Atrial tachyarrhythmia burden modelling by some electrophysiological parameters in pacemaker-recipient patients with Brady–Tachy syndrome

Abstract

Aims (1) To correlate atrial tachyarrhythmia (AT) burden of pacemaker-recipient Brady–Tachy syndrome (BTS) patients with a number of diagnostics-derived parameters after 1 month of DDD pacing; (2) to asses whether the activation of atrial overdrive or conventional rate-responsive pacing may affect relevant variables and their correlation.

Methods and results After 1 month of DDD pacing, rate-responsive function or persistent atrial overdrive was randomly activated for 3 months, in 92 BTS patients. Some pacemaker diagnostics parameters collected at 1- and 4-month follow-ups were included in multiple linear regression models, whose dependent variable was the Log transformation of AT burden and compared. With 1-month data, the only variables significantly correlating with Log AT burden were average (with a regression coefficient estimate of −0.07, P=0.02) and standard deviation (0.10, P=0.007) of atrial rate, mean premature atrial contraction (PAC) coupling interval (CI) (−0.005, P=0.001), frequency of PACs with CI<500 ms (1.30, P<10⁻⁶). Atrial pacing percentage (APP) and ventricular pacing percentage (VPP), PACs with CI>500 ms did not significantly correlate. Four-month data largely confirmed these results, except that in DDDR atrial rate average and standard deviation no longer correlated. Overdrive significantly increased APP and reduced PACs with CI>500 ms.

Conclusion AT burden showed significant dependence in DDD and during overdrive on atrial rate average and standard deviation. Highly premature PACs always significantly correlated with AT burden. Though increasing APP, which unexpectedly never correlated, overdrive could only reduce less premature PACs.

Introduction

A critical point in the debate on pacing prevention of atrial tachyarrhythmia (AT) recurrences in patients with sinus node dysfunction is the controversial results obtained so far with the dozen atrial overdrive or AT trigger suppression pacing algorithms currently used on top of conventional dual-chamber pacing modes.^1–5 Compared with conventional DDDR pacing, preventive algorithms have been shown to increase atrial pacing percentage (APP),^1,3,4 reduce the mean number of premature atrial complexes per day,³ and prevent symptoms of atrial fibrillation recurrences.⁴ Despite these results, a net decrease in the frequency of AT episodes was observed only in very specific patient subgroups,^1,2 whereas a significant reduction of AT burden has rarely been reported.⁶

Some approaches based on alternative pacing sites^7–9 have been shown to improve slightly the preventive effect of pacing. However, as it has been convincingly suggested,¹⁰ cumulative AT burden remains the main unattained goal of studies on AT pacing prevention, and lack of clear evidence in favour of sophisticated pacing algorithms is prompting reappraisal, in an attempt to explain why they have substantially failed.¹¹

Current pacemaker technology has evolved sufficiently to provide a considerable amount of information that could be exploited to give additional indications to resolve the remaining controversies. AT burden, atrial high rate episode electrogram (EGM) recordings, atrial pacing percentage (APP) and ventricular pacing percentage (VPP), premature atrial complex frequency and coupling interval (CI), mean atrial rate and standard deviation, for example, are presented simultaneously by the device at a normal follow-up visit and, also, these data are placed in a long-term context.

The specific aim of this study was to model AT burden with a number of parameters derived from the pacemaker diagnostics, to select those which significantly correlate, and to investigate whether correlation may be affected by atrial overdrive and rate responsive pacing activation. This analysis was performed on a subset of patients with Brady–Tachy syndrome (BTS) already enrolled in a previously published study.⁶

Methods

Main study design

The main study,⁶ approved by an independent institutional board, enrolled 149 patients, after obtaining written informed consent and presenting an indication for rate responsive dual-chamber pacing, due to BTS with at least one documented AT episode within the last 6 months. Exclusion criteria were permanent atrial fibrillation, left atrial size >50 mm, angina pectoris, congestive heart failure, prior atrioventricular (AV) node ablation, 1 year or less life expectancy, and pregnancy. The main objective of the study was to evaluate the AT burden associated with three different pacing approaches, 3 and 6 months from their activation: (a) Closed Loop Stimulation, artificially restoring physiological heart rate modulation, but with no specific AT suppression algorithms; (b) persistent atrial overdrive, called DDD+, attempting to suppress spontaneous atrial events, but introducing no sinus-rhythm-independent rate modulation; (c) conventional accelerometric-sensor-based DDDR pacing, as a control arm.

At implant, atrial leads were placed in the right atrial appendage in all patients, and to ensure reliable and consistent data, all the pacemakers were programmed according to a uniform standard, apart from the pacing mode, including bipolar atrial sensing with 0.5 mV threshold, lower, upper pacing, and upper tracking rates at 70, 130, and 140 bpm, respectively.

After implant, the protocol required a 1-month run-in period of DDD pacing at 70 bpm, before activation of one of the three pacing algorithms, randomly assigned. Next follow-ups were scheduled at 4 and 7 months. A decision to add antiarrhythmic therapy, if necessary, had to be taken within the first month; any change in antiarrhythmic drug regimen during the following 6 months became a dropout criterion. Only patients randomized to DDD+ or DDDR groups received a Philos DR pacemaker model (Biotronik GmbH, Berlin, Germany), provided with a specific premature atrial contraction (PAC) detection algorithm and diagnostic features.

Patient selection criteria

A sub-group of 92 patients remained in the study 1 month after implant and these patients were randomized to the DDD+ and DDDR arms (48 in the former, 44 in the latter, respectively). Baseline characteristics of these patients are shown in Table 1, and their antiarrhythmic drug therapy in Table 2. To avoid a biasing effect of different antiarrhythmic regimens between the first and second periods, only 76 patients who reached the 4-month follow-up with unchanged antiarrhythmic therapy from implant entered the 4-month data analysis: 40 patients were in the DDD+ group, 36 in the DDDR group.

Further dropout criteria were documentation of false-positive AT episodes in the EGM recordings and 100% AT burden in any observation period.

Pacing algorithms

Basically, the DDD+ algorithm used in the present study does not differ from other atrial overdrive pacing algorithms commercially available: it is designed to suppress all atrial-sensed events between basic and upper pacing rates. Overdrive pacing algorithms were already described in detail and extensively discussed elsewhere:^3–5,12 the pacing dynamic is established in a circular fashion, decreasing and increasing the escape interval, according to whether atrial events are sensed or not. The objectives are to reduce refractoriness dispersion, suppress PACs, and prevent short-long sequences by permanent atrial pacing occurring slightly above sinus rhythm and closely following spontaneous heart rate modulation.

In the conventional DDDR pacing mode, atrial overdrive is generated by rate-responsiveness, the pacing rates of which depend on the physical activity level detected. Though conventional DDDR mode is not designed to suppress or prevent the most common AT trigger mechanisms, it generally increases atrial pacing percentage, introducing an exercise-dependent heart rate modulation. In this study, rate-responsive function was provided by a standard accelerometric sensor programmed with nominal settings in all patients.

Objectives of the analyses

The objectives of the present sub-analysis were (1) to correlate simultaneously the AT burden, observed during the first month of DDD pacing, with a set of parameters: APP and VPP, long-term average and standard deviation of atrial rate in AT free periods, mean atrial rate before PAC occurrence, PAC CI, and frequency of PACs; (2) to evaluate the impact of overdrive and rate-responsive pacing on those parameters and their correlation level, three months after activation.

Description of the data set and measurement methods

AT burden

A mean rate of the last four atrial intervals equal to or >150 bpm was chosen as the detection criterion for AT episodes, so as to include long lasting clinically relevant atrial fibrillation episodes and, also, other slower sustained or non-sustained atrial tachycardias, which constitute an important portion of AT burden.¹³ In this analysis, AT burden was defined as the percentage of atrial beats satisfying that detection condition. Data for computation were derived from the mean atrial rate histogram provided by the pacemaker. As the presence of false-positive episodes represents a critical issue, which could not be tolerated for the purpose of this analysis, an expert committee reviewed samples of up to nine 20-s atrial and ventricular EGMs and marker channel recordings¹⁴ stored in the pacemaker memory at the onset of atrial high rate (≥150 bpm) episodes and collected at each patient's follow-up. Patients presenting one or more episodes classified as false-positive were excluded from the analysis.

Atrial and ventricular pacing percentages

APP and VPP were directly calculated as the ratio of atrial/ventricular paced events to the total number of atrial/ventricular events recorded by the pacemaker. As under normal conditions atrial pacing is inhibited during AT episodes, any correlation between APP and AT burden might be misleading. For this reason, as far as APP was concerned, calculations were performed excluding atrial events sensed in the AT detection zone. So, herein APP was always assessed in AT free periods.

Long-term average and standard deviation of atrial rate

Long-term average and standard deviation of atrial rate were also derived from the mean atrial rate histogram in the pacemaker diagnostics and calculated by standard procedures. All the atrial intervals recorded between two consecutive follow-ups, except for those detected during AT periods, were included in the computation.

PAC frequency, mean CI, and mean atrial rate at PAC onset

PAC frequency was defined as the mean number of PACs daily detected. The pacemaker uses a specific algorithm for PAC detection: any atrial-sensed event the cycle length of which is shorter than the average of the last consecutive four intervals by more than a programmable factor, called ‘prematurity’, is classified as a PAC. On the basis of previously published data, prematurity index was always set at 20%.¹⁵ For each detected PAC, the CI and the mean rate of the previous four intervals are logged into the pacemaker memory. So for each patient, calculation of the mean CI and atrial rate at PAC onset, as well as PAC classification based on CI, were made possible.

Statistical analysis

A number of multiple linear regression models were devised and tested. Ordinary least squares methods were used to estimate the regression coefficients and their standard errors. Statistical analyses were performed by means of Statistica software package (StatSoft, Inc.). Analysis of variance and calculation of multiple correlation coefficients were performed to evaluate adherence of the models to data. Cook's distances were examined, searching for possible anomalous points during residual analysis. Minimum tolerance for covariates was set at 0.1. To identify the most relevant variables, an automatic forward step-wise procedure was implemented with an F-to-enter=1.0 and F-to-remove=0.0. AT burden was always treated as the dependent variable, while APP, long-term atrial rate average and standard deviation, mean atrial rate at PAC onset, mean PAC CI, and PAC frequency were included as covariates. Log transformation was needed to normalize AT burden and PAC frequency distributions. Paired Student's t-test was used to compare 1- and 4-month data, reported as mean±standard deviation. Yates-corrected χ² test or Fisher's exact test was applied to compare categorical variables. Significance level for all the hypothesis tests was set at P=0.05.

Results

EGM recording classification and patient selection

Nine hundred and seventy eight atrial high rate EGM recordings were collected at 1- and 4-month follow-ups. Seven hundred and ninety of them (81%) were true AT episodes: 202 (25%) were characterized by an irregular cycle length and atrioventricular (AV) conduction, sudden onset and an atrial rate ≥250 bpm; 23 (3%) by a regular cycle, a fixed atrioventricular ratio, and a rate at or above 200 bpm; 503 (64%) by regular cycle length, ventricular conduction, and a rate between 150 and 200 bpm; 62 (8%) were non-sustained ATs, most of them consisting of runs of five or more consecutive PACs. Of the 188 EGM recordings classified as false-positive, 124 (66%) were due to R-wave far field oversensing, 41 (22%) to competitive atrial pacing, 18 (9%) to retrograde atrioventricular conduction, 5 (3%) to failed atrial capture. False-positive episodes were found in 19 of 92 patients (21%) at 1-month follow-up and in 8 of 76 (10%) at 4-month follow-up. As stated earlier, these patients were excluded from further analysis. Also, four more patients had to be excluded due to some missing data at 1-month. Ultimately, 69 patients at 1-month and 68 at 4-month follow-ups (35 in the DDD+ group, 33 in the DDDR group) presented useful data. All these patients had a non-zero AT burden <100% in both follow-up periods.

Regression analyses

The Log AT burden collected after 1 month of DDD pacing at 70 bpm was considered as the dependent variable in a multiple regression linear model including APP, VPP, long-term average and standard deviation of atrial rate, mean atrial rate at PAC onset and mean CI of PACs, as covariates. Owing to a significant correlation detected with APP, PAC frequency was initially excluded from the model, for redundancy reason, and subsequently analysed in a separate model. The results of this first analysis are displayed in Table 3, where the regression coefficients, their standard errors and significance level for each variable are reported. The model fairly fitted sample data with a multiple correlation coefficient of 0.65 (P<10⁻⁵). Neither anomalous points nor critical values of tolerance were found. The same model was again used for 4-month data (Table 3), initially failing to distinguish between ‘DDD+’- and ‘DDDR’-paced patients. The pacing group was introduced as a new dummy variable, to take into account the possible different effect of the two pacing modes on Log AT burden. The group variable had no significant coefficient (P=0.91), reflecting no evidence of a clear difference between the two pacing strategies in reducing AT burden. On the other hand, the model was again fairly representative of the underlying relationships (R²=0.73, P<10⁻⁵) and showed good stability, basically confirming the results of 1-month analysis, with similar coefficient estimates, VPP excepted.

The regression model was separately investigated in the DDD+ and DDDR groups at 4 months, after undergoing an automatic forward step-wise procedure to eliminate potentially irrelevant variables (as imposed by the reduced sample sizes). Only APP, long-term average and standard deviation of atrial rate and mean PAC CI survived in the model after the step-wise procedure. This reduced model was applied separately to the 4-month data of the two groups and the results are reported in Table 4. Multiple correlation coefficients were 0.73 (P<10⁻⁵) in the DDD+ group and 0.77 (P<10⁻⁵) in the DDDR group.

In the DDD+ group, the regression coefficient estimates reached the significance level again for the mean PAC CI (P=0.0001) and the average (P=0.007) and standard deviation (P=0.003) of atrial rate; the signs and estimates of these coefficients always confirmed the results of the complete model. Of note, APP coefficient approached but did not reach significance (P=0.08).

On the other hand, in the DDDR group, APP, long-term average, and standard deviation of atrial rate coefficients were not statistically significant (P=0.80, P=0.45 and P=0.10, respectively), whereas the mean PAC CI remained the only variable significantly correlating with Log AT burden, with a negative coefficient estimate of −0.007 (P<10⁻⁵).

PAC frequency analysis

AT burden dependence on PAC frequency was investigated by distinguishing PACs in two groups based on their CI: PACs with CI <500 ms were defined as ‘short’-PACs; PACs with CI >500 ms as ‘long’-PACs. The limit of 500 ms was arbitrarily set, based on the consideration that a rate of 120 bpm is the maximum pacing rate commonly programmed in pacemakers, thus representing the upper limit below which PACs can be mostly affected by pacing. Hence, relationship between Log AT burden and PAC frequency was investigated by a linear model including Log short-PAC and long-PAC frequencies as the only covariates. This model was applied to data from the complete group of 1-month patients, and to the 4-month data, in the DDD+ and DDDR groups, separately. Table 5 displays the results of this analysis. The regression coefficient estimate of Log short-PAC frequency was always positive and highly significant (P<10⁻⁵ in all follow-ups and groups). On the contrary, Log long-PAC frequency did not appear significantly correlated with Log AT burden: the regression coefficient reached significance (P=0.001) only at 1-month, but unexpectedly with a negative value; with 4-month data, coefficients were not significant in both the DDD+ (P=0.28) and the DDDR group (P=0.32).

Variable comparisons between follow-ups

Looking at the effect of pacing algorithms on the variables considered, Table 6 reports paired (within patients) comparisons between 1- and 4-month follow-ups, in each study arm. Besides Log AT burden which was not significantly affected by pacing algorithm activation, DDD+ and DDDR showed different effects on some variables. In particular, DDD+ significantly increased APP (from 67.2±27.5 to 91.9±20.1%, P=0.002) and long-term mean atrial rate (from 79.8±5.0 to 85.1±10.4 bpm, P=0.005) with respect to DDD pacing, and reduced Log long-PAC frequency from 2.92±0.50 to 2.26±0.70 (P=10⁻⁶). On the other hand, DDDR only significantly increased long-term mean atrial rate from 78±7 to 81±7 bpm (P=0.009), but there was no relevant effect on the other variables.

Discussion

A set of different parameters was simultaneously correlated with AT burden, to ascertain the most relevant variables during conventional DDD pacing, observing whether and how they are affected by the activation of permanent atrial overdrive or rate-responsive pacing. Discussion of each variable will follow individually, although outcomes of a multiple regression analysis should be comprehensively evaluated, as each variable plays its own role in the specific model considered.

Atrial pacing percentage

In this analysis, APP never significantly correlated with Log AT burden, both during DDD pacing and after pacing algorithm activation. No more than a trend could be observed only in the DDD+ group, as shown by the negative regression coefficient approaching but not fully reaching significance (P=0.08). This result may appear quite surprising but is not inconsistent, as no effect was observed in terms of AT burden reduction with both DDDR and DDD+, despite the huge increase of APP obtained with DDD+. This further confirms that increasing APP may not be sufficient for controlling the AT burden, especially considering that, in this analysis, APP was calculated only during AT free periods. Schuchert et al.¹⁶ found that APP and AT duration were inversely related, with a highly significant correlation coefficient of 0.95. But after APP was corrected for AT duration, correlation was lost (r=0.08) and the authors concluded that APP should be carefully interpreted in relation to AT duration. Actually, data published to date have not demonstrated a clear reduction in AT burden associated with specific prevention algorithms designed to increase APP up to 100%, in comparison with conventional DDDR or DDD pacing.^1–3 A possible explanation may be, in part, related to the pacing site: conventional sites, inducing intra- and inter-atrial conduction delays, may partially abolish benefit of pacing. Alternative sites (particularly, interatrial septum) may overcome this problem, but this still has to be fully demonstrated: Boriani et al.¹⁷, in a small BTS population including both patients paced at the right atrial appendage and interatrial septum, in comparable percentages, performed a similar regression analysis involving only PAC frequency and APP as covariates, but never reported a significant regression coefficient for APP with the preventive algorithms tested.

Ventricular pacing percentage

In this analysis, VPP did not show a significant linear correlation with AT burden in both the complete models applied to the whole population at 1- and 4-month data. This result is in contrast to previous findings^2,18,19 and may be explained by the following two considerations: firstly, the observation period (4 months, on the whole) was probably too short to observe a significant effect on AT burden; secondly, the mean VPP in all the periods and in all the groups was generally too high (82–96%) to sample a sufficient range of variability. This high mean VPP was obtained despite no patient having III degree AV block and very few (12%) having I/II degree AV block reported before implantation. The programmed AV delay at the lower rate was 180 ms or longer in all the patients without AV block. Therefore, the percentage of patients with I degree AV block could have been underestimated, most likely due to the effect of the antiarrhythmic drug regimen established after implant. As expected, DDDR and DDD+ activations did not affect VPP: only in the DDDR group, a modest though significant reduction of VPP was observed.

Long-term average and standard deviation of atrial rate

A significant linear correlation was found between Log AT burden and long-term average and standard deviation of atrial rate during sinus rhythm in the complete models with 1- and 4-month data, indicating that sinus rate modulation may be directly involved in AT initiation and/or maintenance. In particular, regression coefficients for atrial rate average and standard deviation were statistically significant and their estimates were negative and positive, respectively. This can be interpreted as follows: the lower the mean atrial rate during sinus rhythm (or closer to the pacemaker lower rate) and, simultaneously, the larger its variability, the longer the time in AT tends to be, this trend assumed statistical significance. Interpreting low mean atrial rates as being due to sinus bradycardia and large variability being due to frequent ectopics, it easily follows that the association of these two factors plays a primary role in increasing AT burden.

It was of interest that when the reduced model was separately applied to the DDD+ and DDDR patient groups, a clear difference appeared: with DDD+ outcomes of the complete model being fully confirmed, whereas with DDDR long-term average and standard deviation of atrial rate were no longer significantly correlated with AT burden. A straightforward explanation of this result is difficult to offer but it may possibly be related to the difference between the working principles of DDDR and DDD+ modes. A large body of published data supports the view that AT initiation is promoted by variations in autonomic tone.^20–22 DDDR may have some effect on sympatho-vagal balance preventing bradycardia during exercise; on the other hand, a pure atrial overdrive is not expected to provide any heart rate modulation during exercise, working basically as a DDD mode unless spontaneous beats are detected, especially in BTS and/or chronotropically incompetent patients. A hypothesis may be advanced that the association between AT burden and bradycardia plus intense ectopics is still present in both DDD+ and DDDR, but in the latter, this mechanisms may be partially hidden by rate modulation of rate-responsiveness. Such an explanation would also be consistent with the results obtained comparing long-term average and standard deviation of atrial rate between DDD and preventive algorithms: both DDDR and DDD+ significantly increased mean atrial rate, as a result of their different working function, while standard deviation was unchanged.

PACs and AT burden relationship

PAC CI always significantly correlated with Log AT burden, in all the regression models evaluated, regardless of the pacing modes. The regression coefficient estimates were found negative, reflecting an increasing pro-arrhythmic effect of PACs bearing on a reduced cycle length. More explicitly, when modelling Log AT burden by Log long-PAC and Log short-PAC frequencies, as the only covariates, the former resulted in having no effect on Log AT burden, which on the contrary turned out to depend only on the frequency of PACs with CI shorter than 500 ms. Only with the 1-month follow-up data the Log long-PAC regression coefficient reached statistical significance, but its estimate was negative. These results further confirm previous observations based on 24-h Holter recordings^23,24 and electrophysiological studies.²⁵ However, to our knowledge, this was the first long-term in vivo analysis of the relationship between AT burden and PAC CI performed in a BTS-pacemaker-recipient population, and probably represents the major feature of this work.

The threshold chosen to distinguish ‘short-’ from ‘long-PACs’ is particularly of interest to evaluate the role of pacing in PAC suppression: 500 ms or less corresponds to 120 bpm or more, a limit that is usually not exceeded by pacing, for obvious tolerability reasons. Inevitably, these results raise some concerns about the efficacy of pacing and may, in part, explain the disappointing results obtained so far with sophisticated overdrive algorithms whose objective is to reach a high degree of APP while merely following sinus rhythm. The paired comparisons between 1- and 4-month data showed that, whereas no difference was observed in the DDDR group, DDD+ overdrive could only significantly reduce PACs with CI longer than 500 ms with respect to DDD pacing, but did not affect frequency of PACs with a higher prematurity. This result, pertaining to conventional atrial pacing sites, is in contrast to the hypothesis that pacing may promote a reversal of electrical remodelling. On the contrary, one might argue that the beneficial effect of reduced dispersion of conduction and refractoriness exerted by pacing is not sufficient to reduce ectopics occurring above the upper limit of the pacing rate. In fact, the mean atrial rate before PAC onset never correlated, positively or negatively, with AT burden, not even during pacing algorithm activation. Moreover, overdrive significantly increased mean atrial rate before PAC onset and even reduced the mean PAC CI, compared with DDD. These somewhat unexpected results were obviously due to the increased atrial rate imposed by pacing, but are not in favour of the view that pacing may suppress ectopics also with a high degree of prematurity by reducing the escape interval so as to increase refractoriness and prevent AT recurrences.

Study limitations

Some limitations of this analysis are worthy of note:

• About 60% of the selected patients were taking antiarrhythmic therapy. This could have affected variables and also their regression on Log AT burden in the linear models. However, within the observation period, every effort was made to keep relevant boundary conditions constant, including drug regimen: therefore, any variation observed between 1- and 4-month follow-ups should mainly be ascribed to activation of the pacing algorithm.

• The observation period was relatively short and the results could depend in part on the brief follow-up. It might be interesting to assess whether the conclusions reached in this analysis would be confirmed over longer periods.

• The short observation period and the high mean VPP observed could have led to unreliable results on correlation between Log AT burden and VPP. Interestingly, a recent report²⁶ also acknowledged that in patients who mostly benefited from a 3-month activation of preventive pacing algorithms, experienced an unexpected increase in VPP. Our data appear to be consistent with this finding which merits further investigation.

• Some patients presenting with EGM recordings of false-positive high atrial rate episodes had to be excluded from the analysis. In total, inappropriate detection occurred in 19% of the EGM reports. This rather high percentage was probably due to the low setting (0.5 mV) of the atrial sensitivity threshold (two thirds of these episodes were due to far-field R-wave oversensing). That setting was chosen in order to avoid undersensing of atrial events during tachyarrhythmias. Generally, there is no way to handle true-negative episodes, potentially leading to an underestimation of AT burden. A more specific detection algorithm and an automatic sensitivity adjustment system would have improved accuracy of results and reduced patient dropout rate.

Conclusions

This analysis investigated the association between AT burden and a number of parameters, in a group of BTS patients, paced at conventional atrial sites, with DDD mode for 1 month and with atrial overdrive or DDDR for 3 months. AT burden appeared substantially unrelated to APP, and mean atrial rate before PAC onset. Average and standard deviation of atrial rate and mean PAC CI were the only relevant covariates. The model estimate was basically unaffected by the overdrive algorithm activation. Frequency of only those PACs with a CI shorter than 500 ms significantly and positively correlated with AT burden. Overdrive pacing could only reduce frequency of PAC with CI longer than 500 ms, which had no effect on AT burden.

These results may help to refine the role of pacing in the prevention of AT recurrences and to permit better evaluation of approaches merely aimed at permanently pacing the atria 100% of time.

Acknowledgement

The authors are indebt to Alessio Gargaro for statistical data-analysis and interpretation of the results.

References