Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia, occurring in 2%–4% of the general population with increasing prevalence among the elderly [1]. It is predicted that 10%–15% AF patients will require percutaneous coronary intervention (PCI) for coronary artery disease (CAD) during their life, while patients with AF and acute coronary syndrome (ACS) will be more likely to experience adverse outcomes than ACS patients without AF [2, 3]. Concomitant risks of subsequent ischemic events, in-stent thrombosis and treatment-related bleeding need to be carefully considered before initiating appropriate antithrombotic therapy.

Several risk clinical scores have been proven to enhance the assessment of thrombo-embolic risk in AF. CHA${}_2$DS${}_2$-VASc score [congestive heart failure (CHF), hypertension, age $\ge$75 years, diabetes mellitus, stroke, vascular disease, age 65–74 years, sex category (female)] is a recognized tool to stratify stroke risk and recommended by guidelines [4, 5]. Several studies have showed that a higher CHA${}_2$DS${}_2$-VASc score was independently associated with a poor outcome in CAD patients with sinus rhythm [6, 7, 8].

Early risk stratification is also important for ACS patients to help clinicians to determine prognosis and therefore guide management strategies. A number of prognostic models have been developed including Global Registry of Acute Coronary Events (GRACE) [9], thrombolysis in myocardial infarction (TIMI) [10], and platelet glycoprotein IIb/IIIa in unstable angina: receptor suppression using Integrilin therapy (PRUSUIT) [11]. Among these, the GRACE risk score has been externally validated and proved to display the best discriminative performance [12]. The GRACE score at admission [age, systolic blood pressure (SBP), heart rate, serum creatinine, cardiac arrest at admission, elevated cardiac biomarkers, ST-segment deviation and Killip class at presentation] is established and widely accepted for assessing death or myocardial infarction from admission to six months after discharge [13]. Meanwhile, another prediction model, the GRACE score at discharge [age, history of CHF, history of myocardial infarction (MI), heart rate, SBP, ST-segment depression, serum creatinine, elevated cardiac enzymes and no in-hospital PCI] also derived by the implemented GRACE registry, has been established a robust tool for predicting 6-month post-discharge mortality in patients with ACS [14].

For patients with AF and ACS, it still remains unclear whether CHA${}_2$DS${}_2$-VASc or GRACE score may be useful to assess the risk of major adverse cardiovascular and cerebrovascular events (MACCEs) and thus guide the anticoagulant regimens. Therefore, in the present study we aim to compare the prognostic values of CHA${}_2$DS${}_2$-VASc, GRACE score at admission and GRACE score at discharge in predicting long-term MACCEs in patients with AF and ACS or undergoing PCI.

In the present study, we assessed prognostic values of the CHA${}_2$DS${}_2$-VASc, GRACE at admission and GRACE at discharge scores in adverse outcomes among AF patients with ACS or undergoing PCI during 3-year follow-up. We demonstrated that higher CHA${}_2$DS${}_2$-VASc and GRACE at discharge scores were independently associated with increased risk of MACCEs, but the GRACE score at admission was not. The fact that CHA${}_2$DS${}_2$-VASc and GRACE at discharge demonstrated comparable discriminative capacity meanwhile GRACE at admission provided relatively lower discrimination further supported this viewpoint. For prediction of all-cause mortality, three models displayed good discriminative capacity. The GRACE at discharge showed better predictive ability. A significant discrimination improvement of GRACE at discharge compared to the CHA${}_2$DS${}_2$-VASc was detected. In addition, none of the three scores had an acceptable value in predicting major or minor bleeding during the follow-up.

The co-existence of AF and the need for PCI is a much more complicated situation compared to suffering from AF or CAD alone. Existing evidence reports the incidence of ACS with concomitant AF between 6% to 22%, with an increased incidence in elderly and female patients [17, 18]. AF is a well-established marker of poor short- and long-term prognosis in patients with ACS and is associated with an increased risk of overall mortality. An analysis derived from 1558205 ACS patients observed that patients with AF had significantly longer and more complicated hospital stays with nearly double adjusted in-hospital mortality [19]. Pilgrim et al. [20] showed that among patients with CAD undergoing revascularization with drug-eluting stents (DES), AF conferred a rising risk of both all-cause mortality and ischemic stroke during four-year follow-up. Similar results were obtained from sub analysis of the Global Registry of Acute Coronary Events (GRACE) study where ACS patients with concomitant AF were more likely to have a complicated in-hospital course than those without AF [21]. Meanwhile, in a large-scale, prospective registry including 29,679 consecutive patients presenting with AF, a prior ACS conferred higher adjusted risks of stroke, systemic embolism, all-cause mortality and CV mortality [22]. In the present study, we reported that the 3-year incidences of composite MACCEs, all-cause mortality, CV mortality reached 15.6%, 9.7% and 7.0% respectively. In the EPICOR (long-tErm follow-uP of antithrombotic management patterns In acute CORonary syndrome patients) Asia study, 6.2% patients experienced the composite endpoint of death, MI and ischemic stroke and 3.6% suffered all-cause death (including 1.3% cardiovascular-related) within 2 years. Although our analysis was from a different group of patients in EPICOR Asia, it could be predicted that ACS combined with AF had a numerically higher relative risk in long-term adverse events compared to ACS alone. Whether AF contributes to the onset of ACS or if ACS leads to AF is beyond the scope of this paper as we lack the precise information about the time of appearance of these diseases. However, previous studies observed that AF could promote inflammation that could cause a prothrombotic state and eventually coronary artery occlusion [23]. In addition, AF with high heart ventricular rates might yield symptoms of myocardial ischemia characterized by an imbalance between demand and blood supply [24]. Conversely, CAD affecting the atrial branches could result atrial scarring and remodeling to form a substrate conducive for consequent persistent AF [25]. In the past three decades, catheter ablation has evolved to a well-established treatment option for AF patients to obtain rhythm control. The safety and effectiveness of ablation in increasing freedom from recurrences and lowering AF burden during one year follow-up has been documented in multiple clinical trials [26, 27]. While the study lacked information on catheter ablation, there is no randomized controlled trial sufficiently large to properly evaluate a reduction in thromboembolic events compared with antiarrhythmic dugs [5].

The CHA${}_2$DS${}_2$-VASc score has been widely used for the assessment of thrombo-embolic risk and guiding antithrombotic therapy in AF or atrial flutter. A growing number of studies have assessed the predictive accuracy of CHA${}_2$DS${}_2$-VASc score in patients with CAD. Podolecki et al. [28] included 2647 consecutive acute myocardial infarction (AMI) patients without AF and found that the risk of ischemic stroke and all-cause death in CHA${}_2$DS${}_2$-VASc $\ge$4 increased 4-fold compared to CHA${}_2$DS${}_2$-VASc = 1. Besides, every point in CHA${}_2$DS${}_2$-VASc score was independently associated with 41% increase in stroke risk and 23% increase in mortality. Both CHADS${}_2$ and CHA${}_2$DS${}_2$-VASc scores were evaluated in a study of 929 AF patients referred for PCI. A high CHA${}_2$DS${}_2$-VASc score was predictive of all-cause mortality and MACCE at 12-months while the CHADS${}_2$ score could only predict MACCEs. CHADS${}_2$ and CHA${}_2$DS${}_2$-VASc were not associated with major bleeding [29]. These findings were further supported by another survey enrolling 13,422 ACS patients which demonstrated that a higher CHA${}_2$DS${}_2$-VASc score was associated with an increased risk of 1-year mortality after adjusting for in-hospital treatments [30]. However, the available studies mainly targeted ACS or ACS undergoing PCI as research population and there is little evidence evaluating the association between risk models and stable coronary artery disease (SCAD) patients undergoing elective PCI. Also, previous study often calculated c-statistics or utilized Kaplan-Meier curves to illustrate the predictive performance of CHA${}_2$DS${}_2$-VASc score. There are no sufficient data regarding the impact of post-discharge antithrombotic regimens as it plays an important role in long-term outcomes. In the current study, we included 433 (30.8%) SCAD patients who were eligible for elective PCI, among whom 54 suffered MACCE and 18 died within 3 years. We found that CHA${}_2$DS${}_2$-VASc score had a good predictive performance in MACCE (c-statistic: 0.677 as continuous) and all-cause death (c-statistics: 0.750 as continuous) throughout the entire range of the score. Of note, the thromboembolic risk was approximately 1.184 times greater for each point increase in the CHA${}_2$DS${}_2$-VASc score, while low and intermediate category reclassified by CHA${}_2$DS${}_2$-VASc had comparable risk of MACCEs after adjusting for covariates. These findings could not be attributed to poor predictive performance for the cutoffs were chosen on the basis of previous literature subjectively. It is the first to consider not only clinical presentations and laboratory results but also post-discharge medications as potential risk factors. Patients who experienced MACCE were less likely to be prescribed Aspirin and Ticagrelor but more frequently to receive anticoagulant therapy at discharge. Multivariable Cox regression analysis including antithrombotic regimens as covariates showed that CHA${}_2$DS${}_2$-VASc score was an independent risk factor for MACCEs and all-cause mortality. Besides, CHA${}_2$DS${}_2$-VASc $>$4 could increase the risk of MACCEs by double and all-cause death by 5 times compared to CHA${}_2$DS${}_2$-VASc equal to 1 or 2.

The GRACE ACS score was derived from an international registry of ACS patients and has been a well-recognized risk system to stratify patients according to their estimated risk of future death or ischemic events. Several studies have also tried to evaluate and compare the predictive performance of CHA${}_2$DS${}_2$-VASc and GRACE scores to determine which system is more suitable for AF patients combined with ACS or coronary stenting. Fauchier et al. [31] aimed to find out the most appropriate score among CHA${}_2$DS${}_2$-VASc, GRACE at admission, REACH [32], SYNTAX [33] and Anatomical and Clinical Syntax II Score (ACSS) [34] to use in the setting of 845 AF with coronary stenting. The results indicated that CHA${}_2$DS${}_2$-VASc was the best predictor of stroke and thromboembolic events with a c-statistics of 0.604 and SYNTAX was better to predict MACE with a c-statistics of 0.612. GRACE at admission was the best to predict all-cause mortality with a c-statistics of 0.682 [31]. Another retrospective study consisting of 1452 consecutive patients undergoing PCI with a diagnosis of AF demonstrated that the GRACE at admission but not the CHA${}_2$DS${}_2$-VASc score was associated with the incidence of MACEs within 1 year. However, the two scores showed similar predictive performance in the prediction of all-cause mortality [35]. Although the above researches came to controversial conclusions, they adopted GRACE score designed for predicting cumulative six month risk of death or MI other than the GRACE score developed for predicting post-discharge outcomes. The variables used by GRACE at discharge involved history of CHF, history of MI and in-hospital PCI suggesting that more weighting is given to chronic conditions in the post-discharge system. To the best of our knowledge, this is the first report comparing CHA${}_2$DS${}_2$-VASc, GRACE at admission and GRACE at discharge scores in the same cohort. We found that the CHA${}_2$DS${}_2$-VASc and GRACE at discharge scores showed significant prognostic values in long-term MACCEs according to multivariable analysis as well as c-statistics, while the GRACE at admission had no impact on predicting MACCEs. Notably, the prognostic ability of GRACE at discharge score could be weakened after grouping in accordance with recommendations of the GRACE system. On the other hand, GRACE at discharge demonstrated significant superiority in predicting all-cause mortality with a c-statistics of 0.846 (as continuous) to CHA${}_2$DS${}_2$-VASc score with a c-statistics of 0.750 (p 0.001). The results were consistent with the differences of purpose when the two risk scoring systems were established. The CHA${}_2$DS${}_2$-VASc score was developed for AF in stratifying high-risk patients who would sustain thromboembolic events. Nevertheless, the main outcome measured in designing GRACE at discharge score was all-cause mortality during 6 months follow-up. As discussed above, any risk score has to balance simplicity and practicality against precision. In the present study, the GRACE at discharge score should undoubtedly be advocated for evaluating the long-term survival if conditions permit. The CHA${}_2$DS${}_2$-VASc only performed modestly in predicting all-cause mortality, but it could be utilized rapidly if biomarkers or electrocardiogram information necessary for calculating the GRACE were difficult to obtain. Furthermore, it was suitable to combine CHA${}_2$DS${}_2$-VASc with GRACE at discharge to improve the accuracy of risk stratification, leading to more effective clinical decision-making and prolonged survival of AF patients with ACS or undergoing PCI.

The following were several limitations in the present study. First, this is an observational, prospective, single-center registry and has its inherent residual confounding bias. Our findings should be carefully interpreted when applied to external validation cohorts. However, we analyzed a total of 1408 AF patients with ACS or undergoing PCI and the sample size was comparable to those of similar researches. Second, we did not assess the prognostic values of the three models according to the subtypes of AF for we were unable to determine the accurate order of the presence of AF and ACS. Previous evidence suggested that only permanent AF was an independent predictor for death in AMI patients treated invasively [36]. Third, the post-discharge antithrombotic regimens were collected and treated as a covariate in our study. During 3-year follow-up, the medication adherence of participants and the possible transitions in dual-antiplatelet therapy after coronary stenting were unable to be obtained. The changes in anticoagulant or antiplatelet therapies might have significantly affected the incidence of ischemic or bleeding events. Further well-designed clinical trials are needed to compare and validate the prediction performance of several risk stratification systems for AF patients with ACS or undergoing stent implantation.

In the setting of coexistence of AF and ACS or coronary stenting, higher CHA${}_2$DS${}_2$-VASc and GRACE at discharge score were independently associated with increased risk of MACCEs and they had comparable discriminative capacities. Both CHA${}_2$DS${}_2$-VASc and GRACE at discharge scores demonstrated good prognostic values in all-cause mortality. A significant discrimination improvement of GRACE at discharge was detected compared to CHA${}_2$DS${}_2$-VASc. The GRACE at admission score could not identify patients at high risk of MACCEs. Further studies are needed to validate the clinical significance of these scores externally or help build a more accurate and practical risk score.

RM acquired the data, performed statistical analysis and drafted the manuscript. YMY designed the research and revised the manuscript. HZ and NS collected data and provided help in interpreting. JYW and SQL collected the data and helped in endpoints adjudication. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript.

The study was approved by the ethics committee of Fuwai Hospital and was conducted in accordance with the Declaration of Helsinki (No. 2017-922).

We wish to thank all individuals who agreed to participate in this study. The authors also thank the staff in Emergency Department in Fuwai Hospital for patients’ recruitment.

This research was funded by Capital’s Funds for Research and Application of Clinical Diagnosis and Treatment Technology, grant number Z171100001017193.

The authors declare no conflict of interest.