Incidence of Post-operative Atrial Fibrillation Secondary to Abdominal Surgery

Background and rationale Post-operative AF (POAF), i.e. atrial fibrillation (AF) that occurs secondarily to surgery is considered a phenomenon rather than a definite diagnosis, and continuous anti-coagulation is not recommended in AF of short duration and non-persistent AF.1,2 The most recent guidelines from the European and American Societies of Cardiology have no specific recommendation regarding the management of secondary AF, though referral to an outpatient cardiology clinic can be considered if the condition is not self-limiting.3,4 Despite this, recent data from the Framingham population suggest that secondary AF is not a benign condition,5,6 and some studies have found POAF in relation to non-cardiac surgery to be associated with increased post-operative mortality and stroke.7-9 Currently few prospective studies have been performed in non-cardiac conditions and the consequences of secondary AF in patients without known heart disease are not well described. POAF in relation to non-cardiac surgery is common with an average incidence of 10.94%.10 However, the true incidence of POAF in relation to non-cardiac surgery is uncertain, as the available studies are few, heterogeneous, and often methodologically inadequate.10 Different risk-stratification models exist, however it is still challenging to accurately predict which patients will develop AF both perioperatively and in the years following surgery.11-14 Important risk factors for developing POAF are: age (age 65-74, OR 2.08; age >85, OR 3.56), hypertension (OR 3.66), heart failure (OR 1.64), thyroid disease (OR 6.29), laparotomy vs laparoscopy (OR 3.30) and duration of surgery (> 600 min, OR 1.38).10 The use of electrocardiogram (ECG) in risk stratification for AF and stroke is not yet incorporated in clinical guidelines. However, short and prolonged p-wave and p-wave terminal force in lead V1 (PTFV1) is correlated with AF.15 Myovista (HeartSciences, Southlake, Texas, USA) is a novel ECG utilizing continuous wavelet transform signal processing (wavECG) and can be used to prove left ventricular diastolic dysfunction (LVDD).16 LVDD is in other studies associated with AF.17-20 These new biomarkers might improve the identification of patients at risk of developing POAF.

Purpose The study aims at reporting the incidence of AF, and associated complications, in relation to acute abdominal surgery. Concurrently, patients will be characterized by specialized ECG analysis and wavECG.

Methods Prospective, single-centre, cohort study of consecutive patients undergoing acute* abdominal surgery. Patients will be identified among acute referrals for the Department of Abdominal Surgery at Bispebjerg-Frederiksberg Hospital. The Investigator (Department of Cardiology at Bispebjerg-Frederiksberg Hospital) will ensure signed informed consent prior to any study-related examinations. Surgery will be performed at the Department of Abdominal Surgery at Bispebjerg-Frederiksberg Hospital independently of study enrolment. A subset of patients will be enrolled in a feasibility study of continuous cardiac rhythm monitoring (C3+, Cortrium, Copenhagen, Denmark). Heart rhythm monitoring will be worn from admission and until discharge (preferably ≥72 hours post-operatively) in patients admitted to the Department in the time interval of 7 AM to 11 PM$. Additional follow-up will be based on a review of patient charts at discharge and one# month postoperatively.

No randomization or placebo will be used. ECG measurements will be compared to existing data from a healthy population.

Measurements

• ECG: AF, PR-interval, p-wave area, PTFV1
• wavECG: LVDD
• Heart rhythm monitoring: AF

Statistical considerations Consecutive patients attending the Department of Abdominal Surgery at Bispebjerg-Frederiksberg Hospital will be invited to participate in the pre- and post-operative examination programme. The study's aim is to identify the incidence of peri- and postoperative AF. Hence, sample size calculations are uncertain. The expected incidence of AF for patients undergoing non-oesophageal abdominal surgery is at least 7.63%.10 Pre-operative p-wave indices have shown the ability to predict POAF in a cardiac surgery population of 105 patients (with a POAF incidence of 11%).21 wavECG has been used to show LVDD in a study population of 188 patients referred for computed tomography angiography of the heart.16 The study will include 400-500 patients of which an estimated 2/3 will wear heart rhythm monitoring corresponding to 264-330 patients. Follow-up will be one month.

Normality was inspected visually by histograms and tested by Shapiro-Wilk. Nonparametric statistics were used. Wilcoxon rank-sum (continuous variables) and Chi-square and Fisher's exact test (categorical variables) were used as appropriate. Data were presented as median with interquartile range (IQR) for continuous variables and frequency with percentages for categorical variables. Odds ratios (OR) with 95% confidence intervals (CI) were calculated for dichotomous values by univariate logistic regression. Multiple logistic regression was used to adjust for variables' influence on the primary outcome of POAF and the secondary outcome of severe adverse events. Based on an á priori knowledge of risk factors of atrial fibrillation and expected important surgical factors, variables were chosen for the full multivariate model. The following variables were included: age, sex, prior atrial fibrillation, prior stroke, hypertension, heart failure, heart valve disease, chronic renal disease, diabetes mellitus, vascular disease, chronic obstructive pulmonary disease, cancer, surgery invasiveness, and surgery urgency. For multivariate modeling of the secondary outcome, POAF was added to the full model as it was expected to influence the development of the outcome. Subsequently, the full multivariate logistic regression model was reduced to a prediction model. Variables with p<0.05 remained in the prediction model, while variables with p>0.05 were eliminated. Multicollinearity was checked and a Variance Inflation Factor level <5 was accepted. McFadden's R-squared test was applied to evaluate model fit, while Analysis of variance (ANOVA) tests were applied to compare the components of the model, to reduce the risk of overfitting. A receiver operating characteristic curve (ROC) with area under the curve (AUC) was calculated for the multivariate model. The Closest Topleft decision rule was applied to define the optimal sensitivity and specificity values. For AUC, sensitivity, and specificity, 95% confidence intervals were produced through bootstrapping with 2000 replicates. All tests were two-sided, and a p< 0.05 was considered statistically significant. RStudio (version 1.4.1717) was used for statistical analyses.

Footnote(s): *The study was originally designed to include both acute and elective patients, but due to limited availability of heart rhythm monitoring devices, prior to study initiation (June 5, 2020), we decided to enroll patients referred for acute abdominal surgery only. $The study was originally planned to apply heart rhythm monitors all 24 hours of the day. This was not possible due to logistics, and prior to study initiation, we agreed to apply monitoring to all patients admitted between 7 AM to 11 PM. #The study was originally planned for 12 months follow-up but was reduced to one month, due to limitations of the permissions achieved prior to study initiation. In addition, an outpatient clinic visit 3 months after surgery was planned where ECG and heart rhythm monitor examinations were to be repeated. The visits were canceled due to local COVID-19 restrictions.