Predicting irregularity: computational modelling of haemodynamics in atrial fibrillation

Atrial fibrillation (AF) is the most common cardiac arrhythmia and significantly increases the risk of stroke and heart failure, while also impairing patients’ quality of life (Charitakis et al., 2015). AF is characterised by chaotic, high-frequency excitation of the atria, resulting in uncoordinated atrial contractions and rapid, irregular ventricular activity. While an increased heart rate alone can reduce cardiac output in AF patients, the variability and irregularity of ventricular activity associated with AF have been shown to further reduce cardiac output independent of heart rate (Plappert et al., 2025). In normal sinus rhythm, atrial contraction (‘atrial kick’) contributes up to 20–30% of ventricular filling. During AF, the loss of this coordinated atrial boost and the highly variable beat-to-beat filling times both act to reduce cardiac output (Anselmino et al., 2022). These AF-induced haemodynamic abnormalities lead to a range of clinical symptoms, including palpitations, dizziness and fatigue. An important unresolved issue in AF is the wide variation in symptom severity (Charitakis et al., 2015), which is closely tied to the underlying haemodynamic variability of the disease. Understanding and predicting the haemodynamic impact of AF is challenging because it arises from a complex interplay of factors. Clinical data on the haemodynamic consequences of AF are also limited and often inconsistent. This is largely due to the difficulty of obtaining precise measurements in specific vascular regions and the presence of numerous clinical confounders, particularly hypertension, that limit the ability to investigate arrhythmia-specific effects (Adeodu et al., 2025; Anselmino et al., 2022). In this context, computational models provide a valuable tool for simulating AF in a controlled manner, which allows for the examination of its effects on the circulation and to guide personalised treatment strategies. Through in silico modelling, it becomes possible to isolate the arrhythmia's direct impact without the influence of other comorbidities and to simultaneously observe multiple haemodynamic variables that are challenging to measure in vivo. Despite these advantages, developing an effective computational model of AF remains difficult. The challenge stems from the need to integrate electrophysiology and haemodynamics across multiple spatial and temporal scales during AF. Finite element models are computationally expensive and typically not suited for simulating the many heartbeats required for AF-related investigations, while reduced-order models lack an electrophysiological description of the uncoordinated and irregular activation times characteristic of AF. In this issue of The Journal of Physiology, Plappert and colleagues present a computational model that addresses key concerns for effective AF simulation (Plappert et al., 2026). Their framework includes a description of atrial and ventricular activation integrated with a Windkessel circulatory model, enabling continuous haemodynamic simulation over many heartbeats to capture beat-to-beat variation and to efficiently support patient-specific parameter fitting. In their framework, the authors incorporate an electrical subsystem that generates the uncoordinated atrial and irregular ventricular activation patterns characteristic of AF, alongside a mechanical subsystem that simulates circulatory haemodynamics. Building on their previous work (Oomen et al., 2022), they introduce an enhanced electrophysiological component that produces realistic activation sequences and adapt the representation of the atrial walls to capture the behaviour of fibrillating atria. Specifically, the model employs a TriSeg formulation with a MultiPatch approach, using a relatively high number of patches for the atria to enable uncoordinated activation through a stochastic timing strategy, which serves as a surrogate in the absence of detailed cardiac activation maps. This integrated approach enables the model to reproduce the rapid, disorganized atrial activity seen in AF and to predict ventricular rate and rhythm, including heart rate, RR-interval variability, and irregularity. The authors also use patient-specific data to calibrate both the electrical and mechanical subsystems. To replicate AF behaviour, they performed patient-specific model fitting using electrical activation patterns derived from patient electrocardiogram and intracardiac electrogram data. In this study, they modelled cardiac performance in both normal sinus rhythm (NSR) and AF, and compared the simulation results, including pressure and volume trends, with the corresponding clinical values. Their results demonstrate that the proposed model closely reproduced arterial and intracardiac pressure measurements in both NSR and AF. The model also accurately captured ejection fraction metrics in NSR, showing strong agreement with clinical observations. When compared with models incorporating alternative atrial contraction patterns during AF, the proposed framework consistently produced the lowest pressure simulation errors and generated atrial pressure variations within a cardiac cycle that aligned well with measured clinical trends. Although a model featuring regular and synchronised atrial contraction yielded slightly smaller ejection fraction errors, it failed to reproduce physiologically realistic atrial pressure behaviour and resulted in pressure variations larger than those observed clinically. Overall, these findings highlight the superior ability of the proposed model to predict haemodynamics during AF. While the proposed model successfully reproduces several well-established AF-induced haemodynamic changes, the authors acknowledge some limitations that present opportunities for future development. One key limitation is the absence of haemodynamic reflex mechanisms, such as baroreceptor-mediated responses. Considering their substantial influence on cardiovascular dynamics, the authors intend to incorporate baroreflex feedback in their future work to further improve the physiological accuracy of the simulations. Additionally, the authors acknowledge that future modelling based on patient-specific electrophysiological mapping would enhance the representation of atrial activation heterogeneity and enable direct comparison with the stochastic activation patterns used in this study, providing valuable insights into when spatial heterogeneity becomes haemodynamically significant. In conclusion, this work presents a computational model with carefully defined simplifications of the cardiovascular system that make it efficient and suitable for simulating AF, particularly given the temporal resolution required to capture beat-to-beat variability. The model enables the investigation of patient-specific haemodynamics through appropriate model fitting. By capturing irregular activation patterns on a beat-to-beat basis, it provides a computational framework for exploring haemodynamic responses during arrhythmic heart rhythms and serves as a valuable tool for better understanding how AF affects cardiovascular function. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article. The authors declare they have no competing interests. Both authors have read and approved the final version of this manuscript and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed. No funding was received for this work.