Twitter Facebook LinkedIn
Center for Global Research Data

Impact of atrial fibrillation and the effects of comorbidities and treatment on disease progression and outcomes

Lead Investigator: Gregory YH Lip, University of Liverpool
Title of Proposal Research: Impact of atrial fibrillation and the effects of comorbidities and treatment on disease progression and outcomes
Vivli Data Request: 7074
Funding Source: None
Potential Conflicts of Interest: Co-Chair for the GLORIA-AF registry.

Summary of the Proposed Research:

Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia that is characterised by an irregular heart beat. In 2010, the global prevalence of AF was estimated at 33.5 million. The condition is associated with an increased risk of stroke, heart failure and mortality, and reduced quality of life. Additionally, it poses a significant healthcare economic burden.

Despite years of research, there remains many aspects of AF that are poorly defined, partly due to the complex nature of this disease. There have been much improvement in the care of patients with AF over the past decade. Nonetheless, further progress needs to be met with greater understanding of the condition. Here, we identify several important topics that warrant further investigation and have good potential to enhance direct patient care by improving risk stratification and the choice of treatment in AF.

This research will be conducted using patients with AF from the GLORIA-AF registry.

Statistical Analysis Plan:

Normality of continuous variables will be assessed using Kolmogorov-Smirnov test. Continuous variables will be described with mean and standard deviation (SD) or median and interquartile range (IQR), and tested for differences with Student’s t-test and Mann-Whitney U test, respectively. If more than 2 groups are present, between-group differences will be tested with Kruskal-Wallis test.

Categorical variables will be described as count and percentage, and tested for differences with chi-squared or Fisher’s exact test.

Plots of Kaplan-Meier curves will be performed and survival distributions were compared using log-rank test.

Cox and logistic regression analyses will be performed for the outcomes of interest. Multivariable models will include a combination of the following covariates: age, gender, body mass index, estimated glomerular filtration rate, left atrial size, left ventricular ejection fraction, left ventricular hypertrophy, type of AF, European Heart Rhythm Association classification, chronic obstructive pulmonary disease, coronary artery disease, diabetes mellitus, heart failure, hypercholesterolaemia, hypertension, peripheral artery disease, prior haemorrhagic event, prior thromboembolism, sleep apnoea, and use of anticoagulation, angiotensin-converting enzyme inhibitor, angiotensin receptor blocker, diuretics, aldosterone blockers and statin.

Propensity score matching will be performed where patients will be matched in a 1:1 ratio across each cohort on a propensity score generated by logistic regressions using the nearest-neighbour technique without replacement with a maximum caliper of 0.2, thus avoiding at least 98% of the bias due to the measured confounders.

A two-sided p value of less than 0.05 will be considered statistically significant.

The outcome elements that will be assessed include stroke, TIA, non-CNS arterial embolism, PE, MI, bleeding and death.

For the machine learning analysis, we will use non-linear machine learning classification methods. Random forests, support vector machines, k-NN classification, and gradient boosted decision trees (amongst others) will be used.
The dataset will be split into a train (70%) and test (30%) set, where the non-linear machine learning algorithms will be applied on the training set using 5-fold cross-validation. If any imbalance exists in the outcome of interest, both undersampling and oversampling will be used on the training set to improve the predictive accuracy of the algorithms. The best trained algorithm will be used on the withheld test set to determine predictive accuracy. The c-index (AUC), sensitivity, specificity, positive predictive value, and negative predictive value will be reported.

Statistical analyses will be performed using SPSS version 27 (IBM Corp, Armonk, NY) and R Studio.

Requested Studies:

GLORIA – AF: Global Registry on Long-Term Oral Anti-thrombotic Treatment In Patients With Atrial Fibrillation (Phase II/III)
Data Contributor: Boehringer Ingelheim
Study ID: NCT01468701
Sponsor ID: 1160.129

GLORIA – AF: Global Registry on Long-Term Oral Anti-thrombotic TReatment In PAtients With Atrial Fibrillation (Phase II/III – EU/EEA Member States)
Data Contributor: Boehringer Ingelheim
Study ID: NCT01671007
Sponsor ID: 1160.136

GLORIA-AF: Global Registry on Long-Term Oral Anti-thrombotic Treatment In Patients With Atrial Fibrillation (Phase II/III-India and Switzerland)
Data Contributor: Boehringer Ingelheim
Study ID: NCT01937377
Sponsor ID: 1160.171

Public Disclosure:

Ding, W.Y., Calvert, P., Gupta, D. et al. Impact of early ablation of atrial fibrillation on long-term outcomes: results from phase II/III of the GLORIA-AF registry. Clin Res Cardiol (2022). doi:10.1007/s00392-022-02022-1