Pitfalls and potential for the use of computational modeling to guide the treatment of atrial fibrillation

Atrial fibrillation (AF) is the most common arrhythmia, with an increasing incidence in the developed world due to concomitant aging of the population and an increased incidence of obesity. Obesity promotes an array of cardiovascular conditions that includes hypertension, sleep apnea, autonomic dysfunction, heart failure and atrial fibrillation. In the past 30 years, conceptual advances in understanding AF led to two major interventional approaches for treating AF. With the work of Moe, Allessie and others, the re-entrant wavelet hypothesis of AF focused attention on the relation of atrial size with the wavelength of electrical activity, where wavelength is the product of refractory period and conduction velocity¹. Shortening of wavelength facilitates reentry, thus increasing arrhythmia persistence. This theory contributed directly to the development of the Maze procedures in which the goal of the surgery was to provide a conduction path from the sinoatrial node to the atrioventricular node while creating lines of block that both isolated the pulmonary veins and minimized the area in which reentry could occur. In the first iteration of the Maze procedure, surgeons created a lesion in the cavotricuspid isthmus (CTI, the region between the vena cava and the tricuspid valve in the right atrium). Some patients treated in this manner could not adequately increase heart rate in response to exercise. In subsequent iterations of the Maze procedure, this lesion was omitted and patients retained their ability to increase heart rate. While the Maze procedure has had a significant impact on AF burden in treated patients, it never gained widespread adoption due to the significant time and surgical expertise required.

Catheter ablation of AF was developed based on the observations of Haissaguerre and colleagues who found that ectopic activity originating in the pulmonary veins was very often the trigger of AF episodes². This observation led to the development of circumferential isolation of the pulmonary veins (PVI) as a first-line strategy for the treatment of AF. As PVI alone seems to have an upper limit of efficacy of ~70–80%, and rates of success in patients with persistent AF have had lower efficacy (~50%). Efforts to enhance procedural efficacy have been widely sought.

In this issue, Lim and colleagues report on efforts to improve AF ablation outcomes using radiofrequency catheter ablation³. This study included a single center clinical series of 846 AF cases in which pulmonary vein isolation (PVI) was performed in all cases; supplemental interventions also performed ablation of interatrial conduction with or without isolation of the cavo-tricuspid isthmus. The authors reported a larger change in the dominant frequency of activation, a higher rate of AF termination, and lower recurrence of AF at ~2 years of followup in the patients whose procedure included CTI ablation. In addition to the clinical analysis, the team used a combination of 3D modeling of patient-specific atrial electro-anatomic maps from 10 paroxysmal AF patients, plus analysis of CT imagery and analysis of clinical ablation results. Modeling enabled the authors to systematically test varied ablation strategies in silico.

This is a timely paper that addresses a complex topic in a logical and sophisticated manner. It builds on the literature which has documented a higher dominant frequency in the LA than RA. Given the history of paroxysmal AF in all 10 of the mapping patients used as inputs for modeling studies, it is not surprising that the atria were nearly normal in size. However, study of only atria of normal size limits the generalizability of the reported analyses, as most AF patients have enlarged left atria. The authors attempt to address this limitation by use of propensity matching of their larger clinical dataset which compares PVI alone with PVI plus cavotricuspid isthmus (CTI) ablation. In both their propensity matched analysis and in silico modeling, the authors report that additional targeting of the CTI improved both immediate and longer term (2 year) AF ablation success.

The results of this study are discordant with much of the literature. A recent meta-analysis of 1400 patients undergoing PVI vs. PVI+CTI isolation identified no additional benefit of the CTI ablation⁴. This analysis included a small study in which the prevalence of persistent AF and enlarged left atrial volume was higher in the CTI than in the PVI group⁵. The authors of that study concluded that “CTI block ablation in addition to PVI for AF patients without a history of AFL or inducible AFL during ablation may not improve the clinical outcome of AF ablation in the patients with larger LA volume, nonparoxysmal AF, or post-PVI inducible AF.” Thus, differences in patient selection in the Lim study may account for the differences in outcome from other published analyses and meta-analyses.

Many of the current clinical approaches to treating AF are empirically driven, seeking to use the same approach and endpoints for all patients. However, computer modeling of individualized atrial electrogram recordings and structure (size, fibrosis burden and locations) are already being systematically implemented to guide personalized AF ablation treatments⁶. The study by Lim and colleagues suggests that personalized treatment in the subset of patients with atria of normal or smaller than normal dimensions may benefit from CTI ablation during catheter ablation. It is important to recognize that many of the risk factors for AF (obesity, hypertension, sleep apnea) are associated with left atrial enlargement, and thus, use of adjuvant CTI ablation may not be broadly applicable. The thoughtful combination of individualized imaging and computational modeling has potential to guide treatment approaches, improve treatment outcomes and shorten procedure times.