Ablation Index for Catheter Ablation of Atrial Fibrillation　― Clinical Applicability and Comparison With Force-Time Integral ―

Abstract

Background: Key determinants for lesion formation in catheter ablation are contact force, radiofrequency (RF) power and time. The aim of this study was to evaluate the clinical applicability of ablation index (AI), a novel non-linear formula based on these components, and to compare AI with the conventional linear force-time interval (FTI) in pulmonary vein isolation (PVI).

Methods and Results: Target AI ranges were defined for anatomical segments of the ipsilateral pulmonary veins. The operator was blinded to AI during PVI for the initial 11 patients (group A), and was unblinded for the remaining 23 patients (group B). We assessed (1) the clinical value of AI to avoid excessively high and low values with an operator blinded vs. non-blinded to AI; and (2) the relation of AI and FTI in predefined ranges. In group A, 235/564 lesions (41.7%) were in the predefined target range as compared with 1,171/1,412 lesions (82.9%) in group B (P<0.001). A given AI may correspond to a wide range of FTI, as reflected by a quartile coefficient of dispersion for AI of 0.11 vs. a quartile coefficient of dispersion for FTI of 0.36.

Conclusions: Incorporating RF current power, the non-linear AI provides more comprehensive information during PVI compared with FTI. Given that the FTI for a given AI varies widely, the value of FTI in clinical practice is questionable.

Precise lesion formation is crucial for successful catheter ablation of atrial fibrillation (AF). Key determinants for lesion formation using radiofrequency (RF) ablation are contact force (CF), time and RF power.¹ Insufficient RF delivery may result in ineffective lesion formation, whereas excessive RF current (RFC) delivery may result in potentially life-threatening complications such as atrio-esophageal fistula formation or cardiac tamponade.² Reliable and clinically applicable feedback parameters are essential for ablation monitoring and to adequately regulate RF delivery to avoid complications and to keep AF recurrence rates as low as possible.³

CF and the force-time integral (FTI), the product of CF and ablation time, are the most widely used parameters to control RFC applications.⁴ Low CF and FTI during AF ablation are associated with a higher incidence of pulmonary vein (PV) reconnection and recurrence of atrial arrhythmia.^5,6 But even after initially successful PV isolation (PVI), the incidence of reconduction due to gaps in ablation lines is still high.^5,6

CF sensing catheters are now routinely adopted for catheter ablation of cardiac arrhythmias,⁷ but FTI has several limitations. First, it does not take into account RF power. The effect of increased power has been shown to be more relevant compared with increased CF.⁸ Second, FTI assumes a linear relationship between CF and time. In recent studies, however, the ablation depth graphs reflected a non-linear relationship with time for diverse CF and RF power settings. The novel ablation index (AI; Biosense Webster, Diamond Barr, CA, USA) is intended to overcome these limitations. According to animal models, AI can precisely predict lesion size.⁹

AI is a commercially available parameter that combines CF, RF time and RF power in a non-linear formula in real time. It therefore allows for intra-procedural visualization of RFC application in a 3-D electroanatomic mapping system (Carto 3, Biosense Webster). Recent adoption of AI in clinical practice has produced a significant improvement in procedural outcome¹⁰ and reduced 1-year recurrence rates.³ In initial clinical studies low AI has been associated with acute and late PV reconduction.^11,12 AI-guided PVI is associated with higher impedance drop and AI is used as a surrogate parameter for effectiveness of lesion formation.^13,14 AI, however, is not yet widely used in clinical practice. Recently, we have adopted the intra-procedural use of the parameter as an institutional standard for RFC ablation of cardiac arrhythmias. In contrast to FTI, target values for AI adapted to the respective atrial segment can be deduced from previous studies.¹¹

The aim of this study was to (1) assess the clinical utility of AI by comparing AI during PVI with the operator blinded to real-time AI vs. non-blinding of the operator to AI during the procedure, as a retrospective assessment of ablation quality; and (2) compare the widely used FTI with the non-linear AI with respect to distribution.

Methods

Patients

The subjects consisted of 34 patients diagnosed with symptomatic paroxysmal AF defined as >1 episode of AF <7 days duration with ≥1 episode documented on 12-lead electrocardiogram.¹⁵ Patients with prior left atrial (LA) catheter ablation were excluded. Data were obtained from the institutional ablation registry and analyzed retrospectively. Written informed consent was obtained from all patients. The study was approved by the institutional ethics committee. All patients underwent PVI for paroxysmal AF according to the institutional standard.

PVI

To exclude LA thrombus, all patients underwent transesophageal echocardiography 1 day prior to the procedure. Intake of novel oral anticoagulation was discontinued 24 h before the procedure. Oral anticoagulation with phenprocoumon was continued and the international normalized ratio was maintained at 2.0–3.0. All patients underwent PVI in deep sedation applying a continuous i.v. infusion of propofol with additional boluses of midazolam and fentanyl. As per the institutional standard, no esophageal temperature monitoring was performed.

A decapolar catheter was advanced into the coronary sinus. After double trans-septal puncture, 2 SL1 sheaths were advanced into the LA. A commercially available navigated decapolar spiral mapping catheter (LASSO, Biosense Webster) was used for PV mapping. A 3-D electroanatomic map of the LA was obtained using either a point-by-point technique with a CF-sensing catheter (ThermoCool SmartTouch Surround Flow [STSF], Biosense Webster) or a fast anatomic mapping technique using the spiral catheter. Anatomical segments of the PV were arbitrarily defined as anterior, roof and posterior segments of the ipsilateral PV. Wide area circumferential PVI for treatment of symptomatic paroxysmal AF was performed in power-controlled mode. RFC application was limited to 35 W at the anterior and to 25 W at the roof and posterior segment, respectively, to take into account differing thickness of the myocardial wall and to avoid potential complications such as steam pop formation or esophageal injury. Target AI was 380 for the posterior segment, 450 for the roof segment and 500 for the anterior segment. Importantly, no dragging technique was used. RFC energy was delivered with a target CF of 10–40 g until a signal reduction of 70% was achieved. Applications <5 s were excluded from analysis. All procedures were performed by 1 single operator.

Force-Time Power Index Measurement and Data Acquisition

For ablation and AI acquisition we used the STSF catheter (7.5-F sensing catheter with a 3.5-mm irrigated tip, Biosense Webster. Calibration of the catheter was performed before 3-D electroanatomical mapping and thereafter every 20 min. CF was displayed to the operator throughout all procedures. Target range for CF was 10–40 g. AI was not displayed to the operator for 11 patients (group A) but was registered using the CARTO3 SMARTTOUCH Software (Biosense Webster) module. For each ablation point the following parameters were recorded: ablation time, maximum temperature, maximum CF, maximum power, baseline impedance, impedance drop (defined as minimum impedance subtracted from baseline impedance) and FTI. All data were exported from CARTO3 after the procedure. Settings of the automatic lesion annotation software (Visitag, Biosense Webster) were as follows: stability, 3 mm; minimum CF, 3 g; minimum time, 3s; force over time, 25%.

We performed a retrospective analysis of 11 routine PVI procedures (group A; AI blinded) in which AI was recorded but not displayed to the operator. For the prospective cases, AI target values were based on previous clinical studies,¹¹ and the median values of AI recorded during the blinded cases resulting in target values of 380, 450 and 500 for the posterior, roof and anterior segment, respectively. During subsequent procedures, AI was displayed to the operator (group B; AI unblinded). RFC delivery was then stopped on reaching AI target thresholds. Ranges of low, optimal, and high AI were defined as target AI±15%. AI values were analyzed according to 3 different anatomical segments of the ipsilateral PV (anterior, posterior and roof). The proportion of AI in the target range was determined for cases with the operator blinded to AI and for cases with AI displayed to the operator. For AI in very low (0–300), low (300–400), high (400–500), very high (500–600) and excessively high (>600) ranges, we assessed the corresponding FTI in order to evaluate the usefulness of AI and FTI in detecting insufficient or excessive RFC delivery.

Statistical Analysis

Statistical analysis was performed with R (R: A Language and Environment for Statistical Computing, R Core Team, R Foundation for Statistical Computing. Vienna, Austria, 2017. https://www.R-project.org). Data sets from both patient groups were filtered for a minimal duration of 5 s. Non-normally distributed continuous variables are represented as median (IQR). Kolmogorov-Smirnov test was used to test for normality of distributions. To assess the proportion of values in a target range as a categorical variable, Fisher’s exact test was used. Statistical significance was set at P<0.05. To estimate dispersion of non-normally distributed data, the quartile coefficient of dispersion was applied. Pearson correlation coefficient was used for correlation analyses of AI and FTI.

Results

Thirty-four patients with symptomatic paroxysmal AF were enrolled in the study. Patient characteristics are listed in Table.

Table. Patient Characteristics

Characteristics	Patients (n=34)
Age (years)	63.7±11.5
Male	64.7
BMI (kg/m2)	28.3±5.5
CHA2DS2-Vasc score (no. patients)
<2	13
2–4	6
>4	4
Hypertension	76.5
CAD	32.3
Diabetes mellitus	17.6
COPD	5.9
LVEF (%)	54.5±3.5
LAD (mm)	37.5±3.8
Medication
β-adrenergic blocker	85.3
Amiodarone	26.5
ACEI	29.4
Calcium-channel blocker	32.3
Sodium-channel blocker	44.1
Oral anticoagulant	100.0

Data given as % or mean±SD, unless otherwise indicated. ACEI, angiotensin-converting enzyme inhibitor; BMI, body mass index; CAD, coronary artery disease; COPD, chronic obstructive pulmonary disease; LAD, left atrial diameter; LVEF, left ventricular ejection fraction.

Procedural Data and Outcome

PVI was successfully achieved in all patients and no major periprocedural complication occurred. Mean procedure duration was 144±38 min. Mean dose-area product was 3,070.6±2,034.6 cGy·cm² and mean radiation duration was 15.5±6.2 min. Median procedure time did not differ between groups: group A, 144 min (IQR, 40 min); group B, 143 min (IQR, 46.5 min; P=0.70). There was no difference in total ablation time, maximum contact force, maximum temperature, baseline impedance or impedance drop between blinded and non-blinded AI (Figure 1). After a mean follow-up of 639.0±92.9 days, with 1 patient lost to follow-up in group B, the number of recurrences was 4/11 in group A and 7/22 in group B (P=1.0, Fisher’s exact test). There were no complications in group A, as opposed to 2 complications in group B (1 post-procedural transitory ischemic attack, 1 groin hematoma not requiring surgical therapy; P=0.54, Fisher’s exact test).

Figure 1.

Ablation parameters for procedures where the operator was blinded vs. non-blinded to ablation index (AI) during the procedure. There was no difference between groups for procedural (A) total ablation time [min], (C) maximum contact force [g], (D) maximum temperature [℃], (E) maximum power [W], (F) baseline impedance [Ω] or (G) impedance drop [Ω]. (B) Ablation time per lesion [s], however, was significantly shorter when AI was displayed.

Blinded vs. Non-Blinded

A total of 1,967 ablation points (group A, 564 points, 28.5%; group B, 1,412 points, 71.5%) were included in the analysis.

For group A, median AI at the anterior, the posterior and the roof segment was 457 (IQR, 175–707); 426 (IQR, 175–670) and 469 (IQR, 160–787), respectively, for the left PV and 498 (IQR, 166–833); 425 (IQR, 107–722) and 490 (IQR, 237–731), respectively, for the right PV. Median AI in group B (non-blinded) at the anterior, the posterior and the roof segment was 461 (IQR, 261–665); 390 (IQR, 361–420) and 442 (IQR, 439–474), respectively for the left PV and 510 (IQR, 479–547); 381 (IQR, 358–422) and 452 (IQR, 435–481), respectively, for the right PV.

Display of AI (group B) resulted in a significantly higher rate of ablation points in the predefined target range as compared with blinded ablation: group A, 235/564 (41.7%) vs. group B, 1,171/1,412 (82.9%; P<0.001), so that very low or very high AI outside the target range were avoided. There was no difference in AI between group A and group B (P=0.631, Wilcoxon test). Median ablation parameters did not differ between groups with respect to ablation time (24.2 min vs. 24.2 min), maximum contact force (19.9 g vs. 18.4 g), maximum temperature (26.9℃ vs. 26.9℃), maximum power (25.8 W vs. 25.9 W), baseline impedance (154.0 Ω vs. 154.8 Ω) or impedance drop (7.3 Ω vs. 8.2 Ω). Median ablation time per lesion was longer in group A (24.5 s) versus group B (22.0 s; Figure 1). AI below the target range was registered for 90 points (16%), 25 points (4.4%) and 32 points (5.7%) for the anterior, roof and posterior segment, respectively, in group A, and for 97 points (6.9%), 46 points (3.3%) and 35 points (2.5%) for the anterior, roof and posterior segments, respectively, in group B. AI above the target range was registered for 49 points (8.7%), 44 points (7.8%) and 89 points (15.8%) for the anterior, roof and posterior segment, respectively, in group A and for 18 points (1.3%), 16 points (1.1%) and 29 points (2.0%) for the anterior, roof and posterior segment, respectively in group B (Figure 2).

Figure 2.

Proportion of ablation points in the ablation index (AI) target range (±15% of target values, i.e. 500, 380 and 450 for anterior, posterior and roof segment, respectively) when the operator was (Left) blinded or (Right) non-blinded to AI during the procedure.

AI vs. FTI

For very low AI (0–300 AU) FTI ranged from 174 to 512 gs; for low AI (300–400 AU) FTI ranged from 331 to 572 gs; for high AI (400–500 AU), from 357 to 604 gs; for very high AI (500–600 AU) from 215 to 1,220 gs; and for excessively high AI (>600 AU), FTI ranged from 477 to 4,369 gs. AI and FTI did correlate (R=0.70); distinct AI may exhibit a wide range of corresponding FTI (Figure 3). For 74/1,976 points (3.7%) FTI was >400 and AI was below the target range of the respective segment.

Figure 3.

Relation between force-time integral (FTI) and ablation index (AI) when the operator was (A) blinded or (B) non-blinded to AI during the procedure. (B) When AI was displayed, ablation was performed to AI target values of 500 for anterior, 380 for posterior, and 450 for roof segments, resulting in (C, Lower) distinct peaks of AI in the respective segments. (C, Upper) Corresponding FTI does not exhibit distinct peaks, but is widely distributed. (D) Consequently, corresponding FTI for grouped AI has a wider range.

Generally, FTI varies more widely compared with AI, which is reflected by the quartile coefficient of dispersion, which for AI was 0.11 and 0.36 for FTI.

FTI increases linearly whereas AI is characterized by a non-linear growth reflective of lesion formation.¹⁶ Figure 4 shows representative lesions and the increase of FTI and AI over time, illustrating the wide range of FTI for a given AI.

Figure 4.

Force-time integral (FTI) vs. ablation index (AI) for 2 example lesions demonstrating a wide range of FTI for a given AI. Ablation on the (A) right and (B) left anterior wall until a maximum AI of 550 is reached; the corresponding FTI is (A) 1,098 and (B) 424.

Discussion

In this study we have demonstrated that AI is a clinically applicable tool that allows for more standardized RFC applications. When compared with the established parameter, the dispersion of FTI is larger, whereas RFC application can easily be adjusted to a narrow range of AI. As demonstrated by Figure 4, FTI increases linearly whereas AI is characterized by a non-linear growth, reflective of lesion formation. Especially for high AI, FTI can vary widely. For CF as a feedback parameter, display of CF resulted in greater stability of the CF values.¹⁷ A higher proportion of AI values were in the target range when AI was displayed to the operator. The ablation parameters ablation time, maximum CF, maximum temperature, maximum power, baseline impedance and impedance drop did not differ between group A and group B, although ablation time per lesion was shorter in group B. Shorter ablation time per lesion may be explained by a more rigorous point-by-point ablation technique. Outcome and complication rates did not differ between the groups, but the sample size was a limited, and hence the study was not sufficiently powered for an assessment of recurrence rates and incidence of complications. A major shortcoming of CF is its high variability related to anatomical and respiration-caused instability, also when visible to the operator.^7,18 In contrast to the corresponding AI ranges, FTI has a large proportion of overlapping values: FTI considered acceptable may correspond to AI in the target range but also to AI below the accepted target range. This demonstrates that FTI may be misleading, notably because high CF can correspond to small lesions.¹⁹ Therefore, when compared with FTI, as an established parameter to control RFC application, AI is more stable and reliable. Similarly to the current findings, impedance drop is a poor predictor of adequate RFC delivery due to its large variability.^1,7 Additionally, for high AI – around 500 as recommended for the anterior wall of the LA – FTI has a wide dispersion, thereby limiting the clinical applicability of FTI.

The principal shortcoming of FTI is its linearity, which does not correspond to the mechanism of lesion formation, which is best approximated in a non-linear formula, incorporating power.¹ These requirements are met by AI, with the respective target values according to the anatomical segment.^11,20

Prevention of low AI has the potential to improve efficacy, whereas avoiding high AI may improve safety. Integration of AI into clinical routine may be useful to achieve more homogeneous lesion formation and allow for more standardized ablation procedures.

We therefore suggest the adoption of AI as a real-time feedback parameter as a clinical standard.

Study Limitations

The present evaluation is a retrospective and observational non-randomized study with a limited number of patients. Randomized studies with a larger sample size would be necessary to assess PVI outcome when adopting AI as a procedural feedback tool. In the present study, an assessment of conduction gaps during waiting time or of dormant conduction after PVI was not performed.

Conclusions

When displayed to the operator, a larger proportion of ablation points are located in the defined target range, thereby avoiding low AI, potentially resulting in ineffective ablation, as well as excessively high AI, potentially resulting in complications such as steam pop formation, cardiac perforation or collateral damages such as atrio-esophageal fistula formation.

Given that the non-linear AI incorporates RFC power, it provides more information on lesion formation during PVI compared with the linear FTI formula. AI and FTI do correlate, but, given that the FTI for a given AI ranges widely, the value of linear FTI in clinical practice is questionable.

Acknowledgment

We would like to thank Marion Stahl and Niklas Scholz for their support with the Carto 3D mapping system.

Particularly, we would like to thank Dr. Roza Meyer-Saraei for her support in the management of our institutional database.

Disclosures

The University Heart Center Hamburg has received a travel grant from Biosense Webster.

References

• 1.   Nakagawa H, Jackman WM. The role of contact force in atrial fibrillation ablation. J Atr Fibrillation 2014; 7: 1027.
• 2.   Dagres N, Hindricks G, Kottkamp H, Sommer P, Gaspar T, Bode K, et al. Complications of atrial fibrillation ablation in a high-volume center in 1,000 procedures: Still cause for concern? J Cardiovasc Electrophysiol 2009; 20: 1014–1019.
• 3.   Phlips T, Taghji P, El Haddad M, Wolf M, Knecht S, Vandekerckhove Y, et al. Improving procedural and one-year outcome after contact force-guided pulmonary vein isolation: The role of interlesion distance, ablation index, and contact force variability in the ‘CLOSE’-protocol. Europace 2018; eux376, https://doi.org/10.1093/europace/eux376.
• 4.   De Potter T, Van Herendael H, Balasubramaniam R, Wright M, Agarwal SC, Sanders P, et al. Safety and long-term effectiveness of paroxysmal atrial fibrillation ablation with a contact force-sensing catheter: Real-world experience from a prospective, multicentre observational cohort registry. Europace 2018; eux290, https://doi.org/10.1093/europace/eux290.
• 5.   Neuzil P, Reddy VY, Kautzner J, Petru J, Wichterle D, Shah D, et al. Electrical reconnection after pulmonary vein isolation is contingent on contact force during initial treatment: Results from the EFFICAS I Study. Circ Arrhythm Electrophysiol 2013; 6: 327–333.
• 6.   Kautzner J, Neuzil P, Lambert H, Peichl P, Petru J, Cihak R, et al. EFFICAS II: Optimization of catheter contact force improves outcome of pulmonary vein isolation for paroxysmal atrial fibrillation. Europace 2015; 17: 1229–1235.
• 7.   Makimoto H, Lin T, Rillig A, Metzner A, Wohlmuth P, Arya A, et al. In vivo contact force analysis and correlation with tissue impedance during left atrial mapping and catheter ablation of atrial fibrillation. Circ Arrhythm Electrophysiol 2014; 7: 46–54.
• 8.   Chelu MG, Morris AK, Kholmovski EG, King JB, Kaur G, Silver MA, et al. Durable lesion formation while avoiding esophageal injury during ablation of atrial fibrillation: Lessons learned from late gadolinium MR imaging. J Cardiovasc Electrophysiol 2018; 29: 385–392.
• 9.   Koruth JS, Iwasawa J, Enomoto Y, Bar-Tal M, Ultchin Y, Sigal A, et al. Chamber-specific radiofrequency lesion dimension estimation using novel catheter-based tissue interface temperature sensing. JACC Clin Electrophysiol 2017; 3: 1092–1102.
• 10.   Hussein A, Das M, Chaturvedi V, Asfour IK, Daryanani N, Morgan M, et al. Prospective use of ablation index targets improves clinical outcomes following ablation for atrial fibrillation. J Cardiovasc Electrophysiol 2017; 28: 1037–1047.
• 11.   Das M, Loveday JJ, Wynn GJ, Gomes S, Saeed Y, Bonnett LJ, et al. Ablation index, a novel marker of ablation lesion quality: Prediction of pulmonary vein reconnection at repeat electrophysiology study and regional differences in target values. Europace 2016; 19: 775–783.
• 12.   El Haddad M, Taghji P, Phlips T, Wolf M, Demolder A, Choudhury R, et al. Determinants of acute and late pulmonary vein reconnection in contact force-guided pulmonary vein isolation. Circ Arrhythm Electrophysiol 2017; 10: e004867.
• 13.   Hussein AA, Barakat AF, Salimba WI, Tarakji KG, Bassiouny M, Baranowski B, et al. Persistent atrial fibrillation ablation with or without contact force sensing. J Cardiovasc Electrophysiol 2017; 28: 483–488.
• 14.   Ullah W, Hunter RJ, Finlay MC, McLean A, Dhinoja MB, Sporton S, et al. Ablation index and surround flow catheter irrigation. JACC Clin Electrophysiol 2017; 3: 1080–1088.
• 15.   Calkins H, Hindricks G, Cappato R, Kim YH, Saad EB, Aguinaga L, et al. 2017 HRS/EHRA/ECAS/APHRS/SOLAECE expert consensus statement on catheter and surgical ablation of atrial fibrillation: Executive summary. J Interv Card Electrophysiol 2017; 50: 1–55.
• 16.   Nath S, Haines DE. Biophysics and pathology of catheter energy delivery systems. Prog Cardiovasc Dis 1995; 37: 185–204.
• 17.   Makimoto H, Heeger CH, Lin T, Rillig A, Metzner A, Wissner E, et al. Comparison of contact force-guided procedure with non-contact force-guided procedure during left atrial mapping and pulmonary vein isolation: Impact of contact force on recurrence of atrial fibrillation. Clin Res Cardiol 2015; 104: 861–870.
• 18.   Shah DC, Lambert H, Nakagawa H, Langenkamp A, Aeby N, Leo G. Area under the real-time contact force curve (force-time integral) predicts radiofrequency lesion size in an in vitro contractile model. J Cardiovasc Electrophysiol 2010; 21: 1038–1043.
• 19.   Calzolari V, De Mattia L, Indiani S, Crosato M, Furlanetto A, Licciardello C, et al. In vitro validation of the lesion size index to predict lesion width and depth after irrigated radiofrequency ablation in a porcine model. JACC Clin Electrophysiol 2017; 3: 1126–1135.
• 20.   Chikata A, Kato T, Sakagami S, Kato C, Saeki T, Kawai K, et al. Optimal force-time integral for pulmonary vein isolation according to anatomical wall thickness under the ablation line. J Am Heart Assoc 2016; 5: e003155.