OUP user menu

Conventional and dedicated atrial overdrive pacing for the prevention of paroxysmal atrial fibrillation: the AFTherapy study

A.J. Camm, N. Sulke, N. Edvardsson, P. Ritter, B.A. Albers, J.H. Ruiter, T. Lewalter, P.A. Capucci, E. Hoffmann
DOI: http://dx.doi.org/10.1093/europace/eum253 1110-1118 First published online: 27 November 2007


Aims This investigation was conducted to determine the effectiveness of several conventional overdrive pacing modalities (single rate and rate responsive pacing at various lower rates) and of four dedicated preventive pacing algorithms in the suppression of paroxysmal atrial fibrillation (AF).

Method and results In this multi-centre, randomized trial, 372 patients with drug-refractory paroxysmal AF were enrolled. Patients received a dual-chamber pacing device capable of delivering conventional pacing therapy as well as dedicated AF prevention pacing therapies and to record detailed AF-related diagnostics. The primary endpoint was AF burden, whereas secondary endpoints were time to first AF episode and averaged sinus rhythm duration. During a conventional pacing phase, patients were randomized to single rate or rate-responsive pacing with lower rates of either 70 or 85 min−1 or to a control group with single rate pacing at 40 min−1. In the subsequent preventive pacing phase, patients underwent pacing at a lower rate of 70 min−1 with or without concomitant application of four preventive pacing algorithms. A substantial amount of data was excluded from the analysis because of atrial-sensing artefacts, identified in the device-captured diagnostics. In the conventional pacing phase, no significant differences were found between various lower rates and the control group receiving single rate pacing at 40 min−1 or between single rate and rate-responsive pacing. Patients receiving preventive pacing with all four therapies enabled had a similar AF burden compared with patients treated with conventional pacing at 70 min−1 (P = 0.47).

Conclusions The results do not demonstrate a significant effect of conventional atrial overdrive pacing or preventive pacing therapies. However, the observations provided important information for further consideration with respect to the design and conduct of future studies on the effect of atrial pacing therapies for the reduction of AF.

  • Atrial fibrillation
  • Overdrive pacing
  • Preventive pacing


With respect to prevention of atrial fibrillation (AF) by pacing, a number of questions are still not fully answered, such as the additional benefit of atrial overdrive pacing over single rate or rate-responsive dual-chamber pacing. Several studies suggest that patients with sick sinus syndrome are at a lower risk for AF when they receive atrial-based pacing compared with ventricular pacing alone.15 Conventional atrial pacing has been applied in patients with recurrent paroxysmal AF with favourable or neutral results with respect to the reduction of the total time in AF and/or the number of AF episodes.68 Atrial pacing at an increased lower rate, exceeding the patient's mean intrinsic rate (atrial overdrive pacing), has been used to suppress paroxysmal AF. Mixed results with regard to the reduction of paroxysmal AF were reported from this approach.912 Besides atrial overdrive pacing at a single rate, more advanced continuous overdrive algorithms that dynamically set the pacing rate slightly above the patient's intrinsic atrial rhythm have become available. Various results were reported from such algorithms, ranging from an effect on the number of premature atrial contractions only to a 25% relative reduction of the AF burden associated with symptomatic episodes.1315

This study was conducted to evaluate the effects of conventional overdrive pacing and a set of four dedicated atrial overdrive algorithms on the suppression of AF.



For inclusion, patients were required to have an atrial arrhythmia history of at least 12 months (not necessarily documented), experienced at least three AF episodes (symptomatic or asymptomatic) during the 3 months prior to enrolment and ECG documentation of at least one AF episode within the 12 months prior to enrolment. In addition, patients should be refractory to at least two antiarrhythmic drugs. All patients gave their written informed consent at the time of enrolment. Implantation of a pacemaker (Selection, model 900, Vitatron B.V., Arnhem, The Netherlands) was performed. Device diagnostics and preventive pacing therapies available in this device are described below. Except for the device settings related to AF episode detection and the conventional or preventive atrial pacing therapies under study devices were programmed to the investigator's discretion.

Trial design

The trial was conducted in compliance with the Declaration of Helsinki at 33 study centres located in 11 European countries and Canada under a protocol approved by the Ethics Committees at all participating study centres.

This was a single-blinded, randomized, prospective study. A flowchart of the study is provided in Figure 1. Before entering the study phases in which the effects of conventional and preventive atrial pacing were assessed, patients were monitored for 2 months. The results of this monitoring phase have been published separately.16 Following this monitoring phase, patients entered a 2 month conventional pacing phase, for which they were randomly assigned to either a control group (DDD pacing at a lower rate of 40 min−1) or one of the following four conventional pacing groups: DDD pacing at lower rates of 70 min−1 (DDD70), 85 min−1 (DDD85), rate-responsive pacing with lower rate settings of 70 min−1 (DDDR70), or 85 min−1 (DDDR85). The lower rate of 40 min−1 in the control group was assigned with the assumption that this pacing mode would result in a very low amount of atrial and ventricular pacing. During the subsequent 2 months, preventive pacing phase patients were randomized to DDD pacing with a lower rate setting of 70 min−1 with or without all four preventive pacing therapies activated concurrently.

Premature transitions from the conventional pacing phase to the preventive pacing phase were allowed when a patient experienced two severe symptomatic AF episodes. A premature phase transition followed by a 2 week stabilization period was required in case of a change in medication or cardioversion.

Device diagnostics

The devices used in this study were capable of storing data on the duration and frequency of AF, as well as antecedent events such as PACs. The time of onset of a maximum of 32 AF episodes and their duration was stored. In addition, the device stored a programmable number of AF onset reports, showing the rate profile prior to and following the AF onset, as well as the event classification by the device. It should be noted that these onset reports did not include stored electrograms but, instead, provided a graphical representation of sensed and paced events, associated intervals, and the classification of each event by the device. In the evaluation of the endpoints for this study, the device diagnostics were used to determine the AF burden and to evaluate atrial sensing.

The criterion applied in establishing the occurrence of an AF episode was an atrial rate >200 min−1 persisting for more than six ventricular beats. Episodes were judged to have terminated if the atrial rate was <200 min−1 for at least 10 ventricular beats.

Preventive pacing therapies

Four dedicated preventive pacing therapies including pace conditioning, PAC suppression, post-PAC response, and post-exercise response were intended to suppress AF by means of atrial overdrive pacing, activated separately or in any combination with each other.

The pace conditioning therapy is an atrial overdrive mode, intended to achieve a high percentage atrial pacing at a rate slightly exceeding the underlying intrinsic rate. The PAC suppression therapy increases the atrial pacing rate upon increased PAC activity. The aim of the post-PAC response therapy is to reduce the pause after a PAC and to provide a smooth transition from the PAC coupling interval into the underlying heart rate. The post-exercise response therapy provides a smooth transition between an elevated rate, for instance after stopping exercise, and the lower resting rate.

Atrial sensing

In nominal settings, the device was programmed to unipolar atrial sensing with a sensitivity of 1.0 mV, a post-ventricular atrial blanking of 150 ms, and a sensed/paced AV delay of 140/180 ms. The most sensitive setting for the atrial sensitivity available in the devices used in the study was 0.5 mV. The post-ventricular atrial blanking could be adjusted between 50 and 300 ms to blank possible far-field R-wave sensing. Bipolar atrial sensing was recommended. To reduce the chance of blanking of tachyarrhythmia events, the AV delay was programmed relatively short and the post-ventricular blanking was to be set to the shortest possible duration without the presence of far-field sensing.


The primary trial endpoint was AF burden, defined as the proportion of time the patient is in AF, expressed in hours per day. Secondary endpoints were the time to first AF recurrence, i.e. the duration of the first episode of sinus rhythm after the start of a follow up phase, and the averaged sinus rhythm duration (ASRD), i.e. the duration of sinus rhythm divided by the number of sinus rhythm episodes. These endpoints were determined on the basis of diagnostic data provided by the device as a result of continuous monitoring of the atrial rhythm.

False-positive detection of AF episodes may arise because of sensing of far-field R-waves. False-negative findings may result from atrial undersensing or partial blanking of an atrial arrhythmia because of which the sensed atrial rhythm appears to be physiologic (2:1 blanking). AF onset reports, captured by the device used in the present study, allow such spurious AF observations to be identified. Far-field R-wave sensing is characterized by short, constant VA intervals, resulting in an atrial rate exceeding the detection rate (200 min−1). These artefacts will result in an over-estimation of both AF burden and the number of episodes. False-negative onset detections resulting from 2:1 blanking or atrial undersensing can be identified from device captured onset reports on the basis of a short apparent absence of the arrhythmia and a false-positive detection of a new onset upon the cessation of undersensing or 2:1 blanking. Data required for assessing the primary and secondary endpoints were retrieved at the conclusion of each 2 months study phase and reviewed in order to discriminate between genuine and spurious AF episodes.

Sample size

On the basis of a mean AF burden of 3.3 h/day (SD 4.5 h/day), observed during the monitoring phase, a minimum sample size of 117 patients per group was required in the conventional and preventive pacing phase to show a significant effect of the atrial overdrive pacing algorithms on AF burden (95% confidence and 80% power). For this calculation, it was assumed that the atrial overdrive pacing algorithms would result in an actual 50% reduction of AF burden.


For all phases of the study, AF onset reports were inspected with regard to spurious AF episode detection and possible false-negative AF detection. During this study, the devices were programmed to capture a maximum of 12 AF onset reports, broken down into the first three and last nine onsets during the follow-up period. Data were excluded from the analysis of a study phase if spurious AF observations were identified in more than one of the onset reports.

Data from the conventional pacing phase were analysed to evaluate the effects of the lower rate setting and rate-responsive pacing separately. For the analysis of the effects of the lower rate, data from randomization arms with the same lower rate setting were pooled irrespective the activation of rate-responsive pacing (data from the DDD70 and DDR70 group were pooled, and also data from the DDD85 and DDDR85 group were pooled). The pooled data sets were both compared with the data from the DDD40 group. To analyse the effect of rate-responsive pacing pooled data from the DDD70 and the DDD85 groups were compared with the pooled data from the DDDR70 and DDDR85 groups. Data from the DDD40 group were not included in this analysis in order to obtain balanced groups with respect to the programmed lower rates.

Data from the preventive pacing phase were analysed as two independent samples (parallel design). Data were analysed using SAS statistical software (version 8.2, SAS Institute Inc., Cary, NC, USA). Since the endpoints were expected to have a skewed distribution, non-parametric statistics was used to evaluate primary and secondary endpoints. The criterion of significance was an α-level of 0.05 by two-sided comparison.


Patient accountability

A total of 372 patients were enrolled for this study with a mean age of 65 ± 11 years; of which, 206 (55%) were male. The mean left atrial diameter at enrolment was 41 ± 7 mm and the mean LV fractional shortening was 0.36 ± 0.10 mm. Baseline demographic characteristics of the entire enrolled population are summarized in Table 1. Atrial leads were implanted in the right atrial appendage (73%), the atrial septum (12%), the free wall (11%), and other locations (3%). In all but seven patients, a bipolar atrial lead was implanted. A total of 130 patients (35%) had a standard indication for pacing. Of 372 patients enrolled, 16 were not followed beyond the monitoring phase and 26 left the study during the conventional pacing phase. Most commonly, patients requested to leave the study prematurely for non-medical reasons or reasons not related to their AF status. In addition, 21 patients were removed from the study because of the development of persistent or permanent AF and/or of a scheduled ablation. Other common reasons for premature study termination were the requirement of device settings not allowed by the study protocol and lead dislocation. Six fatalities occurred in the study population. Sudden death because of unknown reasons occurred in two patients; four other patient deaths were attributable in one case each to end-stage renal disease, acute pulmonary edema, hepatic cirrhosis, and an automobile accident. None of the deaths were judged to be related to the pacing system.

View this table:
Table 1

Baseline demographic data

CharacteristicNumber of patients (%)
Documented history of AF
 <6 months28 (7.5)
 6–12 months43 (11.6)
 >12 months300 (80.6)
Number of AF episodes in preceding 3 months
 <10145 (39.0)
 10–100185 (49.7)
 >10038 (10.2)
Concomitant arrhythmias
 None179 (48.1)
 Supraventricular86 (23.1)
 Ventricular5 (1.3)
 Sinus node disease118 (31.7)
 AV Block18 (4.8)
 Other12 (3.2)
Concomitant cardiovascular diseases
 None257 (69.1)
 Congestive heart failure6 (1.6)
 Prior myocardial infarction17 (4.6)
 Rheumatic heart disease7 (1.9)
 Ischaemic heart disease50 (13.4)
 Cardiomyopathy13 (3.5)
 Valvular disease41 (11.0)
 Other16 (4.3)
Other concomitant diseases
 None155 (41.7)
 Diabetes23 (6.2)
 Stroke21 (5.6)
 Pulmonary emboli5 (1.3)
 Chronic obstructive pulmonary disease7 (1.9)
 Pulmonary hypertension8 (2.2)
 Systemic hypertension120 (32.3)
 Chronic renal failure4 (1.1)
 Other123 (33.1)
 No medication40 (10.8)
 Digitalis80 (21.5)
 Beta blocker73 (19.6)
Anti-arrhythmic agents
 IA18 (4.8)
 IC84 (22.6)
 III212 (57.0)
 IV50 (10.8)
  • Notes: Categories are not mutually exclusive. Other arrhythmias include WPW syndrome, asymptomatic sinus arrest, intercurrent asystole, accelerated junctional rhythm and RBB block. Other cardiovascular diseases include pericardial disease and congenital heart disease. Other diseases include (amongst others) peripheral vascular disease, hypercholesterolemia, and systemic emboli.

Randomization and inclusion in the analyses for the conventional and preventive pacing phases is documented in Tables 2 and 3, respectively. AF onset reports, recorded by the implanted devices, were used to identify the occurrence of spurious AF observations. Spurious AF episodes attributable to far-field R-waves, atrial undersensing or blanking of arrhythmia senses were detected among 147 patients in the conventional pacing phase and 149 in the preventive pacing phase. Eventually, the total number of patients with valid data was 154 for the conventional pacing phase and 153 for the preventive pacing phase.

View this table:
Table 2

Conventional pacing phase: randomization and inclusion in analysis

DDD 40 min−1DDD 70 min−1DDDR 40 min−1DDD 85 min−1DDDR 85 min−1Total
No valid data1071051345
Inappropriate atrial sensing2628303429147
Included in analysis3338293024154
  • Notes: 16 out of the 372 enrolled patients were not followed beyond the monitoring phase. Of the 356 patients who entered the conventional pacing phase, randomization for this phase was not documented for 10 patients. The group of patients with no valid data includes patients with no data stored and patients with data from a monitoring period <2 days.

View this table:
Table 3

Preventive pacing phase: randomization and inclusion in analysis

Preventive pacingConventional pacingTotal
No valid data16723
Inappropriate atrial sensing7277149
Included in analysis8271153
  • Notes: Of the 356 patients who entered the conventional pacing phase, 330 entered the preventive pacing phase. Of these 330 patients, randomization was not documented for 5 patients. The group of patients with no valid data includes patients with no data stored and patients with data from a monitoring period <2 days.

In 46% of patients, pacemaker implantation was performed at the time of enrolment. In 25%, implantation preceded the enrolment by a median of 2 days (range, 1–258 days); whereas, in 30%, implantation followed the enrolment by a median of 8 days (range, 1–297 days).

Conventional pacing phase

The number of patients per randomization arm included in the analysis of the conventional pacing phase ranged from 24 in the DDDR85 group to 38 in the DDD70 group (Table 2). As expected, the amount of atrial pacing increased with increasing lower rate setting; the mean percentage of atrial pacing was 21, 73, and 76% for lower rates of 40, 70, and 85 min−1, respectively. Activation of rate-responsive pacing in comparable groups did not yield substantially different amounts of atrial pacing. Similar observations were made with regard to the mean percentage of ventricular pacing (71, 82, and 90% for lower rates of 40, 70, and 85 min−1, respectively). Throughout the randomization groups in this study phase, the mean AV delay was similar (range, 212–217 ms).

The results from the conventional pacing phase with regard to the primary endpoint are summarized in Figure 2. In all groups, a considerable number of patients had very low AF burden values. The percentage of patients with a burden <3 min/day were 58, 50, 31, 47, and 42% in the DDD40, DDD70, DDDR70, DDD85 and DDDR85 group, respectively. A higher median AF burden was observed for higher settings of the lower rate as well as for the rate-responsive pacing compared with single rate pacing. AF burden in patients with lower rates of 70 and 85 min−1 did not significantly differ from the AF burden in the control group (DDD40) (P = 0.25 and 0.12 for lower rates of 70 and 85 min−1, respectively). No significant difference was obtained between single rate pacing and rate-responsive pacing (P = 0.21).

Figure 2

Results from the conventional pacing phase. The bars represent the mean AF burden, and the error bars indicate the standard deviation. Median AF burden values are indicated by the horizontal lines. (A) AF burden for various lower rates. For this graph, the data from patients with the same lower rate settings were pooled. The median AF burden was 0, 0.17, and 0.24 h/day for lower rates of 40, 70 and 85 min−1, respectively. (B) AF burden for patients receiving DDD and DDDR pacing. For this graph, data were pooled from patients receiving single rate pacing, excluding data from the DDD40 group, and from all patients receiving rate-responsive pacing. Median AF burden was 0.11 h/day for the DDD group and 0.45 h/day for the DDDR group.

Secondary endpoints are listed in Tables 4 and 5. A significant difference in time to AF was found between lower rates of 40 and 85 min−1 (P = 0.03) with median times to AF of 1343 and 168 h for lower rate settings of 40 and 85 min−1, respectively. All other secondary endpoint comparisons yielded non-significant results.

View this table:
Table 4

Secondary endpoints for various lower rates programmed during the conventional pacing phase

40 min−1 (DDD)70 min−1 (DDD + DDDR)85 min−1 (DDD + DDDR)
ASRD (h)
Time to AF (h)
Mean % atrial pacing217376
Mean % ventricular pacing718290
  • Note: P-values refer to comparison with the DDD40 group.

View this table:
Table 5

Secondary endpoints for single rate pacing and rate-responsive pacing during the conventional pacing phase

DDD (70 min−1 + 85 min−1)DDDR (70 min−1 + 85 min−1)
ASRD (h)
Time to AF (h)
Mean % atrial pacing7476
Mean % ventricular pacing8686

Preventive pacing phase

Of the 153 evaluable patients in the preventive pacing phase, 71 were randomized to the control group and 82 to the therapy group with all four preventive pacing therapies activated. A higher amount of atrial and ventricular pacing was observed in the therapy group compared with the control group. The mean percentage atrial pacing was 89 and 66% in the preventive pacing group and the conventional pacing group, respectively. The mean percentage ventricular pacing was 88% in the therapy group vs. 79% in the control group. The mean AV delays were similar in both groups (182 ms in the preventive pacing group and 181 ms in the conventional pacing group), but shorter than in the conventional pacing phase, mainly in response to observed under-detection of atrial arrhythmias as a result of longer AV delays.

Figure 3 summarizes the results with regard to the primary endpoint for the preventive pacing phase. The proportion of patients with an AF burden of <3 min/day was 46 and 54% for the control and the therapy groups, respectively. Median AF burden was 0.18 h/day in the control group vs. 0 h/day in the therapy group. No significant difference in AF burden was found between the control and the therapy groups (P = 0.47).

Figure 3

Results from the preventive pacing phase, showing mean AF burden (bars), standard deviation (error bars), and median AF burden (horizontal lines). Median AF burden was 0 h/day in the preventive pacing arm and 0.18 h/day in the conventional pacing arm.

Secondary endpoints for the preventive pacing phase are presented in Table 6. Time to AF and ASRD did not significantly differ among the control and the therapy groups (P = 0.84 and 0.81, respectively).

View this table:
Table 6

Secondary endpoints for the preventive pacing phase

Preventive PacingConventional pacing
ASRD (h)
Time to AF (h)
Mean % atrial pacing8966
Mean % ventricular pacing8879


This study was one of the first large-scale trials in which preventive atrial pacing therapies were applied to suppress AF.

None of the study phases yielded significant effects of atrial pacing with regard to the primary endpoint of AF burden. It should be noted that the analysis of the primary endpoint was underpowered because of the low number of patients, eventually included in the analysis. The interpretation of the study data as a randomized trial was hampered by the large amount of exclusions because of atrial-sensing artefacts. Consequently, the results do not allow conclusions to be drawn as to the effective suppression of AF by the pacing modalities used in this study.

Effect of atrial-sensing artefacts

This investigation was made feasible by the sophisticated continuous diagnostic monitoring capabilities of the implanted device, which are essential for an improved understanding of the electrophysiology of AF as well as sensitive and specific detection of AF episodes. A major strength of this study was the ability to identify genuine AF episodes accurately and to exclude findings that may have been generated by various sensing artefacts. During this study, atrial-sensing artefacts showed to be a dynamic phenomenon. Some patients, excluded from the analysis of the conventional pacing phase, could be included in the preventive pacing phase and vice versa. These observations underline the need for an accurate and continued monitoring of the atrial-sensing performance as well as adequate measures to reduce the incidence of sensing artefacts when using device-derived endpoints. In addition, they emphasize the fact that studies without clearly described methods regarding monitoring and handling of atrial-sensing artefacts should be interpreted with caution.

In the conventional pacing phase and the preventive pacing phase, ∼50% of the patients had to be excluded from the analysis as a result of atrial-sensing artefacts. In programming devices during this study, often a subtle balance had to be established between a high atrial sensitivity, required to detect low-amplitude atrial tachyarrhythmia events and an appropriate atrial blanking, sufficiently long to prevent far-field R-wave over sensing and short enough to avoid blanking of atrial tachyarrhythmia events. Similar to the observations in this study, the far-field R-wave sensing has been observed in other investigations17,18 in ∼20% of the patients. It should be noted that in these studies the occurrence of the far-field R-wave sensing was assessed by in-office tests during follow-up, whereas the observations from our study reflect ambulatory conditions monitored over a relatively long period. This suggests that far-field R-wave sensing may not be fully resolved by an assessment during the follow-up, but requires dynamic ambulatory adaptation supported by the implanted device. For instance, Israel et al.19 observed atrial-sensing artefacts in ∼40% of their patients, although the atrial sensitivity and post-ventricular atrial blanking were optimized after implantation to avoid oversensing. Several measures may be considered to reduce the incidence of atrial-sensing artefacts. Obviously, the atrial sensitivity and blanking period should be appropriately programmed to avoid atrial over- or undersensing. In a recent study, Kolb et al.20 found that, compared with the nominal setting, the post-ventricular atrial blanking had to be shortened in about one-third and extended in two-third of the patients in order to avoid inappropriate mode switching. It should be noted that the nominal setting for the atrial blanking in the devices used in their study was 100 ms, which is substantially shorter than in the pacemakers used in our study. They also noted that adjustment of the atrial sensitivity is usually not adequate to differentiate between far-field sensing and true atrial events, because of the overlap in amplitudes between far-field senses and atrial arrhythmia events. Furthermore, recently reported studies show that bipolar leads with tip-to-ring distances of 9 and 10 mm may substantially reduce atrial oversensing.21,22 Atrial lead placement should be optimized to minimize sensing artefacts, and be verified during the implantation procedure. Finally, new intelligent sensing methods applying digital signal processing techniques may discriminate between far-field R-waves and true atrial signals.23

Effect of the lower rate setting in conventional pacing

The results from the conventional pacing phase of this study do not indicate that the increased lower rate settings with the application of conventional pacing reduce AF burden compared with DDD pacing at a lower rate of 40 min−1. Although Garrigue et al.10 reported a significant reduction in AF burden, total duration and maximum episode duration from 22 patients with DDD overdrive pacing at 10 min−1 above the mean ventricular rate, more recent trials using dual-chamber overdrive pacing found less beneficial results.11,12 Similar to our results, AF burden increased with increasing lower rate settings. In contrast to these results, Wiberg et al.24 observed a significant reduction in symptomatic AF episodes during medium and high overdrive AAI pacing compared with no pacing.

It was assumed that DDD pacing at a lower rate of 40 min−1 would act as a passive mode resulting in (almost) no atrial pacing and therefore not providing any therapy. Despite the lower rate of 40 min−1, a mean atrial pacing percentage of 20.7 was observed in this group. In the conventional pacing phase, the mean amount of atrial pacing did not substantially differ between patients with and without a bradycardia-related pacing indication. This conventional atrial pacing therapy may have contributed to the prevention of bradycardia-induced AF episodes. In addition, conventional pacing at a lower rate of 70 min−1 may have contributed to a reduction in AF in the preventive pacing phase, irrespective the application of other concurrent therapies.

Effect of rate-responsive pacing

In this study, a higher mean and median AF burden was observed for the rate-responsive pacing, compared with single rate pacing, but the observed differences were not statistically significant.

After the introduction of rate-responsive pacing, several studies were conducted to investigate the arrhythmogenic nature of this pacing modality. However, little experience has been reported as to the ability of rate-responsive pacing to suppress atrial tachyarrhythmias. Haywood et al.25 did not find significant differences between AAI and AAIR pacing with respect to the frequency of AF episodes. In contrast to these results, Belocci et al.7 reported a significantly lower frequency of mode switches with DDDR pacing than with DDD pacing, suggesting a lower tachyarrhythmia frequency with this pacing mode. This reduction was found only in patients with sinus node disease and chronotropic incompetence. The number of mode switches in this report does not clearly reveal the total AF burden. Bradycardia has been supposed to be one of the factors triggering AF episodes, as a result of an increased dispersion in atrial repolarization and conduction. Rate-responsive pacing could reduce bradycardia, but may also add an arrhythmogenic factor by means of competition with the intrinsic atrial rhythm, especially when undersensing of AT occurs. The eventual result of rate-responsive pacing on AF burden and the multitude of circumstantial aspects influencing this result still remain to be unravelled.

Effect of preventive pacing

No significant differences in AF burden were observed between the therapy group and the control group in the preventive pacing phase. Patients receiving preventive pacing showed a 37% lower mean AF burden than patients in the control arm, although this difference did not reach statistical significance.

Several other authors have reported results from preventive pacing therapies with respect to the suppression of AF.

The onset of AF is often associated with the occurrence of premature atrial contractions.2629 Algorithms that accelerate the pacing rate following the detection of a premature atrial contraction were reported to significantly reduce the total time in AF or the number of AF episodes.14 Lee et al.30 studied the effects of three preventive pacing therapies on AF burden. From a population of 324 patients they reported a higher, but not significantly different AF burden in the therapy group compared to the control group. It should be noted that in this study, the preventive pacing therapies were combined with a therapy providing antitachycardia pacing. Carlson et al.15 evaluated the effectiveness of dynamic atrial overdrive pacing, an algorithm that increases the pacing rate upon detection of intrinsic atrial rhythm. In a study with 319 patients, they found a reduction in the percentage of days with symptoms and AF in both the control and the therapy groups. The actual arrhythmia burden as indicated by the device was not reduced by overdrive pacing. Neither study scrutinized the presence of sensing artefacts.

The effect of ventricular pacing

Several authors have suggested that ventricular pacing may be associated with an increased risk of AF31,32 and that the effect of preventive pacing therapies to suppress AF may be counteracted by ventricular pacing.33 Moreover, the comparison of the results from overdrive pacing therapies in dual chamber and atrial mode24 indicates that possible beneficial effects of atrial pacing may be adversely influenced by ventricular pacing.

In this study, relatively high percentages of ventricular pacing have been observed, ranging from a mean percentage of 71% in patients in the DDD40 group in the conventional pacing phase to >90% in the DDDR85 group. The high amount of ventricular pacing is a result of the policy followed in this study to programme a relatively short AV delay in an attempt to reduce the total atrial blanking period and thereby the risk for blanking of atrial arrhythmia events. In the conventional pacing phase, many devices were programmed to an AV delay of 220 ms. In response to the observation of blanked arrhythmia events in this phase, the AV delay was shortened to 180 ms in many patients during the preventive pacing phase, resulting in a considerable amount of ventricular pacing. A sub-analysis of the preventive pacing phase did not show a significant effect from ventricular pacing on AF burden, but it is likely that this analysis was hampered by the unbalanced distribution of the percentage of ventricular pacing. An adverse effect of ventricular pacing on AF, reducing a possible effect of the preventive pacing therapies cannot be denied, given the earlier reports on these effects. Similar findings have subsequently confirmed these effects.

Study limitations

Using the endpoint of AF burden detected by an implanted device is the most accurate method to assess treatment success. Studies in which no device is implanted often report on the proportion of patients without arrhythmia recurrence, usually assessed by means of less stringent monitoring than possible with an implanted device. Furthermore, the episode detection criteria used in this study (atrial rate >200 min−1 for six or more ventricular beats) included short arrhythmias with limited clinical relevance.

The chief limitation of this study was the high rate of attrition, as a consequence of which conclusions rested upon the results obtained in 41% of the enrolled population. The analyses were therefore underpowered. In addition, the observed effects as a result of various pacing therapies were smaller and the standard deviations in AF burden were often higher than assumed for the calculated sample size.

The devices used in this study included stored onset reports to evaluate appropriate atrial sensing, but no stored atrial EGMs were available. Although a conservative approach has been followed in excluding data with suspected atrial-sensing artefacts, especially undersensing of AT may not always have been noticed.

In the preventive pacing phase, all four AF preventive pacing therapies were activated in patients randomized to the therapy arm. The effectiveness of AF prevention pacing as observed in this study may be influenced by the fact that the activation of individual AF prevention pacing therapies was not based on the observed type of AF onsets and the patient's AF history. The results reported by Lewalter et al.34 show that the selection of specific pacing therapies based on onset patterns may be more effective than enabling all therapies, as was done in the preventive pacing phase of this study.

To be enrolled for this study, patients were required to have an AF history of at least 1 year and have experienced at least three AF episodes during 3 months prior to enrolment. Despite these demanding criteria, 46 and 50% of the evaluable patients had an AF burden of <3 min/day in the conventional and preventive pacing phase, respectively. For future clinical trials on the prevention of AF, it may be considered to exclude patients from further randomization if their baseline AF burden is below a pre-defined threshold.

The high number of patients without AF episodes may also be explained by the temporal pattern of AF episodes. Ziegler et al.35 used diagnostic data from dual-chamber pacemakers to simulate the effect of the frequency and duration of atrial rhythm monitoring. They showed that the representation of the patient's true AF status improves with more frequent monitoring and longer monitoring periods. In the awareness that AF episodes often appear in clusters, the observations from our study suggest that future studies on this topic should include substantial longer monitoring periods to capture a representative assessment of AF burden.


Data from this study do not demonstrate a significant effect of conventional or dedicated atrial overdrive pacing in the reduction of AF. On the basis of observations from this study, the following aspects should be considered for future studies on the reduction of AF by means of pacing therapies:

  1. the effect of ventricular pacing;

  2. the effect of pacing in the control group;

  3. optimization of atrial sensing by appropriate programming, short-sensing bipoles, appropriate atrial lead location and intelligent sensing algorithms;

  4. minimum baseline AF burden as a requirement for inclusion in a randomized trial;

  5. longer monitoring periods during which device-based AF burden is determined.


This work was financially supported by Vitatron B.V., Arnhem, The Netherlands.


AFTherapy study investigators and institutions

J.H. Bennekers, Martini Hospital, Groningen, The Netherlands; M. Brignole, O.O.R.R.V. Leonardi E. Riboli, Lavagna, Italy, A.J. Camm, St. George's Hospital, London, UK, P.A. Capucci, Ospedale Civile, Piacenza, Italy; J. Clémenty, Hosp. Cardiologique du Haut Leveque, Marseille, France; F.G. Cosio, Hosp. Universitario de Getafe, Madrid, Spain; H.J.G.M. Crijns, University Hospital, Groningen, The Netherlands; P. Dini, Ospedale San Camillo, Rome, Italy; N. Edvardsson, Sahlgrenska Hospital, Göteborg, Sweden; M.D. Gammage, Queen Elizabeth Hospital, Birmingham, UK; A. Grande, Hosp. Severo Ochoa, Madrid; Himmrich, Joh. Gütenberg Universität, Mainz, Germany; E. Hoffmann, Klinikum München Bogenhausen, Munich, Germany; L. Jordaens, University Hospital, Rotterdam, The Netherlands; C. Kirchof, University Hospital, Maastricht, The Netherlands; T. Lewalter, Rheinische Friedrich Wilhelms Universität, Bonn, Germany; P. Mabo, CHR Universitaire, Rennes, France; A. Marshall, Derriford Hospital, Plymouth, UK; T. Meinertz, Universitätskrankenhaus Eppendorf, Hamburg, Germany; C. Menozzi, Arcispedale Santa Maria Nuova, Reggio Emilia, Italy; F. Molin, Hôpital de Sacre Coeur, Montreal, Canada; J.M. Morgan, Southampton General Hospital, Southampton, UK; M. Møller, Odense Universitetshospital, Odense, Denmark; H. Oddson, Orebro Medical Center, Orebro, Denmark; S. Osswald, Kantonnspital, Basel, Switzerland; P. Ritter, Centre Chirurgical de Val d'Or, Saint-Cloud, France; D.J. Roda, Hosp. General Universitario, Valencia, Spain; J.H. Ruiter, Medisch Centrum, Alkmaar, The Netherlands; A. Spampinato, Casa di Cura Villa Tiberia, Rome, Italy; G.Steinbeck, Klinikum der Universität München-Groshadern, Munich, Germany 8; A.N. Sulke, Eastborne Hospital, East Sussex, UK; R. Tavernier, University Hospital, Gent, Belgium; K.H. Tscheliessnigg, Vorstand der Universitätsklinik für Chirurgie, Graz, Austria; R. Yee, London Health Sciences Centre, London Ontario, Canada.

Conflict of interest: A.J.C. is a speaker and occasional consultant to Vitatron. B.A.A. is an employee of Vitatron.


View Abstract