Blood coagulation modelling to inform stroke risk stratification and intervention planning in atrial fibrillation patients with non-cardiac comorbidities 

Project reference:  SIE_03_21

Co-supervisor: Adelaide De Vecchi
Co-supervisor: Oleg Aslanidi 

Start date:  February 2022 or October 2022

Project summary:  Atrial fibrillation (AF) is a common heart condition that promotes blood clotting, significantly increasing risks of a thromboembolic event. Non-cardiac comorbidities affecting blood rheology can catastrophically worsen these risks in AF patients. In particular, COVID-19 has been linked to a 7.5-fold increase in the odds of stroke due to blood hypercoagulation both the short and in the long term, due to the so-called “long COVID”. Treatment options include anticoagulation therapy and surgical occlusion of the left atrial appendage (LAA), the anatomical region more prone to thrombus formation. However, current stroke risk stratification - suboptimal in itself - does not account for the presence of comorbidities where blood parameters are altered, exposing the urgent need for better-informed, personalised intervention planning. We will address this issue using novel computational models of blood flow and coagulation to identify crucial prognostic factors for thrombus formation and personalise treatment planning in high-risk patients. 

 

Project description: 

We hypothesise that stroke risks are particularly high in AF patients with co-morbidities where blood clotting is driven by the simultaneous presence of superimposed risk factors, e.g. blood stagnation and hyperviscosity. In these cases, stroke risk stratification, and subsequently treatment strategies, are suboptimal. As over 90% of thrombi in non-valvular AF are localised in the narrow region of the LAA, current treatment options include surgical intervention to occlude the LAA and prevent blood from entering this critical area, and anticoagulant drug therapy. However, it is uncertain in which patient categories therapy is warranted, and whether anticoagulants alone are sufficient to prevent thrombus formation in the LAA or if surgical occlusion is a necessary preventative step in high-risk patients. A complete assessment of the best course of action relies on the mechanistic understanding of the patho-physiological changes induced by the alteration of blood properties in the underlying thrombogenic substrate of AF. 

We propose to build multi-scale models of (patho-)physiological blood flow and coagulation by combining patients’ imaging data with clinical data from blood samples to create a virtual cohort of AF patients with non-cardiac comorbidities. The study will initially focus on the effects of COVID-19 on AF due to the large availability of blood samples from COVID-19 patients at our institution. However the approach will be generalised to model any comorbidities where prothrombotic factors are present, either by alterations in the coagulation cascade or by blood stagnation (e.g. mitral stenosis). Blood flow models of the left atrium will be based on imaging datasets from 12 AF patients equally split between sinus rhythm (normal state) or fibrillation (arrhythmic state), updated with values from blood samples of 40 COVID-19/AF patients from an ongoing study on COVID-19 at St Thomas’ Hospital (Dr Williams). 

These models will be used to achieve three main objectives:

  1. Elucidate the interplay between AF and comorbidities in terms of blood coagulation.

  2. Predict disease-specific thrombus formation dynamics.

  3. Test new blood biomarkers to tailor stroke risk stratification and inform treatment.

 

Our existing models of blood flow and coagulation in the left atrium will provide a starting point. The candidate will develop them to include disease-specific coagulation models, and to simulate the surgical LAA occlusion and the thrombolytic action of varying doses of heparin. A virtual cohort of 480 patients will be generated and used as a testbed for stroke risk predictions based on blood biomarkers using principles of Bayesian inference and machine learning. 

 

Given the rising AF prevalence (1-2% in the general population) and the long-term effects of COVID-19, this disease combination carries and will continue carrying a significant socio-economic burden if effective treatment strategies are not devised in a timely manner. This study will enable a better understanding of the mechanisms of both disease and treatment – a necessary step for an effective patient stratification. Perhaps even more importantly, it will establish a computational framework for assessing the effect of different comorbidities on stroke risks in AF patients, contributing to plan tailored intervention for specific sub-groups of high-risk patients.

Adelaide Oleg project image edited