Atrial fibrillation is frequent but does not affect risk stratification in pulmonary embolism

Although prior studies indicate a high prevalence of atrial fibrillation (AF) in patients with pulmonary embolism (PE), the exact prevalence and prognostic impact are unknown.


Introduction
Atrial fibrillation (AF) is the most common cardiac arrhythmia, affecting approximately 3% of the adult population [1]. A limited number of cohort studies have demonstrated an association between AF and venous thromboembolism (VTE) that appears to be stronger for pulmonary embolism contraction as well as high and irregular ventricular rates impair ventricular filling and lead to a reduction in cardiac output by up to 25%, thus potentially worsening haemodynamic instability in acute PE [5,6]. Although AF is a frequent finding in acute PE affecting 12% to 24% of patients [7][8][9], the effects of AF on the outcome of PE patients have not been conclusively answered. Studies investigating the influence of AF on the prognosis after PE provided conflicting results regarding in-hospital [7,10,11] and 30-day [7,8] mortality.
Therefore, in the present study we evaluated the prevalence and prognostic impact of AF on inhospital adverse outcomes and one-year mortality of patients with acute PE. In addition, we investigated whether existing risk stratification tools are affected by the presence of AF and should be adapted considering the heart rhythm at presentation. Furthermore, we compared the characteristics and outcomes of patients with newly diagnosed AF to patients with known AF and patients without AF.

Study design and definition of outcomes
In the present cohort study, patients with objectively confirmed PE ≥ 18 years of age prospectively enrolled in the Pulmonary Embolism Registry of Goettingen (PERGO) at the University Medical Center Goettingen, Germany, between September 2008 and September 2017 were included. The study protocol has been described in detail previously [12,13]. We excluded patients (i) withdrawing previously given consent for participation in PERGO, (ii) included a second time in PERGO because of recurrent PE, (iii) with missing electrocardiogram (ECG) on admission, (iv) who received cardiopulmonary resuscitation (CPR), electrical cardioversion or suffered from a sustained ventricular arrhythmia before admission and (v) with subsegmental PE in combination with significant acute cardiorespiratory illness responsible for clinical presentation and symptoms (Fig. 1). All patients were followed for the in-hospital stay, 635 patients included in PERGO between September 2008 and September 2017 107 patients excluded: -7 included a second time in PERGO because of recurrent PE -57 with missing ECG on admission -26 received CPR, electrical cardioversion or suffered from a sustained ventricular arrhythmia before admission -17 with subsegmental PE in combination with significant acute cardiorespiratory illness responsible for clinical presentation New AF diagnosis (n = 34) ) and laboratory testing, treatment and in-hospital outcomes were obtained using a standardized case report form. Heart rhythm was assessed from admission ECGs and independently adjudicated by two blinded authors (N.I.J.R. and M.L.), and disagreement was resolved by a third author (A.S.P.). RV dysfunction on CTPA was defined as right-to-left (RV/LV) diameter ratio ≥ 1.0. RV dysfunction on TTE was defined as RV dilatation (end-diastolic diameter >30 mm from the parasternal view or a RV/LV diameter ratio ≥ 1.0 from the subcostal or apical view) combined with right atrial hypertension (the absence of inspiratory collapse of the inferior vena cava) [16]. Active cancer was defined as known disease, treatment with antitumor therapy within the last 6 months, metastatic state or haematologic cancer that was not in complete remission [17].
Individual risk stratification was performed according to the algorithm proposed by the ESC 2014 guidelines [15], the simplified Pulmonary Embolism Severity Index (sPESI) and the modified FAST score [18]. For calculation of algorithms and scores, missing values were considered to be normal [18].
As shown in Fig. 1, study patients were stratified (i) according to the heart rhythm on admission (AF vs. sinus rhythms) and (ii) based on whether AF was known or newly diagnosed on admission.
An in-hospital adverse outcome was defined as PErelated death, cardiopulmonary resuscitation or administration of catecholamines. Further study outcomes include in-hospital all-cause death, duration of the in-hospital stay (days) and oneyear all-cause mortality. Death was determined to be PE-related if either confirmed by autopsy or following a clinically severe episode of acute PE in the absence of an alternative diagnosis. All events and causes of death were independently adjudicated by two of the authors (M.E. and N.I.J.R.), and disagreement was resolved by a third author (M.L.).

Statistical analysis
Categorical variables are presented as total numbers and percentages; continuous variables not following a normal distribution if tested with Kolmogorov-Smirnov test are presented as medians with interquartile ranges (IQR). Associations between binary and categorical variables were analysed using Fisher's exact test or chi-squared test, as appropriate. For comparison of continuous variables, the Mann-Whitney U-test was employed.
To allow comparison of scores, the four-level ESC 2014 algorithm was dichotomized as low risk and intermediate-low risk ('low risk') versus intermediate-high risk and high risk ('high risk') [18]. The prognostic relevance of AF, patient characteristics and comorbidities, biomarkers and risk assessment strategies/scores with regard to study outcomes was tested using univariable logistic regression analyses, and results are presented as odds ratios (OR) with the corresponding 95% confidence intervals (CIs). To investigate the prognostic role of different risk stratification markers in the presence of AF, we conducted multivariate logistic regression analyses. Kaplan-Meier analysis was used to compare the probability of one-year survival in subgroups stratified (i) according to heart rhythm on admission and (ii) based on whether AF was known or not; the log-rank test was used for comparison. Cox regression analysis was used to identify predictors of one-year survival in patients discharged from hospital alive; results are presented as hazard ratios (HR) with the corresponding 95% CIs.
A two-sided significance level of a < 0.05 was defined as appropriate to indicate statistical significance. As this was an explorative testing, no adjustments for multiple testing were carried out. P-values were provided for descriptive reasons only and should be interpreted with caution and in connection with effect estimates. Statistical analysis was performed using Statistics Package for Social Sciences (IBM SPSS Statistics, version 25, IBM Corp. Armonk, NY).

Of 635 patients enrolled in PERGO between
September 2008 and September 2017, 107 patients (16.9%) were excluded from analysis ( Fig. 1). Information on baseline characteristics and risk stratification of the 528 study patients are presented in Table 1, left column. On admission, 57 patients (10.8%) presented with AF; of those, 34 (59.6%) had a first documented AF episode. Of 52 patients (9.8%) with known AF, 23 (44.2%) patients presented with AF on admission ( Fig. 1).
As shown in Table 1, right columns, patients presenting with AF were older and more frequently had chronic heart failure, coronary artery disease, renal insufficiency and overt hyperthyroidism compared with patients presenting in SR (Table 1, right columns). In addition, AF patients more frequently presented with tachycardia, elevated cardiac biomarkers and were hence more often stratified to higher risk classes by the ESC 2014 algorithm, sPESI and modified FAST score.

Prognostic impact of AF on admission
Overall, 44 patients (8.3%) had an in-hospital adverse outcome and 24 patients (4.5%) died during the in-hospital stay, of those 13 (54.2%) due to PE. Interestingly, AF on admission was neither associated with an increased risk of an inhospital adverse outcome nor in-hospital all-cause mortality (Table 2A). In the subgroup of 501 normotensive patients, a higher rate of an adverse outcome was observed in patients with AF on admission compared with patients presenting in SR (9.6% vs. 5.1%); however, this finding did not reach statistical significance (OR 1.93 [95% CI 0.54-6.89]; Table 2B). Further, AF was not associated with adverse outcomes focussing on other subgroups of interest (women, patients with chronic heart failure, RV dysfunction on TTE/ CTPA or intermediate-high risk/high risk according to risk stratification algorithms/scores; data not shown). However, patients presenting with AF on admission had a longer median duration of inhospital stay compared with patients presenting in SR (Table 1).
Of 504 patients discharged from hospital alive, information on the survival status at one year was available for 496 patients (98.4%). During the first year after PE, 53 patients (10.7%) died after hospital discharge. AF on admission was not associated with an increased risk (Table S1 of  Relevance of AF on admission for the prognostic performance of risk assessment strategies As shown in Table 3A, left column, elevated levels of hsTnT, NT-proBNP and MR-proANP, tachycardia and classification to higher risk classes according to the sPESI, modified FAST score and ESC 2014 algorithm were identified as predictors of an inhospital adverse outcome. Importantly, adjustment for AF on admission using multivariate models did not affect their prognostic performance (Table 3A, right column). Although AF on admission was not predictive of in-hospital adverse outcome or death, AF was associated with elevated cardiac biomarkers, tachycardia and higher risk classes in algorithms and scores (Table S2 of the supplementary material). The strongest effect of AF on admission was observed on MR-proANP ≥ 120 pmol L À1 (OR 45.3 [95% CI 6.2-331.0]). Furthermore, risk assessment models were able to predict in-hospital mortality, whilst single parameters such as tachycardia or cardiac biomarkers were not of predictive value in our cohort (Table 3B). Similar results were obtained focussing on normotensive patients only (Table S3 of the supplementary material). Prevalence and prognostic relevance of newly diagnosed AF on admission To assess the prognostic importance of a newly diagnosed AF at the time of admission for acute PE, we compared baseline characteristics, initial risk stratification and outcome of patients with a newly diagnosed AF to patients with known AF and patients without AF. Whilst relevant cardiovascular comorbidities were more often present in

Discussion
In the present real-world single-centre cohort investigating 528 PE patients included consecutively over a 9-year period, 10.8% of patients presented with AF on admission; of those, 59.6% had newly diagnosed AF. The prevalence of known hsTnT, high-sensitivity troponin T; LV, left ventricle; MR-proANP, mid-regional pro-atrial natriuretic peptide; NT-proBNP, N-terminal pro-brain natriuretic peptide; OR denotes odds ratio; RV, right ventricle; sPESI, simplified pulmonary embolism severity index.    or newly diagnosed AF of 16.3% in the present study was higher compared with the German population (4.7% and 7.6% for patients aged 65-69 and 70-74 years, respectively) [22]. Patients with AF on admission had more comorbidities, presented more frequently with tachycardia and elevated cardiac biomarkers and were hence stratified to higher risk classes. Importantly, AF on admission had no impact on in-hospital adverse outcomes and did not affect the prognostic performance of biomarkers and risk assessment strategies. Baseline characteristics of patients with newly diagnosed AF and patients with known AF differed. Whilst cardiovascular comorbidities were more frequent in patients with known AF, patients with newly diagnosed AF had more often overt hyperthyroidism.

Prognostic impact of AF on admission
The few previous reports that investigated the prognostic relevance of AF on admission for acute PE provided contradicting results. An analysis of 508 PE patients derived from a prospective registry published in 2005 reports that nonsurvivors more frequently had atrial arrhythmias on admission compared with survivors (25% vs. 12%, P < 0.001) [8]. However, since this registry only included patients with 'major PE' (defined as haemodynamic instability or RV dysfunction or signs of pulmonary hypertension on TTE or right heart catheterization), results are not generalizable. Koracevic et al. conducted a smaller study with 140 PE patients and found no impact of AF on in-hospital mortality, similar to the findings of our study [11]. However, important methodological and outcome information are missing and limit interpretability of these findings. An analysis by Krajewska et al. investigating the effect of AF during hospitalization for acute PE (rather than on admission) in 391 patients reported that paroxysmal AF had little effect on all-cause mortality compared with sinus rhythm (mortality rate 6.5% and 5.0%, respectively), but observed a higher in-hospital mortality rates in patients with permanent AF (25%) [10]. The remarkable high mortality rate in the latter group might be partially explained by an uneven distribution of other relevant prognostic factors, such as a lower median LV ejection fraction and renal function in the permanent AF group. Hence, the observed differences in mortality might not be exclusively due to the effects of atrial fibrillation. Our study considerably adds to these previous investigations. We report on the yet largest population of well characterized and consecutive PE patients. In contrast to Krajewska et al., we did not investigate the effects of different AF types occurring over the course of hospitalization, but focused on the heart rhythm on admission, the critical time-point for risk assessment and therapeutic decision-making.
In contrast to our study hypothesis, AF on admission was no predictor of an in-hospital adverse outcome and mortality. Nevertheless, two findings might hint towards a prognostic impact of AF in acute PE: First, normotensive patients with AF on admission had a numerically higher rate of an adverse outcome compared with patients presenting in SR (9.6% vs. 5.1%). Secondly, AF was associated with tachycardia and elevated cardiac biomarkers; thus, patients with AF on admission were stratified to higher risk classes by the ESC 2014 algorithm, the sPESI and the modified FAST score. Therefore, the prognostic impact of AF in acute PE more likely appears to be small rather than absent and a larger sample size would have been required to demonstrate statistical differences. However, as patients with known AF receive therapeutic anticoagulation for prevention of stroke and are thus protected from developing acute PE, inclusion of a large number of patients with PE and AF is challenging.
MR-proANP, secreted from the atria as a result of increased wall tension and stretch [23], was strongly associated with the presence of AF in our PE patients. MR-proANP levels ≥ 120 pmol L À1 were found in 97.9% of patients with AF on admission compared with 50.1% in patients presenting in SR. Despite this fact, elevation of MR-proANP (as well as elevated hsTnT and NT-proBNP levels) was associated with an increased risk of an in-hospital adverse outcome regardless of the presence of AF (Table 3A) indicating the MR-proANP integrates different prognostic relevant information from comorbidities.
Importantly, we are the first to demonstrate that the prognostic performances of established risk assessment strategies and biomarkers are not affected by the presence of AF. This finding supports the use of risk stratification for patients with acute PE irrespective of heart rhythm on admission.

Differences of PE patients with newly diagnosed AF and known AF
The incidence, risk factors and prognostic implications of newly diagnosed AF on admission for acute PE have not been investigated so far. In our cohort, as many as 59.6% of PE patients presenting with AF on admission had no history of AF. These patients with newly diagnosed AF differ from patients with known AF in several important aspects: Not surprisingly, patients with known AF had a higher prevalence of chronic heart failure, coronary artery disease, chronic pulmonary disease, arterial hypertension, diabetes mellitus and renal insufficiency compared with patients without AF. In contrast, the prevalence of these comorbidities was lower in PE patients with newly diagnosed AF. However, patients with newly diagnosed AF more often had overt hyperthyroidism, a condition known for its pro-arrhythmogenic potential [1]. Further studies are needed to investigate to which extent PE might trigger AF or whether patients with newly diagnosed AF actually suffered from undiagnosed paroxysmal AF prior to PE. Further, the implications of newly diagnosed AF on the optimal duration of long-term anticoagulation remain unclear. Thus, studies that explore the long-term risk of ischaemic stroke after discontinuation of anticoagulation in patients with newly diagnosed AF at presentation for acute PE are warranted.
Of note, 9.8% our PE patients had known AF. Although all but four of these patients had a CHA 2 DS 2 -VASc score ≥ 2 points and therefore should have been treated with therapeutic anticoagulation for prevention of arterial thromboembolism [1], only 23.1% of patients with known AF received therapeutic anticoagulation at the time of PE diagnosis. However, the large proportion of nonanticoagulated AF patients in our cohort most likely reflects the effective prevention of VTE in AF patients who receive guideline-recommended anticoagulation treatment.

Conclusion
Atrial fibrillation is a frequent finding in patients with acute PE, present in more than 10% of cases. Of those, more than 50% had no previous AF diagnosis. These newly diagnosed AF patients had a distinct pattern of risk factors compared with patients without AF or patients with known AF.
Although not predictive of in-hospital adverse outcomes in our cohort, patients with AF on admission were more frequently classified to higher risk classes due to tachycardia and elevated cardiac biomarker levels. Importantly, the prognostic performance of risk assessment strategies was not affected by AF. Thus, our data support the use of risk stratification tools for patients with acute PE irrespective of the heart rhythm on admission.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Figure S1. Probability of one-year all-cause mortality in PE patients discharged alive from hospital. Table S1. Predictors of one-year mortality in 496 patients discharged alive from hospital. Table S2. Odds of AF on admission for having elevated biomarkers, RV dysfunction or higher risk scores in PE patients. Table S3. Prognostic value of biomarkers and risk assessment strategies in normotensive patients.