Advanced analytic methods to refine phenotyping and risk prediction among patients with atrial fibrillation in the RE-LY trial

Advanced analytic methods to refine phenotyping and risk prediction among patients with atrial fibrillation in the RE-LY trial

Dr. Desai was the recipient of a career development award from the Agency for Health Care Research and Quality (AHRQ) which will support this analysis.

Potential conflicts of interest will be disclosed when the research is presented and published.

Atrial fibrillation (AF) is the most common cardiac arrhythmia requiring medical attention. The prevalence of AF in the United States was estimated to be as high as 6 million in 2010 and projected to rise to 12 million in 2050. Much of the morbidity related to AF arises from consequent systemic embolic complications, including a 5-fold increase in cerebrovascular accidents (CVA) or transient ischemic attacks (TIA). Aside from the significant medical burden, it is estimated that the incremental cost of AF exceeds $26 billion annually.

Anticoagulation therapy is highly effective at reducing the risk of stroke. The decision for whether a patient with AF should receive an anticoagulant is based, in part, on the risk of stroke which is estimated using a risk stratification tool. Among the most commonly used in clinical practice is the CHADS2 score.

However, Current risk stratification tools to estimate the risk of stroke or major bleeding in patients with atrial fibrillation are inadequate for personalized decision-making. Both the CHADS2 and CHADS2-VASc models were derived from historical cohorts with short follow-up and small numbers of events and offer a relatively crude estimation of risk; c-statistic for predicting a thromboembolic event of about 0.60, and accounting for only about half the attributable risk.20 Moreover, though simple for providers to memorize, these models are predicated on a reductionist approach that ignores the heterogeneity that is required for personalized decision-making. The risk of stroke or SE is likely to vary across different subpopulations and the relationship between risk factors and outcomes is unlikely to be binary, monotonic or linear as modeled in CHADS2 and CHADS2-VASc. Novel analytical methods, including sequential pattern mining, machine learning, and graph visualization, applied to large, complex, multidimensional datasets can illuminate complex, non-linear, inter-dependent relationships and help produce powerful, personalized, risk-prediction tools enabling patients and providers to make truly informed, patient-centered decisions.

Our objectives for this project are to:
Objectives:
1. Define a novel taxonomy and/or identify distinct subtypes of atrial fibrillation by using baseline demographic, biophysical, clinical, and laboratory measurements through the application of cluster analysis.
2. Apply novel analytic methods (network analysis, sequential pattern mining, machine learning, and graph visualization) capable of capturing non-linear and inter-dependent relationships to develop more powerful tools to predict stroke and major bleeding for patients with AF as compared with contemporary risk stratification tools.

We believe that the findings of this project would be of interest to patients and providers around the world and improve the quality of care for patients with AF. The results would be communicated through publication and seminars.

[{ "PostingID": 2599, "Title": "BI-1160.26", "Description": "Randomized Evaluation of Long Term Anticoagulant Therapy (RE-LY) With Dabigatran Etexilate

Medicine: dabigatran etexilate, Condition: Atrial Fibrillation, Stroke, Phase: 3, Clinical Study ID: 1160.26, Sponsor: Boehringer Ingelheim" }]

The publication citation will be added after the research is published.