Incidence, epidemiology, diagnosis and prognosis of atrio-oesophageal fistula following percutaneous catheter ablation: a French nationwide survey

Abstract

Aims

Rate, incidence, risk factors, and optimal management of atrio-oesophageal fistula (AOF) after catheter ablation for atrial fibrillation (AF) remain obscure.

Methods and results

All French centres performing AF ablation were identified and surveys were sent concerning the number of procedures, eventual cases of AOF, and characteristics of such cases. Eighty-two of the 103 centres (80%) performing AF ablation in France were included, with a total of 129 286 AF ablations since 2006 (93% of the whole procedures in France). Thirty-three AOF were reported (reported rate 0.026% per procedure) with a stable reported annual incidence despite the increasing number of procedures. Sensitivity of computed tomography (CT) scan for AOF was 81%. Mortality was 60%, significantly lower in case of surgical corrective therapy (31 vs. 93%, P = 0.001).

Conclusion

The reported rate of AOF after AF ablation in this nationwide survey was 0.026%, with a stable reported annual incidence over time. A normal CT scan does not rule out the diagnosis and should be repeated in case of suspicion. Prognosis remains poor with a mortality of 60% and crucially dependant of immediate surgical correction. No clear protective strategy has been proven effective.

Introduction

While recent trials demonstrate some reduction in mortality after catheter ablation (CA) for atrial fibrillation (AF) compared to drug therapy,^1,2 the procedure still entails significant complication and mortality rates.³

Atrio-oesophageal fistula (AOF) represents an exceptional but dramatic complication. First cases of AOF were described in 2004 for radiofrequency (RF)⁴ and 2012 for cryoablation⁵ and around 50 cases have been published to date.⁶ Rate of AOF occurrence is reported to range between 0.02% and 0.1% of CA^7–10 in non-exhaustive registries, but true rate and incidence remains obscure. Risk factors for AOF are still undetermined or debated as well. Mortality is high, ranging from 50% to 70%,^6,9,11 and even more in the absence of surgical therapy.^6,12

The aim of this study was to evaluate the rate, yearly incidences, epidemiology, clinical presentation, diagnosis and prognosis of AOF after CA, based on a nationwide retrospective analysis of AF CA procedures performed in France over the last 13 years.

Methods

This nationwide centre-based survey aimed to collect the most comprehensive data on declared cases of AOF for the whole French territory between 2006 and 2019. The following steps were performed.

Establishing full list of centres performing AF catheter ablation in France

All French public and private centres practising (or having practised) percutaneous CA for AF or left atrial tachycardia (AT) were identified via professional networks and PMSI databases (see below).

Assessment of total and annual number of AF/AT ablation procedures in France and in each centre from January 2006 to June 2019

The total and annual number of AF/AT ablation procedures from January 2009 to December 2018 throughout the French territory was obtained from the French National Hospital Database (PMSI). These are official data established by the Ministry of Health from declarations of each centre, accessible in open-access (https://www.scansante.fr/open-ccam-2018), providing an ‘as exhaustive as possible’ data, since linked to reimbursement. Internal codings (DEPF033, DEPF014, DENF014, and DENF018) were used for selecting procedures of AF or left AT ablations. Number of AF/AT ablations performed between 2006 and 2009 was estimated from the available PMSI data for 2009 taken into account the increase in activity between 2006 and 2009. The number of CA for the first 2019 semester was estimated from the 2018 data as well.

The number of annual ablations for each centre was estimated from the surveys (see below) and additionally verified using PMSI data. The annual number of AF/AT ablations for each centre was available from PMSI database between 2012 and 2018. Between 2006 and 2012, this annual number for each centre was estimated using data from 2012 to 2018 assuming similar trends for each centre. For the first 2019 semester, this number was estimated for the included centres as 93% of the whole French AF/AT ablations (see further).

Evaluation of ablation techniques and counting of AOF

Two surveys were sent in June 2019 to all centres practising AF/AT ablations as previously identified. The first survey concerned the local centre habits for AF/AT ablation procedures (number, technique, and perioperative management), while the second collected any case(s) of proven AOF the centre had eventually to deplore. Proven cases of AOF were defined by formal identification of AOF using computed tomography (CT) scan, autopsy, or surgery.

According to the French law, studies based on the exploitation of PMSI data do not need an ethics committee but have to be covered by reference methodology of the French National Commission for Informatics and Liberties (CNIL). After evaluation, this study completing all the criteria is registered at the Toulouse University Hospital (number RnIPH2020-35) and cover by the MR-005 (CNIL number: 2206725 v 0).

Results

Total number of AF/AT ablation procedures in France between 2006 and June 2019

One hundred and eleven centres practising AF/AT ablation were identified based on professional networks and PMSI databases, taken account of some gatherings over the studied period of time. From these, eight were excluded, either because they stopped any activity before 2015 or because they began AF/AT ablation activity in 2018, so their activity was considered minor (total 222 procedures for the eight centres). Thus, a total of 103 centres have been contacted (63 public and 40 private hospitals).

Of them, 82 fulfilled the surveys (80%) and were finally included in the study: all French university hospitals (n = 34), 78% of general hospitals (n = 18), 83% of private hospitals with public mission (n = 5), and 62.5% of private institutions (n = 25). Ninety-seven percent of centres performing >200 AF ablation yearly (n = 35) were included, as well as 86% of those with 100–200 annual procedures (n = 24), 72% of centres with 50–100 ablations per year (n = 18), and 36% with <50 ablations annually (n = 5). The full list of centres included in this study is available in Supplementary material online, Files. Among the 21 centres which did not answer, only one performed >200 ablations yearly and nine performed <50 procedures per year.

Based on PMSI database, a total of 124 924 AF/AT ablations have been identified in France from January 2009 to December 2018. AF/AT ablations performed between 2006 and 2009 were estimated at n = 1500 from the available number for 2009 (n = 668) (see Methods section). Although rough, this estimation was anyway negligible when compared with the whole number of procedure (>120 000) performed thereafter. The estimated ablation number from January to June 2019 was n = 13 000 based on 2018 data, leading to a total of 139 424 AF/AT CAs in France between 2006 and mid-2019.

The number of procedures for the 82 included centres was estimated through the number of AF/AT ablations declared in the surveys (see Methods section). The 82 centres declared a total of 129 286 ablations between 2006 and mid-2019, which was used for calculations and analysis. This number represented 93% of the whole number of procedures performed in France during the same period of time as evaluated above. This number was much close to the total number of AF/AT ablation procedures for the 82 included centres based on available PMSI data and additional estimations (see Methods section) (n = 129 152 procedures). This second calculation reinforced the robustness of our estimations.

Reported rate and incidence of atrio-oesophageal fistula

Thirty-three AOF were declared by 24 of the 82 included centres (29%) between 2006 and mid-2019. The reported rate of AOF was therefore 0.026% [33/129 286; 95% confidence interval (CI) (0.02–0.03)], i.e. 25 for 100 000 procedures. All reported AOF occurred between 2006 and 2019.

The reported incidence of AOF between 2009 and 2018 is displayed in Figure 1. The incidence appears stable since 2010, despite the exponential increase in the number of procedures, with values ranging between 0.007% and 0.043% [95% CI (0.00–0.06%)]. The reported rate of AOF was similar regarding centre volumes (0.024% for centres with 100–200 ablations per year and 0.029% for centres with >200/year). There was no reported case of AOF for centres with <100 ablations per year.

Patient and procedural characteristics

The detailed clinical characteristics of patients who developed AOF were available for 30 cases (Table 1). Eighty-three percent were men, with a mean age of 60 ± 11 years old. Structural cardiac disease was present in 40% (n = 12). Mean left ventricular ejection fraction was 55 ± 11%. Ten patients presented with paroxysmal AF (33%), 19 with persistent AF (63%), and one with a left AT (3%). Only two patients (7%) had undergone a previous AF ablation. Most of them had a dilated left atrium (81%). A previous history of oeso-gastric pathology was noted only in four patients (13%) and 20% were on proton pump inhibitor (PPI) before ablation.

Characteristics of ablation procedures are detailed in Table 2. All cases of AOF occurred after RF ablation, no case was reported after cryoablation. Catheter ablation was performed under general anaesthesia in 22 (73%) and a thermal probe was used in 6 (20%). Among them, excessive heating was reported only in two (left pulmonary veins in one and right inferior pulmonary vein in one). CA was performed using 3D navigation systems in all, with contact-force ablation catheters in 13 (43%), remote magnetic navigation in one and robotically steered sheath in two. Pulmonary vein isolation (PVI) was performed in each case. Ablation within the left atrial posterior wall was performed in 7 (23%) and additional linear ablation (roof or more posterior) in 11 (37%). Proton pump inhibitors were prescribed in 16 (53%) after the procedure.

Patient presentation and outcome

The mean delays between procedure and onset of symptoms and between symptoms and diagnosis were 21 ± 11 days (1–43) and 6 ± 6 days (0–23), respectively. One AOF occurred at Day 1 (robotically steered sheath, cardiogenic shock, and intra-cardiac air bubbles) and one oesophageal perforation happened at Day 4 evolving towards oesophago-pericardial fistula. AOF was immediately suspected and confirmed in 15 patients (50%). In the other cases, the diagnosis was delayed because other diagnoses were initially suspected, mainly endocarditis or digestive pathology. The main initial symptoms at admission were fever (n = 19, 63%), chest pain or odynophagia (n = 14, 47%), and neurological signs (stroke or seizures) (n = 7, 23%) either isolated or associated. Other initial symptoms were AF recurrence (13%), pericardial effusion (13%), peripheral embolism (7%), or confusion (7%) (Figure 2). Digestive bleeding was never observed at admission. Delayed complications were stroke or cerebral haemorrhages (n = 19, 63%), severe sepsis (n = 16, 53%), coma (n = 13, 43%), cardiac arrest (n = 5, 17%), tamponade (n = 3, 10%), and minor digestive bleeding in 2 (7%) (Figure 3).

Investigations allowing AOF diagnosis were thoracic CT scan in 22 out of 30 AOF cases (73%), CT scan with oesophageal opacification in 4 (13%), oeso-gastric endoscopy in 2 (7%), cerebral CT scan in one (3%) showing intracranial air bubbles. In one patient, air bubbles were seen in the left atrium using transthoracic echocardiography (TTE). Computed tomography scan was positive in 22 out of the 27 performed CT (sensitivity 81%) and CT scan with oesophageal opacification in four out of six such investigations (sensitivity 67%). There was no abnormality at initial TTE in 21/22 patients (95%), at first cerebral imaging in 18/19 (95%), and at initial oeso-gastric endoscopy, thoracic CT scan and CT scan with oesophageal opacification in 6/8 (75%), 5/27 (18%), and 2/6 (33%) patients, respectively. Repeated thoracic CT was performed in three instances at 3, 6, and 7 days after the initial negative one and were positive in two cases and doubtful in the latter. Two patients underwent abdominal surgical investigation and one transoesophageal echocardiography. Final diagnosis was AOF in 26 patients and oesophago-pericardial fistula in 4.

Twelve patients (40%) could not beneficiate from interventional therapy because of their critical condition and 11/12 died (the only survivor had severe neurological sequellaes). Isolated oesophageal stenting was realized in two patients (7%) who died afterwards. Surgery alone was performed in nine patients (30%) and seven survived (78%), while combined therapy with stenting followed by surgery was performed in seven patients (23%) and four survived (57%) (Figure 4). Total mortality was 60% (n = 18). The four patients with isolated oesophago-pericardial fistula underwent surgery (two with stenting also) and none died. Mortality was significantly lower in case of surgical therapy (31%) compared to no therapy or isolated stenting (93%) (P = 0.001) (Figure 5). The mean hospitalization duration was 39 ± 42 days (2–180).

Discussion

Epidemiology of atrio-oesophageal fistulae

This nationwide survey provides important data to better understand the epidemiology and outcome of AOF occurring after AF/AT ablation. The survey response rate was important, corresponding to 80% of centres and representing 93% of all AF/AT ablations performed since 2006 in France, providing as exhaustive as possible data at a national level.

Thirty-three AOF have been declared. We found an AOF reported rate of 0.026%, close to the rate of 0.016% reported in the 2016 international survey (including 191 215 ablations)⁹ and in the 2008 US survey (20 245 ablations) which reported an AOF rate of 0.03%.⁸ Higher rates were reported in the Canadian survey (0.07%) and Korean one (0.15%) but they included only 7016 (10) and 6724 procedures from three centres.¹³

We should acknowledge that AOF rate should only be considered as a rough estimation, because of the inherent bias linked to the retrospective design, with a risk of underreporting and/or underdiagnosis. Underreporting is probably not the main issue but underdiagnosis is likely more relevant because of the poor knowledge of such complication in the medical community. An AOF incidence of 0.2–0.4% in 1000 patients with regular follow-up was reported in 2009.¹⁴ This means that true incidence would be 5–10 times more compared to surveys. However, even if underestimated, AOF remains a very rare complication of AF ablation but most probably the common cause for early death after AF ablation. In this aspect, even if few cases were probably missed, this would not have dramatically changed the rate and incidence.

The reported annual incidence of AOF appears stable over time, ranging from 0.007% to 0.043% between 2012 and 2018. The large majority of AOF occurred in the course of the first ablation, which is concordant with the Korean registry,¹³ whereas redo procedures accounted for 55% of AOF cases in the 2016 international registry.⁹ This was not reported in other surveys or they include only first ablations. All but one occurred after AF ablation (only one after AT ablation). Occurrence of AOF after left AT ablation has not been studied before. This may mean that the lesions performed during ablation of left AT are less prone to AOF, or that AT ablations were mainly redo procedures, which are probably less risky regarding occurrence of AOF, possibly due to late fibrosis isolating the vulnerable structures involved in AOF. It is also possible that redo procedures imply less amount of ablation at the posterior wall. Anyway, AOF may happen after redo or AT ablation as well, even if more exceptional.

Patient presentation and outcome

We report a high mortality rate (60%), which is concordant to what was reported in a meta-analysis on 57 AOF cases (57%),⁶ but somewhat less than the 80% mortality reported in the 2016 international survey.⁹ Our data clearly show that mortality is significantly lowered when patients were treated with surgery, which is concordant with previous data.^9,11 Oesophageal stenting for AOF alone is clearly insufficient and should not be considered without subsequent surgery. Prognosis of oesophago-pericardial fistula seemed quite better than AOF when operated, and we do not have sufficient data supporting a simple stenting for oesophago-pericardial fistula.

The delay of apparition of the first symptom(s) was highly variable (from 1 to 43 days). Such early or late delays were already mentioned before.^6,8 Two cases happened during the first week and even the day after the procedure in one case. Thus AOF may happen at any time during the two months after ablation. If the initial presentation mainly included fever, chest pain, or odynophagia, secondary most frequent complications are neurological with stroke, coma, and septic shock.

One important finding is the excessive delay between onset of symptoms and correct diagnosis (mean 6 days, up to 23 days). An initial wrong diagnosis was made in half of the patients. This highlights the fact that this rare—but often fatal—complication is probably under-recognized by physicians. Warning signs such as fever, neurologic, or embolic complications, especially if occurring in combination with chest pain, should raise suspicion and should lead to additional investigations including thoracic CT scan with/without scan with oesophageal opacification. Achieving an early diagnosis is probably crucial in the management of AOF to propose early surgery before severe complications occur putting the patient beyond all therapeutic options. Most diagnoses were made using thoracic CT without (73%) or with (13%) oesophagus opacification. Oesogastric endoscopy is contraindicated in case of suspected AOF because of the risk of air embolism. Sometimes, CT had to be repeated to achieve the correct diagnosis, thus sensitivity of CT is far from perfect. In case of unexplained fever/chest pain days/weeks after AF ablation, this diagnosis should be considered and thoracic CT and even CT scan with oesophageal opacification should be performed¹¹ excluding any other investigation, and quickly repeated in case of high suspicion or lack of a clear different diagnosis.

Risk factors associated with atrio-oesophageal fistulae

Pathophysiology of AOF after CA remains poorly understood. If oesophageal lesions following AF ablation are frequent (18% of patients, 6% with ulcerations)¹⁵ only a minority of lesions (10%) will evolve to oesophageal perforation and even less to AOF. Some hypothesis have been proposed to explain the formation of AOF, such as direct thermic lesions of the oesophagus, ischaemic oesophageal injury after occlusion of the anterior oesophageal arteries, gastro-oesophageal reflux exacerbation, decrease in oesophageal motility by nervous lesions, or general anaesthesia.¹¹

Due to the very low incidence of this severe complication, the identification of risk and protective factors remains very difficult to assess. Several risk factors have been reported as the use of general anaesthesia, type of energy, power and temperature settings, amount of ablation, and ablation within the posterior wall.¹¹ In this survey, analysis of risk factors could have only been based on comparisons between centres with/without reported AOF, suffering evident limitations, but some factors are worth to be discussed in view of the reported cases of AOF.

Considering the type of energy, all reported AOF occurred after RF ablation and not any case was reported with cryoablation in this survey. Possible underreporting by low-volume centres practising cryoablation may explain this apparent lack of cases. In the Canadian and the US survey, all reported AOF occurred after RF,^8,10 but cryoablation was responsible for 7% of oesophageal perforations in the 2016 international survey.⁹ Cases of AOF after cryoablation have been reported.^5,11 Oesophageal lesions are found in up to 20% of cases after second-generation cryoballoon ablation,¹⁶ which is similar to the rate observed after RF ablation.¹⁵ Rates of AOF were evaluated to 0.009% with the first and 0.014% with the second-generation cryoballoon.¹⁷ Eleven cases of AOF were identified from a worldwide experience since 2011 (rate 0.001%).¹⁸ Interestingly, they were always related to the left inferior pulmonary vein.¹⁸ Thus, there is no reason to believe that cryo protects against fistula. If the risk of AOF is lower than with RF remains to be demonstrated and this complication has to be considered with cryoablation with the same attention as with RF.

Another debated point is the efficacy of oesophageal thermal probe to prevent AOF. To date, the efficacy of monitoring oesophageal temperature has not been proven and is still subject to controversies¹¹ and has even been suspected to increase the risk for AOF.¹⁹ Two meta-analysis did not find any significant relationship between the occurrence of oesophageal lesions after AF ablation and the use of a thermal probe.^20,21 In the 2016 multicentre survey, 75% of AOF occurred despite oesophageal temperature monitoring, suggesting a lack of efficacy of this strategy.⁹ In the present survey, a thermal probe was used in only a minority of reported cases, while not altering reported incidence of AOF (Figure 1). However, specificities of the probes and temperature cut-offs used were not available and hinder any formal conclusion in this aspect. In our survey, half of AOF occurred despite PPI, showing incomplete efficacy in preventing AOF. Although largely prescribed, we lack evidence on the benefit of systematic prophylactic prescription of these drugs.

Another interesting finding was the association between the use of catheters without contact-force monitoring and the occurrence of AOF. When looking to the incidence of AOF over time, we observed a decrease in the reported annual incidence after 2009, which is concomitant with the advent of contact-force catheters on the French market. Contact-force monitoring is expected to avoid excessive force on atrial walls during RF application, but may at the same time optimize and increase catheter contact to atrial tissues. However, this is only speculative and the apparent decrease in incidence by using contact-force catheter may simply be explained by the lower number of ablations before 2009 rendering any AOF more significant regarding the incidence. Again, it is difficult to conclude if contact-force catheter may be associated with better safety regarding the risk of AOF. RF power is another parameter associated to contact-force for explaining lesion size and collateral injury, but the mean power used at the posterior wall in reported cases of AOF in this survey was unremarkable (26 ± 3 W). Finally, we noted two AOF cases after robotically steered AF ablation. Even if the total number of such ablations was not available, two AOF can be regarded as a high incidence due to the assumed relatively low number of these procedures.

The type of lesions performed during the ablation procedure and individual factors could also be associated with AOF occurrence. Most of AOF cases were persistent AF (63%) with dilated left atria (88%). Around 40% of patients had a roof/posterior line ablation in addition to PVI, which is similar to what is reported in the Korean Registry.¹³ However, in previous registries, only a minority of patients had additional linear ablation and most AOF occurred after PVI only. In the Korean registry, no association between additional lines/defragmentation was found and all AOF were located near the left inferior pulmonary vein,¹³ suggesting that this area, just like for cryoablation,¹⁶ could be the more vulnerable. Due to the lack of precise data, we could not confirm or infirm this point in our survey. Finally, the use of general anaesthesia has been reported as a risk factor in previous studies^10,21 and had been used in 73% of AOF cases in this survey, but it is impossible to draw any conclusion about the true risk of general anaesthesia from this study.

Finally, the rates of AOF did not change according to the centre volumes (and were even absent in very low-volume centres probably because of its rarity), meaning that this dramatic complication is probably not dependent on the physician experience and that everybody may experience AOF if doing AF ablation.

Limitations

Although as exhaustive as possible, this survey was retrospective and based on voluntary declaration by the centres/physicians. Although currently undetermined, the true rate of AOF is probably higher, since it is likely that some cases were ignored, either because AOF occurred at a time when physicians were not aware of this diagnosis, or either patients were simply not referred to cardiologists. The amount of such non-declared cases is of course impossible to estimate, but, as already discussed above, unreported lethal complications after CA ablation are nowadays uncommon and most AOF have been probably counted. Thus, reported incidence and rate as determined in this survey, even if approximate, seems to constitute a reasonable appreciation of the risk of AOF.

Conclusion

This French nationwide survey reports a rate of 0.026% of AOF after AF/AT ablation, which is probably lower than the true incidence. This unpredictable but often fatal complication needs to be better recognized at earlier stage to improve prognosis by proposing early surgery. No clear protective strategy has been proven to be effective. The development of new alternative energies, like electroporation, offers promising perspectives to avoid this complication.

Supplementary material

Supplementary material is available at Europace online.

Acknowledgements

The authors thank all the physicians and institutions having answered to the surveys. Hervé Poty, Michael Peyrol, Linda Koutbi, Fabrice Extramiana, Eloi Marijon, Vincent Algalarrondo, Nicolas Lellouche, Julien Laborderie, Gabriel Lactu, Yann Dagher-Hayeck, François Jourda, Antoine Dompnier, Laurent Liprandi, Pierre Socie, David Huchette, Antoine Deplagne, Alexandre Guignier, Olivier Barthez, Noura Zannad, Nicolas Delarche, Georges Nadji, Jérôme Taieb, Alain Lebon, Sophie Lepage, Marc Badoz, Jacques Mansourati, Grégoire Massoulie, Paul Milliez, Jean Sylvain Hermida, Frederic Sacher, Gabriel Laurent, François Brigadeau, Fabrice Demoniere, Jean Marc Sellal, Jean Baptiste Gourraud, Guillaume Theodore, Rodrigue Garcia, François Lesaffre, Raphaël Martin, Laurence Jesel, Laurent Fauchier, Pascal Defaye, Benoit Guy Moyat, Jean Luc Pasquie, Pierre François Winum, Frederic Anselme, Gael Clerici, Antoine da Costa, Samir Fareh, Agustin Bortone, Babé Bakouboula, Franck Raczka, Stéphane Combes, Jean Louis Lloret, François Xavier Hager, Philippe Ricard, Zida Khoueiry, Franck Halimi, Olivier Piot, Claude Mariottini, Bruno Cauchemez, Walid Amara, Yves Guyomar, Paul Bru, Philippe Chevalier, Henri Broustet, Julien Seitz, Géraldine Vedrenne, Jérôme Lacotte, Nicolas Mignot, Daniel Gras, Antoine Roux, Philippe Poret, Alexis Mechulan, Raphael Sandras, Nicolas Detis, Thierry Chalvidan, Frederic Victor, Denis Raguin.

Conflict of interest: none declared.

Data availability

Data available on request at [email protected].

References

Author notes