates, pacing modes, and the activation or deactivation of different preventive pacing algorithms did not show any impact in suppressing AF recurrences. This is, unfortunately, in perfect harmony with the results of most other large-scale studies on pacing to prevent AF. 10–15 Some arguments are easily found: atrial pacing was at best achieved for only 76% (not >80–90%) of the time in all five conventional pacing modes, and the programmed short AV delays have caused unnecessary right ventricular pacing which in turn may trigger AF as suggested in MOST and DAVID. 16 , 17 However, several lessons can be learnt from the AFTherapy study: the most impressive result of this study may in fact be the necessity to exclude >50% of randomized patients from the analysis. This is considerably unusual, has been used as a major criticism but impressively documents the unforeseen problems with the use of device memory to assess the success of rhythm control therapy. Implanted devices can—in contrast to 24 h Holter ECG, external event recording, or transtelephonic monitoring—continuously monitor the atrial rhythm over a period of years and document any detected arrhythmia. This renders them incomparably more sensitive for the detection of asymptomatic AF 18 and makes them the gold standard for the evaluation of the success of any rhythm control therapy. It is predictable that future studies on rhythm control in AF will be challenged to use implantable devices to test therapy efficacy whenever possible. However, this requires specific attention. As Tables 2 and 3 of the article of the AFTherapy study 1 show, ∼10% of patients were excluded because no valid data were retrieved from devices. This reflects a proportion of error that may be representative for physicians using a new diagnostic tool. If pacemaker memory functions are not adequately activated, interrogated, and documented, they cannot be used for diagnostic purposes. More importantly, ∼50% of patients were excluded from the analysis because the AF-related diagnostics showed atrial sensing artefacts and these patients by study design were not allowed to enter the final analysis. To optimize sensitivity of AF detection, the study recommended to programme atrial blanking periods as short as possible. The investigators were as convinced as most other experts and pacemaker manufacturers at that time that far-field oversensing of ventricular signals in the atrium is not an issue in bipolar atrial leads. This has proved to be the biggest misconception in AFTherapy. Bipolar atrial leads are extremely effective in suppressing any kind of atrial oversensing—except for ventricular far-field signals. Other studies have documented that ventricular far-field oversensing is the major cause for inappropriate AF detection by pacemakers. In the BEATS study, 19 ∼50% of all AF detections by pacemakers were inappropriate, almost entirely due to ventricular far-field oversensing. This occurred despite the protocol recommendation to check for atrial oversensing at the time of hospital discharge, showing that the detection of ventricular far-field oversensing may not be trivial. Therefore, a dedicated ventricular far-field oversensing test 20 should be applied in patients who received a dual-chamber device to optimize not only the specificity of AF detection but also the function of dual-chamber pacing itself. Optimized programming of atrial sensing (sensitivity, blanking periods), ideal atrial leads with narrow bipolar ring spacing, 21 and ventricular far-field oversensing rejection algorithms 22 allow a specificity of AF detection close to 100%. Unfortunately, the questions which the AFTherapy study attempted to address still remain unclear. Frequently, the potential of pacing to prevent AF is considered to be low. However, all studies on pacing for AF prevention suffered from the surprisingly high number of patients without any AF recurrence or with a cumulative time in AF (‘AF burden’) below 3 min per day or 1%. In most studies, ∼40–50% of patients were free of AF with simple antibradycardia pacing, usually programmed in the ‘control arm’. 10–15 Therefore, any additional pacing feature in the ‘treatment arm’ has to jump over this hurdle to demonstrate its superiority and requires a considerably large amount of patients with appropriate stored data. In addition, a parameter such as AF burden is correctly regarded as one the most objective and representative endpoints in a study on rhythm control therapy in AF. 23 However, the AF burden is distributed non-normal, and AF recurrences occur in clusters with a tremendous variability over time, 24 , 25 making any statistical analysis complex. It is highly desirable that future studies consider the lessons learnt from AFTherapy. In steadily growing populations with AF respectively an implanted pacemaker or defibrillator system (and a significant overlap), it is one of the most important issues to clarify the ability of atrial pacing to maintain sinus rhythm and to define settings and device features that are likely to stabilize the patient’s rhythm. Conflict of interest : C.W.I. is a member of the advisory board and speaker's bureau and a participant in clinical studies sponsored by Boston Scientific, Medtronic Inc., Sorin Group, and St Jude Medical.